University of Southampton Research Repository
Copyright © and Moral Rights for this thesis and, where applicable, any accompanying data are retained by the author and/or other copyright owners. A copy can be downloaded for personal non-commercial research or study, without prior permission or charge. This thesis and the accompanying data cannot be reproduced or quoted extensively from without first obtaining permission in writing from the copyright holder/s. The content of the thesis and accompanying research data (where applicable) must not be changed in any way or sold commercially in any format or medium without the formal permission of the copyright holder/s.
When referring to this thesis and any accompanying data, full bibliographic details must be given, e.g.
Thesis: Author (Year of Submission) "Full thesis title", University of Southampton, name of the University Faculty or School or Department, PhD
Thesis, pagination.
Data: Author (Year) Title. URI [dataset]
UNIVERSITY OF SOUTHAMPTON
FACULTY OF SOCIAL, HUMAN AND MATHEMATICAL SCIENCES
GEOGRAPHY AND ENVIRONMENT
Social-ecological tipping points in world deltas: designing a safe and just operating space for the Chilika lagoon fishery, India
by
Gregory Stephen Cooper
Thesis for the degree of Doctor of Philosophy
March 2018
UNIVERSITY OF SOUTHAMPTON ABSTRACT
FACULTY OF SOCIAL, HUMAN AND MATHEMATICAL SCIENCES Geography and Environment Thesis for the degree of Doctor of Philosophy
SOCIAL-ECOLOGICAL TIPPING POINTS IN WORLD DELTAS: DESIGNING A SAFE AND JUST OPERATING SPACE FOR THE CHILIKA LAGOON FISHERY, INDIA
Gregory Stephen Cooper
Systems of feedbacks, delays and cross-scale interactions can undergo surprising dynamics. Climate change and globalisation are magnifying social-ecological complexities across regional systems, questioning policies that prioritise stability over variability, predictability over adaptability, and optimisation over persistence. Similar concerns extend to modelling techniques that explore a limited number of scenarios framed by stationary external conditions and an absence of human- natural feedbacks. Such methods are ill-equipped to identify safe and just operating spaces for sustainable development, characterised by interacting environmental limits, tipping points and regime shifts. To this end, this study develops and evaluates a system dynamics model to identify the safe and just operating spaces of a natural resource system with a legacy of collapse. Based on the Chilika lagoon fishery of the Mahanadi delta, India, the model explores how decision-makers can influence internal resilience to a spectrum of plausible driver trajectories and interactions.
The principal contribution of this study is the operationalisation of the safe and just spaces concept as a forward-looking tool to identify interacting pathways to sustainable futures. Specific to Chilika, periodically dredging the tidal outlet desensitises the fishery to the hydroclimatic processes causing collapse under do nothing governance. However, stable resource availability facilitates fishing effort growth that can trigger overexploitation by 2050. Amongst a suite of social- ecological trade-offs, fishing bans and alternative livelihoods widen the safe spaces but require decision-makers to forgo Chilika’s common-pool status. Normative safe spaces are found to have properties of social-ecological resilience, including latitude, resistance and precariousness. These characteristics help identify “core” safe and just spaces, representing interacting trajectories with the highest chances of reaching the sustainable future. Contrastingly, futures of fishery overcapacity and livelihood loss associate with deeply undesirable dynamics, including ecological surprise, tipping points and hysteresis. This study is transferable to social-ecological system of stocks and flows, feedbacks and future uncertainty, highlighting considerations for how we view sustainability and shape regional systems to avoid boundaries of safe and just spaces.
i
Table of Contents
Table of Contents ...... ii List of Tables ...... ix List of Figures ...... xi Academic Thesis: Declaration Of Authorship ...... xxiii Acknowledgements ...... xxv Definitions and Abbreviations ...... xxvi Chapter 1: Introduction ...... 1
1.1 Background and rationale ...... 1 1.2 Thesis structure ...... 3
1.2.1 Literature review ...... 3 1.2.2 Model construction, evaluation and plausible futures ...... 4 1.2.3 Model results ...... 4 1.2.4 Discussion ...... 5 1.2.5 Conclusion ...... 5
Chapter 2: Literature review ...... 7
2.1 Complex social-ecological systems ...... 7
2.1.1 Social, ecological and social-ecological systems ...... 7 2.1.2 Characteristics of complex social-ecological systems ...... 10
2.2 Social-ecological sustainability, resilience and adaptability ...... 14
2.2.1 Concepts of sustainability and resilience ...... 14 2.2.2 Beyond definitions: successes, traps and adaptability ...... 18 2.2.3 Plausible and desirable social-ecological futures ...... 21
2.3 Social-ecological regime shifts ...... 23
2.3.1 Thresholds, tipping points and regime shifts ...... 23 2.3.2 Threshold detection methods ...... 26 2.3.3 Modelling social-ecological systems and thresholds ...... 29 2.3.4 Putting theory into practice: threshold-based management ... 34
2.4 Deltaic social-ecological systems ...... 36
2.4.1 Concepts of deltaic sustainability ...... 36
ii
2.4.2 ‘Safe and just’ futures for deltaic and wider social-ecological systems ...... 40
2.5 The Chilika lagoon social-ecological system ...... 43
2.5.1 Biophysical settings ...... 44 2.5.2 Anthropogenic settings ...... 45 2.5.3 Social-ecological dynamics ...... 49
2.6 Aim and objectives ...... 52
Chapter 3: A system dynamics model of the Chilika lagoon fishery system ...... 55
3.1 Chapter introduction ...... 55 3.2 Conceptual modelling ...... 58
3.2.1 Fundamental reference modes ...... 58 3.2.2 Historical and plausible future reference modes ...... 60 3.2.3 Setting system boundaries ...... 62
3.2.3.1 Natural processes ...... 63 3.2.3.2 Anthropogenic processes ...... 67
3.3 Interviews ...... 70
3.3.1 Key informants ...... 71 3.3.2 Fishers ...... 73
3.4 Module parameterisation ...... 75
3.4.1 Hydroclimatic ...... 76
3.4.1.1 Lower Mahanadi freshwater yield ...... 76 3.4.1.2 Evaporative losses ...... 80 3.4.1.3 Additional removals ...... 80 3.4.1.4 Flow routing at Naraj ...... 81 3.4.1.5 Western catchments ...... 81 3.4.1.6 Future climatic inputs ...... 81 3.4.1.7 Fluvial sediment yields ...... 83 3.4.1.8 Chilika’s sediment stock ...... 84 3.4.1.9 Outlet closure ...... 85
3.4.2 Ecohydrological ...... 86
iii
3.4.2.1 Salinity ...... 87 3.4.2.2 Dissolved oxygen ...... 90 3.4.2.3 Water temperature ...... 91 3.4.2.4 Aquatic vegetation coverage ...... 92 3.4.2.5 Habitat quality ...... 94
3.4.3 Core fishery ...... 99 3.4.4 Socioeconomic ...... 103
3.4.4.1 Traditional fishers ...... 104 3.4.4.2 Non-traditional fishers ...... 107
3.4.5 Statement of circularity ...... 108
3.5 Chapter conclusion ...... 109
Chapter 4: Model performance analysis ...... 111
4.1 Chapter introduction ...... 111 4.2 Non-behavioural analysis ...... 111
4.2.1 Unit consistency ...... 112 4.2.2 Qualitative parameter assessment ...... 112 4.2.3 Qualitative extreme conditions ...... 113
4.3 Behavioural analysis ...... 114
4.3.1 Quantitative extreme conditions ...... 114 4.3.2 Behavioural reproduction ...... 117 4.3.3 Time-step dependency ...... 123 4.3.4 Structural analysis through alternative pasts ...... 124 4.3.5 One-at-a-time sensitivity analysis ...... 127 4.3.6 Multivariable sensitivity analysis ...... 130
4.3.6.1 Introduction and purpose ...... 130 4.3.6.2 Parameter selection, sampling technique and behaviour metrics ...... 131 4.3.6.3 Global sensitivity analysis outputs ...... 135 4.3.6.4 Isolating critically sensitive drivers ...... 139
4.4 Chapter conclusion ...... 142
Chapter 5: Plausible future scenarios and normative definitions ...... 145
iv
5.1 Chapter introduction ...... 145 5.2 Exploratory external scenarios ...... 145 5.3 Internal governance scenarios ...... 151 5.4 The safety of plausible futures ...... 153 5.5 Chapter conclusion ...... 157
Chapter 6: Model outputs – future outcomes, safe and just operating spaces, and social-ecological complexities ...... 159
6.1 Chapter introduction and structure ...... 159 6.2 Plausible future catch ...... 160 6.3 ‘No governance’ scenario ...... 164 6.4 Probabilistic resilience analysis ...... 166 6.5 Future driver trajectories ...... 170
6.5.1 External drivers ...... 171 6.5.2 Internal drivers: anthropogenic ...... 173 6.5.3 Internal drivers: biophysical ...... 175
6.6 Driver based safe and just operating spaces ...... 177
6.6.1 Design and visualisation ...... 177 6.6.2 External drivers ...... 180 6.6.3 Internal drivers: anthropogenic ...... 182 6.6.4 Internal drivers: biophysical ...... 185 6.6.5 Safe and just space interactions ...... 188
6.7 Core safe and just spaces ...... 191 6.8 Wider social-ecological outcomes ...... 195
6.8.1 Rationale and indicators ...... 195 6.8.2 Future dynamics ...... 196 6.8.3 Outcome trade-offs ...... 198
6.9 The sustainability of contemporary trajectories ...... 201 6.10 Driver-response relationships ...... 204
6.10.1 Driver-response nonlinearities ...... 205 6.10.2 Chilika: a hysteretic system?...... 206
6.11 Defining ‘safe and just’ by catch predictability ...... 213
6.11.1 Rationale and approach ...... 213
v
6.11.2 Outcome trajectories ...... 216 6.11.3 Alternative driver thresholds ...... 218
Chapter 7: Discussion ...... 223
7.1 Chapter introduction ...... 223 7.2 Future of the Chilika lagoon ...... 225
7.2.1 Through the system archetype lens ...... 225 7.2.2 Implications for users and decision-makers ...... 227 7.2.3 Implications for the wider system ...... 231 7.2.4 Subchapter synthesis ...... 233
7.3 Governing regional social-ecological systems ...... 234
7.3.1 Systems of leverage points ...... 234 7.3.2 Sustainable fishery governance ...... 236 7.3.3 The roles of resilience, adaptability and transformability ..... 238 7.3.4 Subchapter synthesis ...... 241
7.4 Undesirable shifts in social-ecological systems ...... 242
7.4.1 Thresholds, tipping points and regime shifts ...... 242 7.4.2 Operationalising threshold-based management through safe and just operating spaces ...... 246 7.4.3 How just are safe and just operating spaces? ...... 249 7.4.4 Subchapter synthesis ...... 253
7.5 Modelling social-ecological systems ...... 253
7.5.1 Using models as satellite-navigation systems ...... 254 7.5.2 Evaluating the role of system dynamics modelling in sustainability science ...... 256 7.5.3 Study transferability ...... 260 7.5.4 Subchapter synthesis ...... 264
7.6 Model uncertainties, multiple worldviews and indeterminacy ..... 265
7.6.1 Uncertainties and limitations that we roughly know ...... 266 7.6.2 We do not know if we know the effects of excluded processes268 7.6.3 A brief word on indeterminacy ...... 271 7.6.4 Subchapter synthesis ...... 272
vi
Chapter 8: Conclusions and future research ...... 273
8.1 Five-point synthesis ...... 273 8.2 Future research directions ...... 279 8.3 Closing thoughts ...... 281
Appendix A – Interview information sheet and question matrices .. 284
A.1 Participant information sheet ...... 284 A.2 Key informant ...... 287 A.3 Stakeholder ...... 289
Appendix B – Model equations and constants ...... 290 Appendix C – Model evaluation ...... 309
C.1 Qualitative parameter assessment ...... 309 C.2 Quantitative extreme scenarios ...... 312 C.3 External trajectory combinations ...... 313 C.4 Early model evaluation ...... 314
Appendix D – Model outputs ...... 315
D.1 Colour-blind friendly graphs ...... 315
D.1.1 Normative catch ...... 315 D.1.2 Catch predictability ...... 316 D.1.3 Driver pathways ...... 317 D.1.4 Wider social-ecological outcomes ...... 320
D.2 Mean and variability of catch pathways ...... 321 D.3 Safe and just spaces correlation analysis ...... 322 D.4 Wider core spaces ...... 325 D.5 Autocorrelation and partial autocorrelation functions ...... 326
List of References ...... 331
vii
List of Tables
Table 4.1: Scenarios used to assess model integrity under extreme conditions. Outputs: 1- outlet closure; 2- salinity; 3- macrophyte coverage; 4- catch; 5- fisher population...... 115
Table 4.2: Characteristics and performance metrics of important modelled drivers and outputs, alongside respective observed averages. MAPE – mean absolute percentage error; PE – percentage error...... 118
Table 4.3: One-at-a-time sensitivities of critical outputs to 17 social and ecological drivers. Highlighted cells represent critically sensitive driver- response relationships, where generalised sensitivity >0.1. ... 128
Table 4.4: The 25 model variables given Monte Carlo error magnitudes alongside a summation of their model process. See figure 3.11 for hydroclimatic abbreviations...... 132
Table 4.5: Outputs depicting the separation between the error distributions producing catch within the corridor of constraints and those causing catches outside. See figure 3.11 for the hydrological acronyms. Significance levels: ‘1’ – p < 0.01; ‘2’ – 0.01 ≤ p ≤ 0.1; ‘3’ – p > 0.1. K-S – Kolmogorov-Smirnov...... 137
Table 5.1: The plausible trajectory ranges of the five external drivers used to drive the model’s internal dynamics into future; current values derive from the baseline period 1986-2005...... 147
Table 6.1: Summary statistics of the annual catch time-series per governance scenario for the period 2015-2060, unless otherwise stated. See chapter 5.3 for governance scenarios...... 161
Table 6.2: The 95% confidence intervals of the external SJOSs, alongside Kolmogorov-Smirnov statistics (D) signifying the degree of separation between the safe and the dangerous spaces. Significance (S) levels: ‘1’ – p < 0.01; ‘2’ – 0.01 ≤ p ≤ 0.1; ‘ 3’ – p > 0.1. Dashed cells are where distinct SJOSs are undefinable...... 180
ix
Table 6.3: The 95% confidence bounds of the internal socioeconomic SJOSs, alongside Kolmogorov-Smirnov statistics (D) signifying the degree of separation between the safe and the dangerous spaces. Significance (S) levels are as per table 6.2...... 183
Table 6.4: The 95% confidence bounds of the ecohydrological SJOSs, alongside Kolmogorov-Smirnov statistics (D) signifying the degree of separation between the safe and the dangerous spaces. Significance (S) levels are as per table 6.2...... 187
Table 6.5 The proportions of different future states within the deterministic funnels of the four socioeconomic drivers (figure 6.19). The risk of a driver exceeding its safe operating space is inversely proportional to the percentage of safe and just outcomes within the funnel...... 203
x
List of Figures
Figure 2.1: Aggregated version of the social-ecological systems framework describing resource units, users and nested governance structures within SESs (Ostrom et al., 2009). Social, economic and political settings influence contemporary systems through legacy effects, whilst related ecosystems may provide external stresses. Reprinted with permission from the American Association for the Advancement of Science (AAAS)...... 9
Figure 2.2: Stability landscapes depicting the two regime shifts of Scheffer et al. (2001): (a) the system appears resilient to stresses (D), remaining within the desirable basin when the maximum stress (C ) occurs; max (b) the system undergoes parameter changes (∆P) causing normal stresses (D ) to induce state changes; (c) parameters remain norm constant, but the system is forced beyond its desirable basin. 24
Figure 2.3: The deltaic domains (black) of Brondizio et al. (2016) arranged as a conceptual model and labelled with cross-domain interactions (blue)...... 39
Figure 2.4: A sequential approach to operationalise regional SJOSs: (A) construct a model to investigate problematic behaviors; (B) simulate future dynamics under arrays of external driver trajectories; (C) compare outcome dynamics to plausible normative futures relative to today’s norms (Dearing et al., 2014), like safe and just (green), cautionary (orange) and dangerous (red) outputs; (D) trace outcomes to drivers to depict permissible pathways...... 42
Figure 2.5: Annual fish catch data from Chilika covering 1929-2015. Source: Iwasaki and Shaw (2009) and CDA (2016)...... 46
Figure 2.6: Chilika’s historical catch per unit boat (top) and catch per unit fisher (bottom) from 1958-2009. Source: Kumar and Pattnaik (2012).49
Figure 3.1: The overarching methodological framework of this study – from initial problem identification to communicating insights to regional decision-makers...... 56
xi
Figure 3.2: Causal-loop diagram depicting feedbacks relating the fish stock size, fishing effort and fish catch. Short-term catch may increase in response to intensified fishing effort but cause long-term stock depletion. Two horizontal lines represent a system delay longer than the average interaction...... 58
Figure 3.3: Plausible reference modes for Chilika’s fishery: (A) catch exponentially increases due to infinite fish stock; (B) stock exponentially grows or declines if the extraction rate is respectively above (red) or below (green) sustainable (blue); (C) boom and bust dynamics; (D) boom and bust influenced by variable catchability...... 59
Figure 3.4: Potential futures resulting from the management of fundamental components and feedbacks (figure 3.2). The question mark refers to uncertainty associated with the timing of future resource decline...... 61
Figure 3.5: Causal-loop diagram illustrating the components, links and feedbacks underlying the fish stock. Italicised variables refer to stresses external to the fish population. Feedbacks: R – reinforcing, B – balancing...... 64
Figure 3.6: Ecohydrological influences on the aggregated fish population of Chilika. Variables such as ‘total sediment influx’ consisting of both LMC and WC sediment loads are aggregated for visual clarity. Dashed lines are for visual clarity. The feedback directions are as figure 3.5...... 65
Figure 3.7: Observed fisher population (red) (Noack et al., 2013) versus the trend which may have occurred if the link between fishery productivity and the fisher population was removed, making the population only driven by Odisha’s birth and death rates (black)...... 68
Figure 3.8: (A) Causal-loop diagram illustrating drivers of fishing boat numbers and days fished; (B) secondary diagram underlying the transition from traditional fishing practices to more intensive non-traditional practices. The feedback directions are as figure 3.5 ...... 69
Figure 3.9: The seven key informants interviewed placed within the social- ecological systems framework structure of Ostrom (2009)...... 72
xii
Figure 3.10: Locations of the four fishing villages where stakeholder interviews were conducted. Chilika’s sectors: NS - northern sector; CS - central sector; SS – southern sector; OC – outer channel...... 74
Figure 3.11: Stock and flow structure driving LMC rainfall-runoff generation. Acronyms correspond to the runoff production and routing variables of Ghashgaei et al. (2013): Rc – runoff coefficient; Tg – groundwater delay; Bc – baseflow contribution; Co – proportion of discharge removed; Te – evaporation delay...... 77
Figure 3.12: The spatial extent of district-wide rainfall and temperature datasets relative to the Tikarapara gauging station and Chilika lagoon. 78
Figure 3.13: Time-series comparing the effects of a static (blue) and dynamic (red) rainfall-runoff coefficient against the observed LMC freshwater yield (black) over the calibration period...... 79
Figure 3.14: The empirical cumulative distribution functions used to derive future December (blue) and April (red) rainfall across the LMC. A random number is generated between 0 and 1 (x-axis), which corresponds to a given monthly rainfall total (y-axis)...... 82
Figure 3.15: The nonlinear relationship between observed sediment flow rates, derived from the function of Mishra and Jena (2015), and observed sediment concentrations obtained from CWC data. The logarithmic trend (r2 = 0.50, df = 118, p < 0.05) forms the graphical function converting between sediment rate and yield. The grey shading represents the 95% confidence interval of the regression model.84
Figure 3.16: Stock and flow diagram depicting the inputs and outputs driving the sediment stock dynamics. Feedback labels are as per figure 3.5.85
Figure 3.17: The seasonal trend of seasonal lagoon-wide salinity from pre- monsoon 2002/2003 to post-monsoon 2009/2010...... 88
Figure 3.18: Functions converting (A) monthly monsoon and post-monsoon and (B) pre-monsoon freshwater deliveries to deviations from the annual mean salinity. The deviations insert into equations 3.8a and 3.8b...... 89
Figure 3.19: Seasonal lagoon-wide dissolved oxygen values from pre-monsoon 2002/2003 to post-monsoon 2009/2010...... 91
xiii
Figure 3.20: Time-series coherence between: (i) observed surface air temperatures (black) at Chandraput on the western bank of Chilika (CDA, n.d-a) ; (ii) averaged water temperatures (blue) observed from 30 ecohydrological monitoring stations uniformly distributed across Chilika (CDA, n.d-c); (iii) modelled water temperatures (orange)...... 92
Figure 3.21: Modelled and observed aquatic macrophyte coverage from 1973- 2009. Observations used in model calibration are grey...... 94
Figure 3.22: Graphical functions relating (a) salinity (b) dissolved oxygen and (c) water temperature to the rate of fish survival per time-step. .... 96
Figure 3.23: The fish survival parameter is a function of the underlying population density and level of ecosystem degradation. In essence, fish survival for a given density declines from default as ecological degradation (fresher, deoxygenated and warmer waters) increases...... 97
Figure 3.24: Modelled fish regeneration and total fish population in response to the hypothetical degradation scenario. Other stresses are switched- off (e.g. catch and migration) to isolate the effects of ecological degradation on population change...... 98
Figure 3.25: Graphical functions determining the number of days fished each month as a function of the perceived fish density, based on the type-II functional response (Holling, 1959b)...... 102
Figure 3.26: Comparison of the observed fisher population (black) (Noack et al., 2013) against the modelled fisher population with a permanent carrying capacity (red)...... 105
Figure 3.27: Graphical function determining the proportion of traditional fishers that can afford upgrades each month. The function output feeds into equation 3.20A...... 106
Figure 4.1: System responses to (A) extreme high (red) and low (blue) rainfall inputs: (B) outlet closure expressed as a percentage of the lagoon’s sediment capacity; (C) lagoon-wide average salinity; (D) macrophyte coverage; (E) catch and (F) active fisher population...... 116
xiv
Figure 4.2: Modelled (non-black) versus observed (black) driver time-series. (A) Annual average lagoon-wide salinity (B) annual catch per unit boat; (C) active fisher population and (D) annual per capita income.120
Figure 4.3: Modelled (red solid) versus observed catch (black dashed) from 1973- 2009, overlain with modelled (red dotted) and observed (black long-dash) time-series breakpoints...... 122
Figure 4.4: Time-series depicting the difference between fish catch outputs from default (red) and halved (blue) time-steps against observed catch (black)...... 123
Figure 4.5: (top) time-series depicting seven alternative pasts, each derived by the removal of one system stress; (bottom) time-series calculated by the cumulative sum of the absolute errors between the annual catches of the alternative pasts and the reference mode...... 125
Figure 4.6: Observed catch (black) and its upper and lower 99% confidence intervals (red). Observed catch is smoothed (n=3) to remove short- term volatility and accentuate the long-term trend...... 134
Figure 4.7: The 2000 catch time-series resulting from randomly perturbing 25 model variables by ±50% of their parameterised values. Prop.inside – proportion of annual catches inside the constraint corridor. 136
Figure 4.8: (A-D) density plots depicting the relative frequency of runs inside and outside the constraint corridor across different error magnitudes; (E) an example of a driver with insignificant difference between error magnitudes producing runs inside and outside the constraint corridor...... 138
Figure 4.9: Graphical outputs from sensitivity analysis where the six critically sensitive variables are simulated under four error ranges. Observations (black) are plotted against model generated time- series of four social-ecological processes: (A) salinity; (B) macrophyte coverage; (C) catch (99% confidence intervals) (D) catch per unit boat (99% confidence intervals)...... 140
Figure 5.1: The five external funnels generating dynamics until 2060: (A) mean annual near-surface air temperature across the Lower Mahanadi and Western catchments; (B) mean annual rainfall across the same
xv
areas; (C) fish price; (D) diesel price and (E) net of birth and death rates (‘r’)...... 148
Figure 5.2: Histograms of the five external trajectories, plotted in the same order as figure 5.1. The socioeconomic trajectories are described by annual gradients from 2015-2060; the climate variables are described by their 2050-2060 annual average relative to the baseline...... 150
Figure 5.3: Schematic representation of plausible catches resulting from the spectrums of internal and external dynamics. The challenge here is to understand which futures are safe and just, and in turn, discover the causal pathways and interactions to guide system governance...... 155
Figure 6.1: Catch time-series generated under the four governance scenarios (colour), plotted alongside pre-2015 catch values (black). The pie charts depict the proportions of different normative futures per governance scenario. Green – safe and just, orange – cautionary, and red – dangerous. See appendix E for colour-blind friendly version...... 162
Figure 6.2: Relationships between the trajectories of (A-C) natural and (D-F) human drivers and the dates of catch disappearance under NO governance. All degrees of freedom equal 998...... 165
Figure 6.3: The decadal evolution of stability landscapes under OM, OB and OL. Catch between MSY estimates (green) is symptomatic of a resilient fishery under the range of human and natural driver trajectories. Note: the scale of the y-axis across the top panel is different to the other five due to the high density of MSY catches...... 167
Figure 6.4: Time-series depicting the annual probabilistic resilience score (the proportion of catches inside maximum sustainable yield estimates) under OM, OB and OL governance...... 169
Figure 6.5: External trajectories under OM, OB and OL governance, classified as safe and just (green), cautionary (orange) or dangerous (red) depending on their normative future from 2050-2060 (figure 6.1). Climate trajectories are plotted as decadal anomalies from the
xvi
1986-2005 baseline (blue dashed). Historical – black. See appendix E for colour-blind friendly versions...... 172
Figure 6.6: Internal fishing effort trajectories driving fish catch under OM, OB and OL governance. Time-series colours are the same as figure 6.5. See appendix D.1.3 for colour-blind friendly version...... 173
Figure 6.7: Chilika’s ecohydrological trajectories classified as safe and just, cautionary or dangerous, depending on their respective fishery outcomes. Time-series colours are as per figure 6.5...... 176
Figure 6.8: (A) SJOSs of current outcome values relative to environmental limits and social foundations, represented by single threshold values (Dearing et al., 2014, p.233). (B) driver based thresholds defined here by associations between driver trajectories (horizontal axis) and normative outcome states (vertical axes); soft thresholds (green line) are due to multivariable interactions causing a driver trajectory to associate with multiple outcome states. See chapter 6.7 for core SJOS explanation...... 178
Figure 6.9: Conditional probabilities relating the magnitudes of external drivers to the proportion of ‘safe and just’ (solid green), ‘cautionary’ (dashed orange) and ‘dangerous’ (dot-dash red) futures from 2050- 2060...... 181
Figure 6.10: Conditional probabilities relating the magnitudes of four internal socioeconomic drivers to the proportion of ‘safe and just’, ‘cautionary’ and ‘dangerous’ futures from 2050-2060. Colours and line types are as per figure 6.9...... 184
Figure 6.11 Conditional probabilities relating the magnitudes of internal ecohydrological drivers to the proportion of ‘safe and just’, ‘cautionary’ and ‘dangerous’ futures from 2050-2060. Colours and line types are as per figure 6.9...... 186
Figure 6.12: Hypothetical interactions between two SJOSs, whereby the safety of ‘driver a’ is conditional on the safety of ‘driver b’. In actuality, SJOSs exist within hyperdimensional space (e.g. more than three dimensions), but two dimensions are shown here for visual clarity...... 188
xvii
Figure 6.13: The strongest cross-scale SJOS interactions under (A) OM and (B) OB governance scenarios, suggesting system resilience to local drivers is partly conditioned by external processes . See appendix D.3.1 for correlation coefficients...... 190
Figure 6.14: The core SJOS trajectories of the system drivers found to have statistically distinct sustainable and dangerous configurations (chapter 6.6). Key: black – core SJOS, green – wider SJOS...... 192
Figure 6.15: The negative correlations between external and internal drivers inside the core space suggests that the system must follow multidimensional pathways where the permissible number of motorboats is conditioned by the decadal magnitudes of (left) rainfall change and (right) freshwater delivery change...... 194
Figure 6.16: Time-series of five wider social-ecological outcomes under OM, OB and OL governance (chapter 5.3). Key: dark green – ‘core safe and just’, green – 'safe and just', orange – 'cautionary', red – ‘dangerous’. See appendix D.1.4 for colour-blind friendly version...... 197
Figure 6.17: The negative associations between short-term and mid-century catch under (A) OM, (B) OB and (C) OL. The GLMs (black lines) account for the variance of all 1000 points per plot. Coefficients of determination: r2 = 0.33, r2 = 0.34, r2 = 0.10 (all p < 0.05, df = OM OB OL 998). Colours are as per figure 6.16...... 199
Figure 6.18: Time-series charting cumulative annual fishery value under (A) OM and (B) OL governance. Colours are as per figure 6.17...... 200
Figure 6.19: The deterministic trajectory funnels of four critical drivers of fish catch (2015-2050), created by linearly extrapolating their 5 (dot- dash) and 35 year (dashed) trends...... 202
Figure 6.20: Phase spaces comparing the driver-response trajectories of safe and dangerous futures under OM and OB. Total fish population is compared with active fishers (top) and motorboat numbers (bottom)...... 206
Figure 6.21: Phase spaces exploring the coupled relationships between fish population and (A) the active fisher population and (B) the number
xviii
of motorboats. The forward switch covers until 2060; the reverse switch covers 2061-2100...... 207
Figure 6.22: Time-series of (A) motorboats, (B) active fishers, (C) catch and (D) fish population (percentage of average 2015 population) classified by the four types of driver-response relationships. The bottom panel shows the phase space portraits of the fish population against (E) motorboats and (F) active fishers...... 209
Figure 6.23: (Main) three-dimensional phase space depicting the number of motorboats, fish population and the year in which the fish population recovers to 90% of its 2015 value; the forward shift has been removed for visual clarity. (Inset) the negative relationship between the duration of recovery and the reversal magnitude (difference between the number of motorboats in the year of recovery and 2060)...... 211
Figure 6.24 Driver and outcome time-series of trapped simulations (colour), plotted for reference on top of the wider spectrum of dangerous futures (grey)...... 212
Figure 6.25: A method to classify the safety of outcomes based on ARIMA forecasting: (A) catch exiting (orange dot) and remaining outside the envelope (green and blue dashed) are unsafe. ARIMA models (grey fill – 95% confidence interval, blue fill – 80% confidence interval) are then projected for each unsafe trajectory from the year after envelope exit. (B) Catch is classified as ‘cautionary’ if the next five records (red) fall inside the 95% confidence intervals of the ARIMA forecast (predictable), but (C) ‘dangerous’ if catch falls outside the forecast at least once (unpredictable)...... 215
Figure 6.26: Future catch under business-as-usual OM governance plotted alongside pre-2015 catch values (black). Whether catch is ‘safe and just’ (green), ‘cautionary’ (orange) or ‘dangerous’ (red) is determined by the predictability of envelope departure. See appendix D.1 for colour-blind friendly version...... 216
Figure 6.27: Temporal distributions of predictable (orange) and unpredictable (red) departures from the historical envelope...... 217
xix
Figure 6.28: The landscape of predictable envelope departures for five socioeconomic drivers (A-E) and one ecological driver of catch (F). Two-dimensional density is proportional to the frequency of thresholds for a given driver trajectory at a given time. The salinity threshold landscape reflects the simultaneously occurring human- induced thresholds...... 220
Figure 6.29: The landscape of dangerous envelope departures for the six drivers depicted in figure 6.28...... 221
Figure 7.1: Causal-loop diagram depicting the ‘success to the successful’ archetype found to drive unsustainable fishery intensification over the next decade; adapted from Senge (1990). Feedbacks: R – reinforcing, B – balancing...... 225
Figure 7.2: Causal-loop diagram of the ‘shifting the burden’ archetype, whereby periodic tidal outlet opening promotes fishery intensification until catch surpasses the sustainable limit; adapted from Senge (1990). Feedbacks: R – reinforcing, B – balancing...... 226
Figure 7.3: Phase space diagram depicting the trajectories of safe and just (solid green), cautionary (dotted orange) and dangerous (dashed red) futures across the wise use matrix of Kumar et al. (2011). .... 228
Figure 7.4: Qualitative scenarios from today’s window of opportunity summarising the adaptability, transformability and stability landscapes of the plausible futures...... 240
Figure 7.5: Shift from a desirable to an undesirable basin under OM and OB governance: (A) a resilient system varies within its desirable basin during unremarkable disturbances (D ); (B) outcome variability norm increases with stress magnitudes, pushing the system into non- analogous regions of the basin; (C) slow parameter changes encourage basin shallowing, reducing outcome resistance to internal variability...... 243
Figure 7.6: Hysteretic trajectories coloured by a composite metric summing the standardised temporal and driver reversal dimensions (figure 6.23– inset). Red colours represent trajectories where the number of
xx
motorboats are reversed relatively rapidly per unit time. The forward shift is removed for visual clarity...... 245
Figure 7.7: (Top) annual fish catch and (bottom) motorboat use ranked by the average per capita income (INR/year) associated with each simulation. Please refer to chapter 6 for historical dynamics. 252
Figure 7.8: The probabilistic resilience landscape of OM governance involving all outputs from 2015-2060, as per the method of Bitterman and Bennett (2016), rather than the annual method used in chapter 6.4...... 256
Figure 7.9: The 1000 future catch simulations of OM and OL governance scaled by z-score standardization: 푧 = 푥 − 휇휃 , where 푥 equals annual catch in tonnes, 휇 equals the mean of all catches and 휃 equals the standard deviation of all catches...... 260
Figure 7.10: Time-series of (A) future lake depth under the spectrum of (B-D) precipitation, evaporation and export changes; (E-G) the driver based SJOSs depicting the proportion of future states after 50 years under the external trajectories. Colours are as per figure 6.11...... 263
xxi
Academic Thesis: Declaration Of Authorship
I, Gregory Cooper declare that this thesis and the work presented in it are my own and has been generated by me as the result of my own original research.
Social-ecological tipping points in world deltas: designing a safe and just operating space for the Chilika lagoon fishery, India
I confirm that:
1. This work was done wholly or mainly while in candidature for a research degree at this University; 2. Where any part of this thesis has previously been submitted for a degree or any other qualification at this University or any other institution, this has been clearly stated; 3. Where I have consulted the published work of others, this is always clearly attributed; 4. Where I have quoted from the work of others, the source is always given. With the exception of such quotations, this thesis is entirely my own work; 5. I have acknowledged all main sources of help; 6. Where the thesis is based on work done by myself jointly with others, I have made clear exactly what was done by others and what I have contributed myself; 7. None of this work has been published before submission
Signed: ......
Date: ......
xxiii
Acknowledgements
My sincerest thank you to Professor John Dearing and Professor Stephen Darby; without your knowledge and continuous encouragement, this work would not have been possible. As a supervisory team you have dedicated the perfect amount of support to challenge and inspire, whilst allowing enough space for me to pursue my interests. A special thank you to you John for our frequent meetings, entrusting me with this research and originally sparking my interest in social- ecological systems back in 2014. I hope this work reflects the many hours you have spent guiding, reviewing and sharing novel ideas.
Thank you to my funders: Department of Geography and Environment, University of Southampton; UK Government’s Department for International Development (DFID); Canadian International Development Research Centre (IDRC), and the DEltas, vulnerability and Climate Change: Migration and Adaptation (DECCMA) project. An extra thank you to Jon Lawn for being the go-to-guy for anything travel/funding/DECCMA related. I have thoroughly enjoyed my time as a DECCMA PhD associate and I hope this work does the DECCMA name justice in depth, breadth and insight.
Thank you to the DECCMA India team for helping organise my visit in spring 2016. Special gratitude is extended to Dr AK Pattnaik, Dr RN Samal and SK Mohanty for hosting me in Odisha, as well as sharing invaluable knowledge over the course of my study. Thank you also to the Chilika lagoon Wetland Research and Training Centre for enriching my trip both academically and culturally.
Thank you to my family and friends who have supported me throughout this project… and a special shout out to the Broadlands Boys! A big thank you to my Father Brian, Alistair, Sophie, Phil, Mary and Alexander for all of your help and encouragement over the past three years. A special thank you to my Mother Julia for always being a willing soundboard no matter the hour or achievability of the idea/task. An extra special thank you to my girlfriend Georgie for your inspiration and encouragement every day.
Finally, thank you to Shackleton Geographicals Football Club. We have had some great times being promoted, relegated and promoted (again) together! Up the Shack!
xxv
Definitions and Abbreviations
ABM Agent-based Modelling ARIMA Autoregressive Moving Average CDA Chilika Development Authority CIFRI Central Inland Fisheries Research Institute CPUB/E/F Catch Per Unit Boat/Effort/Fisher SES Social-Ecological System CWC Indian Central Water Commission DECCMA Deltas, Vulnerability and Climate Change: Migration and Adaptation DF Degrees of freedom (statistical) EWS Early Warning Signals FAO Food and Agriculture Organization of the United Nations GLM Generalised Linear Model ICZMP Integrated Coastal Zone Management Project IMD Indian Meteorological Department INR India Rupee IPCC Intergovernmental Panel on Climate Change LMC Lower Mahanadi Catchment MAPM Monsoon and Post-Monsoon months MCA Monte Carlo Analysis MSY Maximum Sustainable Yield OMFRA Odisha Marine Fishing Regulations Act and Rules PPM Parts Per Million PPT Parts Per Thousand S(J)OS Safe (and Just) Operating Space SDM System Dynamics Model TBM Threshold-based Management UN United Nations WC Western Catchments
xxvi Chapter 1: Introduction
Chapter 1: Introduction
1.1 Background and rationale
The Earth system, comprising interacting physical, chemical and biological processes, is thought to exist within a no-analogue state (Steffen et al., 2004). Atmospheric carbon dioxide levels are higher than any other time in the past 800,000 years (IPCC, 2014a). Combined earth and ocean temperatures have warmed by 1°C since 1880 (IPCC, 2014a) and the rate of species extinction is up to 1000 times Earth’s historical baseline (Mace et al., 2005). Termed ‘the Anthropocene’ (Crutzen and Stoermer, 2000), biophysical novelties are entangled with the emergence and acceleration of humanity’s dominance of the Earth system (Vitousek et al., 1997, Steffen et al., 2015a). Humans have transformed up to 50% of Earth’s natural landcover and exploited over 50% of all accessible freshwater (Vitousek et al., 1997). Earth’s human population eclipsed 7 billion in 2011, and even though global gross domestic product increased six fold from 1950 to 2010 (Steffen et al., 2015a), the richest 10% hold 57% of income (Raworth, 2012).
The global challenges of climate change, population growth and inequity have called for Earth system governance (Biermann, 2007, Biermann, 2012). Despite various successes, including the 1987 Montreal Protocol that banned the production of ozone depleting substances, multilateral organisations such as the United Nations (UN) and Intergovernmental Panel on Climate Change (IPCC) are often thought to mismatch the scale of regional systems and decisions (Gunderson and Holling, 2002, Glaser and Glaeser, 2014). The promotion of aquaculture across Southeast Asia as a means of poverty reduction has exposed previously artisanal fisheries to international consumption demands, exploitative pricing and aquatic diseases (Berkes et al., 2006, OECD, 2010). Therefore, governance of ‘vulnerability hotspots’ (Nicholls et al., 2007) must tackle cross- scale stresses, trade-offs and uncertainties associated with future climatic, technological and societal changes.
To this end, computer simulation models provide virtual test-beds to explore the impacts of different assumptions, scenarios and decisions on system functioning (Mulligan, 2013). Models in physical sciences often offer deterministic views of reality, whereby an input produces a predetermined output. Such determinism extends to how models view the future, with conventional predictive scenarios
1 Chapter 1: Introduction forecasting what will happen under a number of assumptions (Börjeson et al., 2006). Arguably, various characteristics of conventional modelling approaches mismatch the complex challenges facing human-natural systems in the Anthropocene (Liu et al., 2007). Systems like coral reefs and fisheries are bistable (Scheffer et al., 2001), whereby positive feedbacks may trigger an abrupt shift to an alternative state (e.g. fishery collapse). A limited number of predictive scenarios based on past trends might leave such thresholds unexplored, keeping decision-makers naive to their occurrence. Awareness of such thresholds reduces ecological surprise (King, 1995), facilitates preparation (Olsson et al., 2006) and enables threshold-based management (Kelly et al., 2015). However, conventional integrated assessment models traditionally used in climate change studies are criticised for their user-defined damage functions built from uncertain knowledge and relationships, like the effect of temperature change on gross domestic product (Pindyck, 2013, Weyant, 2017).
Social-ecological interactions underpin complexity in human-natural systems (Lade et al., 2015), particularly where socioeconomic development is dependent on long-term biophysical constraints (Ostrom, 2009). Models must incorporate the diversity of underlying components to deliver realistic governance options relative to system complexity and uncertainty (Verburg et al., 2016, Cumming and Peterson, 2017); however, models must not be so complex that outputs are incomprehensible. The evaluation of uncertainties must go beyond comparing modelled and observed time-series to assess wider system functions, including model structure, sensitivity and integrity (Jakeman et al., 2006).
There is therefore a growing need for models to move from deterministic forecasters to explorers of multiple interacting pathways steering towards desirable futures (Bai et al., 2016, Verburg et al., 2016, Hughes et al., 2017). This vision aligns with the ‘safe and just operating spaces’ (SJOS) concept of sustainability (Raworth, 2012, Dearing et al., 2014), which is an extension of the Earth system’s ‘planetary boundaries’ paradigm (Rockström et al., 2009). Rather than attributing sustainability to a limited number of predictions, SJOSs identify system configurations achieving social foundations of human wellbeing within biophysical limits (Cole et al., 2014, Dearing et al., 2014). Establishing such desirable spaces can form the first stage of sustainable governance, from which goal-achieving actions can be mapped. Efforts to identify routes to SJOSs are in their infancy, however, there is recognition that sustainable pathways are multidimensional and interactive (Dearing et al., 2014, Scheffer et al., 2015). Such
2 Chapter 1: Introduction interactions causing thresholds to shift are absent from the original planetary boundaries and subsequent modelling efforts of theoretical (Carpenter et al., 2015b) and real social-ecological systems (Hossain et al., 2017).
Regionalising sustainability principles is actively encouraged to positively frame actions, reduce uncertainties of scale and progress the narrative of global to local commons (Dearing et al., 2014, Scheffer et al., 2015, Verburg et al., 2016). Consequently, this thesis explores the concepts of SJOSs, complexity and modelling through the Chilika lagoon fishery of the Mahanadi delta, India: forming one of three deltaic case studies of the ‘Deltas, Vulnerability and Climate Change: Migration and Adaptation’ (DECCMA) programme (Lazar et al., 2015). Fishing activities support livelihoods of 35,000 primary and 200,000 secondary dependants (Iwasaki and Shaw, 2010a), but the system’s future is deeply uncertain due to a legacy of collapse, plausible trajectories of climate change and globalisation, and various alternative governance options.
Therefore, this thesis has three main contributions. With the first mid-range projections of Chilika’s fishery, this thesis provides a necessary extension to the abundant literature on Chilika’s past. Through Chilika, this study seeks to convert the SJOS concept from a retrospective heuristic to a regionalised decision-support tool, with wider implications for how we govern for and view sustainability and resilience in social-ecological systems. Third, this thesis contributes to the disciplines of system dynamics and modelling through the capture of human- natural feedbacks, use of exploratory futures and the integration of qualitative and quantitative data.
1.2 Thesis structure
This thesis adopts the structure of introduction, literature review, methodology, results, discussion and conclusion (Wisker, 2008). This section gives an overview of the purpose and content of each chapter.
1.2.1 Literature review
Chapter 2 aims to discuss key concepts underpinning this study to identify gaps in current knowledge. Concepts of social-ecological systems are first discussed (chapter 2.1), asking what makes a system social-ecological? What makes a social- ecological system complex? What frameworks characterise and distinguish social- ecological systems? Notions of sustainability are introduced (chapter 2.2), including definitions, relations to resilience and adaptability, and implications for
3 Chapter 1: Introduction the future of social-ecological systems. Chapter 2.3 focuses on tipping points and regime shifts, including methods of detection, modelling and operationalisation in the real world. Chapter 2.4 reviews the application of sustainability, thresholds and SJOS concepts to deltaic social-ecological systems. Chapter 2.5 introduces the Chilika lagoon fishery, including ecological settings, the emergence of human operations and the need to find plausible futures. Chapter 2 culminates with the aim and objectives of this study.
1.2.2 Model construction, evaluation and plausible futures
Three chapters describe the methods used to address this study’s research questions. Chapter 3 focuses on model development, including conceptual modelling, interview methods employed during the field visit of spring 2016 and the process of formal model parameterisation, including datasets, equations and structures. Model evaluation (chapter 4) assesses whether appropriate quantitative behaviours are produced for structurally robust reasons. Finally, chapter 5 describes the techniques used to generate future dynamics, governance scenarios and classify future system safety.
1.2.3 Model results
Outcomes are organised into subchapters depicting different elements of Chilika’s sustainability (chapter 6), starting with outcome time-series and their causes, before evaluating the broader implications of SJOSs for Chilika and beyond. Chapter 6.2 presents future fish catch (hereafter catch) trajectories to understand the link between output safety and system governance. The following two subchapters investigate resilience further, before causal driver trajectories are depicted (chapter 6.5). Driver based SJOSs then synthesise the causal pathways (chapter 6.6), including a suite of highly sustainable core SJOSs (chapter 6.7). Wider ecological outcomes and trade-offs evaluate the social justness and ecological safety of safe and just futures (chapter 6.8), before comparing contemporary trajectories to the identified thresholds (chapter 6.9). Chapter 6.10 assesses nonlinear driver-response thresholds, assessing the association between normative futures and characteristics of critical transitions – like tipping points and hysteresis. Finally, nonlinear SJOS definitions provide an alternative method to characterise sustainable outcomes (chapter 6.11).
4 Chapter 1: Introduction 1.2.4 Discussion
Chapter 7 position findings within the context established by the literature review, including: (i) the future use and governance of Chilika; (ii) governance of social-ecological systems; (iii) threshold-based management; (iv) the use of (system dynamics) models in social-ecological studies, and (v) transferability, uncertainties and limitations of this study.
1.2.5 Conclusion
Chapter 8 summarises the main findings and novel contributions of this study, plus implications for further study both within the case study and further afield.
5
Chapter 2: Literature review
Chapter 2: Literature review
This chapter aims to review and identify prominent research gaps in the concepts and practices underpinning this research, including: (i) social-ecological systems, complexity and sustainability science; (ii) social-ecological thresholds and tipping points; (iii) deltaic systems and functions underpinning their regional SJOSs, and (iv) the Chilika lagoon. Chapter 2.6 then outlines the study aim, objectives and secondary research questions this study seeks to answer.
2.1 Complex social-ecological systems
2.1.1 Social, ecological and social-ecological systems
Identified by their constituent stocks and flows, ‘systems’ are structures of inter- connected components producing patterns of behaviour over space and time (Meadows, 2009). For instance, humans form the main actors of human-based systems interconnected by communication, coordination and self-organisation (Zeleny, 2009), whilst biophysical systems are comprised of microorganisms through to geomorphological landforms (Campbell and Norman, 1998). Systems involving abiotic and living entities are ‘ecosystems’ (Tansley, 1935), dependent on processes like reproduction to sustain species abundance and competition for life-supporting resources.
Social systems and ecosystems share various characteristics. Globally, multinational corporations compete to access unexploited natural resources and consumer markets. Geomorphological processes such as sediment erosion and deposition exhibit self-organisation, illustrated by the spatial patterns of river deltas (Fagherazzi, 2008), coastal cliffs (Paredes et al., 2015) and Aeolian dune fields (Kocurek and Ewing, 2005). Likewise, Holling et al. (1996) outline the importance of fires in the self-regulation of light, water and shelter within forest canopies, which in turn influence species abundance and distribution.
However despite these commonalities and the shared dependency on clean water, air and soil, humans have traditionally overlooked human-natural connections (MEA, 2005, Bennett et al., 2009). Vitousek et al. (1997, p.494) use the term ‘domination’ to describe the conversion of between one-third and one-half of Earth’s natural land cover. Economic activities cover 25-30% of Earth’s land surface (Lambin and Geist, 2006), whilst approximately 50% of natural mangrove
7 Chapter 2: Literature review coverage has been lost (Crutzen and Stoermer, 2000). Global atmospheric carbon dioxide levels reached 400 parts per million (ppm) in early 2015, 100-130 ppm above pre-industrial levels (Kellie-Smith and Cox, 2011) and the Holocene limit of variability (Steffen et al., 2011). Hydrological systems are also vastly altered: approximately 70% of the Earth’s rivers are intercepted by large reservoirs (Kummu and Varis, 2007), whilst the number of dams worldwide increased from ~26,000 in 2000 to ~32,000 in 2010 (Steffen et al., 2015a). At the regional scale, deforestation for timber production is degrading pollination and natural flood defences across Bangladesh and India (Barbier et al., 2008, Giri et al., 2014). Overexploitation and the subsequent collapse of cod stocks across the Canadian Atlantic has degraded regional food and employment securities (Hamilton and Butler, 2001) – a scenario mirrored across the Baltic (Lade et al., 2015), Bering (Rodionov and Overland, 2005) and North seas (Hutchings, 2000).
In response to the realisation that poverty alleviation across developing nations is not possible from socioeconomic actions alone, Gallopín et al. (1989, p.19) conceptualise ‘socio-ecological systems’ as “any system composed of a societal (or human) component and an ecological (or biophysical) component”. The term ‘socio-ecological systems’ now has various pseudonyms, including ‘social- ecological systems’ (SESs) (Berkes and Folke, 1998), ‘coupled human-environment systems’ (Turner II et al., 2003a, Turner II et al., 2003b) and ‘coupled human and natural systems’ (Liu et al., 2007). Together these concepts argue that social and ecological systems must be considered as one, because anthropogenic impacts upon nature influence everyday human activities across the globe.
Gallopín et al. (2001) argue that the SESs concept must become policy relevant to achieve poverty alleviation. Consequently, different perspectives have emerged to transform theories into implementable actions. Binder et al. (2013) review ten SES frameworks, aiming to recommend appropriate situations for each. Despite differences in the conceptualisation of social and ecological interactions, the authors conclude that none of the approaches is universally superior. Therefore, despite a common concept, Binder et al. (2013) argue that transforming knowledge of system components and interactions into policy relevant information requires nuanced and case-specific approaches.
Arguably the most prominent SESs framework is the ‘Social-Ecogical Systems Framework’ of Ostrom and collagues (2007, 2009). The social-ecological systems framework (figure 2.1) aims to address the ‘tragedy of the commons’ which hypothesises that humans disregard cooperation in the absence of regulation to
8 Chapter 2: Literature review
Figure 2.1: Aggregated version of the social-ecological systems framework describing resource units, users and nested governance structures within SESs (Ostrom et al., 2009). Social, economic and political settings influence contemporary systems through legacy effects, whilst related ecosystems may provide external stresses. Reprinted with permission from the American Association for the Advancement of Science (AAAS). deplete natural resources for personal gain (Hardin, 1974). The common language usage and data organisation of the social-ecological systems framework tailors to different SESs, inluding urban lakes (Nagendra and Ostrom, 2014), agriculture (Cox, 2014) and montane irrigation systems (Hinkel et al., 2015). Despite considering the diversity of users, socioeconomic attributes and cross- scale interactions, Hinkel et al. (2015) argue that the social-ecological systems framework fails to account for interdependencies between resource users which may lead to cooperation challenges. Therefore, where user actions are highly coupled, additional heuristics such as the symmetry of resource access helps to distangle networks critical to system functioning (Bodin and Tengö, 2012).
Ecosystem services provide an alternative conceptualisation of SESs. Costanza et al. (1997) and MEA (2005) argue that ecosystem services provide essential functions for human wellbeing and development, including food, material and
9 Chapter 2: Literature review energy sources, and the regulation of disease, water and chemical fluxes. Aiming to value the benefits of conservations efforts, Costanza et al. (1997) estimate global ecosystem services to be worth US$33-trillion/year (1995), since revised to US$125-145-trillion/year circa 2011 (Costanza et al., 2014). Similar to the social- ecological systems framework, key actors are recognisable by analysing flows of ecosystem services. However, Koshel and McAllister (2008) argue that such accounting ignores heterogeneous access to ecosystem services, due to physical, political, economic and cultural barriers. Moreover, econometric valuations may overlook intangible benefits of ecosystem services, such as cultural, emotional and recreational significances (Bagstad et al., 2013).
SESs may be vulnerable to human-natural interactions and dependencies, broadly defined as the susceptibility of an agent or system to harm (Adger, 2006, Eakin and Luers, 2006). Through case study analysis, Turner II et al. (2003b) recognise that agricultural production across the Mexican Yaqui Valley is vulnerable to soil salinisation, and reduced water quality and quantity. Similarly, the Driver- Pressure-State-Impact-Response framework (Eurostat, 1999) depicts a chain of events from underlying social and biophysical stresses, to institutional responses and impacts. These frameworks provide hazard-oriented perceptions of SESs, with clear overlaps with the social-ecological systems framework and ecosystem services, namely the recognition of susceptible actors and the potential for natural capital loss.
Understanding that human behaviours affect the environment is crucial for socioeconomic development and the persistence of vulnerable biophysical functions (Gallopín et al., 1989). The review hitherto has charted the development and demand for SESs thinking, but overlooked how notions of complexity affect human-natural interactions in SESs.
2.1.2 Characteristics of complex social-ecological systems
The Oxford English Dictionary states that ‘complex’ is the opposite of ‘simple’ (OED, 2015). Simple systems are associated with the ability to succinctly summarise system behaviours (Hugget, 1985), whilst complex systems have interconnections that are complicated to understand (OED, 2015). Since the late 1980s, concepts of complexity have proliferated through social and physical sciences (Kauffman, 1991, Mitchell and Newman, 2002), including health care (Kauffman, 1991, Sweeny and Griffiths, 2002), economics (Arthur, 2013) and psychology (Friedenburg, 2009).
10 Chapter 2: Literature review
Ladyman et al. (2013) synthesise seven characteristics of complex systems across physical and social sciences: (i) nonlinearity; (ii) feedbacks; (iii) spontaneous order; (iv) robustness and lack of central control; (v) emergence; (vi) hierarchical organisation; (vii) numerosity. This list does not explicate temporal delays or surprises, although these may be encompassed by nonlinear and emergent dynamics. Complexities may be intertwined, for example, nonlinearities (e.g. ‘outcome a’ changes at a different rate to ‘driver b’) may evolve from emergence, describing the cumulative effects of localised components (Dearing et al., 2010). Consequently, from the viewpoints of users and governance, emergent dynamics may produce system surprises.
Self-organised criticality is a key characteristic of complexity. Through their famous sand-pile experiments, Bak et al. (1987) show that natural systems may reach critical points of release both scale-invariantly and without the precise tuning of perturbations during relatively stable states. Relating to the characteristics of complexity outlined by Ladyman et al. (2013), the cumulative effects of relatively micro-scale disturbances on the sand-piles exemplifies emergent, unpredictable avalanches. Nonlinearity occurs during spiked sediment transportation rates and the duration of activity appears spontaneous relative to the duration of the preceding state. Moreover, the sand-pile appears robust to disturbances before reaching the proverbial straw that breaks the camel’s back (Bak and Chen, 1991).
Tainter (1990) conceptualises complexity based on analysis of socio-political structures during societal collapses, arguing that societies become increasingly complex by generating socioeconomic problems, increasing the number and interconnectivity of institutions, rules and regulations. Moreover, socio-political systems become hierarchical, stratified and centralised (Tainter, 1990). Eventually, societies reach a maximum complexity, diverting energy like finances and natural resources that previously solved problems into maintaining normal functions. Tainter and Crumley (2005) argue the debasing of the denarius in 64 A.D unintentionally undermined both the strength of the Roman economy and the motivation of soldiers, which led to heightened vulnerabilities to financial collapse and warfare.
The seminal works of Tainter (1990) and Bak et al. (1987) suggest complexity should be viewed through underlying system structures and outcomes. However, every component and process cannot be captured with infinitely fine detail (Cilliers et al., 2013, Cuddington et al., 2013), making mental, physical and
11 Chapter 2: Literature review computational models simpler than reality. Therefore, the concept of ‘critical complexity’ argues that complexity is related to the capacities of human perceptions (Cilliers, 1998, Audouin et al., 2013) and values (Heylighen et al., 2007). Critical complexity does not argue for the adoption of overly simplistic models, but that their design should acknowledge finite understanding. To this end, Cilliers et al. (2013) outline approaches to incorporate complexity into everyday decisions, including harnessing diversity and embracing invisible regions between components. The latter recommendation relates to the lessons of Tainter (1990), as underappreciating unintended consequences may cause collapse. Cilliers et al. (2013) also encourage monitoring complex behaviours to learn about critical components, interactions and opportunities. However, considering the seminal work of Bak et al. (1987), it must be remembered that surprises may occur without warning.
Building upon the static social-ecological frameworks introduced hitherto, Liu et al. (2007) identify spatiotemporal complexities within SESs, namely feedbacks, thresholds, surprises and temporal nonlinearities. For example, thresholds produce social-ecological surprises, such as the unintentional extirpation of commercially important walleyes from the introduction of smelt fish across Wisconsin’s Northern Highland lakes (Liu et al., 2007). To capture complex social- ecological processes, Liu et al. (2007) recommend using integrated approaches, including remote sensing imagery, ecological sampling techniques and modelling strategies. These approaches are multi-temporal, analysing system responses to cross-scale social and biophysical conditions. The need for such multidisciplinary and multi-temporal approaches is echoed by van der Leeuw et al. (2011), Bai et al. (2016) and Verburg et al. (2016).
Moreover, Cumming and Collier (2005) note that social-ecological identities may be temporally dynamics. The authors advise basing identities on natural system properties, for example, by establishing natural functions prior to anthropogenic modifications. In turn, building contemporary identities systematically acknowledges major events such as changed governance approaches, market shifts or climate change. In relation, the ‘long now’ approach (Brand, 1999) posits that the present is inextricably linked to the past through historical legacies, contingencies and emergence. Dearing et al. (2012) advocate a minimum timescale of 60 years to capture continuously evolving processes of different durations and frequencies, whilst considering practicalities like data availability and resolution.
12 Chapter 2: Literature review
Cumming and Collier (2005) further argue that only components and relationships pivotal to system functioning should be considered in identity definitions. However, SESs often have multiple purposes, such as supporting local livelihoods reliant on different natural resources. Schlüter et al. (2009) model conflicting water demands across the Amudarya Basin of central Asia, finding that upstream agricultural practices economically gain at the expense of downstream fishers. Within the same economic sector, evidence from coastal lagoons in India, Vietnam and Portugal suggests that native fishers compete for the same resources as economic migrants, who often use technologically advanced practices (BenDor et al., 2009, Nayak et al., 2015). This may lead to ‘ratchet mode’ dynamics (Meadows, 2009), where resource use intensification from all parties causes overexploitation ultimately detrimental to all resource users over time.
Made up of nodes (e.g. individuals, companies or landscapes) interconnected by links (e.g. flows of matter or information) (Janssen et al., 2006), networks may represent complex SESs particularly where individual behaviours cannot explain aggregated behaviours (Scheffer, 2009). The heterogeneity of links describes network structure, which influences system functions through the clustering and connectivity of flows (Albert et al., 2000). Globalisation is increasing connectivity across geographical and virtual space (Homer-Dixon, 2006), meaning previously regionalised stresses now transmit across international and continental boundaries. Self-organisation through preferential attachment occurs in international agro-food security (Ercsey-Ravasz et al., 2012), global water trade (Dalin et al., 2012) and international financial institutions (Minoiu and Reyes, 2013). Consequently, although society seeks to build stress tolerant hubs, such centralisation is vulnerable to targeted attacks responsible for blackouts or disease outbreaks within cities (Albert et al., 2000).
Similarly, Young et al. (2006) argue globalisation is increasing social-ecological complexities through cross-scale interactions. The international shrimp trade explicates cross-scale coupling, as global consumption and price growth have caused endemic environmental degradation across regions of Southeast Asia (Primavera, 1997, Berkes, 2011, Bunting, 2013). Top-down pressures have also caused the overexploitation of commercial species, socio-political conflicts and livelihood losses (Nayak, 2014, Nayak et al., 2015). Consequently, defining SESs must consider external processes influencing system functioning even though individual actors are unable to affect their trajectories.
13 Chapter 2: Literature review
Consequently, globalisation and the porosity of SESs boundaries have encouraged the study of SESs from the global perspective. Paradigms such as ‘global system science’ (Jaegar et al., 2013) and ‘Integrated History and future of People on Earth’ (IHOPE - van der Leeuw et al., 2011) support a focus on dynamics with worldwide scope, such as climate change, urbanisation and health. Biermann (2007, 2012) recommends Earth system governance to tackle vulnerabilities to global environmental and population changes, whilst Finnigan (2017, p.31) states that the need to address global challenges with global governance is a “basic fact”. Paradigms such as the ‘planetary boundaries’ (Rockström et al., 2009), ecosystem valuation (Costanza et al., 1997) and Barnosky et al.’s (2012) global biosphere assessment provide tentative metrics of global social-ecological health. In contrast, Zhang et al. (2015), Dearing et al. (2015) and Hanspach et al. (2014) encourage greater regionalisation, partly based on the deficiencies of global evaluations, such as scale disparity between assessments and policy-making, and the lack of truly global long-term datasets. Consequently, Dearing et al. (2014) and Scheffer et al. (2015) promote the regionalisation of the planetary boundary approach, whilst Lang et al. (2012) and Bai et al. (2016) encourage the co- production of knowledge between regional scientists, decision-makers and users.
SESs exist across the world, and regional vulnerabilities to changing biophysical and social settings surmount to global issues. Yet, disagreement exists about how to tackle cross-scale issues: should regional stresses that accumulate to global problems be prioritised? Alternatively, do we need a systematic approach to ensure that SESs remain functional from the top-down? The appreciation of spatiotemporal complexity is critical, as dynamics cannot be captured without acknowledging emergent processes, nonlinearities and historical legacies. However, these properties differ between systems, questioning the plausibility of managing regional SESs from the global level.
2.2 Social-ecological sustainability, resilience and adaptability
2.2.1 Concepts of sustainability and resilience
Sustainability, resilience and adaptability provide further perspectives to analyse SESs based on coping with different biophysical and socioeconomic stresses. The 1960s generally marks the beginning of the sustainability movement (Kramer, 2012), with a proliferation of books, essays and academic literature stimulating
14 Chapter 2: Literature review awareness of man’s impacts on nature. Boulding (1966) argued for a global transition from the conventional capitalistic ‘cowboy economy’ towards a ‘spaceship economy’, whereby measures of economic success switch from flows to accumulations of stocks. Spaceship economics recognises that economic markets not only rely on employment levels, income and amenities, but also long- term relationships between human-natural components.
Holmberg et al. (1996) defines four sustainability principles based on outweighing destructive human actions by biosphere conservation and renewal. Coherent with Boulding (1966), Holmberg and colleagues’ conceptualisation considers natural capitals as stocks influenced by various flows. However, the principles are environmentally focused (Azar et al., 1996), fail to distinguish between short and long-term interactions, and reasons why societies exploit natural resources. Countering such environmentally centric views of sustainability, Hardin (1974) suggests Boulding’s spaceship economy leads to the tragedy of the commons, whereby multiple users claiming rights without responsibilities cause resource exhaustion. Hardin (1974) argues that akin to a lifeboat, natural and social systems have capacities associated with inherent winners and losers. Whilst taking a relatively pragmatic approach to sustainability, ‘lifeboat ethics’ introduce issues of socio-political rights, responsibility and equity to the discussion.
Through development of the World3 model, Meadows et al. (1972) investigate relationships between global socioeconomic development pathways and environmental limits (e.g. cultivated land availability, fertility and maximum yields). The need to consider principles of ‘limits to growth’ when making socioeconomic decisions is argued, developing the notions that economic throughput cannot be infinite without damaging nature (Boulding, 1966) and foreshadowing the recognition of social-ecological carrying capacities (Hardin, 1974).
Such socioeconomic pathways avoiding environmental limits are central to sustainable development, most famously defined as “[meeting] the needs of the present without compromising the ability of future generations to meet their needs” (UNWCED, 1987, p.16). This definition relates to the spaceship Earth, encouraging the preservation of natural resources for future use. However, Lélé (1991) and Sneddon et al. (2006) emphasise the juxtaposition of conservation and development. Similarly, the environmental Kuznets curve observes a parabolic relationship between income per capita and environmental degradation (Stern, 2004), implying historically, developing countries must initially trade-off
15 Chapter 2: Literature review ecological conservation for socioeconomic development. Consequently, Grossman and Krueger (1995) and Yang et al. (2014) argue that sustainable development must coordinate all aspects of SESs, synergistically progressing economic, social and environmental subsystems.
To this end, the last 15 years has witnessed the growth of sustainability science – concerned with understanding the characteristics of interactions between nature and society (Kates et al., 2001). Sustainability science adopts a system-based approach, investigating issues across spatiotemporal domains, governance structures and scientific disciplines (Sala et al., 2013). Sustainability science also closely relates to fundamentals of complex SESs. For instance, sustainability science does not assume linearity (Sala et al., 2013), meaning that resource degradation is not automatically considered unsustainable. Instead, the range and magnitude of impacts, action durations and the potential for reversibility are considered (Kates et al., 2001). Likewise, social and ecological components are treated as equal contributions to system functioning (Clark and Dickson, 2003). Moreover, sustainability science is problem-orientated (Kates et al., 2001, Clark and Dickson, 2003), investigating reasons why social-ecological functions become unsustainable.
Parallel to the concept of sustainability is the multifaceted notion of resilience, traditionally viewed through the lens of ‘engineering resilience’, which is considered inversely proportional to a system’s return time to equilibrium (Holling, 1973). Therefore, engineering resilience is traditionally applied to disaster recovery (Mileti, 1999), aiming to resume normal societal functions after hazardous events (Wang and Blackmore, 2009). Ives (1995) labels this concept ‘deterministic resilience’ due to its quantifiable nature. However, Folke et al. (2002) argue that quantitative comparisons assume that ecosystem responses to human actions are always predictable and manageable.
In contrast, following the alternative stable states in ecology hypothesis (Lewontin, 1969), Holling (1973) introduces the concept of ‘ecological resilience’ to describe an ecosystems’ ability to absorb change whilst retaining desirable relationships and interactions. Holling (1973) finds that despite significant variability, Eastern Canada’s Spruce Budworm population persists through disturbances, as management conditions the system to collapse and recovery. Thus, ecological resilience acknowledges spatiotemporally dynamic drivers and responses. Contrastingly, engineering resilience assumes long-term static averages (Milly et al., 2008), thought to undermine long-term resilience as systems remain
16 Chapter 2: Literature review unconditioned to changing driver-response relationships (Folke, 2006, Adger et al., 2011).
Addressing these divergent concepts of resilience, ‘social-ecological resilience’ is generally considered to have two dimensions: (i) the maximum disturbance a SES can withstand before dynamics become undesirable, (ii) the ability to reorganise to sustainable trajectories once separated from normal functionality (Folke, 2006, Renaud et al., 2010). Similar to ecological resilience, social-ecological resilience constantly evolves in response to underlying relationships, feedbacks and legacies. Therefore, social-ecological resilience is arguably exhibited if desirable outcomes persist across a range of social-ecological drivers, such as novel climatic, hydrological and socioeconomic conditions. Contrastingly, the Northwest Atlantic fishery of Canada is a famous example of social-ecological resilience loss, with the absence of alternative livelihoods leading to socioeconomic decline after the 1970s productivity collapse (De Young, 1999, Hamilton and Butler, 2001).
Operationalising resilience remains a key research demand in hydrological (Parsons and Thoms, 2017), agricultural (Ashkenazy et al., 2017) and broader social-ecological systems (Bennett et al., 2005, Brand and Jax, 2007, Sterk et al., 2017). To this end, Walker et al. (2004) outline four components of resilience, namely: (i) the maximum disturbance a system can withstand (latitude); (ii) system sensitivity to change (resistance); (iii) system proximity to change (precariousness) and (iv) the influence of multi-scale dynamics upon the scale of interest (panarchy). Latitude and resistance combine to form robustness against social and biophysical stresses (Anderies et al., 2004), whilst global processes (e.g. climate change) have been found to influence emergent local cycles (e.g. soil erosion) across panarchic systems of rangeland agriculture (Walker and Abel, 2002) and small island developing states (Holdschlag and Ratter, 2016).
Parallel to the resilience properties of Walker et al. (2004), Carpenter et al. (2001) stress the need to clarify the stressors and outcomes under consideration (‘resilience of what to what?’), such as the resilience of a freshwater lake system to increasing soil phosphorous. Using simple system models, Bennett et al. (2005, p.955) define resilience surrogates based on “the state of the system relative to the location of the threshold, the sensitivity of the system to movement toward the threshold, and the rate at which the system is moving toward the threshold”. Whilst helping to operationalise resilience, the surrogates of Bennett and colleagues overlook system persistence beyond the threshold. Addressing these
17 Chapter 2: Literature review shortcomings within social arenas, Marshall and Marshall (2007) operationalise the resilience of Australian fishers as the level of interest in livelihood change, plus perceptions of risk, reorganisation and coping in times of stress.
Contrasting the component specific resilience described above, general resilience aims to consider the persistence of multiple outcomes to multiple system drivers (Folke et al., 2010, Matyas and Pelling, 2015). Consequently, general resilience relates to the capacity of a SES to persist and adapt in response to novel (e.g. climate change) or surprising (e.g. disasters) stresses (Folke et al., 2016). In support of general resilience, Lebel et al. (2006) argue that specified resilience can degrade resilience elsewhere within the system, as decision-makers may favour particular components. Furthermore, a historically resilient component may become abruptly vulnerable due to modifications elsewhere in the system (Lebel et al., 2006). Consequently, surveying perceptions of resilience during times of prosperity may overestimate resilience to abrupt and unpredictable collapses (Marshall and Marshall, 2007), whilst only considering social components suggests that the loss of ecosystem services is acceptable if alternative livelihoods are available. Despite these limitations, resilience studies have traditionally focused on specified resilience, to the extent that Sterk et al. (2017) argue that studying multiple interacting stresses represents a paradigm shift!
Overall, SESs must exhibit some resilience in the face of stresses and vulnerabilities. The above review highlights a number of knowledge gaps that make studying resilience an exciting venture. For example, can drivers key to sustainability be identified and prioritised (Biggs et al., 2015a, Hughes et al., 2017)? How may gradual externals influence resilience to abrupt internal drivers (Scheffer et al., 2015)? How may internal governance bolster resilience to destructive feedbacks in order to manipulate threshold locations (Hughes et al., 2017)? The following subchapter builds upon the above concepts by reviewing deeper notions and special cases of resilience, including traps where decision- makers and stakeholders lose the capacity to adapt to stress.
2.2.2 Beyond definitions: successes, traps and adaptability
Adaptability is inherent to social-ecological resilience, describing system capacity to adjust to changes (Young et al., 2006). Central to climate change science, adaptability investigates how to adjust social norms to cope with evolving conditions (Smit and Wandel, 2006). Walker et al. (2004) argue adaptability is an
18 Chapter 2: Literature review anthropocentric notion, relating to the abilities of social actors to control resilience. In turn, transformability describes the ability of actors to adapt towards sustainable trajectories – as per the second component of social- ecological resilience (Walker et al., 2004).
Olsson et al. (2004) explore sustainability, resilience and adaptability across the Kristrianstad region of Sweden. Since the 17th century, the region’s wetlands have faced threats from socioeconomic pressures including agriculture, flood defences and recreational uses. Alongside growing awareness of environmental issues across Sweden during the 1980s (Olsson et al., 2004), Kristianstad’s citizens, business leaders and politicians recognised a need to transform away from unsustainable balances between human activities and ecosystem conservation.
Therefore, Olsson et al. (2004) formulated a framework for sustainable governance and transformations. To avoid unintended consequences, SESs must prepare for change across all societal sectors through education and engagement. Systems must then seize windows of opportunity: requiring realisation that a system’s trajectory is unsustainable but adaptable with sufficient effort (Kingdon, 1995). Kristrianstad’s window of opportunity occurred in 1988, as the leader of the region’s sustainability movement met with senior politicians, who later established the Ecomuseum Kristianstads Vattenrike (EKV) governance board for regional decision-making (Olsson et al., 2006). Lastly, Olsson et al. (2004) describe the need for social-ecological resilience post-transformation. Therefore, Kristrianstad’s SES was constructed upon the principles of the social-ecoloigcal systems framework (Ostrom, 2007; 2009), including polycentric governance structures, and financial and knowledge contributions from various funders. Consequently, Olsson et al. (2004) argue that EKV’s self-organisation enhances social-ecological resilience and adaptability, as per resilient systems in nature (Holling et al., 1996).
Lebel et al. (2006) also argue strong leadership builds public trust and participation in governance; otherwise, system governance may become misdirected, reluctant to admit responsibility and learn from mistakes. Olsson et al. (2006) discover that bureaucracy, mistrust of non-governmental groups and weak leadership inhibits adaptive governance across the Mae Nam Ping Basin of northern Thailand. Consequently, the three-steps for adaptive governance arguably favour regions with socioeconomic conditions facilitating efficient exchanges of information, knowledge and finances.
19 Chapter 2: Literature review
The adaptive cycle (Holling, 1986, Holling, 2001) provides an alternative perspective on the role of adaptive governance in resilience and sustainability, arguing that ecosystems (and SESs: Walker et al., 2004, Pelling and Manuel- Navarrete, 2011) cyclically evolve over space and time. The adaptive cycle is comprised of a forward loop occurring as a system increases in socioeconomic potential and connectedness (Holling, 2001), before a reverse loop drives collapse and reorganisation. The adaptive cycle suggests that resilience aligns with the ability of a system to successfully navigate the loops, differing from the relatively unidirectional adaptive governance framework (Olsson et al., 2004). The adaptive cycle implies that SESs are conditioned to boom and bust (Gunderson and Holling, 2002), producing low levels of internal controllability limiting the ability of governance to shape long-term dynamics (Holling, 2001). In contrast, the framework of Olsson et al. (2004) requires relatively high levels of internal controllability to prepare for change and seize a window of opportunity. It is important to note that the adaptive cycle and framework for transformational governance are heuristics for system understanding, rather than predictive models of change (Biesbroek et al., 2017). Therefore, recognising a window of opportunity does not necessitate successful transformational change and decision-making must avoid falling into this trap.
Traps may also hinder the passage of SESs between different adaptive cycle phases (Carpenter and Brock, 2008). ‘Poverty traps’ due to reasons beyond stakeholder control (e.g. recurrent natural disasters) may keep SESs within the low potential ‘exploitation’ stage with unfavourable socioeconomic conditions, including poverty and inequality. Furthermore, SESs may become rigid through institutional arthritis (Olson, 1982, Carpenter and Brock, 2008, Young, 2010), whereby institutions responsible for past successes are reluctant to adapt. To combat rigidity traps, Carpenter et al. (2001), Olsson et al. (2004) and Fath et al. (2015) advocate the diversification of social-ecological components and governance structures. Moreover, congruent with holistic sustainability perspectives (chapter 2.2.1), the authors argue that socioeconomic growth should be controlled to prolong the conservation stage and reduce vulnerabilities to crises. However, constraining variability may increase long-term vulnerability to large shocks and slowly manifesting stresses (Anderies, 2015, Carpenter et al., 2015b), whilst short-term fixes may lead to costlier remediation in future (Sterner et al., 2006). For example, Palmer et al. (2008) argue that towns and cities dependent on low-magnitude flood defences are becoming increasingly vulnerable to significant events under climate change.
20 Chapter 2: Literature review
The ‘complex adaptive systems’ (Holland, 1995) concept provides further perspectives on sustainability and resilience. Social-agents within complex adaptive systems continuously adapt to system conditions, and the role of ecological processes such as exploitation, competition and cooperation in system dynamics is widely acknowledged (van Bilsen, 2010, Levin et al., 2013). Game theory is central to complex adaptive systems (Lansing, 2003), describing how actors rationalise individual decisions (van Bilsen, 2010). Individuals may act selfishly to maximise profit, improving their resilience in the short-term. However, SESs may enter ratchet mode as defector behaviour becomes endemic, leading to unsustainable resource exploitation and tragedy of the commons (Lansing, 2003). Therefore, as with transformative governance, leadership and social responsibility must conserve the negative feedbacks building robustness against actions undermining long-term sustainability – termed ‘viability loops’ by Hjorth and Bagheri (2006).
Sustainability, resilience and adaptability are dynamic concepts with spatiotemporal scales and components malleable to different systems of governance and individuals. However, the different timescales, settings and unintended consequences contribute to the complexity of sustainability and resilience notions (Berkes and Folke, 1998, Nuno et al., 2014). As such, decision- makers cannot rest on their laurels having seemingly averted unsustainable conditions. Instead, long-term sustainability and resilience should be prioritised, plus the avoidance of traps that may appear resilient in the short-term.
2.2.3 Plausible and desirable social-ecological futures
This review has highlighted desirable social-ecological qualities and pitfalls associated with traps, trade-offs and future uncertainties. Traditionally used in environmental impact assessments (Lawrence, 2003), notions of ‘desirable’ and ‘plausible’ futures are increasingly applied to SESs. Bai et al. (2016) define ‘desirable’ as futures increasing the chances of society overcoming current social- ecological crises. Thus, desirable futures offer an action-oriented take on the ultimate goal of sustainability science to maintain both social and environmental favourable conditions. Plausible futures consider the wider spectrum of possible development trajectories, acknowledging uncertainty associated with projecting social-ecological futures (Bai et al., 2016).
Taking sustainability, resilience and adaptability at face-value will not solely achieve desirable futures. Despite global recognition for its need, the sustainable
21 Chapter 2: Literature review development definition of UNWCED (1987) is criticised for lacking jurisdiction and implementable guidelines, leading to the belief that international decision-makers are merely paying lip-service (Mannion, 1997, Egelston, 2013). Gell-Mann (2010) outlines seven transformations needed to improve the sustainability of current anthropogenic trends and attitudes, each requiring divergence from conventional global social and economic processes, which Bai et al. (2016) argue is difficult to imagine, yet alone practice at regional and international scales. In contrast, it is straightforward to imagine futures whereby cowboy economics persist.
Gilberto Gallopín has championed global plausible futures, including multidimensional and multidecadal pathways of energy use (Gallopín, 1997), human population (Gallopín and Raskin, 1998) and material consumption (Gallopín, 2002). Regional applications include Carpenter et al. (2015a) for the Yahara watershed, USA, and Hanspach et al. (2014) for Transylvania, Romania. The studies have commonalities despite differences in location and scale. First, three or four futures are analysed, such as ‘business-as-usual’, ‘barbarisation’ and ‘sustainable transformation’ (Gallopín, 2002), or ‘abandonment and renewal’, ‘accelerated innovation’, ‘connected communities’ and ‘nested watersheds’ (Carpenter et al., 2015a). Pathways generally centre on major branch points, like excessive lake pollution or changes in global governance. Moreover, endpoints and pathways tend to be expert defined. Consequently, the degree to which such human-led horizon scanning can identify truly surprising dynamics is questionable. Moreover, such divergent scenarios creates gaps in our roadmap, as the ‘real’ future may fall somewhere between the storylines.
Rather than branch points, Palmer et al. (2008) and Contamin and Ellison (2009) note that desirable futures should consider sustainability and resilience proactively and continuously, rather than iteratively after perturbations or once unsustainable trajectories are recognised. However, such adaptability is not universally attainable. Barriers trapping SESs within poverty traps must be overcome, perhaps through the diversification of governance structures to engage and harness indigenous system potential (Holling, 2001). Furthermore, plausible futures consider resilience as a function of trade-offs and bounded rationalities (Levin et al., 2013). Boulding’s spaceship metaphor is therefore desirable, with SESs entirely controllable from the inside. Yet, viewing SESs through the lens of Hardin’s lifeboats is more plausible, recognising that sustainability and resilience are emergent properties of processes with different spatiotemporal scales, trade-offs and adaptive capacities. The discussions
22 Chapter 2: Literature review hitherto place the dynamic concepts of sustainability and resilience on a continuum with the desirability of plausible futures. However, ecological theories suggest SESs may have alternative stable states, separated by discontinuous and often abrupt thresholds. Thus, the next subchapter discusses the implications of tipping points and regime shifts on sustainability and resilience.
2.3 Social-ecological regime shifts
2.3.1 Thresholds, tipping points and regime shifts
The canoe metaphor is often used to communicate the complex concepts of thresholds, tipping points and regime shifts. Fundamentally, an upside-down canoe has lost its ability to transport the canoeist through the reorganisation of system functionality (Crépin et al., 2012, Biggs et al., 2015c). Thresholds represent critical levels of system conditions between states of system functionality (Liu et al., 2007). For example, the canoe may flip once waves external to the vessel exceed a critical height. Tipping points are thresholds representing moments when small driver changes trigger changes in system quality, properties or state (Groffman et al., 2006, Lenton et al., 2008). Interactions between lower-level thresholds and system-wide tipping points are complex. For example, the canoe may remain upright if the canoeist can counteract external forces (Scheffer, 2009). Therefore, the likelihood and implications of threshold crossing relate to system robustness and resilience. A heavy boat may resist external drivers, whilst an experienced canoeist may rapidly re-flip the canoe (engineering resilience). The canoe exhibits ecological resilience if desirable functions persist as individual thresholds are approached and passed (Holling, 1973, Renaud et al., 2010). Social-ecological resilience occurs if the above is achieved without damaging the craft’s ability to self- organise and adapt (Folke, 2006, Renaud et al., 2010).
Following Holling (1973), May (1977) analysed breakpoints between alternative regimes. Through ecological modelling, May showed that the rapid transmission of Schistosomiasis results from the gradual accumulation of worms upon an individual. The breakpoints of May (1977) are synonymous with tipping points, whereby continuous changes in the driving variable cause disproportional responses. The connotation of ‘breakpoints’ has since become less catastrophic. Christensen and Krogman (2012) argue that a social breakpoint identifies a new
23 Chapter 2: Literature review
Figure 2.2: Stability landscapes depicting the two regime shifts of Scheffer et al. (2001): (a) the system appears resilient to stresses (D), remaining within the desirable basin when the maximum stress (C ) occurs; (b) max the system undergoes parameter changes (∆P) causing normal stresses (D ) to induce state changes; (c) parameters remain constant, but the norm system is forced beyond its desirable basin. experience. In statistics, a breakpoint represents a significant deviation from a long-term trend (Zeileis et al., 2015).
From coral reefs, lakes and forests, Scheffer et al. (2001) devised the stability landscape metaphor to classify threshold transgressions. The first category describes steady changes in system parameters (figure 2.2b), such as climatic shifts or chemical flows (Scheffer et al., 2001). The depth of the desirable basin of attraction decreases over time, meaning small perturbations can force the system from its desirable regime. Alternatively, abrupt threshold transgressions may result from external impacts, like land-use changes or hydrometeorological events. In reality, exact causes of state shifts are often complex to distinguish (Scheffer et al., 2001), as long-term degradation may undermine resilience to short-lived events.
Regime shifts are traditionally viewed as instantaneous dynamics, as transitions between alternative states are short-lived compared to the preceding and resulting regimes (Biggs et al., 2009, Dearing et al., 2010). Beisner et al. (2003) recognise the existence of non-catastrophic shifts leading to ‘alternative interior
24 Chapter 2: Literature review states’; for instance, population regimes of fish switching between declining and increasing stocks. Management of such internal dynamics is important to discourage system-wide shifts, for example, weakening positive feedbacks driving population decline is key to resource persistence (Hjorth and Bagheri, 2006).
Global tipping points gained attention following Scheffer et al. (2001), leading Alley et al. (2003), Walker (2006) and Lenton et al. (2008) to quantify climatic thresholds. Rockström et al. (2009) continued this global focus in an attempt to operationalise Earth system resilience, hypothesising nine ‘planetary boundaries’ which may destabilise the Holocene environmental conditions that have supported humanity for the past 11,000 years. Expert opinion and palaeoenvironmental records quantified seven of the planetary boundaries, positing that climate change, Earth’s nitrogen cycle and biodiversity loss have exceeded critical thresholds known as “safe operating spaces” (SOS). Criticism predominantly focused on boundary values (Brook et al., 2013), reflected in the inclusion of uncertainty in the updated planetary boundaries (Steffen et al., 2015b). Moreover, the response of boundaries to management, technological innovations and inter-threshold relationships are unknown (Brook et al., 2013). The concept also overlooks the responsibilities of regional governments to tackle the thresholds (Häyhä et al., 2016). Thus, the planetary boundaries concept is a promising heuristic requiring modifications to become operational (Brook et al., 2013); in particular, the boundaries demand downscaling to match the influence of regional governments (Dearing et al., 2014, Scheffer et al., 2015), whilst the causes, locations and impacts of shifts must be considered to guide decision- making and priority setting.
With a social-ecological focus, Walker and Meyers (2004) categorise thresholds by their degree of human-natural coupling. The first two classes match those of Scheffer et al. (2001) (figure 2.2); class-III depicts unidirectional cause and effect between humans and nature, class-IV includes feedbacks and class-V involves reciprocity, whereby ecosystem shifts cause social shifts (and vice versa). The above typology is useful to prioritise shifts within a single system and compare inter-system shifts. Class-V shifts form the worst case scenario, argued responsible for the Easter Island civilisation collapse (Walker and Meyers, 2004). Counterintuitively, class-III shifts may actually enhance social-ecological sustainability; for example, public opinion may undergo a regime shift in response to natural disasters, potentially catalysing long-term remediation efforts (Scheffer et al., 2003).
25 Chapter 2: Literature review
Rocha et al. (2015) identify climate change as the most common cause of regime shifts globally, whilst agriculture, deforestation and fishing dominate regional levels. The connectivity of climate change to regional drivers is problematic for regional system governance, with slow external stresses causing parameter changes across desirable stability basins (figure 2.2b). Cascading effects may occur when drivers are shared, triggering a domino effect of degradation. Kinzig et al. (2006) observe cascading thresholds of land fertility and salinity, which have both shifted due to climate change across Western Australia’s wheatbelt. Soil salinisation also compounds land fertility, synergistically threatening future productivity and livelihoods.
The preceding discussion introduces the interlinked concepts of thresholds, tipping points and regime shifts. Despite generalities, regime shifts in real systems are contextually distinct, owing to specific drivers, manifestations and impacts. Consequently, knowing the existence of alternative states provides limited insights. Instead, managers would ideally foresee shifts in time to implement diversionary policies. However, due to their often abrupt, unpredictable and contextual nature, threshold identification is predominantly limited to a posteriori analysis. To this end, the next subchapter reviews methods of detecting regime shifts to protect social-ecological identities and functions.
2.3.2 Threshold detection methods
The previous subchapters suggest that the consequences of thresholds may range from new socioeconomic experiences to societal collapse. Cumming and Peterson (2017) establish qualifications for detecting collapse, starting with the abrupt loss of key actors, interactions and desirable outcomes, where abrupt equals less than the regeneration duration of key system actors (e.g. <25 years for humans). The losses of social, financial or natural capital must be substantial relative to some normative threshold, which must persist for at least one generation. Despite the unifying framework, quantitative definitions of collapse may vary between systems due to different regenerations rates and definitions of substantial losses, emphasising the need to consider resilience of what to what (Carpenter et al., 2001), and for whom (Cretney, 2014)?
Further seeking to operationalise resilience, Dearing et al. (2014) propose a data- driven typology of social-ecological behaviours to inform boundaries of regional systems. Building on the planetary boundaries concept, linear (type-I) evolutions become shifts once a normative condition is passed, such as daily calorific intake
26 Chapter 2: Literature review or air quality. Although least catastrophic due to smooth gradients of change and reversibility, linear shifts may cause cascading effects (Hughes et al., 2013); for example, incomes below critical levels can destabilise food security, human health and community cohesion (Barnett, 2003). Time-series variability beyond historical patterns represents type-II shifts, relating to weakened engineering resilience and unpredictability (Holling, 1973, Baland and Platteau, 1996). Importantly, long-term variability is different from the encouragement of multiannual variability, argued to aid the adaptability of SESs (Carpenter et al., 2015b). Instead, long-term changes in outcome variability may indicate nonstationarity (Milly et al., 2008), whereby drifting drivers produce new and undesirable experiences.
Early warning signals (EWSs) distinguish type-IV from type-III shifts (Dearing et al., 2014), describing time-series signatures thought to arise as systems approach fold bifurcations (Dakos et al., 2008). Both type-III and IV may be catastrophic, as structural changes initiate positive feedbacks resisting reverse (Troell et al., 2005). At best such transitions are hysteretic (Bestelmeyer et al., 2011), whereby the reverse flip requires greater energy than the initial transition. Dakos et al. (2008) provide a graphical application of EWSs, concluding that critical slowing down is a ‘universal’ warning of large-scale climatic shifts. EWSs have been found in records of marine species abundance (Lindegren et al., 2012), land-use change (Guttal and Jayaprakash, 2008, Zurlini et al., 2014) and common-pool resource systems (CPR) (Lade et al., 2013). Rising skewness and variance form EWSs for ecological networks (Scheffer et al., 2009), although Scheffer et al. (2012) and Suweis and D'Odorico (2014) argue that network robustness is relatively dependent on the arrangement of mutualistic/antagonistic interactions, as opposed to properties of system architecture (e.g. skewness). Furthermore, from modelled and palaeoenvironmental records, Doncaster et al. (2016) recognise the replacement of ‘ecosystem canaries’ with keystones and weedy species to be symptomatic of regime shifts across ecological communities like freshwater lakes.
Whilst presently limited to a posteriori analysis, monitoring techniques of fine temporal resolution may facilitate real-time warning systems, akin to seismometers in earthquake monitoring. Once detected, managers can accelerate conservation, whilst public communication of EWSs may strengthen ecological awareness. However, such reliance on long-term high-resolution datasets currently limits EWSs application. Moreover, EWSs have limited transferability, meaning a variance threshold for one system does not directly transfer to similar
27 Chapter 2: Literature review environments (Boettiger and Hastings, 2013). Thus, systems may have to experience shifts before developing EWSs. Hastings and Wysham (2010) find that modelled catastrophic shifts can occur without warning, whilst, false alarms may arise if indicators detect changes in underlying drivers (e.g. extreme events) rather than weakening resilience. Furthermore, systems rapidly approaching transitions may not exhibit EWS, particularly if driver randomness is high (Perretti and Munch, 2012). Lastly, decision-makers may not act if outcomes remain favourable; for example, high resource productivity and slow recovery from minor perturbations may discourage averting future shifts (Österblom et al., 2013).
Whilst EWSs flag regime shift possibility, various statistical techniques date transitions. At the simplest extreme are limits conducive to type-I and II shifts (Dearing et al., 2014). Fish stocks may be deemed to have collapsed once catches are 10% less than the historical maximum (Pauly et al., 2013) – making fisheries vulnerable to escalation because the productivity loss needed to constitute a collapse increases with every record high. Moreover, outcome oriented metrics (e.g. catch) assume proportionality with underlying conditions (e.g. resource availability). Pauly et al. (2013) argue that factors such as environmental conditions, regulations and unreliable data complicate this assumption. Resultantly, structural changes like the loss of high-trophic species may go unnoticed (Pauly et al., 1998).
Drijfhout et al. (2015) use a limits approach to classify regime shifts in global climate models, whereby outputs falling outside an envelope of four times the long-term standard deviation undergo abrupt change. This method appreciates time-series variability, consistent with type-II changes (Dearing et al., 2014). However, in systems with high natural variability, undesirable conditions falling within the envelope are undetectable. For instance, low monsoonal rainfalls may not fall outside 4xSD due to natural variability in monsoon timings and magnitudes; yet persistently low rainfalls may alter hydrological regimes, agricultural production and regional food security (Loo et al., 2015).
Breakpoint analysis aims to capture statistically significant changes in regression relationships (Bai, 1994), indicating that the previous regime has statistically terminated. Software packages such as “strucchange” (Zeileis et al., 2015) and “bfast” (Verbesselt et al., 2010) are freely available, with the latter suited to seasonal detrending. Importantly however, breakpoint detection does not explicate the causes, duration and reversibility of the new state. Consequently,
28 Chapter 2: Literature review breakpoints are useful when shifts are hypothesised a priori and social-ecological impacts are known.
Bestelmeyer et al. (2011) identify breakpoints in four environmental driver- response relationships, represented by changes in the long-term cumulative sum of outcomes relative to their mean; breakpoint significance was determined by the residual sum of squares, Bayesian Information Criterion and F-statistics. The approach is adaptable to different resolution datasets, albeit less effective with sporadic observations that smooth inflexion points. Rodionov (2004) developed a sequential t-test for shift detection in near real-time, forming a hybrid between EWSs and breakpoint methods. ‘STARS’ searches for differences in mean and variances between moving windows, flagging a potential regime shift once an observation falls beyond the critical boundary of the current regime. Unlike other approaches, STARS does not require shifts hypothesised beforehand.
Wang et al. (2012) and Xu et al. (2015) have applied sequential t-tests to time- series from China’s Erhai and Taihu lakes, respectively, correlating shifts in environmental and socioeconomic indicators to system state. Zhang et al. (2015) applied multiple tests to investigate multidecadal shifts in provisioning and regulating ecosystem services across China’s Lower Yangtze basin. Composite ecosystem service indices underwent sequential t-tests using 10, 20 and 30 year windows, before F-statistic confirmation and comparison to autoregressive moving average (ARIMA) forecasts. Zhang et al. (2015) also use principal component analysis to assess driver relationships before and after shifts, finding that the transmission of stresses is generally aided by increased network connectedness prior to shifts. However, principal components are unsuitable for data limited systems (Andersen et al., 2009), because driver-response relationships cannot be accurately interpolated.
2.3.3 Modelling social-ecological systems and thresholds
The methods reviewed hitherto may be integrated into various modelling techniques that project SES conditions and thresholds, including: (i) system dynamics (SDM), (ii) agent-based (ABM), (iii) statistical, (iv) generalised, (v) coupled component, (vi) integrated assessments and (vii) knowledge-based models (Schlüter et al., 2012, Letcher et al., 2013, Verburg et al., 2016). Filatova et al. (2016) review the capabilities of approaches i-iv, based on the capture of feedbacks, coherence with methods of regime shift detection, data requirements and computational efficiency. The authors conclude that no one modelling
29 Chapter 2: Literature review approach is superior, although SDMs excel at feedback inclusion, ABMs with modelling human decisions and social learning, and statistical models with time- series parameterisation.
Collie et al. (2004) encourage modelling the causes, detection and impacts of marine regime shifts, although their example model only analyses one-way interactions between harvesting and fish population. To build evidence of regime shifts and clearly visualise stability basins, the authors suggest using response variable(s) with bimodality, such as fish stocks vulnerable to Allee effect collapses. These thresholds are then traceable through model structure to assess associated environmental and socioeconomic boundaries.
Models ideally incorporate emergent datasets and processes (Letcher et al., 2013, Filatova et al., 2016). ABMs directly model social learning, where agents within networks adapt behaviours to information and experience, leading to the emergence of patterns outside formal model relationships (Letcher et al., 2013). ABMs are utilised where individual or community decisions steer system functioning, including deltaic landuse decisions based on profitability and user satisfaction (Mialhe et al., 2012), patterns of coordinated irrigation in Balinese subaks (Lansing and Kremer, 1993) and resource use thought to cause the collapse of the Mayan civilisation (Heckbert, 2013). Fishery applications have assessed individual responses to governance interventions (e.g area bans - Soulié and Thébaud, 2006), emergent cooperation between multi-agents (Wilson et al., 2007) and optimal fishing efforts amongst heterogeneous agents (Elliston and Cao, 2006). A shortcoming of ABM is the inability to compare individual outcomes to aggregated time-series for model validation (Filatova et al., 2016). Relatedly, a trade-off of computational efficiency exists between the number of agents (or the resolution of cellular automata grids) and simulation length (Verburg et al., 2016).
Coupled component models have specifically assessed the role of stakeholder cooperation in resource sustainability (Figueiredo and Pereira, 2011, Brandt and Merico, 2013, Lade et al., 2013), alongside phase portraits and eigenvalue analysis to identify tipping points in driver magnitudes (e.g. harvest rate). These studies incorporate bidirectional social-ecological interactions, low data requirements and visual stability landscapes. However, these strengths arguably become weaknesses when modelling regional systems, where decision-makers require tractable indicators for policymaking. The involvement of stakeholders for knowledge sharing is also limited, due to the complexity of model construction and parameterisation.
30 Chapter 2: Literature review
Filatova et al. (2016) consider feedback inclusion vital, particularly if reinforcing loops can trigger cascading thresholds across social-ecological domains. Where reciprocity exists, models should capture social stresses on ecosystems, resulting ecosystem-use feedbacks and consequent adaptive actions. Originating from earlier analysis of industrial efficiency (Forrester, 1961), SDMs traditionally use data-driven parameters and structures to investigate locational specific problems (Sterman, 2000). The graphical interface of SDMs facilitates the incorporation of stakeholder knowledge (Luna-Reyes and Andersen, 2003), whilst model structure is comprehendible to anyone with a basic understanding of stocks and flows. SDMs are now widely applied in sustainability science, including models of lake water quality (Liu et al., 2015a, Zhu et al., 2015) and governance (Ford, 2010), urban water networks (Venkatesan et al., 2011, Zarghami and Akbariyeh, 2012) and forestry (Mendoza and Prabhu, 2006, Purnomo and Mendoza, 2011). SDMs have also explored human and natural drivers of fishery sustainability. Martins et al. (2015) model the socioeconomic drivers of a southern Portuguese bivalve fishery; BenDor et al. (2009) find technological advance and fisher cooperation underpin the sustainability of a contested resource; Moxnes (2000) illustrate that the timing of fishing effort investment is critical to sustainability. With explicit reference to thresholds and tipping points, Bueno and Basurto (2009) find variable environmental conditions promote overexploitation across Mexico’s Cello de Hacha fishery, with fish population recovery lengthening once its death rate is stochastic (proxy for ecological degradation). Moreover, whilst conventionally spatially implicit, geographical information system and cellular automata extensions have built spatially explicit SDMs (Ahmad and Simonovic, 2004, Neuwirth et al., 2015).
SDMs prioritise pattern over prediction through the uncovering of feedbacks and counterintuitive dynamics (Duer-Balkind et al., 2013), so are unsuited to short- term forecasting. As problem-focused tools (Sterman, 2000), SDMs tend to analyse specified-resilience to a given driver, meaning variables elsewhere in the system tend to be stationary [e.g. Martins et al. (2015) hold the number of fishing boats constant]. To detect regime shifts, outputs require the same methods as time-series. Therefore, the capture of thresholds is reliant upon a fine temporal resolution – potentially producing unwieldy models, particularly if the shift occurs decades in future.
It is worth briefly reviewing other modelling techniques applied to SESs. Bayesian networks (BNs) link system variables with conditional probabilities to explicitly
31 Chapter 2: Literature review represent uncertain interactions (Letcher et al., 2013); expert knowledge is used in parameterisation and validation when insufficient data exists. Incorporating public participation explores the sustainability of stakeholder perceptions and intentions under different social and environmental conditions (Lynam et al., 2007). Regarding plausible futures, Monte Carlo simulations can force BNs in highly divergent directions. BNs have evaluated Baltic sea fishery management (Kuikka et al., 1999, Levontin et al., 2011) and metrics of fishery performance (Underwood et al., 2016). Hoshino et al. (2016) model community-based management of the Kei Islands fisheries, Indonesia. Interestingly, the headline finding points to the lack of empirical data to quantify fishing pressure, illustrating that the ability to define key components, interactions and probabilities does not necessitate useful outputs. Further challenges include the omission of feedbacks and the need to discretise continuous variables to generate probabilities (Uusitalo, 2007, Hoshino et al., 2016).
In stark contrast, integrated assessment models in climate change studies involve deterministic assumptions about how climatic, biogeochemical and economic subsystems interact (Verburg et al., 2016). Key examples include the Dynamic Integrated Model of Climate and the Economy (DICE- Nordhaus, 1993) and its regionalised version (RICE- Nordhaus and Yang, 1996); An Integrated Model to Assess the Greenhouse Effect (IMAGE 2.0- Alcamo and Kreileman, 1996) and the Framework for Uncertainty, Negotiation, and Distribution (FUND- Anthoff and Tol, 2013). Integrated assessment models have pioneered efforts to optimise greenhouse gas scenarios, evaluate adaptive global governance, and related societal costs and benefits (Verburg et al., 2016). However, Pindyck (2013, 2015) is particularly critical of uncertainties associated with modeller defined damage functions relating temperature changes to gross domestic product and consumption, as well as the monetisation of human lives, meaning deaths from climate extremes in North America are worth more than deaths in Southeast Asia. Moreover, the complex and often opaque nature of developing integrated assessment models gives a ‘black-box’ reputation (Stanton et al., 2009, Beck and Krueger, 2016). Reinforcing this perception is the view published on the FUND model’s homepage:
“It is the developer's firm belief that most researchers should be locked away in an ivory tower.” (source: http://www.fund-model.org/)
Further criticisms of integrated assessment models include the independence of social and environmental subsystem, including the lack of climate-economy
32 Chapter 2: Literature review feedbacks (Verburg et al., 2016) and the omission of non-climate benefits (Weyant, 2017).
Whilst the above techniques conventionally forecast future dynamics (i.e. establish future scenarios from today from the past), backcasting provides a method to work back from highly divergent futures assumed to be of equal likelihood to find causal pathways (Robinson, 1990, Verkerk et al., 2016). Backcasting approaches commonly couple with integrated assessment models of global carbon emission pathways (Alcamo and Kreileman, 1996, Bruckner et al., 1999, Petschel-Held et al., 1999), for example, constraining global policies keeping atmospheric CO below 550ppm by 2055 and 350ppm by 2100 (Mathias 2 et al., 2017). Social-ecological studies backcasting are rare relative to those using forecasting. Verkerk et al. (2016) backcast landuse models to find future pathways that agree with desirable visions of European landuse by 2040; Robinson et al. (2011) use backcasting principles to help stakeholders envision desirable futures for Delta, Canada, corresponding to model defined scenarios of population growth, greenhouse gases, energy use and economic growth.
If backcasting defines pathways to sustainable futures, why do most models still forecast? Backcasting can be inefficient as it aims to find all driver combinations and policies producing desirable futures. For example, Mathias et al. (2017) consider a total of 3.1x1037 climate policies between 2010-2100! Moreover, normative futures may not be definable for systems with social-ecological trade- offs, meaning sustainability is attributed to future scenarios post-hoc. Third, understanding the most likely outcome may be of interest to decision-makers needing to allocate system resource.
This review shows that the appropriate modelling technique depends on the problem, system and definition of sustainability under study. Seeking to balance cost and benefits of climate change suits integrated assessment models; exploring individual responses to resource regulation suits ABMs. In turn, the appropriate approach to detect thresholds depends on whether the driver- response relationship is known a priori; for example, caused by internal feedbacks or externally forced; caused by dynamics predictable today or emergent future patterns.
33 Chapter 2: Literature review
2.3.4 Putting theory into practice: threshold-based management
Threshold-based management (TBM) has emerged from a growing awareness of regime shifts (Kelly et al., 2015), based on developing SESs in ways non-fatal to ecosystem health (Roe and van Eeten, 2001). TBM proactively identifies components vulnerable to low performance shifts, monitors evolutions and implements remedial actions (Kelly et al., 2015). Kelly et al. (2015) argue TBM is most effective when analysing time-independent thresholds between drivers and responses. However, time-scale specification is critical to resilience (Walker and Abel, 2002), with drivers exhibiting different periodicities and governance systems undergoing change. Combining these two views places desirable futures decades into future, whereby social-ecological trajectories avoid thresholds and traps within an extended timeframe.
To this end, Groffman et al. (2006) note the potential of TBM to identify decadal- scale thresholds, but outline issues of recognising complex reorganisation near and beyond thresholds. Reduced complexity data collection and model building may tackle this limitation. Simple models producing nonlinearities can be built with reference to systematic frameworks, such as the social-ecological systems framework (Ostrom, 2009), through the identification of known components and interaction (e.g. fishers extract 10,000 fish/day) (Bennett et al., 2009). Simple tools producing complex dynamics are more tractable for decision-makers than complex models producing unexplainable behaviours (Summers et al., 2015). Moreover, appropriate limits help to overcome uncertainties surrounding threshold locations. ‘Cautionary’, ‘target’ and ‘critical’ limits (Johnson, 2013) assumes confidence is proportional to the risk adversity of governance, as cautionary limits are closer to known system dynamics. The cautionary zone is associated with increased risk of nonlinear responses (Steffen et al., 2015b), but overly conservative limits may trade-off socioeconomic development.
The prioritisation of system components can be reasoned by relative contributions to desirable dynamics (Cumming and Collier, 2005). For instance, the presence of fish and fishers is paramount to achieving sustainable yields. The risk of fish population collapse increases beyond sustainable levels (Roughgarden and Smith, 1996), making resource conservation salient for the persistence of fishery livelihoods. Crépin et al. (2012) argue that management should focus on leverage points with the greatest impacts and smallest costs. Moreover, the authors advocate maximising the distance between system conditions and thresholds. If required management is too costly, then regime shifts may be
34 Chapter 2: Literature review allowed under the guidance of adaptive governance (Brock and Starrett, 2003). Sufficient preparation and information about the future state must be available to protect the most vulnerable components (Crépin et al., 2012).
These TBM dilemmas directly relate to regional sustainability. Dearing et al. (2014) integrate social-economic data with records of ecosystem services covering 100 years, considering relationships between deprivation and environmental quality across China’s Yunnan and Shucheng regions. Time-series of ecosystem services are then classified by the four-scale typology (chapter 2.3.2) to understand whether environmental thresholds have been exceeded. Different regional ‘safe and just operating spaces’ (SJOS) result (Raworth, 2012, Dearing et al., 2014), assessing whether current outcomes meet social foundations of human wellbeing whilst avoiding ceilings of environmental degradation. To this end, Cole et al. (2014) investigate the evolution of South Africa’s national indicators since 1994 (e.g. crime, jobs, sanitation and so forth), providing a heuristic for national governance to monitor progress towards various social and environmental targets. Whilst the environmental indicators closely relate to the planetary boundaries of Rockström et al. (2009) (e.g. ozone depletion and freshwater use), little coupling exists with the social indicators (e.g. health care, housing and education), flagging a potential future research direction.
Today’s SJOSs are outcome-based, providing little guidance of the pathways leading towards safe and just futures. Consequently, Dearing et al. (2014) and Verburg et al. (2016) state the need to model SJOSs by capturing feedbacks, pathways and bistable variables. Limited examples exist to date. Kwakkel and Timmermans (2012) model global freshwater use, although driver pathways and thresholds are absent. Instead, outputs focus on model validation, reflecting the uncertainty of modelling global scale hydrology with SDM. Carpenter et al. (2015b) deduces generalities from modelled synthetic systems, finding that allowing short-term variance may enlarge the zone of driver-response equilibriums. Hossain et al. (2017) demonstrate that socioeconomic development across the Ganges-Brahmaputra delta can force ecosystem services beyond normative thresholds and envelopes of variability. Whilst the study pioneers the modelling of safe operating spaces, the model only uses statistical couplings vulnerable to breakdown under nonstationary conditions. Second, the safe future misses an environmental ceiling, implying that rice production growth is indefinitely sustainable. Moreover, whilst developing driver scenarios, their predictive definition creates discontinuities between safe and dangerous pathways so that
35 Chapter 2: Literature review outcome sensitivity to multivariable interactions is unknown. Finally, modeller defined damage functions are used; for example, temperatures beyond 28°C are programmed to trigger 10% crop production declines irrespective of other processes.
SJOSs provide a heuristic to define undesirable thresholds according to system function and identity. To this end, modelling interacting pathways to SJOSs will further operationalise TBM as a proactive tool. Models relating to a regional specific problem may be built with knowledge of ecological, social and social- ecological complexities, including feedbacks and interactions, in order to help deliver realistic governance options relative to system complexity and uncertainty (Verburg et al., 2016). In turn, uncertainties are communicated through multidimensional evaluation, including the capture of ex-post complex behaviours. The next subchapter applies these principles to deltaic environments, before the social-ecological settings and problematic behaviours of the Chilika lagoon are introduced.
2.4 Deltaic social-ecological systems
2.4.1 Concepts of deltaic sustainability
River deltas are the environments of interest for DECCMA (Lazar et al., 2015), requiring the contextualisation of deltaic SESs, functions and sustainability. In geomorphological terms, deltas are systems of fluvial and marine sediment accumulation at the fluvial-coastal interface (Seybold et al., 2007). Natural deltaic functions include the regulation of freshwater, sediment and nutrients for the maintenance of marine fish stocks, vegetation, coastal birds and mammals (Scheirmeier, 2004). Over 500 million people currently inhabit deltas, meaning contemporary identities must include modifications of the natural fluvial, sedimentary and biotic processes which have evolved over the past 7000 years (Holmes, 1993, Overeem and Syvitski, 2009).
Biophysical boundary conditions have reached levels unexperienced during the Holocene (Renaud et al., 2013) – threatening the future ability of deltas to sustain human activities. Since pre-industrial times, combined global land and sea surface temperatures have risen by up to 1°C (IPCC, 2014b), driven primarily by anthropogenic greenhouse gases emissions. Projected impacts on DECCMA regions include reduced rainfall across Western Africa (McCartney et al., 2012)
36 Chapter 2: Literature review and the evolution of monsoon timings across the Indian subcontinent (Pervez and Henebry, 2014). Damming for hydropower and irrigation is reducing freshwater and sediment flows to coastal systems (Gupta et al., 2012), compounded by evolving hydrological frequencies, averages and extremes that threaten regional irrigation, water and food securities (Warner et al., 2010).
Sea-level rise adds to terrestrial-borne threats through coastal flooding (Timmerman and White, 1997), salinisation (Renaud et al., 2015) and population displacement (Werners et al., 2013). Global sea level rose 190 mm between 1901 and 2010 (IPCC, 2014b), with conservative estimates predicting 260-550 mm by 2100 (IPCC, 2014b). Populated deltas are vulnerable to effective sea-level rise from enhanced surface compaction from anthropogenic processes. Of the DECCMA deltas classified by Syvitski et al. (2009), the Mahanadi is considered at risk to effective sea-level rise, whilst the Ganges-Brahmaputra-Meghna is ‘in peril’. Ericson et al. (2006) project the livelihoods affected by 2050 across the Mahanadi, Volta and Ganges-Brahmaputra-Meghna deltas to be 102,000, 6,320 and 3.43- million respectively. Whilst neither study considers alternative futures or governance, they explicate the ‘perfect storm’ (Beddington, 2009) of stresses expected to impact deltas over the 21st century.
These stresses are pertinent because deltaic subsystems exhibit alternative states: fisheries collapse, agricultural plains become infertile and mangrove communities may range shift (Rocha et al., 2015). Different deltaic subsystems are dependent on the same supporting and regulating ecosystem services (Brondizio et al., 2016), meaning changes in boundary conditions may disproportionately affect regional provisioning services like agriculture and fish catch. Synchronously failing deltaic subsystems may transfer dependencies elsewhere through domestic and international migration (Dammers et al., 2014). In stability landscape terms, deltas are experiencing shrinking desirable basins, with human activities operating amidst increasing socioeconomic and environmental variability.
Consequently, various frameworks conceptualise deltaic sustainability. Day et al. (1997) argue deltaic sustainability equates to positive surface accretion, positive net primary productivity and economic outputs that outweigh inputs. Whilst appreciating geomorphological, ecological and economic components of deltas, social-ecological interactions are absent. Moreover, the economic classification suits systems with developed secondary and tertiary economies, downplaying the sustainability of systems dependent on primary activities.
37 Chapter 2: Literature review
Overeem and Syvitski (2009) develop an integrated framework of deltaic sustainability, assessing whether the management of natural, human and composite stresses sufficiently limit socioeconomic vulnerabilities. Deltas may be systematically assessed with each stage adaptable to different contexts. Similarities exist with the Sustainable Livelihoods Approach (Scoones, 1998) and the Vulnerability Framework (Turner II et al., 2003a), including identifying sections of society most at risk, classifying human and natural threats, and seeking management opportunities. Moreover, the framework is problem focused, centred on vulnerability mapping to prioritise management. However, such data might be unavailable for systems crossing administrative boundaries (e.g. Ganges-Brahmaputra-Meghna delta), or, where social-ecological problems are recent phenomena.
Coherent with complex adaptive systems theory, Dammers et al. (2014) decompose deltaic SESs into interconnected layers and networks of ecosystem services. Analysing interconnected subsystems appreciates that sustainable agriculture is unachievable without adaptive water management, requiring understanding of geomorphological and socioeconomic networks, for example. However, the application of complex adaptive systems theory to ‘trapped’ subsystems is challenging. Poverty traps may limit access to alternative livelihoods, whilst rigidity traps might delay transitions to sustainable pathways. Systems with legacies of resource variability might be especially vulnerable to rigidity traps by misinterpreting regime shifts as normal functions.
A recent edition of Sustainability Science focused on deltaic sustainability. Renaud et al. (2016, p.522) emphasise the need to understand “multiple interactions and feedback loops between the components of [deltaic] SES at multiple spatial scales”. Brondizio et al. (2016) develop an approach to define deltaic SESs and their holistic sustainability by spatially bounding social-ecological interdependencies and feedbacks, before examining ‘sustainability action situations’ for solutions to conflicts blocking sustainable trajectories. The assumption that sustainability inversely relates to the number of collective action problems is reasonable for systems with unequal resource appropriation. Resource access differentials exacerbate during resource degradation, as the socioeconomically advantaged exercise greater adaptive capacity, like commuting to fish stocks or purchasing heat resistant crops. Consequently, the system may enter ratchet-mode if the disadvantaged intensify activities during low productivity. Therefore, collective actions emphasise the need to explore
38 Chapter 2: Literature review
Social-economic
socio-demographic, economic ecological economic governance
governance Ecosystem Governance -resource
economic governance ecological ecological
Topographic Ocean- -hydrological climate-hydrological, climatic material
Figure 2.3: The deltaic domains (black) of Brondizio et al. (2016) arranged as a conceptual model and labelled with cross-domain interactions (blue). heterogeneities within deltaic subsystems, which is arguably less feasible using top-down vulnerability analysis and indicators.
The model of Brondizio et al. (2016) (figure 2.3) explicates interconnectivity between natural deltaic processes and anthropocentric functions. Yet, Sebesvari et al. (2016) discover indicators of ecosystem susceptibility are underrepresented in 236 vulnerability measures of the Amazon, Mekong and Ganges-Brahmaputra- Meghna deltas. Socioeconomic indicators dominate even for studies with stated social-ecological focus – surprising given the dependency of modern deltaic functions on nature and the ecological origins of resilience. Future studies must address this human-nature imbalance, involving indicators of ecological shifts with socioeconomic feedbacks (Hughes et al., 2013). Sebesvari et al. (2016) additionally note the dominance of single hazard studies. Problematically, simultaneous stresses may be attributed to background noise in single hazard studies, underestimating stresses on both human livelihoods and nature.
Deltaic sustainability has gained traction over the past 20 years, evolving from binary classifications to spatiotemporal complexities. Studies of deltaic sustainability are aligning with SESs principles, including the study of feedbacks
39 Chapter 2: Literature review and cross-domain interactions. Yet deltaic sustainability remains predominantly anthropocentric, focusing on the persistence of human functions over underlying natural conditions. This prioritisation either reflects anthropogenic issues blocking sustainable ecosystem use (Brondizio et al., 2016), or, that social- ecological feedbacks are still poorly understood.
2.4.2 ‘Safe and just’ futures for deltaic and wider social-ecological systems
Given the multidimensional characteristics of deltas and sustainability, it is logical to question methods of persisting desirable social-ecological functions in future. Application of threshold dynamics to deltaic SESs is sparse, with Hossain et al. (2017) providing the only exploration of deltaic SOSs to date. Therefore, the next step is to investigate pathways balancing ecological conservation with just socioeconomic development across deltaic social-ecological systems.
Fabrice Renaud and colleagues have championed deltaic social-ecological thresholds. Renaud et al. (2013) conceptualise three deltaic tipping points, occurring: (i) once an ecosystem service cannot be relied upon, (ii) once stresses are uncontrollable, and (iii) when management is forced, such as the construction of flood defences. These thresholds apply to real SESs with different sustainability implications. For instance, unreliable resource availability suggests a variability driven shift where stress relaxation may trigger recovery (Dearing et al., 2014). Oppositely, loss of internal control implies irreversibility if external stresses persist. Based on these tipping points, Renaud et al. (2013) classify deltas as either ‘Holocene’, ‘mid-Holocene’, ‘Anthropocene’ or ‘collapsed’. Humans have altered the natural Holocene functions of Anthropocene deltas (e.g. GMB and Mekong deltas), whilst humans have abandoned collapsed deltas. To date, no deltas have completely collapsed; however, increased out-migration from deltas like the Ganges-Brahmaputra-Meghna and Indus make such occurrences plausible under future trends of climate change, uninhabitable land, and food and livelihood securities (Szabo et al., 2016).
Renaud et al. (2013) briefly note that transitions from Anthropocene to collapsed states become less likely if SOSs are found. A key challenge is operationalising this truism, perhaps through the prioritisation of bistable variables and the monitoring of wider system conditions relative to critical levels. Alternatively, de- engineering may enhance deltaic sustainability by encouraging fluvial pulses associated with ‘Holocene’ deltas (Renaud et al., 2013). Promotion of natural
40 Chapter 2: Literature review variability matches the findings of Carpenter et al. (2015b) for SOSs; however, the removal of engineering structures becomes less feasible once a SES is dependent on man-made modifications. Consequently, tipping-back deltas may be hysteretic from the financial costs of de-engineering and social-ecological unpreparedness (Renaud et al., 2014). Similarly, van Staveren and van Tatenhove (2016) introduce the concept of delta trajectories to envision alternative hydrological governance. The authors argue that the hard-engineering flood defences of the Rhine-Meuse have caused technological lock-in, blocking implementation of environmentally desirable eco-engineering. Thus, like deltaic tipping points, delta trajectories identify when contemporary policies become unsustainable.
Therefore, envisioning future deltaic conditions requires projecting interconnected social-ecological characteristics, asking questions like: how might future resource availability evolve? How resilient are users to resource declines? These questions involve deep uncertainty around human responses to novel outcome states, threshold locations and system reversibility, meaning best guesses are inappropriate to tackle problems at hand (Lempert et al., 2003). Deep uncertainty is central to the plausible futures paradigm due to multiple environmental trajectories, individual actions and collective choices (Bai et al., 2016, Maier et al., 2016). In turn, concepts such as desirable futures, cautionary limits and critical conditions frame normative futures (e.g. we aim to achieve 10,000 tonnes/year fishery productivity) and emergent dynamics (e.g. how do we achieve 10,000 tonnes/year?).
Plausible normative futures can be bound by maximums and minimums, before incorporating knowledge of past conditions to constrain future trajectories (van der Leeuw et al., 2011, Maier et al., 2016). For a deltaic fishery like Chilika, minimum possible productivity equals zero tonnes/year. Although maximum productivity may be perceived to be limitless, a plausible biophysical limit corresponds to the ecological carrying capacity. In turn, a spectrum of alternative futures between maximum and minimum outcomes assume equal likelihoods of occurrence, overcoming the over-attribution of confidence often labelled at conventional deterministic scenarios (Börjeson et al., 2006, Maier et al., 2016).
A spectrum of plausible futures also implies a continuity of outcomes and driver combinations, where the difference between adjacent pathways is small. Oppositely, Carpenter et al. (2015a, p.2) argue for “strongly contrasting outcomes” when designing plausible futures for the Yahara watershed, USA. Such divergence suits the approach of envisioning futures from the mental models of stakeholders.
41 Chapter 2: Literature review
Figure 2.4: A sequential approach to operationalise regional SJOSs: (A) construct a model to investigate problematic behaviors; (B) simulate future dynamics under arrays of external driver trajectories; (C) compare outcome dynamics to plausible normative futures relative to today’s norms (Dearing et al., 2014), like safe and just (green), cautionary (orange) and dangerous (red) outputs; (D) trace outcomes to drivers to depict permissible pathways.
However, divergent scenarios arguably hamper TBM as tipping points across regional systems like fisheries and reefs arise from continuous driver changes (May, 1977, Scheffer et al., 2001). Therefore, a deltaic SJOS must detect subtle changes in driver-outcome relationships. To this end, simulation models facilitate automatic and gradual parameter adjustment to explore system boundaries – comparatively infeasible with knowledge-based models.
Relating to the earlier debate on the temporal aspects of TBM, Maier et al. (2016) bind plausible futures to time, suggesting normative targets should be achieved
42 Chapter 2: Literature review at an established point in future. However, sporadically observing system conditions may omit periods of undesirability (Thrush et al., 2016). Therefore, a balance must be struck between the number of variables with plausible futures, their divergence and temporal resolution. To this end, normative scenarios can assess SJOSs of bistable variables, for example, exploring driver pathways towards collapsed states of resource availability. In turn, monitoring internal responses to external forcing (e.g. climate change and globalisation) can explore social-ecological resilience to cross-scale interactions.
The above discussions create a framework to design SJOSs (figure 2.4). A model is constructed to simulate key natural, human and coupled interactions underpinning sustainability across a regional system like Chilika (figure 2.4A). Once the model is structurally and behaviourally robust, arrays of plausible driver trajectories explore outcome dynamics across the spectrum of driver interactions (figure 2.4B). On top of a broad range of plausible futures (figure 2.4C), configurations sustaining desirable system outcomes are modest desires for any system, especially those with legacies of past shifts (Crépin et al., 2012). Sustainable guardrails may be derived from a priori conditions, serving as a conservative warning for undesirable futures (Bai et al., 2016), whilst thresholds relate to minimum required conditions, undesirable divergence from long-term trends or nonlinearities (Dearing et al., 2014). Driver-response relationships are then traced through system structure, exploring whether normative futures correspond to distinct environmental or socioeconomic patterns (figure 2.4D).
2.5 The Chilika lagoon social-ecological system
The Chilika lagoon is Asia’s largest brackish water ecosystem (Mishra and Jena, 2015), located on India’s eastern coast in the state of Odisha (figure 2.5). The lagoon formed as sediment deposition enclosed water from the Bay of Bengal approximately 5,000-6,000 years ago (Sahu et al., 2014). As a biodiversity hotspot, Chilika is as a World Heritage site and Ramsar wetland of international importance (Sahu et al., 2014, Mohanty et al., 2015). The lagoon hosts Irrawaddy dolphins (Peetabas and Panda, 2015), 800,000 migratory birds (Kumar and Pattnaik, 2012) and 317 species of fish (Mohanty et al., 2015). The tropical monsoon climate drives area variations from 900 km2 during the non-monsoon (November-June) to 1100 km2 during the monsoon (July-October) (Pattanaik, 2007). Approximately 800,000 people live within the watershed, of which 35,000 are primarily dependant on Chilika’s fishery (Noack et al., 2013). Total fishery
43 Chapter 2: Literature review production was reportedly 12,000 tonnes in 2015, worth US$24 million to the regional economy (CDA, 2016). To begin exploring associated social-ecological thresholds, the system’s natural and anthropogenic structures and interactions must be understood (Ostrom, 2009, Brondizio et al., 2016).
2.5.1 Biophysical settings
Chilika is bordered by agricultural land to the north and west, a sediment spit to the south, and forest and marshland to the east (Kumar and Pattnaik, 2012). The lagoon receives freshwater and sediment from two hydrological sources (figure 2.5B): (i) the Mahanadi, bifurcating 80 km north at the Naraj barrage and contributing ~70% of Chilika’s annual freshwater and sediment influxes (Kumar and Pattnaik, 2012); (ii) western (WC) and northern catchments. The WC originate from the neighbouring Eastern Ghats (Mishra and Jena, 2015), whilst the latter is the southward extension of the Mahanadi downstream of Naraj. The WC is ephemeral, only transporting freshwater and sediment in response to monsoon rains (Santra and Das, 2013). Despite this, Chilika has an annual sediment surplus: 800,000 tonnes of fluvial sediment was deposited in 2013 but only 130,000 tonnes was flushed to the Bay of Bengal (Pattnaik and Kumar, 2013). Moreover, ~100,000 tonnes of littoral sediment enters the lagoon annually (Rao, n.d).
Das and Jena (2008) review Naraj operation, describing the trade-off between supplying adequate downstream irrigation and diverting hydrological flows from Chilika to avoid sedimentation. Concluding pre-barrage conditions (pre-2003) were key to the lagoon’s 20th century sedimentation, the authors recommend diverting Mahanadi flows from Chilika once undivided flow exceeds 15,570 m3/s, occurring up to two days each year (CWC, n.d). Mishra and Jena (2015) argue that sediment delivery to Chilika has unintentionally increased since Naraj because the Mahanadi’s downstream branches now rarely exceed bankfull discharge.
According to Mishra and Jena (2014), evidence for the northwards migration of the “Magarmukh” tidal outlet originates from Stirling (1846). In essence, longshore drift shifts the lagoon mouth north-eastward, dampening fluvial sediment flushing and tidal ingress. Resultantly, average lagoon salinity declined to below 5 parts per thousand (ppt) during the 1990s, triggering the proliferation of freshwater aquatic macrophytes (Rajawat et al., 2007). The resulting marine disconnect is thought to have constrained the migration of anadromous broods entering marine spawning grounds and the in-migration of catadromous juveniles (Mohapatra et al., 2007b, Pattanaik, 2007, Sahu et al., 2014). In total, 70% of the
44 Chapter 2: Literature review fish stock annually migrating to and from the Bay of Bengal via the principal tidal outlet (Kumar et al., 2011, Kumar and Pattnaik, 2012), meaning fluid biophysical settings influence resource availability and the sustainable level of extraction. A new tidal outlet was dredged in 2000 to rejuvenate lagoon-sea connectivity; monthly salinity has since doubled within the central sector (Panigrahi et al., 2007) which hosts ~50% of fish biomass (Sahu et al., 2014). The new outlet is not stationary despite measuring half the length of the traditional outlet; Rajawat et al. (2007) find ~400 metres of northward erosion by 2004, which had increased to ~2000 metres by 2013 (Mishra and Jena, 2014).
In total, 62 species contribute to at least 1% of fish landings each (Kumar and Pattnaik, 2012). The value of catch ranges from less than 100 Indian Rupees (INR)/kg for freshwater juveniles to 1000 INR/kg for large prawn and crab species (e.g. Giant Tiger Prawn); the most profitable catches are also the rarest and most desirable (Kumar and Pattnaik, 2012). Three different estimates exist for Chilika’s resource capacity. Based on the method of Martell and Froese (2013) for data poor stocks, Noack et al. (2013) estimate an ecological carrying capacity of 57,000 tonnes, since revised to 35,000 tonnes by Noack et al. (2015). In contrast, the Chilika Development Authority (CDA) supports a carrying capacity of 27,000 tonnes (Mohanty – personal communication), estimated by the Indian Central Inland Fisheries Research Institute from the carbon productivity of the lagoon (CIFRI – unpublished).
Ultimately, Chilika’s wider biogeophysical system contributes to the regulation of resource mobility and abundance. Freshwater influxes affect habitat quality; marine and fluvial sedimentation influences fish migration; aquatic macrophytes provide refuge from fishing activities. Therefore, the resource stock would still vary without fishing efforts, creating the challenge of matching appropriate anthropogenic parameters to dynamic natural settings.
2.5.2 Anthropogenic settings
Archaeological evidence suggests Chilika has been fished for ~2,300 years (Nayak, 2014). The fishery is caste-based, meaning access rights historically relate to social status (Salagrama, 2006). The total active fisher population divides between traditional and non-traditional users, with the former operating non- motorised, wind-assisted boats, known locally as Naha, Danga and Dingi crafts (Nayak et al., 2001); the latter use crafts with outboard motors, enabling fishing further from landing sites and enhanced catch capacities (Ray and Garada, 2017).
45 Chapter 2: Literature review
Figure 2.5: Annual fish catch data from Chilika covering 1929-2015. Source: Iwasaki and Shaw (2009) and CDA (2016).
Traditional fishers originate from seven sub-castes of the Scheduled Caste (Iwasaki and Shaw, 2009), considered the most socioeconomically vulnerable in Indian society. Throughout the 20th century, resource access was moderated by cooperatives allocating fishing grounds to fishing villages (Adduci, 2009). Operations were artisanal until World War II, when exports were sent to the Indian military (Jones and Sujansingani, 1954). Catch rose abruptly into the 1950s due to commercialisation (Figure 2.6), before India’s ‘blue revolution’ during the late 1970s and 1980s further boosted annual catches via the introduction of aquaculture and technologically advanced practices (Dujovny, 2009). Production supported ~25,000 fishers by the mid-1980s, compared with 7,000 fishers during the 1940s (Mohanty – personal communication). Principally blamed on the unchecked sedimentation of the main tidal outlet (Kumar et al., 2011), catch collapsed during the late-1980s to an average of 3,100 tonnes/year in the 1990s, before eco-restoration in 2000.
Based on surveys within traditional communities, Sekhar (2004), Nayak and Berkes (2010) and Nayak (2014) argue that the marginalisation of traditional fishers can be traced to the first increase of fishing lease prices in 1965. Lease prices were doubled again in 1978, before renewal became annual rather than every three years in 1988 (Nayak, 2014). The arrival of non-traditional fishers
46 Chapter 2: Literature review during the 1980s led to the encroachment of traditional grounds (Mishra and Griffin, 2010), compounded by the state-wide legalisation of shrimp aquaculture in 1991 (Nayak, 2014). Whilst the decision was reversed in 2001, the industrialised exports of ‘pink gold’ led to both political and physical conflicts within previously cohesive fishing communities (Ghosh et al., 2006, Nayak, 2014).
Nayak (2014) notes that Chilika’s traditional fishers exhibit five coping strategies to counteract livelihood stress. It is estimated that 88% of fisher households were indebted by the end of the 20th century, suggesting a universal imbalance between costs and returns, and the dependency upon ‘middle-men’ for loans and marketing (Samal and Meher, 2003). Fishers also intensify or extensity practices when possible. Intensification commonly manifests as adopting more exploitative practices to boost efficiency, such as switching from traditional boats to motorboats; extensification involves spending more time fishing, travelling further distances for catch or becoming less selective with landings. In relation, Ray and Garada (2017) study the perceptions of livelihood automation across Chilika. Of the 450 households surveyed, >60% believe motorisation is linked to increased catch and profits, whilst only 4% perceive overfishing as a demerit of motorisation. The authors hypothesise that motorisation will further fishery commercialisation, widening the resource access disparity between traditional and non-traditional practices.
Diversification through alternative livelihoods is historically underutilised, stemming from a reluctance to give up fishing rights and the lack of alternative opportunities (Sekhar, 2004). Therefore, traditional fishers during the late-20th century contended with environmental degradation and socioeconomic changes. Competition grew for a dwindling resource, whilst liberal state governance promoted intensive practices and aquaculture entrepreneurs. Robson and Nayak (2010) and Nayak (2014) record the occurrence of permanent out-migration from fishing villages during the 1990s and early-2000s, signalling the exhaustion of coping strategies and the relative attractiveness of livelihoods like wage labour.
The devolution of powers from the state to the CDA in 1992 shifted the focus of governance from economic development to ecological maintenance. However, Chilika’s institutional structure remains monocentric and hierarchical (Mishra and Griffin, 2010), with the CDA embedded within Odisha’s Directorate of Fisheries and Directorate of Environment. Fisher cooperatives bridge the gap between formal governance and users, offering training programmes, social activities and local knowledge (Panda – personal communication). However, 95% of cooperatives
47 Chapter 2: Literature review are thought to be inactive (Nayak and Berkes, 2011), indicating the increased individualisation of fishing and centralisation of governance.
Despite the reported five-fold increase in catch since the 1990s, the new tidal outlet at Satapada is associated with numerous unintended consequences. From surveys within 20 eastern villages, Dujovny (2009) find that enhanced tidal exchange has impaired the ability of outer channel fishers to harness fishing nets safely (Dujovny, 2009). In addition, flow redirection has degraded the traditional outer channel, damaging fisher livelihoods along the eastern bank (Dujovny, 2009). Consequently, a physical disconnect exists between the lagoon and disadvantaged fishing communities, adding to perceptions of marginalisation (Nayak and Berkes, 2014).
Softer management schemes also exist. Since 2002, the CDA has managed 30 stations monitoring monthly samples of water quality and planktonic compositions. A further 10 real-time monitoring stations analyse water quality every 15-minutes. Regular data collection is complemented by sporadic efforts, such as fish tagging to understand species specific migratory dynamics (Pattnaik and Kumar, 2013). The CDA with regional non-governmental organisations involve locals in fish species observations and landing assessments. The integration of stakeholders into monitoring activities is hoped to create a bidirectional flow of benefits (Kumar et al., 2011): enhancing local stewardship through scientific knowledge, which feeds back as community generated information about the socioeconomic and biophysical conditions of the system.
Despite CPR status (Nayak, 2011, Nayak and Berkes, 2011), the Odisha Marine Fishing Regulations Act and Rules 1988 (OMFRA) outlaws activities within the migration channels between January and July, bans nets with mesh spacing below 10mm (‘zero nets’) and prohibits the capture of juvenile prawns and certain fin- fishes. However, governors and experts believe OMFRA to be universally neglected (Mohanty – personal communication, Pattnaik – personal communication). Besides OMFRA, the absence of catch quotas means fishers can practice at intensities deemed necessary to cover and develop livelihood opportunities, making the fishery liberal-utilitarian in nature (Sekhar, 2004). At present, the number of active fishers and boats may evolve unconstrained, leading the CDA’s principal fishery expert to believe:
48 Chapter 2: Literature review
“… [the] long term future of [the] stock is very bleak; if current trends continue then we may see collapse of [the] current stock within 15-20 years.” (Mohanty – personal communication)
2.5.3 Social-ecological dynamics
Indicators of production efficiency are commonly applied in fishery management to synthesise social and environmental dynamics (FAO, 2002). Average reported catch per unit effort (CPUE) from 2000-2010 was six times higher than the 1990s (Mohapatra et al., 2007a), suggesting restoration of both principal ecological (resource abundance) and socioeconomic (catch) conditions. However, this perspective changes across a 60-year timeframe: despite record catches, today’s CPU-boat (CPUB) is equal to the 1970-1989 average, whilst CPU-fisher (CPUF) is ~30% lower than pre-1960 levels (figure 2.7). Therefore, despite efficiency gains,
Figure 2.6: Chilika’s historical catch per unit boat (top) and catch per unit fisher (bottom) from 1958-2009. Source: Kumar and Pattnaik (2012).
49 Chapter 2: Literature review fishers perceive declining individual withdrawals from the lagoon (Nayak and Berkes, 2010). CPUF also overlooks asymmetries in resource accessibility and catch value, with non-traditional fishers exerting superior effort and selectivity (Nayak and Berkes, 2010). These trends combine with the unintentional outcomes of lagoon governance to contribute towards the perceptions of marginalisation.
Chilika has a legacy of nonlinearity when the system appears robust – meaning a temporally dynamic viewpoint must be adopted to appreciate that undesirable regime shifts are surprising and unpredictable. Iwasaki and Shaw (2009) identify three dynamics as key to Chilika’s social-ecological sustainability. Technological advance and human population growth increase fishing efforts and regional resource dependency. As catch becomes increasingly valuable to a growing number of livelihoods, future regimes shifts are likely to cause deeper socioeconomic impacts than in past. Iwasaki and Shaw (2009) also note the dependency of socioeconomic functions on the lagoon’s sedimentary dynamics through the regulation of resource mobility. Therefore, the trajectories of these three social and environmental dynamics of different drivers, gradients and periodicities are crucial to any discussion of Chilika’s future.
Nayak et al. (2015) hypothesise the existence of thresholds in Chilika’s resource availability, water quality, livelihoods and institutional settings, arguing that their management underpins sustainability. The authors do not provide quantification, but the location of these thresholds across different deltaic subsystems suggests transgression may trigger thresholds elsewhere in the system. Therefore, governance must appreciate interdependencies, targeting relatively connected leverage points and thresholds. For example, keeping water quality above critical levels may help to sustain resource availability and human livelihoods.
External to Chilika, the economic liberalisation of India in 1991 is thought to have exposed the system to externally driven fish prices and consumption demands, particularly from the burgeoning Fareast Asia market (Nayak and Berkes, 2014). The monocultural drive for high value catches may degrade the flexibility and diversity of practices (Young et al., 2006). In turn, the most vulnerable fishers may become poverty trapped during periods of environmental and resource variability (Ray and Garada, 2017), reducing system resilience to minor perturbations and increasing likelihoods of socioeconomic thresholds. To combat these undesirable futures, the CDA in partnership with Wetlands International have promoted the system’s ‘wise use’ to maintain ecosystem character for the benefit of users (CDA and WI, 2010, Kumar and Pattnaik, 2012). However, Kumar
50 Chapter 2: Literature review et al. (2011) argue that since ecorestorarion, Chilika’s fishery has occupied a win- lose trade-off, as the continued marginalisation of traditional livelihoods means that the socioeconomic benefits of resource recovery are not yet universally felt (figure 8.3). Therefore, finding pathways leading to win-win social-ecological futures presents a key research demand.
To date, models of Chilika have predominantly focused on lagoonal sediment dynamics (Jayaraman et al., 2007, Dube et al., 2008, Sundar et al., 2015). As for social-ecological dynamics, Kadekodi and colleagues integrate social and ecological processes into conceptual (Kadekodi et al., 2000) and simulation models (Kadekodi and Nayampalli, 2005). According to the mental models of Kadekodi et al. (2000), the main drivers of catch are: (i) outlet sedimentation, (ii) water salinity, (iii) aquatic ‘weeds’ and (iv) poverty. To avoid these drivers becoming unsustainable, the authors advise increasing lagoon average salinity to 15 ppt [echoed by Das and Jena (2008)], and also limiting the number of motor boats and shrimp exports. However, the magnitudes and timings of these limits are unsubstantiated.
Operationalising the conceptual model, Kadekodi and Nayampalli (2005) construct a SDM simulating the effects of fisher population density and shrimp production on water salinity and depth. To date, this is the only modelling study for Chilika attempting to simulate social and ecological processes. However, rather than creating a dynamic social and natural environment, future scenarios alter socioeconomic and environmental conditions one-at-a-time. The authors conclude the continuation of baseline trends of population density and shrimp prices would cause 50% salinity declines over 25 years, whilst afforestation may lead to slight ecological recovery.
Despite being the only of its kind for Chilika, the study has numerous limitations. Model construction is an undocumented black-box, besides one diagram detailing variable names and connections. Consequently, data reliability, model sensitivity and the ability to replicate past trends is not assessable. The study also excludes the opening of the Satapada outlet and the banning of shrimp aquaculture, which both occurred before publication. Salinity declines are projected over 25 years, when in reality, hydrological interventions caused a stepwise increase from 4 ppt to 10 ppt in 2000. Moreover, no feedbacks exist between the social and ecological modules, suggesting that underlying socioeconomic forces are only externally driven.
51 Chapter 2: Literature review
In synthesis, the appreciation of natural variability on Chilika’s natural resource dynamics emerged from the 1990s collapse and subsequent ecorestoration. The need for a social-ecological perspective resulted from the subsequent realisation that expert led remediation of the natural system has produced unintended consequences for stakeholders, which may intensify under climate change, population growth and fishery intensification. Therefore, there is a need to build upon the historically focused research to explore the future sustainability of the range of processes and feedbacks, including outlet sedimentation, resource dynamics, and internalised socioeconomic processes and heterogeneities. Incorporating emergent cross-scale interactions would involve evolving hydrological fluxes under climate change and cross-scale issues, like increasing exposure to globalised markets (Nayak et al., 2015). Alternative policies could explore internal resilience against external trajectories – helping to answer questions like may plausible exogenous changes push the system towards thresholds? How may social-ecological feedbacks change over time? What policies guard against thresholds to sustain outcomes and drivers within SJOSs?
2.6 Aim and objectives
The objectives of this study emerge from the literature themes and gaps, methodological possibilities and problematic behaviours of the case study. The future sustainability of the Chilika lagoon fishery is dependent on the deeply uncertain evolutions of the biophysical and socioeconomic feedbacks, external stresses and governance. More generally, there is a need for studies focusing on locational specific issues, regional datasets and context specific metrics of sustainability to build resilient systems in the face of local and global pressures (Hanspach et al., 2014, Dearing et al., 2015, Zhang et al., 2015).
Therefore, the focus of this research is to investigate the future sustainability of the Chilika lagoon fishery system, with specific reference to social- ecological pathways and thresholds. This overarching aim explores the broad topic of sustainability and the recently developed concept of SJOSs for regional SESs. Such open-endedness provides a central focus, whilst also giving the research flexibility to explore the many dimensions of sustainability and novel findings that emerge from model simulation.
To this end, the underlying objectives and associated research questions are:
52 Chapter 2: Literature review
1. To develop and evaluate a model projecting plausible future processes, states and scenarios of the Chilika lagoon fishery.
A. Does available qualitative knowledge and quantitative data permit the design and parameterisation of a useful model? B. Can the model generate future external scenarios assessing internal resilience to nonstationary social-ecological interactions? C. Do the future scenarios cover the boundaries of Chilika’s SJOSs? D. Does the model capture the historical reference mode, including outcome magnitudes and relative influence of drivers? In other words, are the ‘right’ quantitative behaviours produced for the ‘right’ qualitative reasons (Barlas and Kanar, 2000)? 2. To explore the impacts of internal governance on: (i) fishery persistence, (ii) permissible driver pathways and (iii) threshold interactions.
Through the analysis of model outcomes, this second aim seeks to operationalise the SJOSs concept as a forward-looking tool for decision-making. Specific research questions include:
E. What are the permissible limits of individual human and biophysical drivers, and how do these change with governance? F. Do SJOSs involve cross-scale interactions, where a global scale stress (e.g. climate change) undermines resilience to a system specific driver (e.g. fishing effort) (Scheffer et al., 2015, Green et al., 2017)? G. Are core safe and just pathways of critical system drivers identifiable from their wider SJOSs (Steffen et al., 2015b)? H. Can model outputs help judge the safety of the system’s contemporary trajectory?
3. To investigate the extent to which the concept of future SJOSs is useful for: (i) governing with respects to thresholds and tipping points, (ii) assessing social-ecological trade-offs, and (iii) guiding sustainable development.
The third objective helps to evaluate the utility of operationalising SJOSs for regional governance:
I. Do the normatively defined dangerous futures have deeper social- ecological effects beyond fishery livelihood losses? J. What are the trade-offs (e.g. social-ecological, inter and intra- generational) associated with safe and just pathways?
53 Chapter 2: Literature review
K. How does system safety compare when defined by nonlinear outcomes and driver-response relationships, rather than normative thresholds? L. To what extent does modelling SJOSs contribute towards achieving broader sustainability paradigms, such as poverty alleviation, the wise use of wetlands and social-ecological resilience?
Ultimately, the aim and objectives represent a workflow involving: (i) the construction and evaluation of a novel regional model, (ii) simulation to investigate the persistence of historical problems and the emergence of new stresses, (iii) the generalisation of location-specific results to contribute to the wider fields of sustainability and resilience thinking.
54 Chapter 3: Model construction
Chapter 3: A system dynamics model of the Chilika lagoon fishery system
3.1 Chapter introduction
This chapter describes the process of model development (figure 3.1) based on the SDM framework of Ford (2010) for natural systems with human influences. Whilst the subsections appear discrete, the iterative nature of modelling provides porosity; therefore, components traditionally associated with one phase may appear in another (e.g. causal loop diagrams help to communicate conceptual subtleties of parameterisation).
Here, problematic behaviours revolve around catch loss, historically considered to derive from ecohydrological degradation (Iwasaki and Shaw, 2009). Similar trends may reoccur under changing hydroclimatic scenarios, whilst fisher population and consumption growth threaten resource overexploitation. Chapter 3.2 builds upon the review of social-ecological settings by considering historical and future reference modes, envisaging initial future dynamics from fundamental system properties including finite resource availability, overexploitation and sustainability. As SDMs aim to derive internal explanations for problematic behaviours (Sterman, 2000), it is important to include only the internal and external processes critical to system dynamics. External drivers may be beyond the scope of managers (Meadows, 2009), whilst an overly inclusive model may become uninterpretable (Voinov and Shugart, 2013). Causal-loop diagrams trace the causes of problematic behaviours to inform model structure.
Model parameterisation uses a spectrum of available information (Ford, 2010). Physical laws are prioritised, including mass balances of population or water volume within a catchment. Two pre-existing SDMs are adapted to form the cornerstones of the wider model: the Ghashghaei et al. (2013) rainfall-runoff model and the Bueno and Basurto (2009) fishery model. Primary and secondary datasets also inform model parameterisation. Data from stakeholder and key informant interviews adds qualitative insights to model design, performance analysis and scenario formulation. Expert knowledge helps inform variables associated with high uncertainty and stakeholder knowledge supports variables with decision-making elements.
55 Chapter 3: Model construction
Figure 3.1: The overarching methodological framework of this study – from initial problem identification to communicating insights to regional decision-makers.
The SDM is constructed and run in STELLA v10.1 (ISEE, 2015). In a sentence, the model simulates a dynamic fish stock driven by ecosystem influences on the survival rate and migration via the tidal outlet, plus fish catch that is driven by nonstationary fishing efforts. To this end, the model splits into six interconnected modules to simulate human and natural drivers of catch. Four modules are involved in the generation of historical dynamics: (i) hydroclimatic – freshwater and sediment delivery to Chilika; (ii) ecohydrological – influenced by the hydroclimatic inputs, this module models macrophyte growth and the effects of salinity, dissolved oxygen and temperature on fish survival; (iii) core fishery – fish population and catch dynamics, and (iv) socioeconomic – processes underlying fisher population growth and fishing efforts. Two additional modules help
56 Chapter 3: Model construction generate future dynamics (chapter 5): (v) external trajectories – exploratory scenarios of rainfall, temperature, fish price, diesel price, and human birth and death rates, and (vi) governance –decision-making scenarios building internal resilience against external trajectories.
The model runs at monthly resolution as a compromise between two competing demands. Modelling daily time-steps would capture short-lived hydrometeorological events and intra-monthly catch variations; however, catch statistics and measures of fishing effort are published annually (CDA, 2015). Moreover, modelling multidecadal trajectories at daily time-steps would be computationally impractical. Simulating 444 months (January 1973 – December 2009) requires approximately 5 seconds, which roughly triples when simulating from January 1973 – December 2060. Therefore, an approximation for the duration of daily time-steps gives around 10 minutes per simulation, without considering the reduced computational efficiency of daily Monte Carlo climate simulations (chapter 3.4.1.6). Contrastingly, annual resolution potentially overlooks seasonal processes like monsoonal hydrological fluxes with ecohydrological affects, fish migration via the tidal outlet and variations in macrophyte coverage (Mohanty et al., 2001, Pal and Mohanty, 2002).
The model was built and ran on a 64-bit operating system (Intel® Core™ i5 processor), with 3.82 gigabytes of available random-access memory (RAM). Simulating the model forward 1000 times under the external scenario funnels required 20-30% CPU usage. Outputs were recorded using the in-built ‘table-pad’ function in STELLA, before tables were exported and stored as comma-separated values (CSV) for graphing and analysis in the statistical software R (2008). Analysis in R followed the sequential framework of chapter 6: first graphing and classifying future catch as safe and just, cautionary or dangerous, before analysing the causal pathways, wider social-ecological outcomes and trade-offs. No parallel algorithms were used, but noteworthy specialist R packages include: “ggplot2”, “forecast”, “reshape”, “rgl”, “strucchange” and “viridis” (all downloadable from https://cran.r-project.org/web/packages).
Model structures and behaviours are evaluated in chapter 4. Scenarios exploring interactions between future governance strategies, socioeconomic changes and natural conditions depict plausible system evolutions (chapter 5), from which resilience, SJOSs and decision-making principles are developed (chapter 6).
57 Chapter 3: Model construction
3.2 Conceptual modelling
To bridge the stages of problem identification and model parameterisation, Ford (2010) advises: (i) describing reference modes of problematic behaviours; (ii) considering the spatial and temporal origins of problematic behaviours; and (iii) communicating system structures, scope and assumptions – focusing modelling efforts by converting concepts into mathematical expressions.
3.2.1 Fundamental reference modes
Reference modes depict trends of problematic behaviours, derived from qualitative behaviours, time-series and statistical analysis (Saeed, 1998). Irrespective of the approach, behaviour conceptualisation is important to furthering system understanding and model design.
At the system level, the fish population may be considered a stock with inflows and outflows (Bueno and Basurto, 2009). The stock grows through births and in-
Figure 3.2: Causal-loop diagram depicting feedbacks relating the fish stock size, fishing effort and fish catch. Short-term catch may increase in response to intensified fishing effort but cause long-term stock depletion. Two horizontal lines represent a system delay longer than the average interaction.
58 Chapter 3: Model construction
Figure 3.3: Plausible reference modes for Chilika’s fishery: (A) catch exponentially increases due to infinite fish stock; (B) stock exponentially grows or declines if the extraction rate is respectively above (red) or below (green) sustainable (blue); (C) boom and bust dynamics; (D) boom and bust influenced by variable catchability. migration, and declines through natural deaths, out-migration and catch. Therefore, the absolute rate of change is driven by the magnitudes of the inflows and outflows at that given time, and the size of the stock at any given time is a legacy of past inflows and outflows (Sterman, 2002).
Fishing effort describes the volume of fishing gear used per unit time (FAO, 2002). For instance, effort across Chilika in 2003-2004 equalled 5,059 boats for 336 days (Mohapatra et al., 2007a). For a given resource availability, catch is generally positively associated with fishing effort. In reality, coupling often exists between effort and resource availability, with fishers adapting practices to perceived abundance changes (Tracey and Lyle, 2011, Duer-Balkind et al., 2013). Catchability modifies the effectiveness of efforts through fish interactions with gear and environmental conditions (Ricker, 1975, Jul-Larsen et al., 2003). These simple feedbacks integrating human and natural components build the fundamental dynamics of Chilika’s fishery (figure 3.2).
59 Chapter 3: Model construction
Assuming infinite resource availability and constant catchability, catch may increase exponentially as extraction efforts grow (figure 3.3A). However, in reality, the finite nature of the resource produces a negative feedback giving catch an environmental ceiling. Alternatively, a constant fishing effort below (above) the fish stock regeneration rate may cause the stock to increase (decrease) (figure 3.3B), in line with the traditional equilibrium view of fish population dynamics (Seijo et al., 1998). The third reference mode (figure 3.3C) assumes only constant modifying factors. Whilst it appears that the resource availability drives catch via a unidirectional relationship, the scenario reflects the delayed negative feedback between catch and resource availability (figure 3.2). High catch causes the fish stock decline, feeding back to cause catch decline. Eventually, extraction efforts may fall as fishers adapt practices during periods of low productivity – helping to trigger resource recovery. The fourth reference mode arguably best represents reality as dynamic modifying factors and delays in human perception perturb the peaks and troughs of catch and resource availability (figure 3.3D). Fundamentally, interactions between these interdependencies are central to multidecadal persistence, as desirable catches require both abundant resource availability and appropriate extraction forces.
3.2.2 Historical and plausible future reference modes
Considering the fundamental dynamics, Chilika’s catch increase between 1950 and late-1980s likely occurred from fishing effort growth. In agreement, Chilika’s fisher population increased from 19,000 to 25,000 from 1970-1990 (Noack et al., 2013). Therefore, a reference mode until the 1980s would depict stepwise growth due to three consecutive inter-decadal increases. Future forecasts using time- series techniques such as ARIMA may have projected the persistence of the upwards trajectory, unaware of the impending catch decline.
A reference mode until 1999 suggests either overshoot or natural oscillatory dynamics (Ford, 2010). Distinguishing between these modes is difficult when viewing behaviours through a limited time horizon and without understanding the underlying system drivers. Therefore, various dynamic hypotheses exist: (i) the fish stock collapsed having surpassed a biophysical carrying capacity (e.g. food availability, competition, density) (Daily and Ehrlich, 1992); (ii) the rate of fish extraction from human efforts exceeded stock renewal (i.e. overexploitation); (iii) catch declined due to an external driver influencing efforts or catchability, such as the destruction of fishing boats or a natural modifying factor. If
60 Chapter 3: Model construction
Figure 3.4: Potential futures resulting from the management of fundamental components and feedbacks (figure 3.2). The question mark refers to uncertainty associated with the timing of future resource decline. overexploitation is to be blamed, then constraints upon fishing effort may push the system further in the wrong direction, failing to rejuvenate the fish stock whilst affecting fisher livelihoods. Alternatively, modifying factors may be prioritised, such as the clearance of vegetation from fishing grounds, but this policy risks prolonging collapse if overexploitation is the real cause.
The observed catch recovery post-2000 is attributed to the dredging of a new tidal outlet that boosted marine water exchange, decreased aquatic macrophyte coverage and enhanced fish migration (Mohanty et al., 2008, Iwasaki and Shaw, 2009, Sahu et al., 2014). Sedimentation of the traditional outlet had two effects on resource availability (figure 3.1). First, fish stock renewal declined as the migration of Catadromous, Anadromous and Amphidromous fish decreased. Second, fish survival within the lagoon fell as salinity levels freshened and oxygen levels depleted under stagnant water conditions and abundant macrophyte coverage. Various secondary stresses activated, including the exploitation of a dwindling stock, less-selective catch of juvenile species (Ghosh and Pattnaik, 2003, Kumar and Pattnaik, 2012) and irresponsible fishing within the migratory outlet channels (Kumar and Pattnaik, 2012).
61 Chapter 3: Model construction
Akin to qualitative plausible futures (Gallopín, 2002), future reference modes based on fundamental feedbacks provide initial horizon scans (Saeed, 1998, Khan et al., 2004). If the outlet channel is neglected, or the fish stock is overexploited, then Chilika may remain within a boom and bust mode (figure 3.4). Time between peaks and troughs depends on the rate of lagoon sedimentation, the adaptive capacities of fishers and the implementation of remedial governance (Biggs et al., 2009, Meadows, 2009). Therefore, regular outlet maintenance may conserve the fish stock. If fishing efforts remain unregulated and proportional to the fisher population, then annual productivity may gradually increase (figure 3.4). However, with a finite fish stock, susceptibility to overshoot increases as catch surpasses environmental limits, such as maximum sustainable yield (MSY) (McKinney et al., 2007, Maunder, 2008). Moreover, because fish stock and fisher population dynamics are relatively slow, interannual catch variability derives from interactions between efforts and catchability. Plausible management approaches are recognisable, such as limits on extraction efforts and the maintenance of healthy ecohydrological conditions. Hence, a model exploring the social- ecological impacts of policies is valuable to decision-makers, enabling the assessment of alternative pathways.
3.2.3 Setting system boundaries
Whilst top-down dynamics provide an entrance point (Sterman, 2000), underlying causes must be bound to guide model specifications, data collection and performance analysis. Lewis-Beck et al. (2004) and Walker et al. (2012) identify external variables as those unaffected by variables internal to the system. Internal variables must affect the system, and then the resulting change in conditions must affect the same internal variable, possibly through governance responses. External drivers are commonly seen to shock, due to challenges of forecasting beyond system boundaries. However, external stresses may be creeping, such as the timing and magnitude of monsoon rainfalls across Odisha altered by climate change over the 21st century (Ghosh et al., 2010). Moreover, external processes may be located within a system’s physical boundaries. For instance, rainfall is geographically internal to Chilika’s watershed, yet exogenous in a systems sense due to the lack of endogenous control.
Causal-loop diagrams help to organise drivers across system space (Sterman, 2000). Sterman (2000) and Kim (1991) outline various requirements for causal- loop diagrams:
62 Chapter 3: Model construction
Link direction corresponds to the direction of causal effect. Link polarity depicts whether the causal effect is positive or negative. Links read ceteris paribus, meaning other relationships remain constant. The polarity of a feedback equals the product of polarities within the loop. The inability to describe a link as positive or negative probably means additional variables are needed to clarify the relationship. Two horizontal lines flag delays significantly longer than the average interaction.
3.2.3.1 Natural processes
Natural processes influencing fish stock size and catchability include climatic variations, fluvial processes and the lagoon’s ecohydrological conditions. By treating Chilika’s 317 fish species as a single stock within the Bueno and Basurto (2009) model, fish regeneration, maturation, migration and natural deaths establish the population dynamics. Fundamentally, the size of the immature stock is proportional to the regeneration rate; the mature stock is proportional to the number of maturations. A positive feedback forms, as more births produce more immature fish, increasing regeneration and so on (R1 – figure 3.5). However, the potential fish population is not infinite – counterbalanced by an environmental carrying capacity acting as a natural thermostat to constrain population growth (Rose et al., 2001, Huggett, 2004). Therefore, the rate of population growth decelerates as the stock approaches its carrying capacity, assuming competition for food, energy and space increases (Heath, 1992, Rose et al., 2001). As a single species model, ecohydrological degradation is assumed to have a unidirectional influence on stock dynamics (Collie et al., 2016); therefore, fish survival rate for a given population density is also inversely proportional to ecosystem degradation. Moreover, internalising the lagoon-sea migration of ~70% of species is paramount to Chilika’s long-term dynamics (Kumar and Pattnaik, 2012).
Stock aggregation necessitates various assumptions and exclusions. Ideally, trophic interactions between different food-web components would be modelled (Mace, 2001), including inter-species predation, energy consumption and growth efficiencies. Such ecosystem-based modelling is facilitated by models like ECOSIM (Walters et al., 1997) and ATLANTIS (Fulton et al., 2014). Therefore, it is acknowledged that this SDM cannot inform challenges like optimal mortality rates of commercially important species, the effects of fishing down food webs (Pauly et al., 1998) or the emergence of non-native stocks. Increased biological realism
63 Chapter 3: Model construction
Figure 3.5: Causal-loop diagram illustrating the components, links and feedbacks underlying the fish stock. Italicised variables refer to stresses external to the fish population. Feedbacks: R – reinforcing, B – balancing. comes with practical complexities and data demands (Quinn and Collie, 2005, Collie et al., 2016). To avoid becoming overly complex, ecosystem-based models conventionally omit social-ecological adaptive behaviours (Plagányi, 2007, Collie et al., 2016). The SDM developed here seeks to identify multidecadal pathways sustaining a natural resource of Chilika’s size, mobility and historical dynamics. Therefore, the model does not intend to replace ecosystem-based models, but offer a complimentary technique to horizon scan broad dynamics representing dividing lines between sustainable pathways and collapse.
Despite resource aggregation, it is important to recognise that environmental conditions influence the biological traits of fish (Keyl and Wolff, 2008, Keck et al., 2014), including reproduction, growth and mortality. Salinity is considered the critical physiochemical indicator of Chilika’s health (Jayaraman et al., 2007, Parida et al., 2013), with brackish and marine species constituting 80% of total species.
64
Chapter 3: Model construction
The feedback feedback The
. Variables such as ‘total sediment influx’ consisting consisting influx’ sediment as‘total such Variables .
of Chilika of
clarity. Dashed lines are for visual clarity. visual arefor lines Dashed clarity.
visual
population
fish fish
aggregated for for aggregated
rological influences on the aggregated aggregated onthe influences rological
of both LMC and WC sediment loads are loads sediment WC and LMC ofboth 3.5. asfigure are directions
: Ecohyd :
6
.
3
Figure Figure
65 Chapter 3: Model construction
Pattnaik and Kobayashi (2009) also consider air and water temperatures, pH and dissolved oxygen levels critical determinants of Chilika’s habitat quality, driving the development of immature fish and the death rates of mature fish. Salinity, water temperature and dissolved oxygen vary seasonally with monsoonal rainfalls and fluvial inputs. Aquatic macrophyte growth involves three feedbacks (figure 3.6). Macrophytes thrive in fresh environments (R2) and trap sediment within the lagoon (R3) (Kumar and Pattnaik, 2012). In turn, sediment trapping reduces tidal flushing and aggrades the sediment stock – positively associated with nutrients for macrophyte proliferation (R4) (Kumar and Pattnaik, 2012).
The Hirakud dam, located approximately 300km upstream of Chilika, forms a system boundary by controlling downstream discharges. Downstream of Hirakud, the LMC bifurcates into the Kathajodi, Kuakhai and Mahanadi branches at Naraj, at the head of the Mahanadi delta (Mishra and Jena, 2015). Consequently, freshwater fluxes at Naraj equal a relatively stable baseline from Hirakud, plus rainfall inputs across the LMC. According to Indian Meteorological Department (IMD) and Central Water Commission (CWC) data, 75% of annual rainfall across the region occurs between July and October, corresponding to 84% of annual Mahanadi streamflow. Streamflow from the undivided Mahanadi reaches Chilika once it exceeds 2830 m3s-1 (Das and Jena, 2008, Mishra and Jena, 2014).
As a result, hydrological changes along the Mahanadi have significant ramifications for Chilika’s ecology (Mishra and Jena, 2014, 2015). Monitoring streamflows upstream of Naraj may become increasingly important, as climate change is predicted to increase the magnitude and frequency of extreme rainfall events across the middle and lower Mahanadi (Panda et al., 2013, Jena et al., 2014). By 2030, Gosain et al. (2006) propose up to 29% and 117% increases in the magnitudes of the Mahanadi’s high and low flow regimes, respectively. Thus, discharge may exceed 2830 m3s-1 increasingly frequently, increasing freshwater and sediment inputs to Chilika.
The LMC and WC form the western boundaries of the physical system. Naraj at the delta head forms the northernmost component and the lagoon’s southern bank forms the southern boundary. The Bay of Bengal coastline forms the eastern boundary, with ~100,000 tonnes/year of littoral sediment entering the lagoon (Rao, n.d). Drivers such as wind direction, wave height and angle are excluded due to data limitations and the inappropriateness of a monthly resolution aspatial model to capture impacts upon fishery operations (e.g. possible routes for traditional boats on a given day). Future sea-level rise may alter the long-term
66 Chapter 3: Model construction salinity regime. Although sea-level rise scenarios could be linked to modelled temperature changes, the process is excluded here due to a lack of conceptual understanding (e.g. sea-level rise may increase salinity, but reduce flushing). Sub- monthly tropical cyclones and storm surges are also excluded; instead, the model focuses on resilience to longer-term precipitation and fluvial influxes. Further assumptions include no additional tidal outlet locations, no new dams upstream of Chilika and no major agricultural or industrial developments with substantial pollution outputs. Therefore, the model assesses the resilience of current hydrological system structures under nonstationary averages and variance.
3.2.3.2 Anthropogenic processes
Modelling systems with human decision-making is inherently complex, often involving concepts of bounded rationality (Carpenter et al., 1999), optimisation and collective behaviour. The dynamics of the fisher population, fleet sizes and decisions regarding fishing activities are internalised here.
The fisher population could be driven externally by giving the model predefined population scenarios. Conversely, internality assumes that fishers and efforts depend on the persistence of the fishery. If interdependency exists, then population growth may interact with a carrying capacity, such as food availability or the number of viable fishing jobs (Hopfenburg, 2003). In the sustained absence of fish, it can be assumed that the number of active fishers will tend to zero in the search for alternative income sources. Fishers might not migrate from the region, but would retire practices with costs outweighing benefits. This link can be inferred from Robson and Nayak (2010) and Nayak and Berkes (2010), with the first observations of out-migration occurring during the 1990s fishery collapse. Coherently, the fisher record suggests variability is driven by faster dynamics than only regional birth and death rates (figure 3.7).
The number of fishing boats can be internalised from the fisher population. The days boats operate may be constant, internally or externally driven. Mohapatra et al. (2007a) disprove the constant scenario, with days fished increasing from 306 (2000-2001) to 345 (2001-2002). One internal reason for increased fishing days could be the relaxation of regulations, but neither the CDA nor OMFRA limit the number of days fished (chapter 2.5.2). The functional responses of Holling (1959a) describe how predation varies with prey density, akin to fishers’ perceptions of fish abundance (Tracey and Lyle, 2011, Duer-Balkind et al., 2013). Type-II responses (Holling, 1959b) are arguably most suitable to describe the
67 Chapter 3: Model construction
Figure 3.7: Observed fisher population (red) (Noack et al., 2013) versus the trend which may have occurred if the link between fishery productivity and the fisher population was removed, making the population only driven by Odisha’s birth and death rates (black). effort-stock relationship, as in contrast to the linear or sigmoidal trends, the concave response maintains efforts until a point judged to be unsustainable. Therefore, rather than staying constant, decisions regarding fishing days can be internally derived by assuming fishers adapt efforts to the stock conditions (Moxnes, 2000, Duer-Balkind et al., 2013). Similarly, decisions to extract juveniles may be internally driven, as pressure to maintain productivity reduces time to ‘bycatch-clean’ hauls (Alverson et al., 1996).
Thus, three reinforcing and two balancing feedbacks are assumed to underlie fishing effort (figure 3.8). Stability afforded by fishers adapting efforts to the underlying stock (B3- figure 3.8a) is countered by two reinforcing feedbacks driving the fleet sizes of the traditional and non-traditional fishing communities (R5A and R5A – figure 3.8A). More livelihoods are supported as catch increases, boosting short-term catches. The uncapped number of boats potentially increases until catch shrinks from stock exhaustion.
Recorded CPUE includes non-traditional and traditional boats. Therefore, fishing effort may grow despite total boats remaining constant if the proportion of motorboats increases. Such intensification may be driven by traditional fishers deciding to switch practices (R6 – figure 3.8B) (Nayak, 2014). The ability to
68 Chapter 3: Model construction
Figure 3.8: (A) Causal-loop diagram illustrating drivers of fishing boat numbers and days fished; (B) secondary diagram underlying the transition from traditional fishing practices to more intensive non- traditional practices. The feedback directions are as figure 3.5 upgrade depends on various socioeconomic and cultural settings, such as financial capacities and information available for decision-making. The latter factor reflects the Baker criterion (Sterman, 2000), as SDM variables must only relate to information available to decision-makers. The desire to upgrade practices is countered by B3 (figure 3.8), whereby a shrinking traditional fisher
69 Chapter 3: Model construction population reduces traditional productivity, causing incomes to fall. Moreover, not all fishers affording upgrades choose to leave traditional practices (R7 – figure 3.8). Therefore, upgrade desire is assumed inversely proportional to the state of the traditional fishery (B4 – figure 3.8), meaning the perceived benefits of adaptation through intensification are greater during fishery slumps (Coulthard, 2008, Nayak, 2014).
External socioeconomic stresses also influence the fishery. The per unit price of Chilika’s fish is inelastic (Kadekodi and Nayampalli, 2005), with exportation linking fish price to national and international supply and demand. Government subsidies reduce fuel costs, but have limited impact on revenues during catch slumps. Boat maintenance and upgrade costs have also risen over the past fifty years, driven in-part by international prices (Kadekodi and Gulati, 1999, Noack et al., 2013). Financial loans may be internal to the system, as a positive feedback exists between income and loan taking, with the poorest most susceptible to poverty compounding loans. External factors complicate the modelling of loans as an internal process, such as educational attainment, marketing of catch and financial assistance from cooperatives, family and the state (Pattanayak – personal communication). Various other omissions are due to data absences or expert perceptions that a process is unimportant to Chilika’s multidecadal sustainability. These include defector behaviours (e.g. illegal aquaculture, individual fishing, ignoring the regulations of chapter 5.3), differences between daytime and night- time fishing, and technological step changes (e.g. marine trawlers).
The above causal-loop diagrams inform the structure of SDM parameters. Exact parameterisation methods vary depending on the presence of already constructed models, physical equations and data availability. The Mahanadi fieldtrip of spring 2016 gained additional qualitative insights informing holistic system structure, parameter values and future scenarios. Therefore, the next subchapter outlines the interviews with primary stakeholders and governors, before chapter 3.4 details numerical parameterisation.
3.3 Interviews
Visiting the Mahanadi delta in spring 2016 aimed to improve model structure, parameterisation and performance by quizzing the mental models of stakeholders and experts. System governors are key sources of knowledge, balancing different interests across dynamic socioeconomic and environmental
70 Chapter 3: Model construction conditions. Moreover, stakeholders live, work and learn to adapt to system conditions (Voinov and Bousquet, 2010). These qualitative insights consider knowledge beyond quantitative expressions of system dynamics, whilst reminding the modeller that processes important to stakeholders must be included (Meadows, 2009).
The social-ecological systems framework helped inform the choice of participants (Ostrom, 2009), with interactions between ‘governance’ and ‘users’ contributing to the overall dynamics of natural resource systems. Moreover, these groups may provide alternative perspectives of Chilika’s past and future. Fisher perceptions derive from deeply engrained histories, traditions and practices, whilst decision- makers and regional scientists generally hold relatively deterministic perceptions developed from environmental monitoring, economics and social sciences.
Semi-structured interviews were conducted for various scientific and practical reasons. Semi-structured interviews allow the interviewer to set the agenda (Scott and Garner, 2013, Yeo et al., 2014), yet, questions are adaptable to the interviewee’s responses. Compared to structured interviews, the interviewee can diverge from the line of questioning to share relevant knowledge or engage in two-way questioning (Denscombe, 2007). Such flexibility provides opportunities for participant cross-validation by asking questions designed to compare responses (Scott and Garner, 2013).
Structured surveys are comparatively unsuitable due to their inflexibility and reliance on the interviewer’s preconceptions. Time constraints and a 68% adult literacy rate amongst Chilika’s fishers (Kumar and Pattnaik, 2012) make questionnaires impractical. Although group interviews may broaden the opinions captured, the technique is less preferable due to potential power dynamics arising from mixes of caste, class, sex and age (Riley et al., 2003). However, initials observations realised that fishers socialise in groups outside of fishing hours; therefore, the maximum group size was set to four (Eder and Fingerson, 2001). Lastly, focus groups are beyond the scope of this study due to potential data volumes which may confuse rather than inform model construction (Leung and Savithiri, 2009).
3.3.1 Key informants
Key informant interviews aim to understand: (i) the causes and impacts of past problematic behaviours; (ii) future expectations for drivers and outcomes; (iii)
71 Chapter 3: Model construction
Figure 3.9: The seven key informants interviewed placed within the social- ecological systems framework structure of Ostrom (2009). different management strategies shaping the future sustainability of Chilika. Although specific questions differ between interviews, these themes aim to capture system-wide dynamics, with the first two providing strategies to assess the reality of model outputs and the latter constraining governance scenarios. According to Tremblay (1957), key informants must have continuous exposure to the dynamics under study and be willing to share knowledge in a communicable and impartial manner. Practical considerations included travel times and the availability of regional scientists to facilitate meetings. The seven key informants interviewed derive from four institutions (figure 3.9): three informants are from governmental organisations and four are from non-governmental backgrounds.
CDA knowledge derives from 25 years monitoring and managing the fishery, including the protection of the ecosystem and development of socioeconomic conditions. Science and regulations are communicated to fishers through the cooperative societies (Mishra and Griffin, 2010), whilst productivity is communicated to the Directorate of Fisheries to inform state-wide policymaking. The Integrated Coastal Zone Management Project (ICZMP) monitors coastal ecosystems of Odisha and associated socioeconomic development. Based in Bhubaneswar, the ICZMP is geographically external to Chilika, but their research directly affects fishery livelihoods and policymaking.
The key informant interview strategy focused on non-leading direct questions, with opportunities for clarification. For convenience, the interviews took place in places of work. The maximum duration was one-hour, but the median time equalled 45 minutes. The opening question(s) focused on the role of the
72 Chapter 3: Model construction participant in the governance and study of Chilika (appendix A.2). Fact-based questions were designed to ease the interviewee into the discussion and confirm the background of the participant (Denscombe, 2007). Interviewees were then asked about Chilika’s past dynamics, particularly focusing on the 1990s collapse, requiring the hindcasting of mental models. Next, questions about present status and plausible futures asked interviewees to project forward their mental models. These questions often prompted discussion of different management options; for example, interviewees raised successes of the ICZMP’s alternative livelihood programmes.
Lessons from the key informant interviews are highlighted when relevant during descriptions of model parameterisation, scenarios and results, although general insights can be summarised here. First, the interviews flagged vital socioeconomic and biophysical stresses that had been excluded from the model’s structure, like juvenile catch and the perceived short-sightedness of fishers. Second, various feasible and infeasible policy strategies were discussed, of which the former are modelled. Third, interviews facilitated the discussion of issues external to Chilika. These subjects may be considered ‘noise’ or ‘off-topic’ during structured interviews (FAO, 1990, Scott and Garner, 2013), but provide contextual settings for the findings and implications of this study.
3.3.2 Fishers
Stakeholder engagement seeks to understand: (i) bottom-up drivers of fishery operations, (ii) local perceptions of past changes, and (iii) future expectations. The degree of stakeholder participation was primarily constrained by time, limiting opportunities for co-learning and direct stakeholder inputs into modelling. Given the sample size (N = 12), detailed content analysis of interview scripts and socioeconomic differences is not possible. Therefore, fisher interviews here primarily serve information extraction purposes (Pretty, 1995, Lynam et al., 2007), highlighting concerns of stakeholders which had until then been overlooked. Undoubtedly, potential for more holistic involvement exists by revisiting the field in future.
Heterogeneous social and ecological settings were considered when selecting study locations and participants. Locations along Chilika’s central (high) to north (low) productivity gradient were targeted as logistics prohibited visiting villages within the southern and outer channel sectors, so the two productivity extremes were covered in case perceptions significantly differed (figure 3.10). Interviews
73 Chapter 3: Model construction
Figure 3.10: Locations of the four fishing villages where stakeholder interviews were conducted. Chilika’s sectors: NS - northern sector; CS - central sector; SS – southern sector; OC – outer channel. were desired from fishers of different ages, as life experience influences past perceptions and outlooks (Coulthard, 2008). Six elders and six adults were interviewed from the four villages.
CDA staff or primary cooperative members acted as facilitators, selecting suitable participants and quiet locations within villages, which were either community halls or public buildings known as mandaps. Questions were asked in English but translated to the regional language ‘Odia’ by the facilitator. The single matrix of semi-structured questions (appendix A.3) was revised with CDA scientists, as the original questions were perceived to be too complex; however, the three overarching themes were preserved.
Aiming to build discourse and trust between interviewer and interviewee (Denscombe, 2007), opening questions focused on everyday fisher knowledge, namely family histories, commercial species and fish prices. Insights into the socioeconomic backgrounds of the participants were also be gained, for example, whether the fisher had recently started fishing. Furthermore, these initial questions provided context for later answers, for example, a fisher’s lineage may
74 Chapter 3: Model construction explain their attitudes towards alternative livelihoods. Questions then focused on the perceived productivity of the lagoon and its main threats. Consistent with the views of decision-makers, fishers often compared present productivity with past, blaming ‘illegal activities’ in the outer channel for damaging catch within the lagoon. The final questions focused on plans and expectations, encouraging fishers to project their mental models. Asking fishers about the potential need for alternative livelihoods required the fishers to forecast drivers of fishery stress and validate the attitudes of decision-makers.
The fishers interviewed here were relatively affluent, averaging incomes of 7,800 INR/month (n=12), compared to the CDA recorded mean of 4,400 INR/month (2014). Such affluence may reflect their elder status and associated fishing experience, as well as the dominance of central sector interviews. All fishers interviewed conduct their practices within the central and southern sectors, indicating that northern sector fishers adapt to heterogeneous resource abundance.
3.4 Module parameterisation
The rest of chapter 3 documents the formal parameterisation of the SDM, informed by the conceptual modelling efforts, qualitative insights and quantitative datasets documented below and in appendix B. The methods used to derive hydrological flows are described, before the ecohydrological parameters, fish population dynamics, and socioeconomic behaviours are detailed. The model’s internal structure and behaviours are evaluated in Chapter 4.
As a method commonplace in system dynamics but less common to other environmental modelling techniques (Franco, 2007), it is here worth explaining the rationale behind graphical functions. In essence, graphical functions are idealised bivariate relationships mapping an input to a corresponding level of output (Sterman, 2000, Franco, 2007). These individual models act as look-up tables to help incorporate vital but data poor interactions into a model’s structure. However, to avoid their use in scenarios where the presence and form of causal relationships are merely speculative, Sterman (2000) outlines the following principle to parameterise graphical functions: Graphical functions depict generalised covariate relationships. Therefore, only processes with known causalities are included. Inclusion of tentative
75 Chapter 3: Model construction
relationships without physical basis undermines the use of SDMs to investigate regional problematic behaviours. Graphical functions are normalised in case absolute values exceed historical ranges observed in parameterisation. Extreme values and reference points inform graphical function shape. For instance, salinity and dissolved oxygen cannot be negative. Outputs from graphical functions should undergo sensitivity analysis to assess implications of uncertainties on wider system behaviours.
3.4.1 Hydroclimatic
The hydroclimatic module models monthly freshwater and sediment fluxes influencing Chilika’s ecohydrological conditions and fish migration. This module is built on physical laws, statistical relationships, case studies and personal intuition (Ford, 2010).
3.4.1.1 Lower Mahanadi freshwater yield
Three approaches exist to generate freshwater yields within a SDM environment. First, historical distributions could exogenously generate yields. However, this approach would miss insights into the effects of external variables beyond internal control (e.g. rainfall). Second, statistical models could act as black-boxes by relating monthly rainfall to runoff. Although initial analysis found fair predictability (e.g. monthly LMC rainfall versus runoff from Jan 1973 – Dec 1999, r2 = 0.45, df = 320, p < 0.05), the technique poorly captured spiked monsoonal rainfalls, which contribute ~80% of annual freshwater delivery to Chilika (Kumar and Pattnaik, 2012).
The third approach is to build a rainfall-runoff model capturing the variety of processes influencing monthly runoffs. Within a SDM environment, Ghashghaei et al. (2013) model monthly freshwater yields of Iran’s Gamasiab catchment by treating runoff as a stock. Precipitation increases the total volume of water within the catchment, with losses equalling ‘fluvial discharge’, ‘evaporative losses’ and ‘additional removals’ (figure 3.11). Therefore, the rainfall-runoff models of the LMC and WC are ‘lumped’ rather than ‘distributed’ (Beven, 2002). Physical conditions influencing runoff generation (e.g. rainfall, soil porosity, surface slope) are assumed spatially homogeneous within the catchment, as SDMs are unsuited to modelling spatial heterogeneities across landscapes. The module is also
76 Chapter 3: Model construction
Figure 3.11: Stock and flow structure driving LMC rainfall-runoff generation. Acronyms correspond to the runoff production and routing variables of Ghashgaei et al. (2013): Rc – runoff coefficient; Tg – groundwater delay; Bc – baseflow contribution; Co – proportion of discharge removed; Te – evaporation delay.
‘deterministic’ rather than ‘stochastic’ (Beven, 2002), as a single output is generated per time-step.
The modelled LMC freshwater yield starts at the Tikarapara gauging station (figure 3.12), ~200 km downstream of Hirakud and ~115 km upstream of the Mahanadi delta. The Indian CWC monitored daily freshwater discharge and total suspended sediments at Tikarapara from June 1973 – December 2009. Gauging stations exist at Cuttack and Kanas, located approximately 20 km and 60 km upstream of Chilika, respectively. However, only daily gauge is recorded, requiring stage-discharge relationships not readily available. Therefore, rainfall data across the four districts surrounding Tikarapara (Angul, Baudh, Sambalpur and Subarnapur) generates runoff (figure 3.12), aggregated from sub-district datasets of the Indian Meteorological Department (IMD) for June 1973 –
77 Chapter 3: Model construction
Figure 3.12: The spatial extent of district-wide rainfall and temperature datasets relative to the Tikarapara gauging station and Chilika lagoon.
December 2009. The monthly volume of precipitation across LMC equals the sum of the precipitation rates at time t, multiplied by the sub-basin area (Asokan and Dutta, 2008) (equation 3.1):
퐿표푤푒푟푀푎ℎ푎푛푎푑𝑖푅푎𝑖푛푓푎푙푙푡 =
(퐴푛푔푢푙푅푎𝑖푛푓푎푙푙푡 + 퐵푎푢푑ℎ푅푎𝑖푛푓푎푙푙푡 + 푆푎푚푏푎푙푝푢푟푅푎𝑖푛푓푎푙푙푡 +
푆푢푏푎푟푛푝푢푟푅푎𝑖푛푓푎푙푙푡) × 퐿표푤푒푟푀푎ℎ푎푛푎푑𝑖퐴푟푒푎 (Eq 3.1)
IMD rainfall data is aggregated from the block level to the district spatial scale, assuming that monthly rainfall within each of the districts is spatially homogeneous, excluding natural phenomena such as the orographic effect. In support, Mohapatra et al. (2003) do not find significant spatial variability in Odisha’s monsoonal rainfall, with the LMC lying within a homogeneous zone from Chilika to northern Odisha.
Rainfall inputs separate into streamflow and groundwater components (Beven, 2002). The runoff coefficient (R - appendix B) drives flow proportioning for a c given rainfall intensity, fundamental to all hydrological models irrespective of
78 Chapter 3: Model construction
Figure 3.13: Time-series comparing the effects of a static (blue) and dynamic (red) rainfall-runoff coefficient against the observed LMC freshwater yield (black) over the calibration period. complexity (Beven, 2002). Parameterisation over the period 1973-1982 using a static R (as per Ghashghaei et al., 2013) systematically overestimated non- c monsoon dynamics and underestimated monsoon peaks (figure 3.13). Consequently, a graphical function varies R with rainfall intensity in order to c parameterise the positive nonlinear association between antecedent moisture conditions and runoff (Beven, 2002). When surface saturation increases during the monsoon, the contribution of saturation excess overland flow to runoff generation grows. Rising limbs of the monsoonal discharges only trigger once rainfall across the surrounding districts exceeds 250 mm, forming the knee of the R curve (appendix B). The resulting modelled time-series exhibits greater c congruence with reality over the parameterisation period (figure 3.13). The LMC R is subject to sensitivity analysis (chapter 4.3), exploring how parameterisation c uncertainties affect other components of the model.
Quick-flow precipitation instantaneously adds to the surface runoff stock. The baseflow component is modelled by smoothing groundwater flow, ensuring baseflow alone cannot produce stormflow (Burt, 1996). Values of T and B were g c derived as default values were unsuitable, potentially due to differences in
79 Chapter 3: Model construction magnitudes between the wet and dry season flow regimes of the LMC and Gamasiab catchment. The combined effects of the variable runoff coefficient and groundwater parameters limit the contribution of groundwater to LMC runoff, whilst also producing realistic dry season yields. Finally, to capture rainfall-runoff generated downstream of Naraj (figure 2.5B), a proportion of WC runoff and sediment is converted to LMC runoff and sediment, helping to balance the freshwater and sediment delivery proportions of the sources.
3.4.1.2 Evaporative losses
Water evaporation under a warmer climate may influence multidecadal runoff trajectories. Seasonally variable evaporation from each district is simulated by the functions of Linacre (1977), requiring only dynamic monthly air temperatures:
푇 700 푚 +15(푇−푇 ) 퐸 = 100−퐴 푑 (Eq 3.2a) 0 (80−푇)
푇 − 푇푑 = 0.0023ℎ + 0.37푇 + 0.53푅 + 0.35푅푎푛푛 − 10.9°퐶 (Eq 3.2b)
푇푚 = 푇 + 0.006ℎ (Eq 3.2c)
Where: 푇 = mean monthly air temperature; 퐴 = latitude (degrees); 푇 푑 = mean dew- point; ℎ = surface elevation (metres); 푅 = mean daily temperature range; 푅푎푛푛 = difference between mean temperature of warmest and coldest months.
The single variable method of Linacre (1977) is particularly suited to the computational efficiency of SDMs and where data limitations inhibit methods such as Penman or Thornthwaite equations (Anyadike, 1987, Fennessey and Vogel, 1996). The parameterisation datasets for the six districts span June 1973 to December 1982. The latitude and elevation of Chilika were obtained from Kumar and Pattnaik (2012), and catchment values were obtained from regional government portals. All mean daily temperature ranges were estimated from Mohapatra et al. (2007c). Monthly IMD datasets calculated disparities between warmest and coldest months.
3.4.1.3 Additional removals
After evaporation subtracts from total runoff, the remaining stock splits between fluvial freshwater yield and additional removals (C ). Ghashghaei et al. (2013) set o C as 0.64 and 0.71, equalling 36% and 29% streamflow abstractions. However, o these values cause annual freshwater yields 12% below reality when applied here.
80 Chapter 3: Model construction
The estimated total domestic water demand for the LMC is 800 MCM (Asokan and Dutta, 2008), equivalent to approximately 14% of the Mahanadi’s annual yield at Tikarapara. However, this figure includes higher withdrawal rates from the cities of Cuttack and Bhubaneswar located downstream of Tikarapara. A conservative estimate summing the World Health Organisation’s (WHO) 70 litres/day standard with agricultural, industrial and domestic demands estimates 0.1% abstraction rates. Asokan and Dutta (2008) find a similar rate when applying a global level runoff pathways assessment to the LMC.
3.4.1.4 Flow routing at Naraj
Water from the undivided Mahanadi reaches Chilika once discharge at Naraj exceeds 2830 cumecs (Kumar and Pattnaik, 2012). Therefore, the monthly freshwater inflow to Chilika from the undivided Mahanadi equals:
푀푎ℎ푎푛푎푑𝑖푌𝑖푒푙푑푇표푤푎푟푑푠퐶ℎ𝑖푙𝑖푘푎 = 푰푭 푌𝑖푒푙푑퐶표푛푣푒푟푡푒푟 × 퐿푀퐶퐷표푤푛푠푡푟푒푎푚퐹푙표푤푡 >
2830 푻푯푬푵 퐿푀퐶퐷표푤푛푠푡푟푒푎푚퐹푙표푤푡 × 0.12 푬푳푺푬 0 (Eq 3.3)
In order to compare freshwater yields with the threshold discharge, ‘Yield converter’ translates monthly freshwater yields into an average flow rate. Whilst it is possible to probabilistically model alternative dam operations (e.g. reduced flows to Chilika under a wetter climate), the model simulates Naraj dynamics under the business-as-usual regime, whereby ~12% of undivided Mahanadi flow reaches Chilika during the monsoon (Kumar and Pattnaik, 2012).
3.4.1.5 Western catchments
The sporadically gauged Western catchments (WC) contribute ~30% of Chilika’s freshwater input (Santra and Das, 2013). However, this average exhibits interannual variability, suggesting the dynamics require more complex treatment than multiplying Mahanadi inputs by a constant. The system structure is assumed the same as the LMC (figure 3.11), except rainfall and temperature time-series are from the districts of Puri and Khordha. Runoff coefficients and baseflow contributions to surface runoff are downscaled to the magnitude of WC yields, with Santra and Das (2013) finding negligible non-monsoon flows.
3.4.1.6 Future climatic inputs
Future climate conditions run the hydroclimatic module into future. Determined by continental-scale atmospheric and oceanic processes, the major features
81 Chapter 3: Model construction
Figure 3.14: The empirical cumulative distribution functions used to derive future December (blue) and April (red) rainfall across the LMC. A random number is generated between 0 and 1 (x-axis), which corresponds to a given monthly rainfall total (y-axis). affecting Chilika are likely to persist in future. For example, although Panda et al. (2013) find evidence for earlier onset, the period July to October will likely remain monsoonal. Instead, climate change will likely induce nonstationarity in long-term means, such as the projected 33% increase in September rainfalls by 2050 (Asokan and Dutta, 2008).
Rather than attempting to forecast precise conditions like specialist climate models, this model aims to internally produce plausible rainfall and temperature values and observe outcome dynamics. Initial efforts randomly sampling from Gaussian distributions produced rainfall totals significantly greater than observed records. Therefore, the model randomly samples from empirical cumulative distribution functions (figure 3.14) parameterised with past monthly observations from 1973-1999. The improved method samples from nonparametric distributions to capture the infrequency of extreme rainfalls. Using LMC rainfall as an example: 1. Monthly observations of the four districts are summed, as required by the hydrological model;
82 Chapter 3: Model construction
2. The 324 observations are sorted into months, giving 12 subsets of 27 observations each; 3. Empirical cumulative distribution functions are generated for each month and inserted into STELLA as graphical functions; 4. STELLA’s RANDOM function generates a monthly rainfall value from the empirical cumulative distribution function.
To simulate climate change, an additional graphical function (chapter 5.2) transforms the empirical cumulative distribution function outputs – increasing the mean and variance of rainfall and temperature over time (chapter 5.2).
3.4.1.7 Fluvial sediment yields
The model also simulates the volume of fluvial sediment stored within the Chilika lagoon. The feedbacks underpinning outlet closure and macrophyte growth (figure 3.6) means that small changes in the lagoon’s ecohydrological conditions may cause the types of nonlinear dynamics known to destabilise natural resource systems (Peters et al., 2004).
Monthly LMC sediment yields are generated by parameterising the sediment- discharge curve of Mishra and Jena (2015) as a graphical function. For coherence with available data, the model is interested in sediment load rather than rate. However, the observed sediment flow rate is first approximated by inputting observed monthly freshwater discharges (1973-1982) into the function of Mishra and Jena (2015). This disaggregation assumes discharge is constant within each month. Then, the derived sediment flow rates are compared with observed sediment concentrations (CWC data), approximating the relationship between the two measures of sediment delivery. The resulting converter (figure 3.15) produces dynamics consistent with Das and Jena (2008), finding that the Mahanadi transports between 0.4-0.6 g/L of sediment during monsoonal flows. Due to the lack of primary data, the same relationship is assumed for the WC; albeit the functions are downscaled to reflect WC’s contribution. The total monthly sediment yield equals the concentration of sediment multiplied by the freshwater yield for the month. The LMC sediment load is then multiplied by 0.025, equalling the proportion of annual LMC sediment that reached Chilika during the period of available data (1999-2010) (CDA, n.d-b).
Sensitivity analysis explores the various uncertainties arising from simulating monthly sediment. For instance, the magnitude and significance of the driver- response relationship depicted by Mishra and Jena (2015) are unknown. The
83 Chapter 3: Model construction
Figure 3.15: The nonlinear relationship between observed sediment flow rates, derived from the function of Mishra and Jena (2015), and observed sediment concentrations obtained from CWC data. The logarithmic trend (r2 = 0.50, df = 118, p < 0.05) forms the graphical function converting between sediment rate and yield. The grey shading represents the 95% confidence interval of the regression model. relationship depicts a single trend, meaning the degree of variability around the trend line is also unknown. Therefore, any uncertainties in the parameterisation of the sediment flow rate may affect future monthly sediment yields. The assumption that Mahanadi discharge remains constant over a month is also a limitation.
3.4.1.8 Chilika’s sediment stock
In addition to the two fluvial inputs, littoral inputs are assumed constant as an external driver (Rao, n.d). Dynamic underlying processes drive the removal of sediment via the outlet channel. Removal averages 0.13x106 tonnes/year (Kumar and Pattnaik, 2012), but monthly removal is positively associated with freshwater delivery to the lagoon. Sedimentation exceeds removal during non-monsoon months, leading to outlet channel accretion (Mishra – personal communication). Moreover, a feedback exists with macrophyte coverage, as sediment provides
84 Chapter 3: Model construction
Figure 3.16: Stock and flow diagram depicting the inputs and outputs driving the sediment stock dynamics. Feedback labels are as per figure 3.5. nutrients for vegetation growth, which traps sediment within the lagoon. Thus, monthly sediment removal equals:
6 푂푢푡푙푒푡퐶ℎ푎푛푛푒푙푅푒푚표푣푎푙푡 = 0.13 × 10 ×
푀표푛푡ℎ푙푦퐹푟푒푠ℎ푤푎푡푒푟푌푖푒푙푑 푀푎푐푟표푝ℎ푦푡푒퐴푟푒푎 ( 푡) × (1 − ( 푡)) (Eq 3.4) 퐴푛푛푢푎푙퐴푣푒푟푎푔푒푌푖푒푙푑 퐿푎푔표표푛퐴푟푒푎
When vegetation coverage is low, as during the 1970s, the monthly sediment outflow is effectively proportional to the freshwater inflow. Human decisions also drive sediment removal, forming two balancing feedbacks (figure 3.16): BM1 abruptly reduces the sediment stock, as per the opening of the new tidal outlet in 2000. BM2 represents the increased efficiency of removals post-2000: outlet removal is assumed doubled to reflect the new outlet that is half the distance of the defunct Magarmukh channel.
3.4.1.9 Outlet closure
Sedimentation reduces the range of tidal transgressions into the lagoon – heightening freshwater conditions and constraining fish migration to and from
85 Chapter 3: Model construction the Bay of Bengal (Kumar et al., 2011). To capture this latter effect, the late-1990s sediment stock is assumed to equal a near functional closure of the tidal outlet, equalling the level of sedimentation causing the effective cessation of salinity transgression and fish migration. Approximately 76% of lagoon sedimentation occurred post-1950s (Pattnaik and Kumar, 2013), meaning the sediment stock in 1960 equalled ~24% of the capacity. Capacity is approximated by:
퐶ℎ𝑖푙𝑖푘푎푆푒푑𝑖푚푒푛푡퐶푎푝푎푐𝑖푡푦= (Eq 3.5)
퐴푛푛푢푎푙퐹푙푢푣𝑖푎푙푆푒푑𝑖푚푒푛푡퐷푒푙𝑖푣푒푟푦 + 퐴푛푛푢푎푙퐿𝑖푡푡표푟푎푙푆푒푑𝑖푚푒푛푡퐷푒푙𝑖푣푒푟푦 (39 × ( )) −퐴푛푛푢푎푙푆푒푑𝑖푚푒푛푡푂푢푡푓푙푢푠ℎ
0.76
Where 39 equals the number of years from 1960-1999. Therefore, the sediment stock at the start of the modelling horizon (1973) is approximated by:
퐶ℎ𝑖푙𝑖푘푎푆푒푑𝑖푚푒푛푡푆푡표푐푘1973 =
퐶ℎ𝑖푙𝑖푘푎푆푒푑𝑖푚푒푛푡퐶푎푝푎푐𝑖푡푦 − 27 × (퐴푛푛푢푎푙퐴푣푒푟푎푔푒퐹푙푢푣𝑖푎푙퐷푒푙𝑖푣푒푟푦 + 퐶표푎푠푡푎푙푆푒푑𝑖푚푒푛푡퐷푒푙𝑖푣푒푟푦 − 퐴푛푛푢푎푙푆푒푑𝑖푚푒푛푡푂푢푡푓푙푢푠ℎ) (Eq 3.6)
Where 27 equals the number of years from 1973-2000. Outlet closure takes linear form, approximated by the ratio of sediment stock to lagoon sediment capacity.
3.4.2 Ecohydrological
The ecohydrological module simulates the salient biophysical conditions determining both the habitat suitability of Chilika and the ability of fishers to catch at the desired rate (‘catchability’ – figure 3.2). The selected of ecohydrological variables primarily reflects their influence on the system. By definition, brackish ecosystems exist in water salinities between 0-20 parts per thousand (ppt) (Barnes, 1999). As a guideline, sustained salinity below 7 parts per thousand (ppt) is considered deleterious for the reproduction, growth and mortality of brackish species (Barnes, 1999, Martin et al., 2009). Salinity also drives macrophyte coverage and indicates outlet closure (Shaw et al., 2000).
Dissolved oxygen is critical for fish respiration, reproduction and growth (Alabaster and Lloyd, 1980). Within brackish environments, these processes are generally threatened at dissolved oxygen levels below 5 parts per million (ppm) (Boyd, 1982). Intra-annually, dissolved oxygen is antiphase with salinity, increasing during monsoon freshwater fluxes but declining during the dry summer and winter months. However, at the multidecadal scale, salinity and
86 Chapter 3: Model construction dissolved oxygen have shared a positive relationship linked to the closure of the tidal outlet, macrophyte growth and localised eutrophication.
Fish cannot self-regulate body temperature, so water temperature may affect fish fertility, growth, metabolism and mortality (Alabaster and Lloyd, 1980, Thépot and Jerry, 2015). Mean annual temperature during the 2000s equalled 28.6˚C and non-monsoon temperatures reached 33˚C without severely degrading the fish stock. However, persistent temperatures beyond 30˚C may cause ecological damage particularly amongst freshwater populations (Thépot and Jerry, 2015) – a plausible scenario given expected future warming. Datasets charting the above processes available from the CDA cover 27 seasons (monsoon, post-monsoon and pre-monsoon) between pre-monsoon 2002 and post-monsoon 2011.
Further data is available to model pH and transparency, which are important determinants of macrophyte coverage and faunal habitat quality (Kumar and Pattnaik, 2012). However, initial exploratory parameterisation was unsuccessful, as the pH trend is not coupled with underlying drivers. Consequently, pH is considered parsimonious, as natural stock variations are captured by salinity, dissolved oxygen and temperature. Water transparency moderates photosynthetic activity and primary productivity (Kumar and Pattnaik, 2012). However, low transparency does not necessarily equate to an unhealthy ecosystem (Kumar and Pattnaik, 2012), but instead indicates ecosystem quality in conjunction with other metrics, such as sediment removal and dissolved oxygen. Lagoon depth is a driver of transparency and sedimentation, as well as a determinant of fishing gear suitability (Jones and Sujansingani, 1954). Whilst shallow waters are preferable for catch, an absence of bathometric data inhibits the accurate parameterisation of this process.
3.4.2.1 Salinity
Three data-driven steps estimate monthly salinity based on the guidelines above: (i) monthly freshwater inflow is seasonally classified and normalised; (ii) using graphical functions, the normalised freshwater inflows are converted into monthly deviations from the mean annual average, and (iii) monthly salinity combines the mean annual average, monthly deviation and degree of outlet closure.
Annual salinity has two distinct phases (figure 3.17): (i) pre-monsoon months traditionally from March to June, and (ii) monsoon and post-monsoon months (MaPM), from July – October and November – February, respectively. The annual hydrological regime also splits in two, separating the comparatively low flows of
87 Chapter 3: Model construction
Figure 3.17: The seasonal trend of seasonal lagoon-wide salinity from pre- monsoon 2002/2003 to post-monsoon 2009/2010. the pre- and post-monsoon, and the distinctly spiky monsoon. Therefore, freshwater conditions persist for a season once freshwater inputs have subsided. Kumar and Pattnaik (2012) depict monthly freshwater influxes from 1999-2011 in box and whisker form; however, the records are not referenced by year, disabling aggregation to match the seasonal salinity trend. Therefore, graphical functions are used here because monthly salinity can be bound between reference points (Sterman, 2000), and the available data allows for normalisation rather than a statistical model between monthly freshwater inputs and salinity changes.
Approximately 94% of the annual 4906 MCM freshwater delivered to Chilika occurs during MaPM (Kumar and Pattnaik, 2012). According to the historical influx data, pre-monsoon inputs remain less than 350 MCM from 1999-2009. Therefore, the model classifies freshwater inflows as: (i) pre-monsoon and (ii) MaPM. The means of each classification are calculated to equal 73.6 MCM (pre- monsoon) and 576 MCM (MaPM). Each monthly inflow is then normalised with respect to its classification mean:
푁표푟푚푎푙𝑖푠푒푑푀퐴푃푀퐼푛푓푙표푤푡 =
6 6 (푀표푛푡ℎ푙푦퐹푟푒푠ℎ푤푎푡푒푟퐼푛푓푙표푤푡 − 576 × 10 )/ 576 × 10 (Eq 3.7a)
88 Chapter 3: Model construction
Figure 3.18: Functions converting (A) monthly monsoon and post-monsoon and (B) pre-monsoon freshwater deliveries to deviations from the annual mean salinity. The deviations insert into equations 3.8a and 3.8b.
푁표푟푚푎푙𝑖푠푒푑푃푟푒푚표푛푠표표푛퐼푛푓푙표푤푡 =
6 6 (푀표푛푡ℎ푙푦퐹푟푒푠ℎ푤푎푡푒푟퐼푛푓푙표푤푡 − 73.6 × 10 )/ 73.6 × 10 (Eq 3.7b)
From 2002-2009, mean annual salinity equalled 12.6 ppt. With regards to the form of the graphical functions, salinity must be positive, meaning 12.6 ppt represents the maximum possible decline from monsoon freshwater yields. Therefore, the MaPM graphical function decays nonlinearly to avoid producing salinities below 0 ppt (figure 3.18A). The highest monsoonal yields have positive normalised values, producing negative deviations from the annual mean salinity (figure 3.18A). During months when absolute MaPM yield equals zero, the normalised MaPM inflow equals minus one. In these rare cases, salinity does not decline as normal during the MaPM months (figure 3.18A). During pre-monsoon months, the magnitude of salinity deviation is inversely related to freshwater delivery (figure 3.18B). When normalised, completely dry months lead to higher salinities than months with freshwater inflow. Wetter than average pre-monsoon months are assumed to have negligible impacts on mean annual salinity.
Outlet sedimentation modifies the magnitude of marine water exchange (Jayaraman et al., 2007, Kumar and Pattnaik, 2012). It is assumed outlet sedimentation and salinity have a linear relationship to avoid inducing nonlinear behaviours without sufficient evidence. During near-closure, the function prevents salinity recovery seen during healthy tidal exchange. Therefore, depending on the season, monthly salinity equals:
89 Chapter 3: Model construction
푀퐴푃푀푆푎푙𝑖푛𝑖푡푦푡 = (퐴푛푛푢푎푙퐴푣푒푟푎푔푒푆푎푙𝑖푛𝑖푡푦 + (Eq 3.8A)
(퐴푛푛푢푎푙퐴푣푒푟푎푔푒푆푎푙𝑖푛𝑖푡푦 × 푀퐴푃푀푆푎푙𝑖푛𝑖푡푦퐷푒푣𝑖푎푡𝑖표푛푡)) × (1 − 푂푢푡푙푒푡퐶푙표푠푢푟푒푡)
푃푟푒푚표푛푠표표푛푆푎푙𝑖푛𝑖푡푦푡 =
(퐴푛푛푢푎푙퐴푣푒푟푎푔푒푆푎푙𝑖푛𝑖푡푦 +
+ (퐴푛푛푢푎푙퐴푣푒푟푎푔푒푆푎푙𝑖푛𝑖푡푦 × 푃푟푒푚표푛푠표표푛푆푎푙𝑖푛𝑖푡푦퐷푒푣𝑖푎푡𝑖표푛푡)) ×
(1 − 푂푢푡푙푒푡퐶푙표푠푢푟푒푡) (Eq 3.8B)
In summary, the magnitude of monthly salinity suppression from the mean annual average is proportional to MaPM freshwater inflow and inversely proportional to pre-monsoon freshwater inflow. To address the various uncertainties of data limitations and inexactness, the methods undergo uncertainty and sensitivity analysis in chapter 4.
3.4.2.2 Dissolved oxygen
Parameterisation of the dissolved oxygen trend also uses graphical functions to capture the seasonal dynamics. Lagoon-wide dissolved oxygen tends to increase from pre-monsoon to monsoon to post-monsoon (figure 3.19), suggesting dissolved oxygen increase lags freshwater inputs. The parameterisation dataset has the same temporal characteristics as the salinity records.
Consequently, hydrological inflows are split into monsoon and non-monsoon, rather than pre-monsoon and MaPM. Then, modelled freshwater inputs are lagged by four months so that monsoon freshwater delivery causes dissolved oxygen to increase between November and February. According to total monthly inflows (Kumar and Pattnaik, 2012), non-monsoonal yields did not exceed 800 MCM from 1999-2010. Once modelled total inflow surpasses this threshold, normalisation occurs with respect to the monsoon mean (1000 MCM/month); non-monsoon flows are normalised with respect to the non-monsoon mean (113 MCM/month). As with salinity, normalised freshwater yields form the independent variables of the graphical functions. The form of the non-monsoon graphical function decays nonlinearly as freshwater input declines, as dissolved oxygen must be positive (appendix B). Dissolved oxygen dynamics are then dampened as macrophyte coverage increases (appendix B), representing the biological uptake of oxygen for respiration (Engel, 1990, Petr, 2000).
90 Chapter 3: Model construction
Figure 3.19: Seasonal lagoon-wide dissolved oxygen values from pre- monsoon 2002/2003 to post-monsoon 2009/2010.
3.4.2.3 Water temperature
Despite spatial differences in freshwater input and water expanse, Chilika’s vertical and horizontal temperature profiles are considered homogeneous (Kumar and Pattnaik, 2012). The temperature homogeneity reflects the shallow nature of the lagoon, producing a well-mixed environment with incoming solar energy penetrating to the maximum depth of four metres (Nayak et al., 2004). Consequently, without stratification and thermal upwelling, surface temperatures of shallow lakes often closely track near-surface air temperatures (Lister et al., 1998, Schmid et al., 2014). Such coherence exists in monthly observations from September 2006 to December 2009 (figure 3.20) (CDA, n.d-c, CDA, n.d-a), although the difference increases during winter due to the greater thermal inertia of water relative to air. The spiky monsoon water temperatures are associated with the heightened freshwater influxes.
Due to the strong relationship between observed air and water temperatures (r2=
0.89, p< 0.05, df = 37), a generalised linear model (GLM) is parameterised to convert air temperatures into lagoon surface temperatures. Modelled water
91 Chapter 3: Model construction
Figure 3.20: Time-series coherence between: (i) observed surface air temperatures (black) at Chandraput on the western bank of Chilika (CDA, n.d-a) ; (ii) averaged water temperatures (blue) observed from 30 ecohydrological monitoring stations uniformly distributed across Chilika (CDA, n.d-c); (iii) modelled water temperatures (orange).
temperatures are also reduced proportional to the magnitude of freshwater delivery (appendix B). There is good coherence between observed and parameterised water temperatures during calibration (figure 3.20); although this is expected as the air temperature dataset driving water temperatures is the same.
3.4.2.4 Aquatic vegetation coverage
Aquatic macrophyte coverage is modelled as a stock because: 1. Macrophytes are tangible entities, like stocks of boats and fishers. In contrast, salinity and dissolved oxygen are physiochemical properties more difficult to model as stocks. 2. Exponential macrophyte growth (figure 3.21) suggests reinforcing feedback(s) exists between the stock and its drivers, which cannot be captured when modelling macrophyte coverage as a converter.
92 Chapter 3: Model construction
Biswas (1995) and Kumar and Pattnaik (2012) describe multiple ecohydrological drivers of macrophyte growth, including salinity, dissolved oxygen, pH, sediment dynamics, and aquatic nutrients such as nitrates and phosphates. Salinity below 7 ppt is conducive for luxuriant macrophyte growth (Shaw et al., 2000), although species such as Potamogeton pectinatu, Najas flaveolata and the recently emerged genus of Phragmites survive salinities up to 15-18 ppt (Forbes and Cyrus, 1998, Meyerson et al., 2009).
Macrophytes uptake oxygen for respiration during the night (Engel, 1990, Petr, 2000). Respiration declines under oxygen stressed conditions, increasing the biological oxygen demand associated with weed decomposition (Raven et al., 2012). Aquatic vegetation growth is positively linked to freshwater inputs, with fluvial sediment acting as major sources of phosphorous, nitrogen, iron and micronutrients (Barko et al., 1991). As a feedback, macrophytes can enhance sediment deposition within shallow lakes due to enhanced flow resistance around submerged and emerged stands (Schulz et al., 2003).
Macrophytes were absent prior to the 1970s, when annual salinity levels exceeded 10 ppt (figure 3.21). Persistent salinity decline during the 1970s and 1980s likely caused the initial macrophyte increase, eventually becoming nonlinear due to the positive feedbacks with lagoon sedimentation. The model assumes that macrophyte coverage only increases when salinity is less than 18 ppt; coverage only decreases above 15 ppt:
푀푎푐푟표푝ℎ푦푡푒퐺푟표푤푡ℎ푡 = 푀푎푐푟표푝ℎ푦푡푒푆푡표푐푘푡−1 + 퐺푟표푤푡ℎ푡 − 퐷푒푐푙𝑖푛푒푡 (Eq 3.9A)
퐺푟표푤푡ℎ푡 = 푰푭 푆푎푙𝑖푛𝑖푡푦푡 < 18 푻푯푬푵
퐴 × (18 − 푆푎푙𝑖푛𝑖푡푦푡) × 퐷푂푡 × 푂푢푡푙푒푡퐶푙표푠푢푟푒 푡 푬푳푺푬 0 (Eq 3.9B)
퐷푒푐푙𝑖푛푒 = 푰푭 푆푎푙𝑖푛𝑖푡푦푡 > 15 푻푯푬푵
1 × 퐵 × (푆푎푙𝑖푛𝑖푡푦 − 15) × (1 − 푂푢푡푙푒푡퐶푙표푠푢푟푒 ) 푬푳푺푬 0 (Eq 3.9C) (퐷푂) 푡 푡
‘A’ and ‘B’ are numerical constants parameterised against the observed values of 1973, 1977, 1982 and 2007. If the tidal outlet is closing, then coverage will increase at a faster rate than if only driven by salinity. The resulting macrophyte coverage then affects the rate of sedimentation (chapter 3.4.1.8), producing a positive feedback (R6 – figure 3.6). Rising dissolved oxygen causes macrophyte growth, feeding back to reduce dissolved oxygen through organic demand. Rapid macrophyte growth occurs during late-monsoon, caused by decreasing salinity,
93 Chapter 3: Model construction
Figure 3.21: Modelled and observed aquatic macrophyte coverage from 1973- 2009. Observations used in model calibration are grey. increasing sedimentation and rising dissolved oxygen. Yet, these trends dampen during the pre-monsoon, from increasing salinity, stable outlet sedimentation and stable dissolved oxygen.
The modelled vegetation trend generally replicates the directions and magnitudes of vegetation growth pre-2000, peaking around ~580 km2; however, relative to observations, the model tends to generate shallower nonlinear growth until 2000, then shallower decline post-2000 (figure 3.21). Overall, the model demonstrably captures the underlying reference mode, which is more important in systems modelling than point-to-point errors (Sterman, 2000).
3.4.2.5 Habitat quality
By undercutting the resource stock, ecosystem degradation may accelerate extraction-driven declines. In contrast, healthy ecosystem conditions may build robustness against catch by supporting fish renewal. Therefore, interactions between natural and anthropogenic conditions have significant implications for future sustainability.
94 Chapter 3: Model construction
Five variables drive natural fish stock dynamics, namely fecundity, maturation, natural deaths, the ecosystem carrying capacity and survival. As with Bueno and Basurto (2009), the first four parameters are kept constant. Survival rate defines the proportion of fish births per month adding to the juvenile stock. In the original Bueno and Basurto (2009) model, survival is driven by fish density in relation to a carrying capacity, suggesting young fish are outcompeted at high densities, based on the Beverton-Holt equation used in model fish populations (Ricker, 1975). Therefore, a graphical function sets fish density as the independent variable and survival rate as the dependent variable (figure 3.22). Here, ecohydrological conditions modify the survival rate to add a positive relationship between ecosystem health and effective birth rate (i.e. fish that become juvenile).
The default survival function is parameterised to maintain the mature fish population at 90% of the carrying capacity in the absence of anthropogenic extraction (Bueno and Basurto, 2009, Duer-Balkind et al., 2013). Due to the lack of system-wide and/or species specific data, graphical functions then convert monthly temperature, salinity and dissolved oxygen into survival modifications (figure 3.22), built with known reference points and extreme conditions, and knowledge acquired from literature surveys and expert discussions. Then, a converter representing combined habitat quality modifies the default survival rate.
Regarding the form of the graphical functions, fish survival within brackish ecosystems has a parabolic relationship with salinity (Barnes, 1999), with marine species absent in freshwater conditions and vice versa. Chilika’s marine species withstand salinities above 30 ppt (Sylvester et al., 1975, Lee and Menu, 1981), although seasonal averages remained below 25 ppt from 2002-2011. Therefore, across the functional range, habitat quality is effectively positively associated with salinity. The form of the salinity-survival graphical function is justified by four factors (figure 3.22A): 1. Lagoon-wide annual average salinity of 15 ppt is considered optimal for fishery production (Das and Jena, 2008), balancing the preferences of freshwater fish with brackish and marine species (Biswas, 1995). 2. Fish survival decline accelerates beneath 8 ppt – generally considered the lower limit of brackish environments (Martin et al., 2009). This level also represents the lower range of relatively freshwater tolerant mullets (Saraswathy et al., 2015), representing ~14% of recorded catch (Parida et al., 2013).
95 Chapter 3: Model construction
Figure 3.22: Graphical functions relating (a) salinity (b) dissolved oxygen and (c) water temperature to the rate of fish survival per time-step.
3. Survival degradation steepens at 5 ppt as the survival of tolerant brackish species (e.g. shrimps) deteriorates (Biswas, 1995). 4. Zero salinity may cause the disappearance of brackish and marine stocks, but freshwater species constitute ~20% of the total fish stock.
A parabolic relationship also exists between dissolved oxygen and habitat quality because concentrations above air-saturation lead to growth impairments and ill- effects like gas bubble disease (Kemker, 2013). Chilika’s upper optimal limit is ~9.5 ppm, corresponding to 100% air saturation at mean annual air temperature (~29°C). Dissolved oxygen increases during the post-monsoon following monsoon freshwater delivery, with the maximum seasonal recording from 2002-2009 equalling 8.9 ppm. However, this model only currently models the effects of hypoxia, as lagoon freshness will likely lead to the long-term decline of dissolved oxygen from macrophyte proliferation. The dissolved oxygen-survival function factors (figure 3.22B): 1. Optimum dissolved oxygen for brackish environments is 7 ppm (Biswas, 1995, Kemker, 2013). 2. The safe limit for euryhaline (Harris and Vinobaba, 2013) and freshwater species (Boyd, 1982, Bhaumik, 2015) is generally 5 ppm. Whilst suboptimal, survival does not decline significantly. 3. Survival loss accelerates as dissolved oxygen approaches zero. Oxygen starvation often starts around 2-3 ppm amongst young mullets (Sauriau et al., 1994). 4. Survival is 0 when dissolved oxygen equals 0, due to the absence of oxygen for respiration. The lethal limits for mullets and seabass
96 Chapter 3: Model construction
Figure 3.23: The fish survival parameter is a function of the underlying population density and level of ecosystem degradation. In essence, fish survival for a given density declines from default as ecological degradation (fresher, deoxygenated and warmer waters) increases.
5. correspond to 0.3-0.4 (Sauriau et al., 1994) and 0.5 ppm (Tucker Jr et al., 2002), respectively.
Survival of juvenile fish is generally inversely proportional to temperature (figure 3.22C), particularly within unstratified shallow environments (Gehrke et al., 2011). Due to inter-species differences in energy expenditure at different temperatures, a general temperature-survival function is not obvious. Temperatures (°C) between high-20s and low-30s are safe for larval and young development of tropical euryhaline fish (Tucker Jr et al., 2002, Saraswathy et al., 2015). Evidence suggests mortality increases amongst populations of mullets (Walsh et al., 1991, Tucker Jr et al., 2002) and seabass (Thépot and Jerry, 2015) above 32°C. The survival of tolerant species like Penaeus monodon (Giant tiger prawn) declines beyond 40°C (Alabaster and Lloyd, 1980, Gehrke et al., 2011).
The combination of these individual habitat affects cannot be additive as the survival rate is a ratio between 0 and 1. The individual effects also cannot be averaged, as dissolved oxygen and temperature have lethal thresholds, meaning
97 Chapter 3: Model construction
Figure 3.24: Modelled fish regeneration and total fish population in response to the hypothetical degradation scenario. Other stresses are switched-off (e.g. catch and migration) to isolate the effects of ecological degradation on population change. survival can become zero irrespective of salinity. Modelling multiplicative effects is therefore appropriate; survival will not exceed default, but become zero if dissolved oxygen or temperatures reach lethal limits (figure 3.23):
푀표푑𝑖푓𝑖푒푑푆푢푟푣𝑖푣푎푙푅푎푡푒푡 = 퐷푒푓푎푢푙푡푆푢푟푣𝑖푣푎푙푅푎푡푒푡 × 푆푎푙𝑖푛𝑖푡푦퐸푓푓푒푐푡푡 ×
퐷𝑖푠푠표푙푣푒푑푂푥푦푔푒푛퐸푓푓푒푐푡푡 × 푇푒푚푝푒푟푎푡푢푟푒퐸푓푓푒푐푡푡 (Eq 3.10)
The feasibility of this parameter is assessed by simulating a hypothetical degradation scenario, linearly declining salinity from 15 to 1 ppt, dissolved oxygen from 8 to 1 ppt, and increasing temperature from 28 to 40°C over 37 years (equivalent to 1973-2009). The resulting survival modifier is nonlinear (figure 3.24), reflecting the different timings of stresses becoming unsafe. Both fish population indicators remain stable until survival is approximately halved, suggesting a healthy stock provides a degree of resilience against habitat degradation. Nonlinear fish regeneration is triggered after around 30 years, but the accumulation of mature fish during healthy conditions delays population nonlinearity. Whilst the fish population loss is seemingly unremarkable, emerging
98 Chapter 3: Model construction nonlinearity may have significant implications for an operational fishery dependent on a stable resource.
3.4.3 Core fishery
Whilst the previous two modules simulate natural fish stock variability through migration and habitat effects on survival, the core fishery module simulates fish stock and catch dynamics, which are both dependant on biophysical and socioeconomic conditions. Monthly catches feed into ecosystem value, driving the number of dependants and catch efforts. The fishery system module is built on physical laws, statistical relationships, expert judgement, stakeholder perceptions and personal intuition (Ford, 2010).
This module adapts the model of Bueno and Basurto (2009). Stocks of mature and immature fish form the resource base. Of the three ecological carrying capacity (eco-K) estimates (chapter 2.5.1), the most conservative and up-to-date estimate of 27,000 tonnes is chosen (Mohanty – personal communication). As per Bueno and Basurto (2009), the initial mature stock equals 90% of the carrying capacity (appendix B), representing a healthy population with room to grow before reaching its natural limit. The immature (juvenile) stock grows through regeneration but declines due to juvenile catch and maturation:
퐼푚푚푎푡푢푟푒퐹𝑖푠ℎ푃표푝푢푙푎푡𝑖표푛푡 = (Eq 3.11)
퐼푚푚푎푡푢푟푒퐹𝑖푠ℎ푃표푝푢푙푎푡𝑖표푛푡−1 + 푅푒푔푒푛푒푟푎푡𝑖표푛푡 − 푀푎푡푢푟푎푡𝑖표푛푡 − 퐽푢푣푒푛𝑖푙푒퐶푎푡푐ℎ푡
푅푒푔푒푛푒푟푎푡𝑖표푛푡=
(퐵𝑖푟푡ℎ푠푡 × 푀표푑𝑖푓𝑖푒푑푆푢푟푣𝑖푣푎푙푅푎푡푒푡) − (0.7 × 푂푢푡푙푒푡퐶푙표푠푢푟푒푡 × 퐵𝑖푟푡ℎ푠푡 ×
푀표푑𝑖푓𝑖푒푑푆푢푟푣𝑖푣푎푙푅푎푡푒푡) (Eq 3.12)
Approximately 70% of the stock migrates to and from the Bay of Bengal via the tidal outlet in order to complete its reproductive cycle, so the proportion adding to the immature stock is assumed to linearly decline with outlet closure.
퐹푒푐푢푛푑푖푡푦 퐵𝑖푟푡ℎ푠 = 푀푎푡푢푟푒퐹𝑖푠ℎ푆푡표푐푘 × 0.5 × ( ) (Eq 3.13) 푡 푡 퐴푣푒푟푎푔푒푅푒푝푟표푑푢푐푡푖푣푒퐿푖푓푒퐿푒푛푔푡ℎ
It is assumed that half of the mature fish stock is female, whilst average reproductive life remains at the default value of 9 years (Bueno and Basurto, 2009). Fecundity is the number of eggs spawned per month by sexually mature females (Lévêque, 1997). This model assumes fecundity is constant, with
99 Chapter 3: Model construction environmental influences modelled by adjusting the survival rate (chapter 3.4.2.5). Initial trials with fecundity equalling 16x103 based upon the fecundity of Chilika’s mullets (Kumar and Pattnaik, 2012) caused natural collapse dynamics as the stock rapidly exceeded its carrying capacity. Therefore, as with Bueno and Basurto (2009), fecundity is dimensionless, designed to stabilise the fish stock in the absence of catch. The default fecundity value of Bueno and Basurto (=20) is found to be an underestimation for Chilika’s stock; instead, fecundity of 105 stabilises the mature population at 90% of its carrying capacity. Sensitivity analysis in chapter 4 assesses these parametric uncertainties (chapter 4).
Immature fish mature after twelve months. The mature stock is determined by inputs and outputs reflecting later-life processes:
푀푎푡푢푟푒푆푡표푐푘푡 =
푀푎푡푢푟푒푆푡표푐푘푡−1 + 푀푎푡푢푟푎푡𝑖표푛푡 + 퐼푛푚𝑖푔푟푎푛푡푠푡 − 푂푢푡푚𝑖푔푟푎푛푡푠푡 − 푀푎푡푢푟푒퐶푎푡푐ℎ푡 −
푁푎푡푢푟푎푙퐷푒푎푡ℎ푠푡 (Eq 3.14)
퐼푛푚𝑖푔푟푎푛푡푠푡 = 푂푢푡푚𝑖푔푟푎푛푡푠푡−6 × (1 − 푂푢푡푙푒푡퐶푙표푠푢푟푒푡) (Eq 3.15A)
(푀푎푡푢푟푒푆푡표푐푘 ×0.7) 푂푢푡푚𝑖푔푟푎푛푡푠 = 푡 (Eq 3.15B) 푡 12
푀푎푡푢푟푒푆푡표푐푘 푁푎푡푢푟푎푙퐷푒푎푡ℎ푠 = 푡 (Eq. 3.15C) 푡 120
The return of out-migrants is inversely proportional to outlet closure, delayed by six months to reflect the link between migration and seasonal ecohydrological conditions. Fish lifespan driving the number of natural deaths is kept at the default value of 10 years (Bueno and Basurto, 2009). Catch deducting from the mature population equals the fishing effort of each fleet (𝑖) across the vegetation free area, all multiplied by fish density, forming the feedback between productivity and the finite stock (figure 3.5). Mature catch is converted to tonnes by assuming 25 fish per kg. Interviews and observations from Balugaon fish market helped derive this average, considering the weights of the twenty most commercially important species (Kumar and Pattnaik, 2012). it is acknowledged that an average weight smooths stock heterogeneity, ranging from 600 grams for E. tetradactylum to below 40 grams for shrimp (Kumar and Pattnaik, 2012). As of 2010, shrimps constituted 40% of Chilika’s catch by weight (Kumar and Pattnaik, 2012), supporting a positively skewed average weight. Fishing efforts are then restricted by an area ban under the third governance scenario (chapter 5.3).
100 Chapter 3: Model construction
(Eq 3.16) 푀푎푡푢푟푒퐹𝑖푠ℎ퐶푎푡푐ℎ푡,푖 =
푀푎푡푢푟푒푆푡표푐푘 푉푒푔푒푡푎푡푖표푛 푎푟푒푎 ( ) × 퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡 × (1 − vegetation coefficient × ) 퐶푎푟푟푦푖푛푔퐶푎푝푎푐푖푡푦 푡,푖 퐿푎푔표표푛 푎푟푒푎
퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡푡,푖 =
푀표푛푡ℎ푙푦퐶푎푡푐ℎ퐶푎푝푎푐𝑖푡푦푖,푡 × 푁표. 퐵표푎푡푠푖,푡 × % 푀표푛푡ℎ퐹𝑖푠ℎ푒푑푖,푡 (Eq 3.17)
There are six active fishers per boat (CDA, 2015). Traditional fishers operate in weekly cycles, effectively fishing for five days before returning to landing centres (Jones and Sujansingani, 1954, Kumar and Pattnaik, 2012). Therefore, the maximum number of days fished equals 260/year. In contrast, non-traditional fishers operate daily trips, only limited by the number of religious festivals each year (~17, Kumar Sethi – personal communication).
The catch capacities of each fleet are calculated here. In 1973, total catch equalled 6.65x106 kg, with ~19,000 fishers and/or ~3,000 boats (Iwasaki et al., 2009), and less than 20 km2 aquatic vegetation coverage. Assuming the fish stock was relatively healthy and traditional fishers operated for 260 days, the average traditional capture was approximately 6,500 fish/month. Under these assumptions, motorboat catch for the highest catch year (2004) (Mohapatra et al., 2007a) equals:
푁표푛푇푟푎푑𝑖푡𝑖표푛푎푙퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡 ≈ 푇표푡푎푙퐶푎푡푐ℎ − 푇푟푎푑𝑖푡𝑖표푛푎푙퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡
푇표푡푎푙퐶푎푡푐ℎ = 1.41 × 107 × 25
푇푟푎푑𝑖푡𝑖표푛푎푙퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡 = 퐵표푎푡푠 × 푇푟푎푑𝑖푡𝑖표푛푎푙퐶푎푡푐ℎ퐶푎푝푎푐𝑖푡푦 × %푌푒푎푟퐹𝑖푠ℎ푒푑 × (1 푉푒푔푒푡푎푡𝑖표푛 푎푟푒푎 − vegetation coefficient × ) 퐿푎푔표표푛 푎푟푒푎
∴ 푁표푛푇푟푎푑𝑖푡𝑖표푛푎푙퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡 = 3.53 × 108 260 300 − (3047 × 6500 × 12 × × (1 − vegetation coefficient × )) 365 1000
∴ 푁표푛푇푟푎푑𝑖푡𝑖표푛푎푙퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡 = 2.34 × 108
101 Chapter 3: Model construction
Figure 3.25: Graphical functions determining the number of days fished each month as a function of the perceived fish density, based on the type-II functional response (Holling, 1959b).
2.34 × 108 ∴ 푁표푛푇푟푎푑𝑖푡𝑖표푛푎퐶푎푡푐ℎ퐶푎푝푎푐𝑖푡푦 = 348 300 (1937 × 12 × × (1 − )) 365 1000
∴ 푁표푛푇푟푎푑𝑖푡𝑖표푛푎퐶푎푡푐ℎ퐶푎푝푎푐𝑖푡푦 = 10,000 퐹𝑖푠ℎ/푀표푛푡ℎ ≈ 420 퐾퐺/푀표푛푡ℎ
The monthly catch capacities correspond to the maximum number of days fished. In reality, the number of days fished varies inter-annually (Mohapatra et al., 2007a). To reflect the consideration of species abundance when allocating fishing effort (Mohanty- personal communication, Duer-Balkind et al., 2013), graphical functions depict the nonlinear relationships between prey consumed (days fished) to prey density (perceived fish density) (Holling, 1959b):
푃푟푒푦퐶표푛푠푢푚푒푑 = (Eq 3.18)
(푆푒푎푟푐ℎ𝑖푛푔퐸푓푓𝑖푐𝑖푒푛푐푦 × 퐴푣푎𝑖푙푎푏푙푒푆푒푎푟푐ℎ푇𝑖푚푒 × 푃푟푒푦퐷푒푛푠𝑖푡푦 × 푃푟푒푑푎푡표푟퐷푒푛푠𝑖푡푦)
(1 + 푆푒푎푟푐ℎ𝑖푛푔퐸푓푓𝑖푐𝑖푒푛푐푦 × 퐻푎푛푑푙𝑖푛푔푇𝑖푚푒 × 푃푟푒푦퐷푒푛푠𝑖푡푦)
As a spatial component, ‘search efficiency’ equals one, meaning fishers find fish wherever they operate. ‘Predator density’ is also assumed to equal one, as
102 Chapter 3: Model construction common in experiments (Beals et al., 1999). Regarding the forms of the graphical functions (figure 3.25), the different number of ‘off-days’ each month reflects the different ‘handling times’ of each fleet, including the time to offload and process catch. As ‘available search time’ is 12 months, traditional handling time equals 12 – (12 x (105/365)) and non-traditional handling time equals 12 – (12 x (17/365)). ‘Prey density’ is internally derived by lagging mature fish density by one month, reflecting the perception delay between fishers and their environment (Duer- Balkind et al., 2013). From equation 3.18, nonlinear trends are produced representing the number of days fished for a given fish abundance (figure 3.25). Non-traditional fishers operate for longer until prey density falls below 10% of the ecological carrying capacity. The concavity of the non-traditional trend means the modelled non-traditional fishers are comparatively sensitive to underlying stock density; in reality this might relate to the greater costs associated with non- traditional fishing, forcing the decline of efforts once profits fall.
As the stock declines and fishers operate on fewer days, it is assumed that catch selectivity decreases. Therefore, juvenile fishing effort is proportional to the difference between the maximum and the actual days fished per fleet. The selectivity of the non-traditional fleet dramatically falls as the mature population shrinks, consistent with the views of the traditional fishers interviewed who perceive non-traditional fishers as less environmentally conscious, less connected to ecology and more financially driven.
3.4.4 Socioeconomic
The socioeconomic module internalises the dynamics of the fisher population, fishing efforts and catch. Trajectories of fisher population could be externally generated; however, internal interactions establish a number of feedbacks potentially causing unsustainable nonlinearities. The socioeconomic module is parameterised with physical laws, statistical relationships, expert judgement, stakeholder perceptions and personal intuition (Ford, 2010).
Fish catch is a provisioning ecosystem service (Daily, 1997), providing fishers with food and income. The regional significance of Chilika’s fishery is observed by the lack of alternative livelihoods, with 77% of fisher households estimated to be completely reliant on fishing for income (Kumar and Pattnaik, 2012). Similar to ecological populations, a constraint exists on the benefits obtainable from Chilika as a resource limited system. In general, as the number of dependants increases, the average benefits per person declines, as either the resource spreads amongst
103 Chapter 3: Model construction users or those with best access monopolise the resource (Motesharrei et al., 2014). Therefore, the ability of dependants to achieve social foundations may fall, forcing adaptation through livelihood intensification, diversifications or migration.
Chilika’s livelihood carrying capacity must consider the livelihood demands of the population. For example, if each fisher required 10,000 INR/year to continue fishing, then a fishery worth 161 crore (x107) as of 2014 could support 161,000 fishers assuming financial benefits spread evenly. Therefore, the livelihood carrying capacity can be expressed as the total financial output divided by the average costs needed to maintain fishing activities. The result represents a maximum supportable population, applicable to Lotka (1925) style equations to modify human population change. In addition, incomes and costs model fishing effort via upgrades from traditional to non-traditional boats.
3.4.4.1 Traditional fishers
The traditional fisher population in the year 1973 was 19,000 (Noack et al., 2013), with non-traditional fishing absent before the 1980s (Kadekodi and Gulati, 1999). Regional population growth is influenced by more than just birth and death rates (figure 3.7); however, the fisher population in 2009 is underestimated by ~20% (figure 3.26) when population change is parameterised as a logistic growth function (Lotka, 1925). Therefore, an IF THEN ELSE function only switches on the carrying capacity effect once the fishery is overcapacity. This assumes that fishers exercise adaptive capacities for as long as possible (Nayak, 2014), until the point when total fishery value no longer supports the average livelihood cost of the fisher population:
(퐸푐표푛표푚푖푐푉푎푙푢푒푇푟푎푑푖푡푖표푛푎푙퐶푎푡푐ℎ푡) 푇푟푎푑𝑖푡𝑖표푛푎푙퐶푎푟푟푦𝑖푛푔퐶푎푝푎푐𝑖푡푦푡 = (Eq 3.19A) 푇푟푎푑푖푡푖표푛푎푙퐶표푠푡푠푡
퐸푐표푛표푚𝑖푐푉푎푙푢푒푇푟푎푑𝑖푡𝑖표푛푎푙퐶푎푡푐ℎ푡 = 푇푟푎푑𝑖푡𝑖표푛푎푙퐶푎푡푐ℎ푡 × 퐹𝑖푠ℎ푃푟𝑖푐푒푡 (Eq 3.19B)
푇푟푎푑𝑖푡𝑖표푛푎푙푃표푝푢푙푎푡𝑖표푛푡 =
푇푟푎푑𝑖푡𝑖표푛푎푙푃표푝푢푙푎푡𝑖표푛푡−1 + 푁푒푡퐿표푔𝑖푠푡𝑖푐퐺푟표푤푡ℎ푡 − 푈푝푔푟푎푑푒푠푡 (Eq 3.19C)
푁푒푡퐿표푔𝑖푠푡𝑖푐퐺푟표푤푡ℎ푡 = 푰푭 푇푟푎푑𝑖푡𝑖표푛푎푙푃표푝푢푙푎푡𝑖표푛푡 > 푇푟푎푑𝑖푡𝑖표푛푎푙퐶푎푟푟푦𝑖푛푔퐶푎푝푎푐𝑖푡푦푡
푇푟푎푑푖푡푖표푛푎푙푃표푝푢푙푎푡푖표푛푡 푻푯푬푵 (퐵𝑖푟푡ℎ푅푎푡푒푡 − 퐷푒푎푡ℎ푅푎푡푒푡) × (1 − ) ELSE 푇푟푎푑푖푡푖표푛푎푙퐶푎푟푟푦푖푛푔퐶푎푝푎푐푖푡푦푡
퐵𝑖푟푡ℎ푅푎푡푒푡 − 퐷푒푎푡ℎ푅푎푡푒푡 (Eq 3.19D)
104 Chapter 3: Model construction
Figure 3.26: Comparison of the observed fisher population (black) (Noack et al., 2013) against the modelled fisher population with a permanent carrying capacity (red).
Birth rate minus death rate represents the net effect of reproduction and mortality, known as Malthusian’s ‘r’ (Hopfenburg, 2003). Growth is smoothed across 216 months to desensitise population change to short-term fishery production variations; reflecting the desires of fishers interviewed to remain fishing, despite conceding that the attitudes of younger generations are gradually changing. Odisha’s crude birth and death rates were collected from the ‘India Development Indicators Revised, 2012’ dataset, published by the United Nations Statistics Division (appendix B). The 2013-14 Odisha Economic Survey (PCD, 2014) provides a second source for cross validation.
Fish price deriving from CDA market surveys (CDA, 2016) represents the mean fish price weighted by fish contribution to total catch (equation 3.19B). As found during the fisher interviews, this figure hides the overall range of prices, whilst also assuming that traditional and non-traditional fleets catch the same species. Traditional costs refer to both one-off and recurrent investments needed to fish (equation 3.19A). Traditional fishers must first obtain a licence, a boat and nets. As the socioeconomic status of new fishers is unknown, one-off outlays are
105 Chapter 3: Model construction spread across the lifetime of the gear. Therefore, anyone can begin fishing, but monthly incomes must cover the lifetime cost of the gear. Recurrent costs are annual licence renewals (Mohanty – personal communication) and gear maintenance (Noack et al., 2013). Gear maintenance demand is assumed proportional to the frequency of fishing within a month.
Active from 1984, switching from traditional to non-traditional boats forms the second outflow of the traditional fisher stock (equation 19C). The model calculates the number of fishers upgrading as:
푈푝푔푟푎푑푒푠푡 = 푃푟표푝표푟푡𝑖표푛퐴푓푓표푟푑푈푝푔푟푎푑푒푡 × 푊𝑖푙푙𝑖푛푔푛푒푠푠푇표푈푝푔푟푎푑푒푡 (Eq 3.20A)
Vessels need an outboard motor costing 18,000 INR (Noack et al., 2013), mechanical adaptations (3,970 INR), a motorised licence (235 INR) and fuel (variable – chapter 5.2). Due to the lumped populations, SDMs are generally unsuitable for modelling individual decisions. However, probabilities can be estimated by assuming an underlying normal distribution. Therefore, equation
Figure 3.27: Graphical function determining the proportion of traditional fishers that can afford upgrades each month. The function output feeds into equation 3.20A.
106 Chapter 3: Model construction
3.20B calculates the number of standard deviations the profit mean is above the upgrade cost:
(퐴푣푒푟푎푔푒푃푟표푓푖푡푠 −푈푝푔푟푎푑푒퐶표푠푡 ) 푃푟표푓𝑖푡푠푍푆푐표푟푒 = 푡 푡 (Eq 3.20B) 푡 푆푡푎푛푑푎푟푑퐷푒푣푖푎푡푖표푛푃푟표푓푖푡푠
Where ‘standard deviation of profits’ is assumed constant, equalling 1,246 INR (Noack et al., 2013). The z-score lookup table is converted into a graphical function (figure 3.27) to calculate the proportion of traditional fishers with profits above the threshold value.
When interviewed, traditional fishers expressed a desire to remain within the traditional community, blaming ecological degradation on motorisation, finer nets and immature catches. Consequently, an element of disconnect exists between the two fleets (D’Lima et al., 2014, Nayak, 2014). Therefore, traditional fishers do not upgrade if traditional profitability grows inter-annually (perceived profitability).
퐸푞. 24: 푊𝑖푙푙𝑖푛푔푛푒푠푠푇표푈푝푔푟푎푑푒푡 =
푰푭 푻푰푴푬 > 132 푨푵푫 푃푒푟푐푒𝑖푣푒푑푃푟표푓𝑖푡푎푏𝑖푙𝑖푡푦푡 < 0 푻푯푬푵 1 푬푳푺푬 0 (Eq 3.20C)
These competing processes mean increased traditional profits grows the number of fishers able to upgrade, but declines the attractiveness of the alternative fishery practice. Thus, switching to non-traditional practices is a form of livelihood coping strategy for those who can afford it (Nayak, 2014).
3.4.4.2 Non-traditional fishers
Wishing to profit from the lagoon’s productivity during the early 1980s, non- traditional fishers originated from castes with livelihoods other than fishing (Pattanaik, 2007, Nayak and Berkes, 2010). Now, the intensification of practices also drives the non-traditional fisher stock (Nayak, 2014). Non-traditional population growth takes the same form as equations 19A-D, albeit with non- traditional carrying capacity, population and catches. Non-traditional carrying capacity is dependent upon the six traditional costs, plus four specific to motorised fishing (Noack et al., 2013): new engines every 8 years – 18000 INR; annual engine repairs – 5840 INR; motorised licence – 235 INR; Fuel use ~3.7 litres/day.
107 Chapter 3: Model construction
Diesel prices since 1989 for India’s four metropolitan cities are published online from the Indian Ministry of Petroleum and Natural Gas (Verma and Fernandez, 2010); future prices are externally derived (chapter 5.2). These combined costs (except fuel price) increase at an annual inflation rate of 2.28%, as informed by historical fishing gear data (Jones and Sujansingani, 1954, Kumar and Pattnaik, 2012, Noack et al., 2013).
3.4.5 Statement of circularity
Following the model’s structure and equations, it is worth concisely summarising any instances of circularity, where the same data is used in model parameterisation and evaluation (Sterman, 2000). Circularity is generally considered bad practice (Jakeman et al., 2006), as testing the model over the same parameterisation data does not assess the robustness of model equations.
Within the hydroclimatic module, an element of circularity exists between the parameterization of the LMC’s rainfall-runoff coefficient and outcome dynamics from 1973-1982, both dependant on the district level precipitation and temperature datasets. However, hydrological outcomes are evaluated against data reserved for evaluation – from 1983-1999 (chapter 4.3.2). Moreover, although the assumption that sedimentation in 2000 equals a near functional closure introduces an element of circularity, it does not guarantee that the resulting dynamics will replicate the timing and magnitude of historical behaviours assessed in chapter 4. Within the ecohydrological module, salinity is calibrated with seasonal observations of lagoon average salinity from 2002-2009 collected by the CDA, whilst the independent evaluation dataset is 23 observations of annual resolution from 1973-2009. Due to the absence of sub-annual macrophyte observations, the process is calibrated with four observations (1973, 1977, 1982 and 2007) and evaluated over the remaining 10 observations sporadically charting annual macrophyte coverage from 1973-2009 (chapter 3.4.2.4).
Fisher population growth (equation 3.19) is driven by giving the fisher stocks initial values and simulating growth under external birth and death rates. Therefore, this process is evaluated through CPUB and per capita income. The number of fishers per boat (i.e. number of total boats) is derived from annual CDA (2015) data from 1999-2009. Catch per unit boat is informed by two independent measures of maximum catch per boat type, deriving from the annual catches of 1973 and 2004 for traditional boats and motorboats, respectively (chapter 3.4.3). Mean weighted fish price from 1978-2009 also drives internal
108 Chapter 3: Model construction socioeconomic dynamics, from which per capita income derives from internally driven catch and total fisher populations. The fish price dataset was also used to assess the need for a permanent livelihood carrying capacity; however, it does not necessarily mean modelled fish catch, efficiencies and population growth will replicate reality (chapter 4). Therefore, care has been taken to avoid circularity where possible, by evaluating system outcomes either against independent datasets or through composite measures to assess multiple processes.
3.5 Chapter conclusion
This chapter has detailed the conceptualisation and parameterisation of a model designed to investigated Chilika’s problematic behaviours. Rainfall-runoff models adapted from Ghashghaei et al. (2013) generate monthly freshwater and sediment yields, which contribute to the sedimentation of the historical tidal outlet. These hydrological fluxes also affect salinity, temperature, dissolved oxygen and macrophyte coverage. Salinity and dissolved oxygen are parameterised as graphical functions based on seasonal CDA data; water temperature is parameterised as a linear model relating near-surface air temperatures and freshwater inflows; macrophyte coverage is parameterised as a stock to capture feedbacks with salinity, dissolved oxygen and sedimentation. Macrophytes provide ecological refuge by blocking fishing grounds, whilst habitat quality influences multidecadal fish survival based on reference points for key species. The Bueno and Basurto (2009) fishery model underpins fish and catch dynamics, modified by a dynamic juvenile catch driven by fisher adaptations and a variable survival rate driven by lagoonal water conditions. Externally derived birth and death rates principally drive the human populations until the fleets exceed their livelihood carrying capacities. The number of days fished are parameterised as Holling (1959b) functional responses and a relatively wealthy proportion of traditional fishers switch to non-traditional fishing when traditional productivity value declines.
A diverse range of information has contributed to model parameterisation. However, simply collecting data and best available knowledge does not qualify a model to investigate a problem of interest. Therefore, the next chapter conducts multidimensional model evaluation to assess whether the model is structurally robust (Barlas, 1996), captures past trends and complexities, and produces output sensitivities that are coherent with system understanding (Saltelli, 2000b).
109
Chapter 4: Model evaluation
Chapter 4: Model performance analysis
4.1 Chapter introduction
As abstractions of mental models, simulation models are never perfect representations of reality (Sterman, 2000, Sterman, 2002). However, through various structural (chapter 4.1) and behavioural tests (chapter 4.2), a model’s ability to investigate problematic processes and states can be assessed. First, a model should be structurally robust to produce behavioural outputs for the ‘right’ reasons, rather than as a coincidence of tuning (Barlas and Kanar, 2000). Second, driver-response relationships must be traceable, with unexpected dynamics suggesting unexplainable uncertainties in parameterisation. Third, output variability should be apportioned to different contributors to assess the relative influence of modelled processes (Tarantola et al., 2000). For instance, parameterisation can be questioned if highly variable dynamics result from a limited range of input uncertainties (Sterman, 2000).
Non-behavioural analysis precludes simulation, establishing similarities between the descriptions of the real world and reality (Barlas, 1996). Behavioural procedures then compare outputs with observed dynamics (Sterman, 2000), asking to what extent the model captures historical causes, outcomes and variabilities.
4.2 Non-behavioural analysis
Qualitative analysis was rare prior to the 1990s, with system dynamicists favouring behavioural techniques (Barlas and Kanar, 2000). A series of papers by Barlas and colleagues (Barlas, 1989, Barlas, 1996, Barlas and Kanar, 2000) called for structural assessments to ascertain whether studies replicating historical trends were doing so for robust reasons. Non-behavioural analysis has occurred throughout model construction so far; for example, the adequacy of model boundaries is argued during ‘problem familiarisation’ (chapter 2.5) and ‘system visualisation’ (chapter 3.2). This study applies three non-behavioural tests to flag uncertain processes and data sources for behavioural analysis.
111 Chapter 4: Model evaluation
4.2.1 Unit consistency
Each module is assessed using STELLA’s ‘Check Units’ tool, which assesses whether the mathematical expression of each process (e.g. LMC rainfall-runoff) is consistent with its underlying driving variables (e.g. rainfall volume). STELLA flags an error when the units for the driving variables do not match the unit manually assigned to the higher-level process (Bueno, 2014). Manual assessment also checks for implausible unit combinations, such as boat/fish2 or fish2/m4.
Overall, the assigned units are judged to be just and dimensionally consistent, arguing that the processes have plausible connections in both real and virtual space (Forrester and Senge, 1980, Bueno, 2014). The rainfall-runoff variables convert between millimetres, metres, metres squared and metres cubed. The sediment generation variables are concentrations of sediment per unit volume of water and the absolute volume of sediment in the catchments flowing towards Chilika. Unit inconsistencies arise in the ecohydrological module because the habitat suitability variable converts concentrations (“ppm” or “ppt”) into dimensionless effects on fish survival (chapter 3.4.2.5). Similar inconsistencies occur within the fishery module, as Holling type-II functions convert between fish density and the number of days fished. Rather than errors in the underlying reasoning for connection, these errors are flagged as STELLA assumes the higher- level process (e.g. number of days fished) adopts the same units as its driver (e.g. fish density). In such cases, the errors are manually checked to assess the plausibility of driver connections.
4.2.2 Qualitative parameter assessment
Parameter assessment attributes confidence to information used in parameterisation, traditionally involving discussing parameters with other modellers and system experts (Sterman, 2000). Chapman and Darby (2016) devise a formal framework to assign a confidence score to each parameter, based on the reliability, number and spatiotemporal transferability of data sources. Datasets from the area of interest, multiple experiments and the reporting of statistical significance (e.g. p-value < 0.001) equates to the most reliable variables. The total parameter score of each variable equals the sum of individual scores divided by the total possible score. Variables with an element of personal intuition or with total scores less than one standard deviation below the mean are qualitatively uncertain (appendix C.1).
112 Chapter 4: Model evaluation
Information sources around the average score generally only conducted one experiment, with one time-series and unreported significance levels. For example, whilst following protocols for the parameterisation of graphical functions in the absence of multiple data points (Sterman, 2000), the parameter scores of the dynamic rainfall-runoff coefficients are qualitatively uncertain. The LMC and WC are comparable environments to the Gamasiab river catchment of Ghashghaei et al. (2013); however, the areas of LMC and WC are respectively double and half the Gamasiab, giving limited spatial transferability.
Ultimately, parameter assessment identifies at least eleven variables requiring sensitivity analysis to understand whether parametric uncertainties disproportionately affect model processes and outcome states.
4.2.3 Qualitative extreme conditions
Extreme condition tests explore whether model processes uphold beyond normal operating ranges (Forrester and Senge, 1980, Barlas, 1996). Qualitative extreme conditions trace input minimums and maximums through model structure to hypothesise both local and global impacts – which are then checked with behavioural tests. As foundations of past problematic behaviours, the stocks of freshwater runoff, fish and fisher populations receive the extreme conditions.
Ultimately, extreme inputs and outputs are traceable through model structure, hypothesising scenarios leading to fishery collapse. If LMC and WC rainfalls start at zero mm/month, surface runoff of both catchments should equal zero (equation 3.1). If rainfall becomes zero after months of business-as-usual, then the lagged baseflows (figure 3.11) gradually decline runoff (in reality, upper catchment flows from the Middle Mahanadi should maintain LMC runoff). In turn, fluvial sediment transport should cease if freshwater discharges stop. However, littoral inputs would continue to build Chilika’s sediment stock because sediment removal requires freshwater inflows (equation 4). Assuming the tidal outlet is clear and vegetation coverage is low, salinity persists ~21 ppt (chapter 3.4.2.1) and dissolved oxygen equals ~6.9 ppm (chapter 3.4.2.2) under zero freshwater influx. In the short-term, these conditions would not significantly affect fish survival. However, multidecadal sedimentation from littoral drift would eventually close the outlet (equation 3.5), leading to lagoon freshening and macrophyte proliferation. Fish survival and migration would decline, along with catchability due to blocked fishing grounds.
113 Chapter 4: Model evaluation
Conversely, extreme-high monthly rainfall would produce high freshwater deliveries, causing the lagoon to freshen beyond average during and after the monsoon (chapter 3.4.2.1). Despite boosted flushing, extreme sediment deliveries may choke the outlet, accelerating salinity declines and macrophyte growth. These extreme conditions hypothesise equifinality, as extreme-low and extreme-high inputs may cause fishery decline albeit over different time scales.
Fish regeneration becomes zero if fish births or survival cease (chapter 3.4.3). Once all juveniles mature, catch and natural deaths should cause the total fish stock to fall to zero – ceasing financial outputs and causing the livelihood carrying capacities to tend towards zero. Setting fisher birth rates to zero would cause the human population to decline immediately. Conversely, extreme-high birth rates allow the fisher populations to grow until it reaches the livelihood carrying capacity.
4.3 Behavioural analysis
Behavioural assessments explore whether a structurally sound model produces dynamics consistent with reality (Barlas, 1989). Graham (1976) states that the measures used to judge consistency with reality must relate to the model’s purpose. Processes leading to collapse dynamics are of particular interest as the overarching aim is to investigate future fishery persistence. Therefore, dynamics similar to Chilika’s late-1980s collapse should be reproduced, including timing, magnitude and causes (Forrester and Senge, 1980). Furthermore, outputs should not be overly sensitive to changes in driver values; sensitivity suggests that a parameter value sits within a region of uncertainty, potentially over predicting collapse and causing unjustified concern. Behavioural analysis assesses both high-level and emergent behaviours wherever possible, although data availability forms a limitation where datasets completely inform parameterisation (e.g. dissolved oxygen).
4.3.1 Quantitative extreme conditions
This section tests the hypothesis that model integrity remains under extreme inputs by ‘switching off’ or maximising inputs to stocks from 1973 (Forrester and Senge, 1980, Barlas and Kanar, 2000). Five variables receive extreme conditions (table 4.1); pertaining to the three drivers of monthly rainfall, fish survival and human population growth. Only hydroclimatic extremes are traceable across the
114 Chapter 4: Model evaluation
Table 4.1: Scenarios used to assess model integrity under extreme conditions. Outputs: 1- outlet closure; 2- salinity; 3- macrophyte coverage; 4- catch; 5- fisher population.
Extreme scenario
Module Variable(s) Low High Outputs
Hydroclimatic LMC and WC 0 mm/month 0 mm/month 1, 2, 3, 4, rainfall 5
Fishery Survival rate 0 fish/month Max. ratio (0.5) 4, 5
Socioeconomic Traditional 0 fishers/month 1000 4
and non- fishers/month traditional growth entire system because the dynamics of the fish and human populations do not feedback to the hydroclimatic and ecohydrological modules. Therefore, extreme fish survival and fisher population growth are stored in appendix C.2.
Overall, the modelled extreme behaviours match the qualitative hypotheses, implying that realistic and readily explainable behaviours are produced beyond the model’s normal operating range. As hypothesised, persistent maximum historical rainfall overwhelms sediment flushing and the resulting outlet closure (figure 4.1B) causes lagoon freshening (figure 4.1C). Low salinity and high sedimentation provide suitable conditions for macrophytes to cover the whole lake after 10 years (figure 4.1D). In response to an immobile resource, catch (figure 4.1E) and active fishers (figure 4.1F) fall to zero, with the latter lagged by ten years due to the smoothing coefficient representing fishers unwilling to leave fishing (chapter 3.4.4).
Outlet closure increases under the extreme-low rainfall scenario due to littoral inputs (figure 4.1B), but as predicted, closure is shallower than under extreme- high rainfall. Consequently, salinity remains above 15 ppt (figure 4.1C). Catch increases due to effort intensification and unimpeded fish migration, implying that zero freshwater influx fails to decline dissolved oxygen to levels detrimental to fish survival. Unlike the qualitative hypothesis, the rate of outlet closure does not trigger catch collapse; therefore, extreme-low rainfall is preferable of the two extremes for catch persistence over this timeframe. In reality, a forty-year
115 Chapter 4: Model evaluation
Figure 4.1: System responses to (A) extreme high (red) and low (blue) rainfall inputs: (B) outlet closure expressed as a percentage of the lagoon’s sediment capacity; (C) lagoon-wide average salinity; (D) macrophyte coverage; (E) catch and (F) active fisher population.
116 Chapter 4: Model evaluation drought would cause the fisher population to decline through human health impairments and outmigration.
Stored in appendix C.2, zero fish survival causes collapse within 12 years due to resource exhaustion – as hypothesised under qualitative extreme conditions. In contrast, nonlinear catch growth occurs once the density-survival feedback is switched-off, as exponential fish population growth essentially causes the fish to fall into the nets of fishers. However, fisher population growth remains slow, driven by underlying birth and death rates rather than livelihood constraints.
Catch intensification occurs under 1000 new fishers/month, reaching >15,000 tonnes by 1980 (appendix C.2). However, the collapse following the catch spike suggests overexploitation. Here we see the first signs of a SJOS, with catch reversal beginning around 65,000 fishers. Fishers abandon the system for ~10 years until catch disappearance causes resource recovery and livelihood generation. The second case of catch stagnation emerges after 40 years, suggesting that extreme fisher population growth may trigger boom and bust dynamics. A constant fisher population produces a catch reference mode similar to reality (see below), albeit with three main disparities owing to dampened effort intensification: (i) a relatively limited catch spike during the late-1980s, (ii) a shallower historical collapse and (iii) a post-2000 recovery that is approximately half of reality.
4.3.2 Behavioural reproduction
Behavioural reproduction aims to understand the disparity between the model and reality, in terms of the magnitudes and timings of problematic outcomes and drivers (Sterman, 1983). However, due to non-normality, nonstationarity and co- dependence, SDM outputs are ill-suited to the statistical measures (e.g. coefficient of determination) conventionally used to assess goodness of fit with observations (Barlas, 1989, Sterman, 2000). Moreover, a perfect regression coefficient can occur when time-series dynamics are antiphase. Nevertheless, it is useful to understand whether the model tracks the past. To this end, percentage error calculates coherence between the mean averages of the total simulated and observed time-series, checking whether long-term magnitudes of drivers are systematically over/underestimated. Second, mean absolute percentage error assesses the residuals between individual records, albeit dependent on the completeness of the observed datasets.
117
Chapter 4: Model evaluation
drivers and outputs, alongside respective observed observed respective alongside outputs, and drivers
percentage error. percentage
–
modelled modelled
mean absolute percentage error; PE error;PE percentage absolute mean
–
MAPE MAPE
averages.
: Characteristics and performance metrics of important ofimportant metrics performance and Characteristics :
2
.
4
Table
118
Chapter 4: Model evaluation
Table 4.2: (continued). Table4.2:
119
Chapter 4: Model evaluation
wide salinity (B) annual annual (B) salinity wide
-
.
nnual average lagoon average nnual
series. (A) A (A) series.
-
(D) annual per capita income capita per annual (D)
and
black) driver time driver black)
; (C) active fisher population fisher active (C) ;
black) versus observed ( observed versus black)
-
Modelled (non Modelled
catch per unit boat unit per catch
: :
2
.
4
Figure Figure
120 Chapter 4: Model evaluation
Overall, the model overestimates reality by 1.0% across the 17 processes assessed (table 4.2), although the absolute mean is 2.4%. Likewise, the model reproduces the historical reference modes of four drivers (figure 4.2) and annual catch (figure 4.3), including periods of stability, decline and recovery. Hydroclimatic processes have the closest coherence to the long-term averages, with half of the percentage errors below 1%. However, the mean average percentage errors of LMC freshwater and sediment fluxes above 40% are because the model marginally mistimes the rising and falling limbs associated with the start and finish of the southwest monsoon, leading to monthly differences up to nine orders of magnitude! Therefore, the model captures creeping stresses like the closure of the tidal outlet but underperforms with precise intra-annual deviations, representing a trade-off of using non-specialist models.
Overestimating the proportions of monsoon freshwater and sediment fluxes indicates that the runoff coefficients could be dampened at high rainfall volumes. Yet, freshwater and sediment flows to Chilika from the LMC are within 2% of reality, suggesting suitable flow proportioning and routing downstream of Naraj. The proportions of fluvial inputs to Chilika match reality, suggesting both catchments are parameterised adequately in relation to one another.
The model estimates long-term salinity to within 1% and captures the main features of the observed record, including the gradual multidecadal decline until 2000 and the abrupt recovery to 15 ppt post-restoration (figure 4.3A). However, the modelled trend is less variable than reality, primarily due to the dampening effect of outlet sedimentation (equation 3.8). Consequently, the model misses the short-term spikes that might provide a window of opportunity for recovery (Balke et al., 2014). However, history suggests Chilika’s state is comparatively coupled with long-term salinity trends, with temporary improvements (e.g. 1993) producing negligible fishery recovery.
The modelled fisher population captures the magnitude of observed change (figure 4.3C), with mean average percentage error never exceeding 8%. However, the trends are antiphase during the 1970s and 1980s, meaning regional birth and death rates and the livelihood carrying capacity switch are unable to explain all variations in the observed fisher trend. CPUB is particularly informative for model performance by combining the trends of catch and the number of boats. Both CPUB and per capita income trends capture historical reference modes (figure 4.3B and D). Modelled CPUB replicates the period of stability covering the 1970s to the late 1980s, followed by abrupt decline during the 1990s and recovery post-
121 Chapter 4: Model evaluation
Figure 4.3: Modelled (red solid) versus observed catch (black dashed) from 1973-2009, overlain with modelled (red dotted) and observed (black long-dash) time-series breakpoints.
2000. Modelled and observed incomes remain below 10,000 INR until year 2000, before both abruptly increasing to ~20,000 INR from 2002 following ecorestoration.
It is also important to evaluate fish catch as the principal outcome synthesising both human use and ecological health. Overall, modelled catch replicates both the long-term volume (table 4.2) and temporal patterns (figure 4.3), including stable productivity pre-1985 and briefly increased catch pre-collapse. The model also captures the presence and magnitude of the 1990s collapse, with outputs from 1996 closely aligned with reality. Based on the detection of breakpoints in linear regression trends (Zeileis et al., 2015), both time-series optimally split into three periods, suggesting that the model replicates the number of observed time- series phases (figure 4.3). However, modelled collapse lags reality by 3 years, potentially caused by slightly underestimating sediment deliveries over the 1980s and 1990s (table 4.2). This disparity is arguably insignificant when considering multidecadal dynamics, with the model aiming to explore pathways to normative states rather than specific timings.
122 Chapter 4: Model evaluation
4.3.3 Time-step dependency
The model exhibits the ability to replicate the observed periods of system stability, collapse and recovery when simulating at monthly time-steps (chapter 4.3.2). However, common to system dynamics studies, historical dynamics are simulated at default and half-default resolution to assess whether significantly different outcomes emerge. The underlying assumption is that model equations are insensitive to the model’s temporal resolution, so that doubling the frequency with which the fish population migrates to and from the Bay of Bengal should not cause qualitative changes in the model’s reference mode. However, the emergence of different dynamics questions the generality of model processes and signifies a potential dependence on the selected time-step (Sterman, 2000). Where a variable is represented by a temporally discrete dataset (e.g. LMC monthly rainfalls), STELLA splits the value equally across the new time-step.
Simulating under the half time-step reproduces the phases of catch stability, growth, collapse and recovery (figure 4.4). The trends diverge after the introduction of motorboats in 1983, because the finer temporal resolution doubles the number of times traditional fishers can upgrade per month. This
Figure 4.4: Time-series depicting the difference between fish catch outputs from default (red) and halved (blue) time-steps against observed catch (black).
123 Chapter 4: Model evaluation dynamic causes a higher and earlier late-1980s peak catch, and steeper recovery post-2000. As output normality cannot be assumed (Sterman, 1983), a non- parametric test (Wilcoxon-Mann-Whitney) assesses whether a significant difference exists between the output mean averages, with the time-step halved until differences are insignificant (Simonovic, 2010). The Wilcoxon-Mann-Whitney output (p >0.05) implies that halving the time-step does not cause significantly different catch, suggesting that model outcomes are not overly reliant on time- step length.
4.3.4 Structural analysis through alternative pasts
Thus far, the model is able to replicate the historical reference mode under parameterised conditions and a halved time-step, as well as retaining its behavioural integrity under extreme conditions. The remainder of the chapter assesses the contribution of different drivers to the modelled reference mode, in order to check that the model reproduces the past for the right quantitative and qualitative reasons (Barlas and Kanar, 2000). This subchapter helps to test the reality of the model’s structure by removing modelled processes one-at-a-time and assessing the resulting coherence to the historical reference mode. If the model matches our understanding of Chilika’s past, it is expected that removing the ecological drivers of fish catch will cause the greatest disparities from the reference mode (see figure 4.4 – red line).
Seven scenarios are modelled by removing a key driver of catch from the model:
1. Fish migration via the tidal outlet: all mature fish remain inside the lagoon throughout the year, decoupling fish stock dynamics from the sedimentation of the tidal outlet. 2. Ecohydrological conditions: the survival of new-born fish is decoupled from the salinity, temperature and dissolved oxygen of the lagoon. 3. Aquatic vegetation: fishing efforts are decoupled from vegetation coverage, meaning all areas are open to fishing. 4. Fisher population growth: the fisher population remains constant at its 1973 value of 19,000. 5. Motorboat use: the fisher population is dynamic, but all fishers continue to use traditional boats. 6. Days fished: each fleet fishes the maximum possible days per month (see chapter 3.4.3).
124 Chapter 4: Model evaluation
Figure 4.5: (top) time-series depicting seven alternative pasts, each derived by the removal of one system stress; (bottom) time-series calculated by the cumulative sum of the absolute errors between the annual catches of the alternative pasts and the reference mode.
7. Juvenile catch: fishes no longer have the adaptive capacity to compensate for fewer days fished by catching juveniles.
Disparities between the seven scenarios and the reference mode may be broken into three time periods: (i) pre-collapse (1973-1992), (ii) collapse (1993-2000) and (iii) recovery (2001-2009). As the fisher population is constant and motorboats do
125 Chapter 4: Model evaluation not appear until 1984 (figure 4.5), the fish catch decline under scenario 4 is likely due to the linear increase in aquatic vegetation that blocks ~100 km2 by the early- 1980s. The noticeable divergence from the reference mode occurs from 1984, when all runs except for scenario 5 (motorboats) undergo a short-lived catch peak. Catch is further boosted under the scenarios where the ecological drivers are removed, with catch sensitivity increasing from the removal of aquatic vegetation to the removal of fish migration via the tidal outlet. The cumulative disparities between scenarios 1 and 2 and the reference mode begin increasing nonlinearly from the early-1990s, reflecting the removal of tidal outlet sedimentation and ecohydrological degradation, respectively. In contrast, all other scenarios undergo catch collapse, albeit with subtle differences in their timings and magnitudes. For instance, the deepest and steepest collapse is driven by the removal of aquatic vegetation, which allows fishing efforts to reach their full, unblocked potential. In contrast, removing dynamic fishing extent (i.e. fisher population) and intensity (i.e. motorboats) causes shallower collapses with recoveries that are marginally faster than the reference mode, owing to shallower growths in fishing effort. Interestingly, removing ecohydrological degradation (scenario 2) causes a catch downturn no deeper than pre-1980’s catch levels, suggesting that coherent with system understanding (Kumar and Pattnaik, 2012, Mohanty et al., 2015), the collapse in the modelled reference mode was a result of both outlet sedimentation and ecohydrological degradation.
The removal of fisher population growth (scenario 4) and motorboats (scenario 5) causes weaker than observed fishery recovery, suggesting that reference mode recovery is a synergy of fishery intensification, extensification and a renewed fish stock. Scenario 2 also sharply recovers following outlet opening in 2000, whilst the catch stagnation caused by ecohydrological degradation and macrophyte growth disappears from 2000 in scenario 1. Lastly, catch reaches more than 15,000 tonnes in scenario 3, implying that the complete loss of vegetation from the lagoon would have further boosted fishery recovery.
The modelling of alternative pasts has uncovered the different driver contributions to the historical reference mode. Overall, the removal of fish migration from the model produces the most erroneous estimations, due in large to the overestimation of 1980s fishery intensification and subsequent flatline during the 1990s. Variability in the number of days fished and the removal of juvenile catches caused negligible divergence from the reference mode, reflecting
126 Chapter 4: Model evaluation the system’s comparative sensitivity to the lagoon’s sediment dynamics, ecohydrology and fishing efforts (extent and composition).
4.3.5 One-at-a-time sensitivity analysis
Sensitivity analysis investigates errors associated with uncertain model processes (Loucks et al., 2005); parameterisation may need revising if outputs are overly sensitive to a particular driver in comparison to reality. Methods range from perturbing one variable at a time to complex methods decomposing output variance (Saltelli, 1999). Here, a sensitivity degree metric (equation 4.1) first assesses model sensitivity by monitoring output responses to one-at-a-time input changes. Chapter 4.3.6 then explores multiple interacting sensitivities.
The sensitivity degree (푆푄) has been applied to SDMs of lake management (Guo et al., 2001), urban water planning (Zhang et al., 2008), river basin development (Wei et al., 2012) and agricultural production (Chapman and Darby, 2016). The variables identified as qualitatively uncertain during parameter assessment
(chapter 4.2.2) form the inputs(푋(푡)); outputs are variables influencing the top- level dynamics of Chilika’s fishery (푄(푡)):
∆푄(푡) 푋(푡) 푆푄 = | . | (Eq 4.1) 푄(푡) ∆푋(푡)
One limitation of the sensitivity degree is that it is unclear from Guo et al. (2001) and subsequent applications whether 푆푄 compares inputs and outputs at different times (e.g. t = 20, 40…120) to initial values (t = 0), or, to the reference mode of optimal parameters (e.g. figure 4.3). Here, the former approach assesses how system sensitivity changes for different input values, asking whether driver- responses become nonlinear, rather than how outputs change relative to a reference mode assumed to be correct. As per Chapman and Darby (2016), analysis is ran four times per variable, changing input values by +10%, -10%, +20% and -20% per 40 model iterations, whilst holding all other variables constant at their median observed value. The general sensitivity (GS) degree then averages sensitivity across the four runs. Critically sensitive variables are assumed to have GS >0.1 (Zhang et al., 2008).
Overall, model outcomes are sensitive to a variety of hydrological and ecohydrological processes (GS >0.1), but are less sensitive (GS <0.1) to fish population parameters and socioeconomic drivers, with the exception of
127 Chapter 4: Model evaluation
Table 4.3: One-at-a-time sensitivities of critical outputs to 17 social and ecological drivers. Highlighted cells represent critically sensitive driver-response relationships, where generalised sensitivity >0.1.
ecological carrying capacity (table 4.3). In particular, hydrological fluxes to Chilika are sensitive to LMC rainfall-runoff coefficients, reflecting the LMC’s 75% contributions to the annual freshwater and sediment deliveries. As expected, salinity is sensitive to outlet closure, as seen by observed salinities declining from
128 Chapter 4: Model evaluation
~ 18 ppt (1957) to 3.6 ppt (1997). Modelled macrophyte growth is sensitive to relatively short-term ecosystem variations, consistent with the growth of macrophyte coverage from 314 km2 to 540 km2 between October 1994 and January 1995 (Pal and Mohanty, 2002). It is important to note that lagoon sedimentation was at a record high in the 1990s; thus, stepwise macrophyte growth likely reflects nutrient abundance, low salinity and stagnant water. Historical macrophyte coverage also increased from 20 km2 (1973) to 573 km2 (1996) (Kadekodi and Nayampalli, 2005) without significant alterations in long- term freshwater inputs, indicating sensitivity to multidecadal, cumulative sedimentation and salinity.
The modelled fish population is generally insensitive to individual inputs, suggesting that uncertainties in the hydroclimatic and ecohydrological modules do not propagate through model structure. The only exception is the disproportional fish population change relative to the change in eco-K, likely reflecting the process’ control on the population growth rate and maximum supportable population. Change over the first 40 time-steps has an 푆푄 equalling 0.9, as the fish population rapidly adjusts to a dynamic eco-K. Uncertainty surrounding the eco-K estimate requires further investigation with multivariable sensitivity analysis.
All outputs are insensitive to WC rainfall-runoff coefficients (table 4.3), reflecting the 25-30% contribution to Chilika’s total freshwater delivery. Likewise, ecological dynamics are proportional to the input of littoral sediment, supporting the dominant influence of fluvial sediment deliveries on fishery dynamics (Das and Jena, 2008). The insensitivity of fish population growth to processes like fecundity and natural deaths builds confidence that the fish population traits do not endogenously cause fish population collapse. Similarly, insensitivity to socioeconomic parameters suggests that small changes in fishing efforts are insufficient to cause resource nonlinearities.
Therefore, one-at-a-time sensitivity analysis suggests that nonlinear resource change generally requires stresses that are: (i) combinations of human and natural, (ii) multidecadal, and/or (iii) nonlinear. Positively, problematic behaviours are not endogenously produced around parameter estimations changing one-at-a- time (Bueno, 2014). However, this analysis does not assess interacting uncertainties emerging from multiple nonstationary scenarios. Therefore, current knowledge gaps include the combined effects of uncertainties around fishing effort, plus whether simultaneously varying fishing effort and eco-K causes
129 Chapter 4: Model evaluation unrealistic outcomes. To this end, the next subchapter investigates the sources and magnitudes of multiple interacting uncertainties effecting whether the model reproduces historical dynamics and states.
4.3.6 Multivariable sensitivity analysis
4.3.6.1 Introduction and purpose
The main advantage of global sensitivity analysis over one-at-a-time analysis is the ability to explore how combinations of uncertainties [known as ‘sets’ – Loucks et al. (2005)] effect output dynamics (Saltelli, 2000b). Global sensitivity analysis can be intra- or inter-module, with the former producing relatively regional analysis. Sampling-based methods facilitate systematic global sensitivity analysis (Helton and Davis, 2000); whilst individual frameworks vary (for reviews see Helton et al., 2006, Pianosi et al., 2016), sampling-based methods generally consist of five stages (Campolongo et al., 2000): 1. Factor screening identifies particularly sensitive variables from variable- rich models. Here, qualitative parameter analysis (chapter 4.2.2) and one- at-a-time sensitivity analysis (chapter 4.3.5) form this initial step. 2. Probability distribution functions bind minimum and maximum uncertainty values for each variable. 3. Per simulation, random number generation or stratified techniques such as Latin Hypercube Sampling generate uncertainty sets for each variable (Helton and Davis, 2000). Therefore, some simulations have variable sets close to parameterised values, whilst others simulate major deviations. 4. Simulate uncertainties through the model and capture outputs relating to the behaviour of interest. 5. Undertake uncertainty and sensitivity analysis to attribute confidence to parameters, establish acceptable bounds and contributions to outputs (Campolongo et al., 2000, Helton et al., 2006).
The process of randomly deviating a variable’s value is known as Monte Carlo analysis (MCA) (Fedra, 1983, Campolongo et al., 2000). MCA originates from 18th century probability theory, specifically the “needle experiment” of Georges Louis LeClerc (1707-1788) (Harrison, 2010). Modern MCA is commonly traced to the work of John von Neumann and Stanislaw Ulam, through their application of probabilistic techniques to nuclear physics in the 1940s (Robert and Casella, 2011). Models of chemical, health and environmental systems now use MCA
130 Chapter 4: Model evaluation
(Doubilet et al., 1985, Burmaster and Anderson, 1994), facilitated by computers capable of ~10,000 simulations over a matter of hours.
In SDM, Phillips (1976) and Sharp (1976) use MCA to argue that model outputs are robust if qualitative conclusions are consistent across the range of plausible error values. Similarly, after establishing metrics of acceptable behaviours, Fedra (1983) identifies ‘behaviour-giving’ parameter sets for lake nutrient and rainfall- runoff models of the Attersee watershed, Austria. Unjustifiably sensitive process combinations may undermine the confidence of decision-makers to act on model conclusions, especially if the parameterisation datasets also have considerable qualitative uncertainties.
Building on finding acceptable and optimal parameter values, Hekimoglu and Barlas (2010) assess the sensitivity of dynamic outputs to MCA inputs. For a SDM of business project management, the authors formulate GLMs between rates of job completion and plausible error magnitudes, finding uncertainties in total staff, deadline date and job complexity to be significant determinants of desirable outputs. Importantly, this GLM approach is only applicable under certain assumptions, namely independence between variables, input-output linearity and residual normality. However, variable independence is contentious (Ford, 2010), particularly if multiple feedbacks cause the value of a certain variable to be a legacy of another. Here for example, days fished and boat numbers depend on fishery state, albeit over different temporal scales.
Therefore, MCA can provide numerous insights into model dynamics, uncertainties and robustness unobtainable from one-at-a-time analysis. Here, investigating the effects of error magnitudes on model outputs explores: 1. How outputs vary across multiple interacting uncertainties; 2. Which variables produce most sensitive outputs and why; 3. Whether acceptable error bounds can be defined for the most sensitive variables.
The first objective seeks to understand the effects of output uncertainties, before question two apportions their different sources. Question three checks whether these uncertainties affect the ability to simulate past and future dynamics.
4.3.6.2 Parameter selection, sampling technique and behaviour metrics
Parameter screening selects variables known to exert most influence on outputs (Saltelli, 2000b); as such, variables converting between units are discounted
131 Chapter 4: Model evaluation
Table 4.4: The 25 model variables given Monte Carlo error magnitudes alongside a summation of their model process. See figure 3.11 for hydroclimatic abbreviations.
Module Process Variable(s)
Hydroclimatic Rainfall-runoff LMC Rc, LMC Te, LMC Co, WC Rc, WC generation Te, WC Co
Sediment generation LMC sediment rating, WC sediment rating Coastal Littoral sediment, outlet closure sedimentation Eco-hydrological Salinity Monthly salinity
Dissolved oxygen Monthly dissolved oxygen
Macrophyte coverage Salinity effect on macrophyte growth, dissolved oxygen effect on macrophyte growth, sedimentation effect on macrophyte growth Core fishery Fish population Fecundity, reproductive life length, dynamics natural death rate, survival rate, mature fish per unit weight, ecological carrying capacity Social-economic Fishing effort Catch capacities, proportion of fishers upgrading, fishers per boat, days spent fishing immediately. Variables undergoing MCA here are those which underwent one-at-a- time analysis (table 4.4), which were previously screened through qualitative parameter assessment. Disaggregating LMC and WC rainfall-runoff coefficients is to understand their relative contributions, whilst disaggregating the individual drivers of macrophyte growth is due to the sensitivity of the process during one- at-a-time analysis. MCA additionally explores interacting sensitivities associated with socioeconomic drivers, despite not showing one-at-a-time sensitivities.
STELLA lacks a built-in tool to vary every variable automatically, so probability distribution functions are defined manually between minimum and maximum bounds. The probability distribution functions involve density functions representing the relative frequency of randomly selecting any given error (Hammond and McCullagh, 1974). As particular variables are lacking historical data (e.g. eco-K), samples are made from relative error magnitudes rather than absolute values. Thus, the MCA parameter estimate (V) of a given variable for a
132 Chapter 4: Model evaluation given simulation (i) equals the product of the original parameterised value (V ) 0 and the randomly sampled error magnitudes (ɛ):
푉 = 푉0,푖 × 휀푖 0.5 ≤ 휀 ≤ 1.5 (Eq 4.2)
STELLA’s sensitivity analysis tool generates ɛ. Each variable is given a different random seed to remove duplicate random number strings and reduce collinearity between error estimates (Hannon and Ruth, 2009). The model is run 2000 times for global sensitivities (chapter 4.3.6.4), then a further 1500 times for the variables classified as critically sensitive (chapter 4.3.6.4). The probability distribution functions are given uniform distributions as common when historical observations are lacking (Campolongo et al., 2000). Therefore, per simulation, each variable has an equal chance of simulating the effects of underestimation or overestimation, up to 50% in either direction. The assumption that large errors are equally as likely as small errors arguably undervalues parameterisation accuracy, but the uniform distributions facilitate comparisons between input and output distributions to explore error bias (Phillips, 1976, Campolongo et al., 2000). It is assumed that the relationship between error magnitude and output is linear; however, as one-at-a-time sensitivity analysis shows, model outputs have different sensitivities to inputs. Consequently, nonlinearities may occur once multiple errors interact, hence the use of MCA to discover the significant contributors and sensitivities across parameter space.
Many strategies exist to communicate uncertainties, from time-series illustrating outcome dispersion to statistical breakdowns of variance (Loucks et al., 2005). Outputs are classified against a ‘corridor of constraints’ to understand which uncertainties fail to produce Chilika’s historical reference mode (Fedra, 1983, Beck and Chen, 2000). As the model is interested in the long-term state of catch, outputs are considered behaviour-giving if ≥80% of modelled production lies within the 99% confidence intervals of historical catch (figure 4.6). The confidence intervals derive from a locally weighted least squares regression curve fitted in R (R, 2008) and the 80% threshold factors the narrowing of the constraint corridor from 2001-2005. Next, the Spear-Hornberger method of generalised sensitivity analysis uses a two-sample Kolmogorov-Smirnov test to assess equality between behaviour-giving and non-giving input distributions (Spear and Hornberger, 1980, Wade et al., 2001); a variable is critically sensitive if its two distributions are significantly different.
133 Chapter 4: Model evaluation
Figure 4.6: Observed catch (black) and its upper and lower 99% confidence intervals (red). Observed catch is smoothed (n=3) to remove short- term volatility and accentuate the long-term trend.
It is here worth describing the Kolmogorov-Smirnov test in detail, as it is also used later to investigate separation between safe and just and dangerous driver trajectories. The Kolmogorov-Smirnov test is nonparametric, meaning observations do not have to meet assumptions about distribution type, including normality or the same number of observations (Massey, 1951, Mitchell, 1971). Moreover, the Kolmogorov-Smirnov test applies to continuous distributions (e.g. uncertainty magnitudes), rather than discretised data required by the Chi-squared test (Mitchell, 1971). In essence, the Kolmogorov-Smirnov statistic (퐷푚,푛) measures the maximum vertical distance between two empirical cumulative distribution functions, where 푚 is the distribution of the behaviour-giving uncertainties and 푛 is the distribution of the ungiving uncertainties (Massey, 1951, Spear and Hornberger, 1980). The null hypothesis (H ) that the distributions are 0 drawn from the same population can be rejected once the vertical difference between the empirical cumulative distribution functions surpasses a threshold set by the significance level (Massey, 1951). Failing to reject H means that the 0 reproduction of historical catch is insensitive to the error magnitude of a given
134 Chapter 4: Model evaluation driver. One weakness of the Kolmogorov-Smirnov statistic is its increased sensitivity to two distributions with vastly different numbers of observations.
Isolating the variables found to be significant determinants of whether the model reproduces historical catch addresses the third aim of this section. In order to assess the error magnitudes degrading the model’s ability to replicate the past, MCA is then conducted again, only this time varying the significant variables between ±10%, ±20% and ±30% of their parameterised values.
4.3.6.3 Global sensitivity analysis outputs
Applying the Spear and Hornberger method to 25 model variables produces an array of catch time-series over the historical period (figure 4.7). Of the 2000 simulations, only 41 (2.1%) have at least 80% of annual catches within the constraints corridor. Therefore, perturbing parameterised values by up to 50% effectively produces a different model, whereby catches may collapse by the late- 1970s or reach 40,000 tonnes after 35 years (figure 4.7)!
Regarding the second aim of this subchapter, whether catch falls within the corridor of constraints is most sensitive to uncertainties in the sediment effect on macrophyte growth and the eco-K estimate (table 4.5). The sensitivity of outputs to macrophyte growth reflects the double-edged nature of the process. As macrophytes provide refuge from fishing efforts, a strengthened sediment- macrophyte feedback reduces catch relative to the constraint corridor (figure 4.7A). Contrastingly, a weakened feedback provides limited refuge, enabling efforts to fulfil their potential (Bueno and Basurto, 2009), although overexploitation occurs if the feedback is weakened towards 50% of original. The highest frequency of behaviour-giving inputs corresponds to an error magnitude of 75%, suggesting that the feedback strength may be slightly overestimated. The sensitivity to eco-K is unsurprising given the one-at-a-time sensitivity, with positive eco-K errors generally overestimating catch relative to the constraint corridor and negative errors causing earlier than observed collapse.
Whether output dynamics match the constraint corridor is also statistically dependent on the number of fishers per boat (table 4.5). The highest frequency of behaviour-giving sets correspond to 7.5 fishers per boat (figure 4.8B), rather than 6 fishers per boat derived from CDA (2015) records (1999-2014). Less complete data sources also indicate fewer than 7.5 fishers per boat, with Noack et al. (2013) assuming 3.5 fishers per boat in their economic model and Kumar
135 Chapter 4: Model evaluation
Figure 4.7: The 2000 catch time-series resulting from randomly perturbing 25 model variables by ±50% of their parameterised values. Prop.inside – proportion of annual catches inside the constraint corridor. and Pattnaik (2012) observing 4-6 fishers per boat. Therefore, insufficient evidence exists to change the conversion between the stocks of fishers and boats, whilst consistency with the observed trends of boats, CPUB and per capita income supports current parameterisation (chapter 4.3.2).
Catch is also sensitive to LMC sediment rating, which drives the volume of Mahanadi sediment transported for a given streamflow, and WC’s runoff coefficient that determines the proportion of stormflow for a given monthly rainfall volume. These sensitivities are coherent with the perceptions of Chilika’s experts, with outlet sedimentation principally blamed for the historical catch collapse (Pattanaik, 2007, Iwasaki et al., 2009). Whilst the parameterisation of WC’s runoff coefficient produces the highest density of behaviour-giving outputs (figure 4.8C), the input-output distributions of LMC’s sediment rating suggests the volume of sediment transported per unit streamflow is reducible by up to ~20% (figure 4.8D). However, the modelled long-term sediment flux at Tikarapara and multidecadal catch mode align with observations (chapter 4.3.2), supporting the suitability of parameterisation.
136 Chapter 4: Model evaluation
Table 4.5: Outputs depicting the separation between the error distributions producing catch within the corridor of constraints and those causing catches outside. See figure 3.11 for the hydrological acronyms. Significance levels: ‘1’ – p < 0.01; ‘2’ – 0.01 ≤ p ≤ 0.1; ‘3’ – p > 0.1. K-S – Kolmogorov-Smirnov.
K-S Variable p-value Significance score Sediment effect on aquatic vegetation growth 0.255 0.005 1
Ecological carrying capacity 0.252 0.005 1
Fishers per boat 0.238 0.011 2
LMC sediment rate 0.238 0.011 2 WC Rc 0.191 0.070 2 Outlet closure 0.179 0.081 2 Dissolved oxygen effect on aquatic 0.167 0.155 3 vegetation growth
Salinity effect on aquatic vegetation growth 0.164 0.168 3
Fish natural death rate 0.163 0.173 3 Catch capacities 0.161 0.183 3 WC abstraction coefficient 0.150 0.254 3 Fecundity 0.145 0.289 3 WC evaporation coefficient 0.145 0.291 3 evaporation coefficient 0.125 0.467 3 Default survival rate 0.124 0.480 3 Dissolved oxygen 0.119 0.535 3 Mature fish per unit weight 0.113 0.620 3 LMC abstraction coefficient 0.110 0.630 3 Littoral sediment 0.110 0.635 3 Salinity 0.110 0.639 3
LMC Rc 0.106 0.684 3
Days fished 0.102 0.723 3
WC sediment rating 0.102 0.730 3
Traditional upgrade 0.097 0.778 3
Reproductive life 0.077 0.946 3
137 Chapter 4: Model evaluation
Figure 4.8: (A-D) density plots depicting the relative frequency of runs inside and outside the constraint corridor across different error magnitudes; (E) an example of a driver with insignificant difference between error magnitudes producing runs inside and outside the constraint corridor.
138 Chapter 4: Model evaluation
MCA identifies six processes with significant separation between behaviour-giving and non-giving uncertainty magnitudes. Arguably, variables not leading to significant behaviour changes are superfluous because their removal from the model might not affect catch. This argument aligns with the misconception that the representation of reality always improves as variables and processes are added (Oreskes, 2003). ‘Integronsters’ may emerge (Voinov and Shugart, 2013), whereby an overly complex model makes driver-response relationships unexplainable and the model useless for its intended purpose.
Various reasons exist to retain insensitive variables. First, whilst insignificantly affecting output behaviour over the historical range, salinity and dissolved oxygen are two key indicators of ecosystem health for Chilika’s decision-makers; for example, low salinity and dissolved oxygen may signify that processes that are more significant require immediate management, such as unsustainable sedimentation of the tidal outlet. Second, Chilika’s experts (Mohanty – personal communication; Pattnaik – personal communication) argue the system is vulnerable to growing motorised practices, so setting the proportion of motorboats and days fished constant arguably underrepresents the future influence of fishers (Nayak, 2014). Third, variables such as fecundity, survival and reproductive life average are fundamental to modelling fish populations with SDMs (Bueno and Basurto, 2009, Duer-Balkind et al., 2013). In all, finding model outputs are indifferent to fishery parameters suggests wide confidence bounds, providing assurances that parameterised values are appropriate to model Chilika’s multidecadal dynamics.
This section has considered the interactions of 25 uncertain parameters to isolate the critical determinants of historical catch; it is now important to identify permissible error magnitudes inhibiting the replication of the reference mode.
4.3.6.4 Isolating critically sensitive drivers
Only the six most sensitive variables undergo this second stage of MCA (table 4.5). To understand permissible error ranges, the experiment splits into four groups of 500 simulations, simulating ±5%, ±10%, ±20% and ±30% of parameterised values. Time-series of salinity, macrophyte coverage and CPUB provide wider sensitivity assessment alongside catch, representing the key ecological and socioeconomic indicators of Chilika’s sustainability.
All simulations within the ±5% and ±10% error range undergo catch decline during the 1990s, before recovering to at least 5,000 tonnes post-restoration (figure
139 Chapter 4: Model evaluation
Figure 4.9: Graphical outputs from sensitivity analysis where the six critically sensitive variables are simulated under four error ranges. Observations (black) are plotted against model generated time-series of four social- ecological processes: (A) salinity; (B) macrophyte coverage; (C) catch (99% confidence intervals) (D) catch per unit boat (99% confidence intervals).
4.9C). Moreover, the maximum magnitude of recovery does not exceed the 99% confidence interval of observed catch, which peaks around 14,800 tonnes in 2006. The pattern is similar for CPUB, with ±10% errors generally capturing the periods of stability pre-1990, collapse, recovery post-2000 and subsequent stability (figure 4.9D).
The degree of similarity between model outputs and observations decreases as error ranges relax. The second subset (±20%) produces catches with shallow declines by 2000, due to a more resilient fish population under positive eco-K errors. Oppositely, underestimating eco-K and comparatively few fishers per boat causes earlier than observed collapse, whilst also suppressing system recovery post-2000. Lastly, ±30% errors cause the earliest fishery collapses, reaching 0 tonnes around 7 years before the 99% confidence interval of observed catch. Alternatively, a combination of positive eco-K errors and underestimated fishers
140 Chapter 4: Model evaluation per boat can keep catch above 10,000 tonnes, before outlet opening further boosts catches up to 20,000 tonnes (~1.4 times MSY).
Regarding the drivers of catch, discontinuous salinity and macrophyte trends inhibit the use of locally weighted least squares regression (figure 4.9A-B). Unaffected by initial uncertainties in ecological carrying capacity and fishers per boat, the ecohydrological time-series tend to diverge over time until 2000, reflecting the cumulative effects of different hydrological influxes, outlet closure and sediment trapping by macrophytes. All simulations within ±10% error bounds follow the historical mode of linear salinity decline by 2000, rapid recovery post- 2000 and shallow decline until 2009 as the outlet begins to reclose. Both ±20% and ±30% errors may produce salinities of 0 ppt during the late-1990s, which is ~4 ppt lower than the freshest observation (figure 4.9A). Within the ±10% error range, the model captures nonlinear macrophyte growth pre-2000 and decline post-2000 (figure 4.9B). The lagoon is completely covered by macrophytes after 25 years if the macrophyte-sediment feedback, WC R , LMC sediment rating and c outlet closure are overestimated by up to 30%. Contrastingly, parameter underestimation causes coverage roughly four times less than observed by the late-1990s, which later disappears under brackish conditions post outlet opening.
Building on previous tests, this subchapter provides two key lessons from multivariable sensitivity analysis. First, six drivers are found to significantly influence whether model processes reproduce historical catch dynamics (table 4.5). The relative importance of these drivers is consistent with qualitative knowledge, with Chilika’s resource availability historically coupled to fish migration/mobility via the tidal outlet (Rajawat et al., 2007, Iwasaki and Shaw, 2009, Mohanty et al., 2015). Therefore, in combination with behavioural reproduction (chapter 4.3.2), it is argued that the model captures the relative magnitudes and sources of drivers underpinning problematic behaviours. Second, it is argued that the six critically sensitive parameters are robust up to ±10% errors, as conclusions about system outcomes are consistent across the range of uncertainties (Sharp, 1976, Saltelli, 2000b). Robustness extends to the trends of salinity, macrophyte coverage and CPUB – which all form dimensions of the future pathways towards different plausible futures.
141 Chapter 4: Model evaluation
4.4 Chapter conclusion
The first aim of this research is to conceptualise and parameterise a model capable of simulating plausible future processes, states and scenarios for the Chilika lagoon fishery (chapter 2.6). SDMs have previously been left unevaluated, either due to data constraints or time demands (see Leal Neto et al., 2006, Zhu et al., 2015), whilst conventional analysis only involves trend matching (Hossain et al., 2017) or one-at-a-time sensitivity analysis (Sušnik et al., 2012, Wei et al., 2012). Therefore, the nine tests conducted here may seem excessive, but each seeks to build confidence in the model’s ability to simulate futures that are plausible given knowledge of contemporary social-ecological interactions, feedbacks and deep uncertainty about the system’s sustainability.
To this end, the structural tests show that (i) flows, stocks and converters have realistic units of measurement and connected variables are dimensionally consistent; (ii) qualitatively uncertain datasets can be screened for behavioural assessment, and (iii) structural integrity persists under hypothesised extreme conditions. In turn, behavioural assessment finds that: (iv) dynamic extreme behaviours match the qualitative hypotheses made during structural assessment, arguing that the model produces expected outputs once human-natural processes are forced beyond their normal operating range; (v) the model reproduces observed growth, decline and inflection for various processes underlying system sustainability, and (vi) catch is not dependent on the parameterised time-step.
Sensitivity analysis does not assess whether the model is right or wrong, but uncovers unexpected dynamics and uncertainties affecting model usefulness (Saltelli, 2000b). Both one-at-a-time analysis and MCA argue that model sensitivities match qualitative and quantitative knowledge of the system. Consistent conclusions and the replication of observed catch declines when varying parameters up to ±10% suggests that uncertain parameters do not sit within overly sensitive regions of parameter space. Regarding model usefulness, sensitivities show that the model is appropriately responsive to short-term socioeconomic changes and creeping environmental trends.
Sensitivity analysis aids model transparency by explaining the causes of variability (Saltelli, 2000b). Through MCA and generalised sensitivity analysis (Spear and Hornberger, 1980), the most significant drivers of multidecadal catch were found to be: (i) the feedback between lagoon sediment and macrophyte coverage, (ii) eco-K of the fish population, (iii) the number of fishers per boat, (iv) Mahanadi
142 Chapter 4: Model evaluation sediment generation (v) WC R and (vi) the degree of tidal outlet closure. These c processes are consistent with the knowledge that the number of boats is a major component of fishing effort (FAO, 2002); aquatic weeds modify fish catchability by blocking fishing grounds (Bueno and Basurto, 2009, Iwasaki et al., 2009) and outlet sedimentation is considered the leading cause of the 1990s collapse through the constraint of anadromous and catadromous fish migration (Pattanaik, 2007, Rajawat et al., 2007). Therefore, this model captures the sources and magnitudes of the competing grindstones of environmental degradation and socioeconomic development (Pattanaik, 2007).
Model insensitivities are just as insightful as sensitivities (Saltelli, 2000a). Catch insensitivity to uncertainties in evaporation and runoff abstraction provides confidence that uncertainties at lower levels do not propagate disproportionately through model structure. With the exception of survival rate, which varies with population density, fish traits are assumed constant. The ranking of multivariable sensitivities (table 4.5) argues that a percentage change in fishing effort exerts a greater influence on catch than an equivalent change in fecundity, reproductive life length and other life traits. Therefore, finding that slightly different fish trait estimations produce problematic behaviours in the absence of environmental or anthropogenic stimuli would have degraded confidence in the model’s ability to simulate realistic social-ecological interactions and future scenarios.
It is important to note the various limitations of the MCA conducted here. First, the constraint corridor assumes that catch observations are accurate. Dujovny (2009) and Nayak and Berkes (2010) question this assumption, citing fishers arguing that the reported figures are overestimations. Moreover, sampling each landing centre six times per month undoubtedly introduces uncertainties (Kumar and Pattnaik, 2012). Therefore, this study assumes catches are consistent between samples, with the replication of the long-term trend taking precedence over annual observations. Second, the initial screening process may have omitted sensitive variables. However, variables undergoing further assessment were those informed by data considered to be qualitatively uncertain (appendix C.1), plus the processes considered fundamental to fishery dynamics. Moreover, the sensitivities of untested variables like diesel price and Naraj flow proportioning aggregate into evaluated processes, such as the number of fishers upgrading and the LMC rainfall-runoff coefficient, respectively.
In synthesis, the model has undergone a variety of tests to gain insight into the variability and causes of output dynamics. The replication of multidecadal trends
143 Chapter 4: Model evaluation signifies that the model produces the ‘right’ quantitative behaviours (Barlas and Kanar, 2000, Sterman, 2000); the dimensionally robust structure upholding under extreme conditions and uncertainties argues that the behaviours are produced for the ‘right’ reasons (Barlas, 1989). The drivers found to be key determinants of fishery persistence are consistent with regional literature, mental models of regional experts and wider understanding of resource-limited systems. Therefore, the relative importance of each process is consistent with reality, supporting the fitness of the model to simulate Chilika’s plausible futures of creeping natural stresses and human activities.
144 Chapter 5: Plausible futures
Chapter 5: Plausible future scenarios and normative definitions
5.1 Chapter introduction
Model rationale, construction and evaluation are detailed to here. Altogether, the model produces dynamics consistent with observations, plus sensitivities coherent with qualitative and quantitative system understanding. However, an evaluated model does not directly inform decision-makers as to the future safety of their system. Therefore, following the framework of Ford (2010), the model can now explore the impacts of policies and future driver trajectories on problematic behaviours of interest.
Future scenarios are required to project system dynamics. The definition of a scenario varies from qualitative storylines, to strings of numbers ranging from ‘visionary thinking to minor adjustments to “business-as-usual”’ (Swart et al., 2004, p.139). This study uses two types of scenarios: (i) exploratory scenarios defining the future trajectories of five external variables (chapter 5.2); (ii) four governance scenarios exploring how the social-ecological sustainability of internal interactions relate to decisions makeable today (chapter 5.3). Outcome dynamics are then classified by pre-defined conditions of ‘safe and just’, ‘cautionary’ and ‘dangerous’ futures (chapter 5.4), based on system sustainability relative to today’s norms and values (Dearing et al., 2014, Verburg et al., 2016).
Consistent with backcasting, driver scenarios are assumed to have uniform likelihoods and are classified by their normative future by mid-century. In contrast to backcasting, only a limited number of feasible policies are tested, assumed to start today rather than any point between now and 2060. Therefore, the approaches described below form a hybrid between forecasting and backcasting techniques (chapter 2.3.3), where internal dynamics are projected forward under a spectrum of external trajectories to uncover SJOS dimensions and associated complexities.
5.2 Exploratory external scenarios
Three main methods exist to generate future dynamics in SDMs. For closed systems, dynamics derive purely from internal structures. For open systems,
145 Chapter 5: Plausible futures fishery SDMs have traditionally favoured constant external dynamics (Bueno and Basurto, 2009, Duer-Balkind et al., 2013, Martins et al., 2015). Akin to turning a key in an engine’s ignition, external variables offer leverage points to generate dynamics, which is particularly appropriate if externals contribute towards problematic behaviours. This subchapter represents stage B of figure 2.4, describing the identification of maximum plausible external trajectories, their operation in STELLA and any resulting bias.
Chilika’s governors are unable to influence rainfall today or in future. Likewise, atmospheric and near-surface air temperatures externally influence aquatic habitat conditions and hydrological flows. Scenarios of future warming across the region are relatively consistent. Under the steepest greenhouse gas emission scenario (IPCC RCP8.5, used hereafter unless stated), IPCC (2013) predict up to 3°C annual mean surface air temperature change across India’s east coast by 2046-2065 relative to 1986-2005 (baseline hereafter). This was revised up from the previous IPCC assessment (Meehl et al., 2007), which predicted 1.5-2.5°C warming by 2046-2065 relative to 1980-1999. Using an ensemble of 10 IPCC climate models, Kumar et al. (2012) predict Odisha’s annual average temperature will be 2.9-3.1°C warmer in the 2050s than the 1970s; Mitra et al. (2012) predict 2.9°C for the same period.
Projections of future rainfall are relatively inconsistent, owing to the complexities of topography, landcover and regional circulation systems, including Odisha’s southwest monsoon (Houghton, 2009). Synthesising these factors globally, Allen and Ingram (2002) find climate models generally produce precipitation increases of 3% per degree Celsius warming. This rule of thumb gives ~9% precipitation increase under RCP8.5 across Odisha by 2060. By 2081-2100, IPCC (2013) predict increases of 20% and 30% for June-July-August (JJA) and September-October- November (SON) rainfalls across the region, respectively. Rainfall changes for 2046-2065 are insignificant, with the exception of a 20% increase in SON. From ensemble simulations using the Hadley Centre’s PRECIS regional model, Kumar et al. (2012) project 15-20% rainfall increases across India’s coastal regions by 2060, reaching 30% by 2080. Mitra et al. (2012) project increases between 5-20%, depending on whether the 1970s (lower) or 1901-2000 (higher) is the baseline.
Whilst rainfall and temperature drive ecohydrological conditions, future fishing efforts and value are driven by fish price, diesel price and the human population growth rate. Unlike the climate projections, regional fish and diesel price scenarios are lacking. Therefore, historical dynamics inform the socioeconomic
146 Chapter 5: Plausible futures
Table 5.1: The plausible trajectory ranges of the five external drivers used to drive the model’s internal dynamics into future; current values derive from the baseline period 1986-2005.
Driver Current value (±SD) Plausible trajectory range
Min Max
Annual average 1259.3 (±205.3) 0 +20% by 2065 rainfall mm/year
Annual average 25.3 (±1.2) °C 0 +3°C by 2065 temperature
Fish price 133.6 INR/kg 0 30 INR/kg/year
Diesel price 55.9 INR/litre 0 7 INR/litre/year
Birth rate minus 14.0 0 2.20%/year death rate (Malthusian’s ‘r’) trajectories, varying between no change and the highest observed interannual change over the data range.
Regional fish price is considered inelastic (Kadekodi and Nayampalli, 2005) – historically increasing regardless of production and local demand. Fish price increased from 12 to 28 INR/kg between 1980 and 1999. The rate of change increased four-fold in the subsequent nine years, reaching 65 INR/kg by 2009 and 140 INR/kg by 2014 (CDA, 2016). The maximum annual fish price increase from 1979-2014 was 30 INR (2000-2001and 2012-2013) (CDA, 2016). Affecting non-traditional costs, the non-traditional livelihood carrying capacity and the cost to upgrade fishing practices, the highest interannual increase in diesel price from 1989-2014 was 7.0 INR (2012-2013) (Verma and Fernandez, 2010). Whilst it is unlikely that these steepest trajectories persist, their simulation aims to cover the dimensions of the SJOSs by exploring internal resilience to the spectrum of external evolutions.
Without introducing fertility awareness programmes, fisher birth and death rates are external to governors. Whilst increasing ‘r’ contradicts Odisha’s wider multidecadal birth and death rate convergence (PCD, 2011), this model aims to produce a broad range of population trajectories between a stabilising population
147 Chapter 5: Plausible futures
Figure 5.1: The five external funnels generating dynamics until 2060: (A) mean annual near- surface air temperature across the Lower Mahanadi and Western catchments; (B) mean annual rainfall across the same areas; (C) fish price; (D) diesel price and (E)
net of birth and death rates (‘r’). and a doubled population from present, in order to probe the edges of the SJOSs rather than predict most likely outcomes. Importantly, livelihood carrying capacities may also influence the populations during fishery declines.
STELLA operationalises these trajectories by setting the variables as sensitivity parameters in the sensitivity analysis tool. Per simulation, STELLA randomly samples between minimum and maximum trajectories (table 5.1), bound by uniform probability distribution functions assuming each trajectory is equally likely under deep uncertainty (Kwakkel et al., 2010). The random number seeds
148 Chapter 5: Plausible futures are the same for the different governance scenarios to assess resilience under the same external trajectories.
Deriving monthly (푡) values of diesel and fish price involves adding the random annual increase (푥) for a given simulation (𝑖) to the previous value:
퐴푛푛푢푎푙푃푟𝑖푐푒퐼푛푐푟푒푎푠푒푖 = 푅퐴푁퐷푂푀(푥푖) (Eq 5.1a)
퐴푛푛푢푎푙푃푟푖푐푒퐼푛푐푟푒푎푠푒 푃푟𝑖푐푒 = 푃푟𝑖푐푒 + 푖 (Eq 5.1b) 푡,푖 푡−1,푖 12
Future ‘r’ follows the same principle, except changes are percentage and multiplicative rather than absolute and additive. Nonstationary rainfall and temperatures are driven by:
퐹𝑖푛푎푙퐶푙𝑖푚푎푡푒퐶ℎ푎푛푔푒 = 푅퐴푁퐷푂푀(푦푖) (Eq 5.2a)
푛 퐶푙𝑖푚푎푡푒퐶표푛푑𝑖푡𝑖표푛 = 푉푎푙푢푒퐹푟표푚퐸퐶퐷퐹 + 퐹𝑖푛푎푙퐶푙𝑖푚푎푡푒퐶ℎ푎푛푔푒 × (Eq 5.2b) 푖,푡 푡 푖 푁
Where 푦푖 is the sampled rainfall/temperature trajectory by 2060 (table 5.1), 푛 is the number of months passed since 2010 and N is the total number of months between January 2010 and December 2060. For instance, if the output from the temperature empirical cumulative distribution function (chapter 3.4.1.6) for March 2030 is 26°C, under a scenario of 2°C warming, the climate change
255 influenced temperature equals 26.8°C (26 + 2 × ). This method increases 612 climatic means and variances over time (Houghton et al., 2001).
Per governance scenario, 1000 trajectories per external driver are simulated. It is important to note that the resulting scenario funnels have various limitations. Climate trajectories could be linked given that atmospheric warming generally increases atmospheric moisture capacity. However, independent trajectories assess system sensitivity to different rates of rainfall and temperature change, equally insightful considering uncertainty exists around the eventual magnitude of rainfall change. Furthermore, modelling monthly climate dynamics excludes sub-monthly events like tropical cyclones and flash rainfalls. Fish price and diesel price are also unlinked to explore resilience to the spectrum of incomes and costs. Moreover, one Monte Carlo sample per driver per simulation produces linear socioeconomic trajectories. Alternatively, the trajectories could be sampled multiple times per simulation (e.g. annually) to generate spikes (e.g. fuel price hikes) and flatlines (e.g. fish price stagnation). However, this strategy would increase the number of Monte Carlo samples from five to 255 per simulation.
149 Chapter 5: Plausible futures
Figure 5.2: Histograms of the five external trajectories, plotted in the same order as figure 5.1. The socioeconomic trajectories are described by annual gradients from 2015-2060; the climate variables are described by their 2050-2060 annual average relative to the baseline.
It is important to understand the external trajectory distributions and combinations to contextualise the outcome distributions. For example, does the above methodology produce bias conditioning the outcomes to a particular future
150 Chapter 5: Plausible futures
(e.g. overexploitation)? Despite localised peaks and troughs, the future trajectories of temperature change, fish price, diesel price and ‘r’ are uniformly distributed between zero and maximum plausible change (figure 5.2). Consequently, the model simulates an array of socioeconomic scenarios (appendix C.3), with combinations of steep fish and diesel price as likely as any other trajectory. In contrast, rainfall trajectories have a slight positive skew: the mean rainfall trajectory is +4% from 2050-2060, with 95% of trajectories falling between -5% and +15% change. The rarity of steep positive change derives from the positively skewed historical rainfalls (chapter 3.4.1.6), particularly during dry post- and pre-monsoon months. Nevertheless, ~30 simulations per governance scenario explore the effects of ≥ 15% rainfall increase, distributed across the temperature trajectories (appendix C.3).
This subchapter has described the methods used to frame and operationalise the exploratory external scenarios. Rather than constant external variables conventionally modelled by fishery SDMs, this study parameterises nonstationary conditions to force internal dynamics into unchartered regions of parameter space. Moreover, this nonstationary approach aims to assess interacting thresholds, such as whether socioeconomic SJOSs shift under climate change.
5.3 Internal governance scenarios
System governance provides a method to manipulate future sustainability and the location of thresholds (Hughes et al., 2017). Numerous options were possible prior to the fieldtrip, including doing nothing; business-as-usual outlet maintenance (Kumar and Pattnaik, 2012); restrictions on fishing area and/or time; licence limits; motorboat bans; MSY quotas and alternative livelihoods. Regional literature and Chilika’s experts helped identify feasible policies, here described in order of increasing decommonisation whereby CPR rights are traded for command and control (Nayak and Berkes, 2011). In addition, the causal mechanisms are hypothesised to provide context and avoid black-boxing governance (Biesbroek et al., 2017). The four governance scenarios modelled are:
1. No governance (NO): less regulatory than business-as-usual, doing nothing is the default model setting after outlet opening in 2000. Modelling NO assesses whether the previous outlet opening has created a fishery resilient to plausible stresses until 2060. NO governance may be forced if state level decision- makers stop funding regional conservation activities:
151 Chapter 5: Plausible futures
Operationalisation: no further action/variables required.
2. Periodic outlet maintenance (OM): although a new outlet has not been cut since 2000, the CDA has published plans to cut a new tidal outlet “approximately once in 10 years” to maintain resource mobility to and from the Bay of Bengal (Kumar and Pattnaik, 2012, p.115). Therefore, starting in January 2020, OM pulses sediment from the modelled stock every 120 months. The magnitude of the pulse stabilises sedimentation at a pre-1980 levels to maintain the brackish water environment, meaning sediment removal is positively associated with sediment accumulation between openings. Operationalisation: an additional variable (‘outlet maintenance’) removes sediment every 120 months: 퐈퐅 퐓퐈퐌퐄 = 576 퐎퐑 696 퐎퐑 816 퐎퐑 936 퐎퐑 1056
퐓퐇퐄퐍 ChilikaSedimentStock푡 × 0.3 퐄퐋퐒퐄 0 (Eq 5.3)
3. OM plus a permanent fishing ban across an area equivalent to the tidal outlet (OB): expanding upon the six months ban of OMFRA 1988, the permanent restriction here reflects the desire to stop all fishing within the tidal outlet, communicated during both the key informant and stakeholder interviews. The aspatial model assumes fish abundance and fishing efforts distribute uniformly across the lagoon, meaning OB may underestimate the effects of limiting extraction from the main migratory highway. The ban is implemented in January 2015 (Eq 5.4a), but its regulatory effect rises linearly from zero to full cooperation in 2025 (Eq 5.4b) (Mohanty – personal communication). Operationalisation: each fleet’s (𝑖) fishing effort (chapter 3.4.3) is limited once the ban is active (equation 5.4a) and adhered to (equation 5.4b):
퐵푎푛퐴푐푡𝑖푣푒 = 푰푭 푻푰푴푬 < 516 푻푯푬푵 0 푬푳푺푬 1 (Eq 5.4a)
퐵푎푛퐴푑ℎ푒푟푒푛푐푒 = 푮푹푨푷푯푰푪푨푳 푭푼푵푪푻푰푶푵
1 [(0,0) − (1056,1)], (515,0), (516, ) , (635,1) (Eq 5.4b) 120
퐵푎푛퐸푓푓푒푐푡푖 = 퐹𝑖푠ℎ𝑖푛푔퐸푓푓표푟푡푖 ×
푂푢푡푙푒푡퐴푟푒푎 (1 − 퐵푎푛퐴푐푡𝑖푣푒 × 퐵푎푛퐴푑ℎ푒푟푒푛푐푒 × ) (Eq 5.4c) 퐿푎푔표표푛퐴푟푒푎 Where 푂푢푡푙푒푡퐴푟푒푎 equals 70 km2 and 퐿푎푔표표푛퐴푟푒푎 equals 1000 km2 (Srichandan
et al., 2015).
152 Chapter 5: Plausible futures
4. OB plus alternative livelihoods (OL): in addition to OB, alternative livelihoods aim to reduce fishing pressure and human dependence on the fish stock. The CDA, Odisha’s state government and ICZMP currently divert fishers into coir production, agriculture and poultry farming (Kumar and Pattnaik, 2012). However, no detail exists about the number and uptake of alternative livelihoods. Therefore, fishers are removed from their respective populations at an exploratory rate of 1 in 2000 fishers per fleet per month. With ~35,000 fishers in 2015, this rate equates to ~420 alternative livelihoods over the year, split roughly 60:40 between traditional and non-traditional fishers. Fishers with alternative livelihoods disappear from the model, thus assuming an absence of part-time fishing. Operationalisation: From January 2015, an outflow from each fisher stock (𝑖) removes 1 in 2000 fishers/month:
퐴푙푡푒푟푛푎푡𝑖푣푒퐿𝑖푣푒푙𝑖ℎ표표푑푠푖,푡 = 푰푭 푻푰푴푬 < 516 푻푯푬푵 0
퐹푖푠ℎ푒푟푃표푝푢푙푎푡푖표푛 푬푳푺푬 푡,푖 (Eq 5.5) 2000 Importantly, these policies are static as decision-makers are unresponsive to changes in conditions after policy implementation. Alternatively, adaptive governance could open the outlet or implement bans once catch falls beneath
MSY, with robustness analysis finding optimal policy magnitudes and frequencies (Kasprzyk et al., 2013, Kwakkel et al., 2016). However, Chilika’s decision-makers are not seeking to optimise outcomes, instead prioritising wise use and maintenance (Kumar et al., 2011). Moreover, SJOSs of decisions makeable today is consistent with the original SJOS concept (Dearing et al., 2014). Third, simulating relatively few governance scenarios improves the tractability of their effects whilst building foundations to consider additional policies in future.
5.4 The safety of plausible futures
The previous subchapter operationalises spectrums of plausible futures to assess the resilience of Chilika’s fishery to nonstationary natural and human drivers. This subchapter defines what it means to be resilient and/or safe and just, based on present day environmental baselines and socioeconomic norms (Dearing et al., 2014). The SJOSs definitions established by Dearing et al. (2014) have different means of operationalisation. Linear thresholds represent conditions that once crossed produce normatively alternative states. Common examples include the
153 Chapter 5: Plausible futures
World Bank’s international poverty line (Jolliffe and Prydz, 2016) and the FAO’s minimum daily energy requirement (FAO, 2008). Linear thresholds also inform a number of social foundations for humanity (Raworth, 2012), including food security, income and equity. Like the international poverty line, linear thresholds are iterative and show outcome proximity to the threshold. However, normative targets may not be attainable for all outcomes, as evidenced by absent foundations for social voice, resilience and jobs (Raworth, 2012).
Dearing et al. (2014) base safe and cautionary thresholds of sediment and water regulation on nonlinear dynamics; Hossain et al. (2017) flag dangerous social- ecological conditions for the Ganges-Brahmaputra-Meghna delta once rice production falls below its historical envelope of variability. Whilst indicating abrupt and surprising trends, the envelope of variability approach has limitations. First, nonlinear changes or novel outcomes are not always unsafe; for instance, abrupt catch decline might reflect a sudden shift in dependence away from fishing rather than resource scarcity. Second, an envelope parameterised with observations may include recent regime shifts (e.g. Chilika in the 1990s), implying that resource collapse is part of normal functioning. Third, whilst models can explore system configurations leading to nonlinearities, these may be missing for systems without reinforcing feedbacks. Therefore, only basing safety on the presence/absence of nonlinear thresholds implies long-term drifts are always safe.
Ultimately, SJOSs are system and outcome dependent. Fisheries have multiple outcomes which should ideally remain safe and just, including catch, resource availability, incomes and broader social foundations (Raworth, 2012). However, rather than defining SJOSs individually, this study adopts a top-down classification to seek driver trajectories leading to social-ecologically desirable catch (Bai et al., 2016, Verburg et al., 2016), in line with the view of Brucet et al. (2013) that future ecological targets should fundamentally secure the long-term capacity to supply ecosystem services. The sustainable future is normative in the sense that it is deemed to be a good Anthropocene future (Berkhout, 2014) by system experts; however, achieving 100 tonnes less than MSY is unlikely to have substantial quantitative or qualitative social impacts. Moreover, the linear definitions aim to provide an initial proof of concept, so that nonlinear boundaries may be sought in future.
To this end, catch is the principal service generating at least 70% of Chilika’s economic value (Ray and Garada, 2017), employing 35,000 fishers and 200,000
154 Chapter 5: Plausible futures
Figure 5.3: Schematic representation of plausible catches resulting from the spectrums of internal and external dynamics. The challenge here is to understand which futures are safe and just, and in turn, discover the causal pathways and interactions to guide system governance. secondary dependants (Iwasaki and Shaw, 2010b), whilst acting as the main indicator of resource health under the assumption that catch and abundance are positively related. Therefore, the following definitions classify the 1000 future catches per governance scenario (figure 5.3), representing stage C of figure 2.4:
1. Safe and just: from 2050-2060, annual catch averages between MSY lower (11,400 tonnes) and MSY upper (13,900 tonnes) estimations (CIFRI, 2006, Kumar and Pattnaik, 2012).
MSY is conceptualised as the maximum catch that is indefinitely extractable from a fish stock without degrading its long-term renewal (Maunder, 2008). Catch above MSY is arguably socially just by allowing short-term profits, but ecologically unsafe in the long-term. Oppositely, sub-MSY catch is ecologically safe, but comparatively unjust by limiting catch below the maximum sustainable rate.
An alternative definition might consider the fishery safe only if all catches equal MSY. However, averaging MSY from 2050 onwards acknowledges that inherent
155 Chapter 5: Plausible futures interannual variability is not necessarily unsustainable (Andrew et al., 2007) and its suppression may constrain multidecadal resilience (Carpenter et al., 2015b). The 2050-2060 period aligns with goal-seeking SDMs, aiming to balance mid- term governance, mid-century climate projections and uncertainties increasing over time from the potential emergence of processes like nonlinear climate change or shifts from today’s social norms. To this end, the goal represents a checkpoint towards longer-term sustainability, recognising that safe and just outcomes are relatively likely to prepare the system for challenges beyond 2060.
2. Dangerous: in any year, the total fisher population exceeds the total livelihood carrying capacity.
The livelihood carrying capacity represents a hypothetical population whose fishing costs are covered by the fishery value (chapter 3.4.4). Therefore, dangerous futures imply that the system is overcapacity, whereby the principal ecosystem service no longer supports the average livelihood cost. Consequently, a number of fishers proportional to the magnitude of overshoot must quit fishing as the system reorganises towards a sustainable population.
Normative dangerous futures have two key assumptions. First, the system is dangerous at an aggregated level, assuming fishery value and costs split evenly. This aggregation is coherent with population data that does not discriminate between traditional and non-traditional fishers (Noack et al., 2013, CDA, 2015). Consequently, the livelihood carrying capacity declines as non-traditional fishers increase as a proportion of the total population due to the heightened costs of motorised fishing. Experimental simulations flagging dangerous futures once either population exceeded its carrying capacity found that traditional fishers upgrading to motorboats caused temporary overcapacity of non-traditional fishing; these futures would become safe at the next time-step in response to heightened non-traditional catch value. Second, a dangerous future is dangerous across its simulation length. Therefore, any aggregated overcapacity is considered undesirable, irrespective of when it occurs and the speed and magnitude of reorganisation.
3. Cautionary: the system is neither safe and just nor dangerous.
Cautionary futures correspond to two conditions that are neither safe nor dangerous. First, sub-MSY catch is cautionary if the total fisher population is below the livelihood carrying capacity. In other words, catch is socially unjust
156 Chapter 5: Plausible futures relative to MSY, but production still covers the average livelihood cost. Second, representing an upper ceiling to discourage unsustainable fishing efforts, averaging catch above MSY from 2050 is also considered cautionary.
5.5 Chapter conclusion
The first aim of this thesis is to model the future processes, states and scenarios of the Chilika lagoon fishery. Through various structural and behavioural assessments, chapter 4 argues that the model captures the magnitudes, relative contributions and resulting dynamics of key social and ecological drivers. The techniques developed in chapter 5 project these processes into future to identify pathways towards safe and just futures (chapter 5.4), which are partly shaped by how decision-makers (chapter 5.3) influence internal resilience to the plausible external trajectories (chapter 5.2).
The exploratory scenarios, governance scenarios and normative futures aim to contribute towards a number of research gaps in sustainability science and resilience thinking. First, the normative futures provide a ‘good Anthropocene’ target for decision-makers to aim for (Andrew et al., 2007, Berkhout, 2014, Bai et al., 2016), representing maximum human utility before multidecadal resource persistence is affected. Critically, achieving safe and just futures here does not rule out steering towards different trajectories in future (Hughes et al., 2017, Sterk et al., 2017), at least relative to the inflexibility of ecological degradation and livelihood losses under normatively dangerous futures. Second, in line with social-ecological resilience, coupling exploratory futures with normative targets investigates the maximum permissible stress before desirable functions are lost (e.g. self-organisation towards safe and just futures) (Folke, 2006). Third, modelling five external trajectories builds on conventional one-at-a-time resilience assessments to define resilience across multiple and interacting pathways (Scheffer et al., 2015, Sterk et al., 2017).
The next chapter operationalises the above principles to investigate the future sustainability of the Chilika lagoon under alternative governance approaches, as judged by the interacting pathways and thresholds delineating safe and just futures, wider social-ecological outcomes, trade-offs and nonlinearities influencing the usefulness of steering SESs under the SJOS concept.
157
Chapter 7: Results
Chapter 6: Model outputs – future outcomes, safe and just operating spaces, and social- ecological complexities
6.1 Chapter introduction and structure
To provide a holistic assessment of Chilika’s sustainability, this chapter conducts a series of exploratory analyses with ramifications for the governance of Chilika, the SJOSs concept, and wider notions of social-ecological sustainability and resilience. Consequently, it is worth mapping the structure of this chapter to aid its narrative. First, by offering a metric to assess the quality of future scenarios and states, catch time-series synthesise the resilience of system functions under different governance approaches (chapter 6.2). Analysis then switches to the causal pathways, assessing driver-response relationships under NO to understand whether Chilika remains susceptible to historical causes of collapse (chapter 6.3). The focus then shifts to governance scenarios with SJOSs, with probabilistic resilience analysis exploring system attraction to undesirable futures (Peterson, 2002) (chapter 6.4).
Chapter 6.5 analyses the classical system dynamics forming pathways towards normative futures (Verburg et al., 2016), including periods of linear growth, nonlinearities and collapse. However, time-series alone do not fully communicate the resilient interactions underpinning SJOSs (Scheffer et al., 2015, Hossain et al., 2017). To this end, SJOSs synthesise the relationships between driver trajectories and outcome states, helping to depict permissible driver limits (chapter 6.6) and interactions (chapter 6.6.5), including how climate change influences resilience to local human stressors (Scheffer et al., 2015). Chapter 6.7 furthers these principles by designing “core SJOSs” to reduce system precariousness to the identified thresholds (Walker et al., 2004), facilitate tiered governance (Biggs et al., 2015b) and identify the most critical drivers of system safety.
Social-ecological functions extend beyond supporting livelihoods at an aggregated level (Cumming and Collier, 2005, Costanza et al., 2011), whilst sustainability has broader dimensions beyond the persistence of resource productivity (Kates et al., 2001). Therefore, chapter 6.8 helps evaluate the use of normative states through wider aspects of ecosystem health, human livelihoods
159 Chapter 6: Results and social-ecological trade-offs omitted from the SJOS definition. Chapter 6.9 then compares the system’s current trajectory to the identified SJOSs, evaluating whether threshold transgression is a genuine concern or if trends are set to remain sustainable.
The final two subchapters focus on the nonlinear concepts of SJOSs. Nonlinear driver-response relationships suggests vulnerability to abrupt and potentially irreversible degradation (Scheffer et al., 2001), forming deeper perils beyond livelihood losses (Green et al., 2017). To this end, simulating until 2100 explores the time-scales, reversal forces and presence of hysteresis when recovering from overexploitation (chapter 6.10). In line with type-II and type-III dynamics (Dearing et al., 2014), a novel time-series metric is then developed to assess the relationship between the predictability of historical envelope exit and future outcome conditions. Such analysis provides additional future guardrails, and evaluates the use of linear and nonlinear SJOS definitions (chapter 6.11).
6.2 Plausible future catch
This section presents the 1000 future catch time-series per governance scenario. Considering all outcomes from each governance scenario (table 6.1), managed decommonisation generally increases annual average catch from 2015-2060, which is most noticeable in the normative future space from 2050-2060. Moreover, managed decommonisation limits catch variability under the spectrum of driver trajectories, thus creating relatively stable social-ecological interactions.
In turn, governance has three headline effects on the persistence of Chilika’s fish catch (figure 6.1). In general, simulations removing CPR rights (i.e. OL governance) (i) are more likely to reach the normative safe and just space, and are less likely to reach the normative dangerous space, (ii) undergo shallower and later collapses from desirable outcomes, and (iii) exhibit safe and just catch pathways with relatively limited variance between now and 2050 (appendix D.2). Therefore, managed decommonisation enhances system persistence, resistance and robustness to plausible external trajectories.
Regarding scenario specific dynamics, safe and just futures are absent under do nothing governance (figure 6.1A), with the disappearance of all 1000 simulations suggesting that unmanaged sedimentation effectively closes the tidal outlet to Chilika’s migratory fish, as occurred during the 1990s (Rajawat et al., 2007, Das and Jena, 2008). In support of this historical legacy, NO catch takes an average of
160 Chapter 7: Results
Table 6.1: Summary statistics of the annual catch time-series per governance scenario for the period 2015-2060, unless otherwise stated. See chapter 5.3 for governance scenarios.
Catch statistics (tonnes)
Mean Governance Mean Standard average Maximum Range scenario average deviation (2050- 2060)
NO 4,500 5,400 15,000 15,000 0
OM 11,200 4,000 17,800 17,700 5,800
OB 12,000 3,300 17,700 17,400 7,800
OL 13,400 1,400 17,600 13,800 12,700
6.5 years to fall from 8,000 tonnes/year to below 2,000 tonnes/year, mirroring the rate of sedimentation induced collapse from 1988-1995 (figure 2.6).
OM governance suggests that averaging MSY from 2050-2060 is only possible if resource availability and mobility are maintained, explicating the need for periodic tidal outlet maintenance. However, whilst OM inhibits all simulations from collapsing, cautionary catches remain the dominant outcome (figure 6.1B). Consequently, the frequencies of cautionary and dangerous futures suggest merely flushing the tidal outlet builds insufficient robustness against drivers like fishing efforts or long-term undesirable ecosystem conditions.
Introducing fishing bans almost doubles the frequency of safe and just catches relative to OM (figure 6.1C). Therefore, it may be hypothesised that a proportion of OM’s dangerous futures are caused by unsustainable fishing efforts, which are converted to sustainable trajectories under OB’s regulations. As a result, exposing Chilika to overexploitation is an unintended consequence of building resilience against the legacy of outlet sedimentation. Transitioning from 8,000 tonnes/year to 2,000 tonnes/year takes an average of 12.0 and 10.5 years under OM and OB respectively. Therefore, processes under OM and OB are more resistant to change than NO, meaning that collapses with elements of overexploitation are more resistant than collapses caused by resource immobility. Despite the slower transitions, dangerous catch generally occurs earlier under OM,
161 Chapter 6: Results
Figure 6.1: Catch time-series generated under the four governance scenarios (colour), plotted alongside pre-2015 catch values (black). The pie charts depict the proportions of different normative futures per governance scenario. Green – safe and just, orange – cautionary, and red – dangerous. See appendix E for colour-blind friendly version.
162 Chapter 7: Results with the year of fishery overcapacity ranging from 2050-2060 (mean = 2056), as opposed to 2052-2060 (mean = 2058) for OB.
The uptake of alternative livelihoods from 2015 completely avoids collapse and increases the frequency of sustainable catch five-fold over OM (figure 6.1D). Therefore, the safest balance between resource availability and exploitation corresponds to policy approaches that preserve resource mobility whilst dampening the growth of fishing efforts, in terms of total boats and area fished. The proportion of safe and just futures under OL increases to 81% if runs averaging above MSY from 2050 are included; however, these are classified as cautionary to discourage mid-century overexploitation.
Analysis of the catch time-series hypothesises that relatively shallow gradients of socioeconomic development (e.g. motorboat uptake and income) are required to persist desirable catches once the resource mobility is secured. Consequently, governors may be cautious of trading short-term profitability for multidecadal persistence. However, with the exception of NO, average safe and just catches until 2049 are ~25% higher than the observed average catch from 2001-2010 (appendix D.2). Furthermore, OL’s safe and just catches from 2050 are closest to the preceding four decades, reflecting the relative resource conservation afforded by the most regulatory policy (appendix D.2). Arguably, an extra 100 tonnes/year under OM until 2049 is an unsafe trade-off, considering the proportion of sustainable futures is 80% less than OL.
In summary, average catch and the achievement of normative safe and just futures are proportional to the number of governance interventions, representing the process of managed decommonisation. Doing nothing attracts the system towards collapse as outlet sedimentation degrades resource mobility. Decadal dredging of the tidal outlet establishes the possibility of sustainable futures, which is further enhanced by the introduction of fishing bans across an area equal to the new tidal outlet. Dangerous futures disappear once alternative livelihoods remove 1/2000 fishers/month. Therefore, securing resource mobility does not guarantee sustainable development as the system becomes sensitive to anthropogenic actions, which represent novel concerns for decision-makers. The following subchapters develop this hypothesis, first by interrogating NO’s collapses, before exploring the driver pathways influencing sustainability across OM, OB and OL.
163 Chapter 6: Results
6.3 ‘No governance’ scenario
NO is analysed separately here because driver based SJOSs are undefinable. Interestingly however, catches exhibit variability in terms of maximum catch, gradients of collapse and the timing of disappearance (figure 6.1). Consequently, this section ascertains whether variability is due to human and/or natural stresses, in order to understand whether doing nothing retains vulnerability to the historical legacy of collapse.
Pearson’s correlation and linear regression analysis performed in R (2008) assess whether the trajectories of three natural and three human drivers associate with the date of catch disappearance (DCD), defined here as the year when catch first equals zero (tonnes/year). Rather than one-off, potentially extreme values coinciding with DCD, rainfall and temperature changes are represented by their respective 2043-2053 mean averages, covering the range of DCD dates. These means are standardised relative to the IPCC (2013) baseline of 1986-2005 to derive climate change magnitude. If natural conditions drive catch variability under NO, it is hypothesised that DCD should be inversely proportional to the magnitude of rainfall increase. In contrast, future temperatures may exhibit a positive relationship with DCD due to the evaporation of runoff. Annual lagoon salinity indicates the internal effects of hydroclimatic changes, tidal water ingressions and the closure of the annual fish migration route. As a slow process relative to annual rainfall and temperatures, salinity is represented by its 2042 average, hypothesised to be inversely related to DCD.
Regarding socioeconomic drivers: (i) fish price is taken in 2042 to match the year before the first DCD, (ii) the maximum active fisher population per simulation indicates both peak fishing efforts and the number of livelihoods directly dependent on catch, and (iii) the maximum number of motorboats depicts fishing effort composition. These latter two conditions are taken pre-2043. If natural conditions solely drive NO’s collapse then no relationship should be present between socioeconomic drivers and DCD.
Overall, variance in DCD can be significantly explained (p < 0.05) by all drivers with the exception of temperature change (figure 6.2), suggesting heightened evaporation cannot compensate for intensified hydrological flows towards Chilika (figure 6.2A). However, insensitivity to ~2°C warming is a positive discovery, suggesting that warmer waters and associated habitat changes are sustainable as long as resource mobility persists. Conversely, average annual rainfall is
164 Chapter 7: Results
Figure 6.2: Relationships between the trajectories of (A-C) natural and (D-F) human drivers and the dates of catch disappearance under NO governance. All degrees of freedom equal 998. negatively correlated with DCD (figure 6.2B) and salinity is positively correlated with DCD (figure 6.2C), together suggesting Chilika remains vulnerable to changing hydroclimatic inputs and resource immobility in the absence of outlet maintenance.
Interestingly, it is not possible to reject the null hypotheses that socioeconomic drivers influence the DCD (figure 6.2D-F). Two possible reasons exist for the
165 Chapter 6: Results positive association between maximum fisher population and DCD (figure 6.2E). First, increased fishing efforts may temporarily offset catch decline before resource loss. However, this hypothesis is undermined by the inverse relationship between the maximum number of motorboats and DCD (figure 6.2F), implying that intensive fishing efforts encourage collapse. Therefore, the positive correlation between the maximum fisher population and DCD is likely an artefact of prolonged livelihood persistence under shallow hydroclimatic trajectories.
In summary, model outputs suggest that NO governance will cause catch disappearance by the early 2050s, regardless of socioeconomic and biophysical trajectories. The timing of collapse is inversely proportional to the magnitude of rainfall change under climate change, with enhanced freshwater and sediment deliveries undermining resource mobility. Therefore, model outputs suggest that decision-makers must dredge the tidal outlet once every 10 years to have any chance of persisting desirable social-ecological functions to mid-century. The analysis of NO unlocks a suite of additional questions to steer Chilika towards a safe and just future. Can sustainable driver pathways be distinguished, securing resource mobility without facilitating overexploitation? How does proactive governance influence resilience to external and internal stresses? And, do novel ecosystem conditions emerge under OM, OB and OL governance?
6.4 Probabilistic resilience analysis
In essence, a system can be considered resilient if desirable outcomes and functions persist across the plausible range of driving conditions (Holling, 1973, Folke, 2006). However, the fuzzy nature of the catch time-series obscures a number of qualities relating to resilience, including the density and range of outcomes at different times. To this end, probabilistic resilience synthesises the catch time-series by plotting probability distribution functions of the probability of a given output magnitude at a given time (Peterson, 2002, Perz et al., 2013, Bitterman and Bennett, 2016), with inverted vertical axes creating resemblance to ball-and-cup diagrams (e.g. figure 2.2). Plotting the probability distribution functions over time explores how the system’s attraction to different ranges of outcomes evolves. A resilient system will have all outcomes aligned with the desirable region of the landscape, meaning that desirable conditions persist regardless of driver trajectories at that time (Perz et al., 2013).
166 Chapter 7: Results
Figure 6.3: The decadal evolution of stability landscapes under OM, OB and OL. Catch between MSY estimates (green) is symptomatic of a resilient fishery under the range of human and natural driver trajectories. Note: the scale of the y-axis across the top panel is different to the other five due to the high density of MSY catches.
167 Chapter 6: Results
In all, the probabilistic landscapes have two key implications for system persistence. First, drifts to higher catches precede the emergence of undesirable basins, suggesting Chilika’s long-term sustainability is undermined by catches above MSY. Second, retaining CPR rights causes the emergence of an alternative basin by 2050, which is delayed until 2060 under OB, and completely absent under OL.
Explaining these overarching dynamics, modelled catch in 2010 forms a single basin within MSY bounds (figure 6.3) because the model is only driven by divergent scenarios from 2010 onwards; the observed 2010 catch of 11,955 tonnes sits within the basins (Kumar and Pattnaik, 2012). All landscapes then undergo two transformations by 2020. First, whilst remaining inside MSY bounds, modal catch is positively displaced. Second, the upper edges of the desirable basins shallow due to catches above MSY. These dynamics signify fishery intensification by 2020, resulting from a combination of fishing effort growth and renewed resource availability after outlet opening in 2020.
By 2030, the modal catch moves beyond MSY under OM and OB (figure 6.3). Although the lower edges of the basins shallow, the proportions of catch below MSY are <10% and <1% for OM and OB respectively, giving little indication of the undesirable outcomes that emerge over the following decades. OM’s desirable basin flattens by 2040, with the system uniformly attracted to catches from 10,000-14,000 tonnes. A second, deeper basin forms by 2050 covering 4,000- 7,000 tonnes, indicating that certain driver trajectories may tip the system away from sustainable outcomes. OM’s desirable basin further shallows by 2060, whilst the undesirable basin becomes more resilient by steepening, deepening and shifting towards diminished catches. In contrast, substantial basin flattening is not evident until 2050 under OB due to robustness afforded by the fishing ban. The stability landscape of OB resembles OM by 2060, albeit with a weaker tendency for catches below 7,000 tonnes and the persistence of a distinct shallow basin around the MSY lower bound.
Under OL, an alternative basin emerges by 2030 centred on 15,000 tonnes (figure 6.3), representing an alternative trajectory away from MSY. However, this basin disappears by 2040 as the landscape evolves towards lower (more sustainable) catches. Concurrently, the proportion of catches below MSY increases from 0.2% (2030) to 12.4% (2050) as cautionary outcomes emerge. A distinct undesirable basin never emerges under OL, coherent with the absence of normatively defined
168 Chapter 7: Results
Figure 6.4: Time-series depicting the annual probabilistic resilience score (the proportion of catches inside maximum sustainable yield estimates) under OM, OB and OL governance. dangerous futures. Therefore, the chance of a given catch magnitude by 2060 remains inversely proportional to its displacement from MSY.
The need for a desirable state definition aligns with the persistence of ecological functions, as a system with a steep-sided basin of attraction may be trapped in unsustainable configurations. To this end, the probabilistic resilience score is developed here, equalling the proportion of catches per year between MSY bounds; a probabilistic resilience score equal to one means that MSY is the only outcome across the driver trajectories at that time.
Probabilistic resilience score fluctuates across all three governance scenarios until ~2040 (figure 6.4). The troughs observed around 2023 in all three scores correspond to the tendency of catches to exceed MSY after outlet opening in 2020, highlighting sensitivity to the unintended consequences of outlet opening during steep socioeconomic growth. Missed by the decadal resolution landscapes (figure 6.3), all trends recover to ~70% of initial probabilistic resilience score by
169 Chapter 6: Results the late-2020s as catches fall to MSY, before outlet opening in 2030 again boosts catches and weakens the probabilistic resilience score. Hereafter, the cyclical patterns under OM and OB terminate, with subsequent openings unable to remediate catch degradation. In contrast, OL’s probabilistic resilience score retains some sensitivity to decadal outlet openings, partly afforded by relatively weak catch intensification following earlier outlet openings.
In conclusion, probabilistic resilience landscapes provide a method to synthesise the dynamics of 1000 time-series, whilst also charting the emergence of undesirable outcomes. Initial landscape evolution towards catches above MSY suggests short-term dynamics may undermine long-term resource renewal. Interestingly, tidal outlet opening in 2020 and 2030 helps boost catches above MSY, but degrades the ability of subsequent openings to restore sub-MSY catches. So far, the catch time-series and probabilistic resilience analysis black-box the causal interactions between resource availability, governance and fishing efforts. Therefore, to further the SJOS concept as a forward-looking tool, the next subchapter investigates causal trajectories before designing SJOSs for sustainable development.
6.5 Future driver trajectories
This section presents 36,000 driver trajectories (1000 trajectories x 12 drivers x 3 governance scenarios) to assess whether particular system dynamics, such as gradients, cycles and nonlinearities, are symptomatic of different futures. The drivers scrutinised are: External drivers: (i) annual rainfall averaged across the six districts (figure 3.12), (ii) annual temperature averaged across the six districts, (iii) fish price at first sale and (iv) diesel price. The effect of birth and death rates (‘r’) is excluded as it is incorporated into the more tractable fisher population, which is annually measured by regional authorities. Internal anthropogenic drivers: (i) total active fisher population, (ii) number of motorboats (iii) proportion of motorboats within total boats, and (iv) juvenile catch rate – defined here as the percentage of immature catch in total catch. These socioeconomic drivers indicate the extent and intensity of fishing efforts, plus human dependency on the system. Natural drivers affected by internal human actions: (i) total annual freshwater inflow, (ii) total annual sediment inflow, (iii) annual average salinity and (iv) annual average macrophyte coverage. Whilst freshwater
170 Chapter 7: Results
and sediment deliveries are unmanageable short of upstream damming, dredging the tidal outlet internalises ecological effects, including the impacts on macrophyte growth, salinity and fish migration.
Other drivers aggregate into the twelve above. Unsafe dissolved oxygen levels historically couple with undesirable salinity during stagnant freshwater conditions and periods of abundant macrophyte coverage (Biswas, 1995). The total number of boats is assumed to equal the fisher population divided by six. Boat upgrades are captured in the motorboat time-series, providing a relatively tangible indicator of fishing effort composition. Whilst catches are generally linked to stock abundance (Pauly et al., 2013), model performance analysis shows that stocks are stable in the absence of human and natural stresses (chapter 4.3.5). Therefore, the fish population is considered a wider social-ecological outcome of fishing and stock renewal (chapter 6.8).
6.5.1 External drivers
The dynamics of the four external drivers have three main implications for Chilika’s sustainable pathways. First, despite their linear trajectories, fish price and diesel price appear to have pathways symptomatic of the three normative futures (figure 6.5). In general, system safety tends to be inversely proportional to the gradient of fish price change but positively related to the gradient of diesel price growth, potentially due to the combination of high profitability and limited costs that make intensive non-traditional practices affordable. Second, managed decommonisation widens the safe and just trajectories of fish price and diesel price. The maximum sustainable fish price is 690 INR/kg by 2060 under OM, expanding to 800 INR/kg and 1,500 INR/kg under OB and OL, respectively. Likewise, the minimum safe diesel price by 2060 falls from 75 INR/litre under OM, to 65 INR/litre and 60 INR/litre under OB and OL, respectively.
Third, and in contrast to fish and diesel prices, future catch appears insensitive to the trajectories of rainfall and temperature change as sustainable and dangerous outcomes occur across the magnitudes of decadal climate change (figure 6.5). The lack of clear associations between climate change and fishery safety supports the hypothesis that decadal tidal outlet maintenance builds resilience against the climate change magnitudes modelled here, with resource mobility preserved to avoid the type of collapse found under NO.
171 Chapter 6: Results
Figure 6.5: External trajectories under OM, OB and OL governance, classified as safe and just (green), cautionary (orange) or dangerous (red) depending on their normative future from 2050-2060 (figure 6.1). Climate trajectories are plotted as decadal anomalies from the 1986- 2005 baseline (blue dashed). Historical – black. See appendix E for colour-blind friendly versions.
172 Chapter 7: Results
6.5.2 Internal drivers: anthropogenic
Three key findings underpin the safety of internal socioeconomic pathways at the system level (figure 6.6). First, the four internal socioeconomic drivers generally increase in magnitude over time, owing to fishery production that supports a
Figure 6.6: Internal fishing effort trajectories driving fish catch under OM, OB and OL governance. Time-series colours are the same as figure 6.5. See appendix D.1.3 for colour-blind friendly version.
173 Chapter 6: Results growing number of livelihoods and the shift towards non-traditional fishery practices (at least over the next three decades). Second, outcome safety is generally inversely proportional to the gradients of the four drivers, further suggesting that Chilika is vulnerable to intensifying human actions in future. Third, as with the external socioeconomic drivers, the maximum gradient leading to safe and just catches generally increases with decommonisation, indicating that decisions are makeable today that increase cross-scale resilience. For instance, the maximum permissible number of motorboats increases from 7,500 under OM to 9,300 and 10,000 under OB and OL, respectively. The SJOSs are detailed further in chapter 6.6.
Amongst these generalisations are complex dynamics that also influence sustainability. First, model outputs suggest that the active fisher population may abruptly rise then fall over the next 15 years. This spikey dynamic is explained by traditional fisher income increasing three-fold by 2020 under the steepest trajectories of fish price and catch. Consequently, ~75% of traditional fishers may switch to motorboats, causing the rapidly expanding non-traditional livelihood carrying capacity to accommodate steep population growth. However, ~25% of the traditional population cannot afford motorboats, causing the traditional sector to enter a spiral of decline. The remaining traditional fishers are then out- fished by the dominant and expanding non-traditional sector, causing traditional livelihood loss once resource access no longer covers livelihood costs.
The population trends return to a restricted funnel from the mid-2020s, with either a relatively sustainable mix of traditional and non-traditional fishers, or a system dominated by non-traditional fishers. Nonlinearities from higher population growth rates re-emerge by 2040. From 2050-2060, dangerous populations undergo stabilisation or reverse, as livelihood carrying capacities shrink due to catch collapse. Therefore, dangerous futures are associated with livelihood losses for the poorest during a quantitatively healthy fishery, before the total population is vulnerable to resource exhaustion from mid-century.
Safe and just motorboat uptake tends to be linear under OM. Nonlinear trajectories may be sustainable under OB, whilst OL accommodates the stepwise changes and motorboat dominance associated with cautionary and dangerous futures under OM and OB. The eventual state of an unsafe trajectory also depends on long-term population growth, with dangerous curvilinear motorboat trends emerging from the 2040s under the steepest population changes. Therefore,
174 Chapter 7: Results whilst short-term nonlinear efforts may push the system from its SJOS, the dynamic does not alone lead to dangerous futures.
Juvenile catch forms a reinforcing feedback driving both resource degradation and the adaptive responses of fishers to declining resource density. Unsafe juvenile catch emerges first under OM as fishers adapt to the earliest stock declines, but due to two synergistic dynamics, OB’s fish catch may be completely dependent on juvenile catch by 2060. First, a predominantly non-traditional fisher population perceives the decline in mature fish, causing the slow switch to juvenile catch in an attempt to maintain productivity (figure 3.23). Second, the outlet fishing ban limits mature fish catch (chapter 5.3), further encouraging the transition to juvenile catch. Thus, whilst helping to delay catch declines and widen permissible fishing effort trajectories, fisher adaptations under OB may unintentionally cause less selective catches that undermine long-term persistence.
The socioeconomic dynamics of OL are distinct insofar as safe and just futures are possible despite spiked population growth by 2020, nonlinear population growth by 2060 and motorboat domination. Such resilience is because the modelled rate of alternative livelihoods keeps the fisher population (i.e. total fishing effort) below the dangerous zone of OM and OB. Therefore, for at least the next 45 years, OL offsets the creeping causes of overexploitation and builds resilience against short-term dynamics found to destabilise the system under less regulatory governance.
Relative to the external drivers, the safety of internal socioeconomic pathways is complex, dependent on different linear and nonlinear growth rates, multivariable interactions and responses to governance. However, it is not yet known whether the system is heading towards particular ecosystem conditions, and if so, whether associations exist with fishery state that undercut resilience to human processes.
6.5.3 Internal drivers: biophysical
Freshwater and sediment delivery trajectories are plotted as mean percentage changes from 1999-2009, which represents the period of available data (figure 6.7). The extreme deliveries of 2001 are excluded as per Kumar and Pattnaik (2012), giving baselines of 4,300 MCM/year and 790,000 tonnes/year for freshwater and sediment deliveries, respectively. Overall, the gradients of ecohydrological change and fishery safety share fuzzy relationships consistent with the climate drivers (figure 6.5). Therefore, decadal outlet maintenance
175 Chapter 6: Results
Figure 6.7: Chilika’s ecohydrological trajectories classified as safe and just, cautionary or dangerous, depending on their respective fishery outcomes. Time-series colours are as per figure 6.5. appears to desensitise fishery state from both climatic and hydrological dynamics, considered the main causes of the 1990s decline.
176 Chapter 7: Results
Decadal outlet opening produces a distinct salinity funnel. Of the 138,000 annual salinities across the three governance scenarios (2015-2060), 77 are beneath 8 ppt lower limit for brackish species (Martin et al., 2009). All annual salinities remain above 5 ppt, regarded as the lower limit for Chilika’s relatively tolerant brackish species (Biswas, 1995). Brackish conditions also indicate that the migration route for 70% of Chilika’s fish stock remains relatively sediment free, thus helping to conserve resource mobility.
Macrophyte coverage steadily increases in future because salinities are suitable for the growth of species such as Potamogeton pectinatu, Najas flaveolata and Phragmites (Forbes and Cyrus, 1998, Meyerson et al., 2009). Across OM, OB and OL, macrophyte coverage by 2060 has a positive correlation with the degree of decadal sediment delivery change (r = 0.63, r2 = 0.39, p < 0.05, df = 2998). Yet, persistent outlet opening keeps salinity and sedimentation safe relative to the levels that triggered nonlinear macrophyte growth during the 1990s.
Overall, decadal outlet maintenance creates pathways of lagoon sedimentation, salinity and macrophyte growth that are ecologically safe, at least relative to the 1990s decline. However, brackish water conditions and ~500 km2 of ecological refuge do not avert overexploitation driven collapses under nonstationary fishing efforts. The presence of different states along similar driver (multifinality) and outcome (equifinality) pathways produces fuzzy time-series. Therefore, except for the most distinct trajectories (e.g. dangerous motorboats – figure 6.6), it is difficult to distinguish safe and just SJOSs from cautionary and dangerous pathways. Therefore, the next subchapter synthesises time-series complexities and interactions to define driver based SJOSs, beyond which sustainable catch from mid-century cannot be achieved.
6.6 Driver based safe and just operating spaces
6.6.1 Design and visualisation
Since the seminal study of Rockström et al. (2009), SOSs and SJOSs (Cole et al., 2014, Dearing et al., 2014) have mainly been depicted as radar plots to facilitate comparisons between current outcome conditions and thresholds (figure 6.8A). Whilst possible here, adopting this method of visualisation for future thresholds downplays: (i) multidimensional driver pathways, and (ii) dynamic thresholds, which map how safety changes due to cross-scale interactions. Evidenced by the
177 Chapter 6: Results
Figure 6.8: (A) SJOSs of current outcome values relative to environmental limits and social foundations, represented by single threshold values (Dearing et al., 2014, p.233). (B) driver based thresholds defined here by associations between driver trajectories (horizontal axis) and normative outcome states (vertical axes); soft thresholds (green line) are due to multivariable interactions causing a driver trajectory to associate with multiple outcome states. See chapter 6.7 for core SJOS explanation. driver time-series, multifinality arises when near identical driver trajectories associate with multiple output states, whilst equifinality occurs when the same outcome state is reached by different driver trajectories. Of the four dynamics defining SJOSs (Dearing et al., 2014), type-I thresholds are used here to represent normatively defined futures from 2050-2060, identifying how far drivers can evolve before the desirable future is unachievable. Whilst potentially downplaying nonlinearities, linear thresholds are applicable to every driver and readily communicate the time remaining until a current trend exceeds its SJOS.
As in chapter 6.3, each trajectory must be summarised by a descriptive statistic to assess its relationship with the normative future:
Socioeconomic drivers are represented by their maximum annual value by 2049, as dynamics until this date determine whether the system is overexploited from 2050-2060. Macrophyte coverage is represented by its maximum value by 2049, due to its perceived role in undesirably limiting resource access.
178 Chapter 7: Results
As freshwater conditions are undesirable, salinity trajectories equal the minimum annual value reached by 2049. Using an annual value for hydroclimatic trajectories is unsuitable because extreme inputs occur even under shallow multidecadal trajectories. Therefore, trajectories of rainfall, temperature, freshwater and sediment inflows are identified by their 2050-2060 mean averages relative to their 1986-2005 baselines.
Conditional probabilities are plotted in the R (2008) package ‘cdplots’ (Zeileis, n.d) to depict the proportion of different futures for a given driver trajectory (figure 6.8B). Driver ranges only associating with safe and just futures (% = 100) SJOS denote the safest plausible trajectories. Alternatively, a range without safe and just futures (% = 0) means that a driver exists beyond its SJOS. As the pathways SJOS are interlinked, being outside the SJOS edge of a given driver means that all other drivers also transgress their SJOS. Such interdependencies are missing from the original planetary boundaries (Rockström et al., 2009), but their consideration here is critical to exploring cross-dimensional boundary shifts.
When viewed together, the plots form a multidimensional tool to drive systems towards safe and just futures. The conditional probabilities may be also displayed as empirical cumulative distribution functions, plotting the cumulative frequencies of safe and just and dangerous trajectories. Based on the separation between empirical cumulative distribution functions, the non-parametric Kolmogorov-Smirnov statistic (Massey, 1951) assesses whether the safe and dangerous spaces are statistically distinct (chapter 4.3.6.2), helping to gauge the significance of the cautionary window that warns of unsafe trajectories becoming dangerous.
Much thought has been given to the representation of SJOS uncertainties. Arguably the most critical uncertainty exists around the edges of the SJOSs (% = SJOS 0), as the location of this threshold is partly dependent on the number of simulations. Three approaches to communicate uncertainty were considered, with option 3 used here: 1. The standard deviation of the conditional probabilities of each SJOS. However, dispersion around the mean is less important than variability in the point beyond which sustainable futures disappear. 2. Calculate confidence intervals of a GLM between conditional probability and driver trajectory. However, complex probability trends (e.g. OL motorboats – figure 6.10) perturb the use of GLMs.
179 Chapter 6: Results
3. Disaggregate the 1000 simulations into 20 groups of 50 simulations, taking the last (first) safe (dangerous) trajectory of each subset. Then, based on the technique of Paul and McDonald (2005), the samples are bootstrapped 1000 times to ascertain the 95% confidence intervals around the mean trajectory causing the loss (emergence) of safe (dangerous) futures. This technique assesses how the space edges may vary with a different number of simulations, assuming that the population distribution can be determined by resampling the original data.
The driver based SJOSs are displayed below in the same order as the time-series (chapter 6.5).
6.6.2 External drivers
The SJOSs of the external drivers have four key implications for the impacts of internal fishery policies on permissible driver pathways. First, the probability of reaching the safe and just normative future from 2050-2060 does not reach 100% (figure 6.9), meaning a safe and just future cannot be guaranteed by a particular external pathway under the governance modelled here. Second, only fish price trajectories under OM and OB have distinct SJOSs (figure 6.9), respectively
Table 6.2: The 95% confidence intervals of the external SJOSs, alongside Kolmogorov-Smirnov statistics (D) signifying the degree of separation between the safe and the dangerous spaces. Significance (S) levels: ‘1’ – p < 0.01; ‘2’ – 0.01 ≤ p ≤ 0.1; ‘3’ – p > 0.1. Dashed cells are where distinct SJOSs are undefinable.
180
Chapter 7: Results
safe and just’ (solid green), (solid green), just’ and safe
‘
2060. 2060.
- l drivers to the proportion of proportion to the drivers l
dash red) futures from 2050 from red)futures dash
-
relating the magnitudes of externa of magnitudes the relating
Conditional probabilities Conditional
‘cautionary’ (dashed orange) and ‘dangerous’ (dot ‘dangerous’ and orange) (dashed ‘cautionary’
: :
9
.
6
Figure Figure
181 Chapter 6: Results reaching 370-430 INR/kg and 490-550 INR/kg by 2050 before losing safe and just outcomes (table 6.2). Third, as hypothesised from the time-series, the chances of achieving safe and just futures are proportional to the diesel price gradient, particularly under OM and OB (figure 6.9). However, both safe and dangerous futures cover the range of diesel price trajectories, making SJOSs undefinable (table 6.2). Therefore, whilst steeper costs tend to discourage upgrades to non-traditional fishing, the effect is insufficient to avert resource overexploitation.
Fourth, whilst model outputs suggest that safe and just futures are achievable across all modelled trajectories of climate change, the conditional probabilities show mixed relationships between climate change and the proportion of normative futures (figure 6.9). For instance, the proportion of dangerous futures increase by 1.5 and 2.5 times beyond +15% rainfall for OM and OB, respectively, suggesting that increased freshwater runoff and outlet closure undermines resilience to human actions. However, the trend is opposite under OL, as the probability of safe futures is proportional to the magnitude of future rainfall change. The safety of temperature change is generally homogeneous across the plausible trajectories, suggesting that as per NO, increased evaporation and warmer waters do not influence fishery sustainability across the modelled horizon.
Safe and just outcomes across the magnitudes of climate change further supports climate resilience. However, under OM and OB, fish price must remain inside well- defined thresholds to avoid unsustainably intensifying anthropogenic processes. Therefore, these SJOSs point to three ways decision-makers can avoid unsafe futures: (i) do nothing and hope that fish price remains safe, (ii) actively reduce exposure to external pressures, perhaps through price caps or export reductions, or (iii) modify internal dynamics to dampen internal causes of overexploitation, like the switch from traditional to non-traditional fishing.
6.6.3 Internal drivers: anthropogenic
The SJOSs of the internal socioeconomic drivers have three key implications for the sustainable governance of Chilika. First, consistent with the time-series (chapter 6.5.2), the proportion of safe and just futures declines as all four socioeconomic drivers intensify (figure 6.10). Simultaneously, managed decommonisation increases the range of socioeconomic trajectories achieving safe and just futures. Third, and in contrast to the external drivers, only the
182 Chapter 7: Results
Table 6.3: The 95% confidence bounds of the internal socioeconomic SJOSs, alongside Kolmogorov-Smirnov statistics (D) signifying the degree of separation between the safe and the dangerous spaces. Significance (S) levels are as per table 6.2.
number of active fishers under OL does not have distinct safe and unsafe spaces (figure 6.10), reinforcing the interpretation that Chilika’s persistence is more closely related to fishery intensity rather than total extent.
Within these generalities lie interesting dynamics. The probability of dangerous futures tips abruptly from <10% around 53,000 fishers, to >50% around 56,000 fishers under OM (figure 6.10), suggesting that business-as-usual is sensitive to slow anthropogenic processes. Explaining the weakest Kolmogorov-Smirnov scores, dangerous fisher populations emerge inside the edges of the SJOSs under OM and OB. Therefore, despite the association between socioeconomic intensification and unsustainability, the safety of active fishers is at least partly dependent on other drivers. In agreement, whilst safe futures are possible across all OL’s fisher populations, the four-fold loss in safety from 45,000 to 50,000 fishers corresponds to the steepest gradients of population growth and motorboat use. In contrast, with the highest Kolmogorov-Smirnov scores (table 6.3), the SJOSs of motorboats and juvenile catch are well defined. The number of motorboats under OM and OB is the only socioeconomic driver with ranges with at least 75% chances of safety. Decision-makers should prioritise these high safety spaces (“core SJOS”) over SJOS edges to ensure drivers with less spiky conditional probabilities remain safe (chapter 6.7).
183
Chapter 6: Results
2060. Colours and line types are as per figure 6.9. figure as per are linetypes and Colours 2060.
-
of four internal socioeconomic drivers to the proportion of ‘safe of‘safe proportion the to drivers socioeconomic internal offour
magnitudes
: Conditional probabilities relating the the relating probabilities Conditional :
and just’, ‘cautionary’ and ‘dangerous’ futures from 2050 from futures ‘dangerous’ and ‘cautionary’ just’, and
10
.
6
Figure Figure
184 Chapter 7: Results
OL expands the upper edge of the motorboat SJOS by 40% and 14% relative to OM and OB, respectively. OL’s SJOS edge beyond 9,000 motorboats suggests an inherent vulnerability to the absolute magnitude of intensive practices, irrespective of policy approach. Interestingly however, fish catch under OL is resilient to the complete loss of traditional practices by 2049, as long as the number of motorboats remains below 8,500. Moreover, OL’s bimodal motorboat SJOS is because 4,500-5,500 motorboats by 2049 are associated with catches above MSY, resulting from resource conservation during the preceding years.
The expansion of the juvenile catch SJOS under OB implies increased system resilience against lower fish maturation rates (figure 6.10). Although the safe edge contracts under OL, resilience to juvenile catch increases in relative terms as the entire range of juvenile catch only produces safe and cautionary futures. Therefore, the distinctness of the motorboat and juvenile catch SJOSs argue that Chilika’s sustainability is more closely linked to effort intensification than extensification (Nayak, 2014); consequently, even though total boat numbers may appear safe, a dangerous number of motorboats may still cause undesirable futures. As Chilika’s motorboat fleet is only sporadically monitored, these results suggest that the system may be sleepwalking towards its SJOS edge.
The socioeconomic SJOSs synthesise permissible fishing effort pathways from 2015-2049, acting as interacting guardrails to guide human processes and avoid tipping towards unsafe futures from 2050. The multidimensional nature of these SJOSs argues for the prioritisation of spaces with the highest proportion of safe and just outcomes. Before identifying these core SJOSs, it is important to assess whether internal ecosystem conditions must also remain within permissible limits.
6.6.4 Internal drivers: biophysical
Overall, the proportion of sustainable outcomes is generally positively related to the maximum macrophyte coverage reached, meaning blocked fishing grounds are desirable to counteract fishing effort intensification (figure 6.11). Therefore, of the four ecohydrological drivers, only macrophyte coverage exhibits significant differences between safe and dangerous trajectories (table 6.4). Interestingly, the link between outcome safety and macrophyte growth is least noticeable under OM, suggesting unregulated fishing efforts overcome the effect of ecological refuge until macrophytes cover approximately 525 km2.
185
Chapter 6: Results
2060. Colours and line types are as per figure 6.9. figure per as are linetypes and Colours 2060.
-
Conditional probabilities relating the magnitudes of internal ecohydrological drivers to the proportion of ‘safe and and ‘safe of proportion to the drivers ecohydrological internal of magnitudes the relating probabilities Conditional
just’, ‘cautionary’ and ‘dangerous’ futures from 2050 from futures ‘dangerous’ and ‘cautionary’ just’,
11
.
6
Figure Figure
186 Chapter 7: Results
Table 6.4: The 95% confidence bounds of the ecohydrological SJOSs, alongside Kolmogorov-Smirnov statistics (D) signifying the degree of separation between the safe and the dangerous spaces. Significance (S) levels are as per table 6.2.
Hydrological inflows also miss distinct SJOSs as associations with safety remain relatively homogeneous across OM and OB (table 6.4). Higher hydrological inflows tend to increase catch safety under OL, potentially by strengthening the feedback between macrophyte growth and sediment retention in-between outlet openings. Interestingly however, salinity trajectories do not associate with safety from 2050- 2060 (figure 6.11), evidenced by the trendless conditional probabilities and insignificant Kolmogorov-Smirnov scores. Therefore, although increasing macrophyte coverage up to 550 km2 may bolster resilience against overexploitation, the plausible trajectories of natural drivers do not form critical pathways towards safe and just futures.
This chapter has shown that modelling can develop the SJOS concept by designing forward-looking driver-based limits based on the proportion of different futures for given driver pathways. In summary, regarding Chilika’s future sustainability: (i) socioeconomic drivers are generally associated with distinct SJOSs, whilst safe and just outcomes are present across the plausible ranges of internal and external biophysical conditions; (ii) managed decommonisation increases resilience latitude to all socioeconomic drivers; (iii) whether a given driver trajectory (e.g. fisher population) is sustainable may also depend on the concurrent trajectories of human (e.g. motorboats) and natural stresses (e.g. macrophyte coverage). Extending these notions of
187 Chapter 6: Results multidimensionality, the next subchapter explores whether cross-scale interactions influence system safety.
6.6.5 Safe and just space interactions
This thesis explores the permissible driver pathways and spaces leading to safe and just, cautionary and dangerous futures. As highlighted in other studies, interactions between SJOSs are presently unclear, forming a research demand at the regional (Hossain et al., 2017) and Earth system scales (Rockström et al., 2009). Of particular interest are trade-offs where the safety of ‘driver a’ decreases as ‘driver b’ increases (figure 6.12). Critically for decision-makers, such antagonistic relationships mean that the safety of ‘driver a’ is not solely self- determined, but partly dependent on interactions with processes potentially of different spatiotemporal scales.
To identify such interactions, Pearson’s correlation analysis is performed in R (2008) between the safe and just trajectories of the twelve key drivers under OM
Figure 6.12: Hypothetical interactions between two SJOSs, whereby the safety of ‘driver a’ is conditional on the safety of ‘driver b’. In actuality, SJOSs exist within hyperdimensional space (e.g. more than three dimensions), but two dimensions are shown here for visual clarity.
188 Chapter 7: Results and OB. Here, the strongest social-ecological correlation under each governance scenario is plotted to operationalise the theory that global stresses like climate change may influence the safety of local drivers (Scheffer et al., 2015, figure 1). The axes of Scheffer et al. (2015, figure 1) are flipped, as it is argued that drivers at higher spatial scales beyond the reach of regional governance condition the permissible trajectories of local stressors. Thus, the interactions help guide the governance of local stressors in the face of uncontrollable climate change and globalisation. The remaining significant correlations (p < 0.05) are plotted in appendix D.3.1.
Eleven cross-scale social-ecological SJOSs are significantly correlated under OM, of which the strongest antagonistic relationship is displayed in figure 6.13A. The inverse relationship between motorboat use and sediment delivery (figure 6.13A) suggests that faster rates of sedimentation may undermine permissible fishing efforts. Likewise, the fish population becomes increasingly vulnerable to juvenile catch under steeper gradients of climate change (appendix D.3.1). Therefore, whilst not directly causing collapse, resource immobility via the tidal outlet contracts permissible socioeconomic trajectories. As an illustration of this trade- off, the maximum number of safe motorboats in a future with >20% sediment delivery change equals 5400 (mean = 4900, n = 16); in contrast, 5900 motorboats (mean = 5200, n = 10) are permissible under futures with <-15% sediment delivery change. These cross-scale interactions have implications for the adaptive capacities of stakeholders under climate change, meaning governors cannot completely discount biophysical trajectories when aiming for SJOSs.
OB governance builds resilience against these antagonistic cross-scale interactions, with only three social-ecological significant correlations (appendix D.3.1). Whilst the synergistic relationship between fish price and diesel price remains (r = 0.25, r = 0.26), the antagonistic relationship between fish price OM OB and active fishers intensifies (r = -0.26, r = -0.31), although this potentially OM OB reflects OB’s 79 extra safe and just futures (figure 6.13B). Therefore, the safest space involves relatively few fishers sharing a relatively invaluable fishery, meaning fishing effort extent and intensity are constrained. The SJOSs of fish price and macrophyte coverage positively interact under OB (appendix D.3.1), implying that refuge from fishing becomes increasingly important as per unit value rises. Interestingly however, the associations between macrophyte coverage and total fishers and motorboats are insignificant, suggesting that ecological
189 Chapter 6: Results
Figure 6.13: The strongest cross-scale SJOS interactions under (A) OM and (B) OB governance scenarios, suggesting system resilience to local drivers is partly conditioned by external processes . See appendix D.3.1 for correlation coefficients. refuge builds resilience against the individual effects of fishing effort extent and intensity (appendix D.3.1).
This subchapter has shown that driver based SJOSs are interactive, where the safety of a given driver trajectory depends on both its own contribution to system sustainability, plus simultaneous pathways of other drivers with moderating or reinforcing effects. Furthering this notion of multidimensionality, the next subchapter seeks to design “core SJOSs” of the most sustainable configurations under each governance scenario, whereby simultaneously remaining inside multiple limits positions the system away from SJOS edges.
190 Chapter 7: Results
6.7 Core safe and just spaces
This study has found that SJOSs of interacting and interdependent drivers are complex; simply remaining within one driver limit does not guarantee sustainable system outcomes. The concept of a “core SJOS” was introduced by Steffen et al. (2015b), who argue that climate change and biosphere integrity are the two planetary boundaries with independent capacities to drive state changes in the Earth system. In relation, the concept of core SJOSs used here arises from causal complexities where similar driver trajectories produce different outcomes due to subtle variations in feedbacks, nonlinearities and interactions (Young, 2010, Schlüter et al., 2015). Therefore, the cores identified here represent trajectories leading to the metaphorical centre of the SJOSs (figure 6.8), communicating the least precarious pathways towards desirable futures.
The core spaces are designed by subsetting driver ranges corresponding to ≥75% safe and just futures (figure 6.8B). This threshold reflects the desire for better than fifty-fifty chances of sustainability, whilst also recognising the absence of the certainty magnitudes often used in statistics (e.g. 90%, 95% and 99% confidence intervals). The core does not require all drivers to have safety ranges ≥75%; instead as per Steffen et al. (2015b), only the drivers critical to system sustainability form the core’s cornerstones. The remaining dimensions are the corresponding trajectories of the other eleven drivers, exploring whether less influential variables must also follow distinct trajectories. Rather than conditional probability plots, the cores are plotted as time-series to visualise their range, number and responses to governance. The six drivers with significant differences between safe and just and dangerous pathways are plotted here; the remaining drivers are plotted in appendix D.4.
The core SJOSs have three key implications for the governance and sustainable development of Chilika. First, the number of cornerstone variables increases as governance is decommonised, meaning that more sensitive leverage points emerge to shape sustainability. Only the number of motorboats have safety ranges ≥75% under OM, meaning confidently producing sustainable outcomes is dependent on a narrow range of trajectories. Under OL, fish price, the number of motorboats and decadal rainfall change all have individual safety ranges ≥75%. Second, the number of simulations within each core increases from nine under OM, to 28 under OB and 712 under OL! Thus, the likelihood of a plausible future following the core increases under decommonisation. Third, the permissible
191 Chapter 6: Results
Figure 6.14: The core SJOS trajectories of the system drivers found to have statistically distinct sustainable and dangerous configurations (chapter 6.6). Key: black – core SJOS, green – wider SJOS.
192 Chapter 7: Results range of core driver trajectories expand under decommonisation, evidenced by core trajectories covering the wider SJOSs of active fishers, fish price, motorboat proportion and macrophyte coverage (figure 6.14).
Regarding specific dynamics, OM’s core depends on the shallowest plausible gradients of socioeconomic drivers (figure 6.14). The current number of motorboats may permissibly double by 2050, although linear or stabilising trajectories must be followed, and motorboats must not constitute over 60% of total boats. Fish price may increase by up to 2.4 INR/kg/year from present; worryingly however, fish price has increased by an average of 7.6 INR/kg/year since 2005. The total number of active fishers must not increase by more than 58%. Moreover, the core space requires steep macrophyte growth relative to the wider SJOS, illustrating the need to counterbalance multidecadal-fishing intensification with ecological refuge (figure 6.14).
The core foundation of OB is formed between 4800-5000 motorboats and fish price between 140-400 INR/kg by 2050 (figure 6.14). The maximum permissible fish price gradient (6.0 INR/kg/year) increases 250% relative to OM. Despite accommodating up to 400 more motorboats than OM, OB’s core tightens due to the tendency of trajectories with less than 4800 motorboats to produce catches above MSY. OB’s core also includes curvilinear population growth up to 60,000 total fishers by 2050 (OB simulation #111). Interestingly however, the fish price corresponding to this future remains below 140 INR/kg, causing the fisher population to relax from the early 2040s as the non-traditional population exceeds its livelihood carrying capacity. Under OB, the system is resilient to macrophyte coverage as sparse as 390 km2 – with fishing bans replacing the need for ecological refuge.
Whilst OL’s core generally covers the wider SJOSs trajectories, an upper limit on the number of motorboats remains (figure 6.14), indicating the persistence of a driver that is able to degrade sustainability irrespective of others. Drivers without distinct safe and dangerous trajectories may also exhibit subtle core trajectories (appendix D.4). Most noticeably, OM’s core requires diesel prices of at least 130 INR/litre by 2050 (figure 6.15), reflecting the anchoring effect of steeper costs on fishery intensification; under OB, the core diesel price trajectory relaxes to 75 INR/litre by 2050.
As per the wider SJOSs, various correlations exist between core socioeconomic and natural trajectories (appendix D.3.2). The number of permissible motorboats
193 Chapter 6: Results
Figure 6.15: The negative correlations between external and internal drivers inside the core space suggests that the system must follow multidimensional pathways where the permissible number of motorboats is conditioned by the decadal magnitudes of (left) rainfall change and (right) freshwater delivery change. within OM’s core displays antagonistic relationships with decadal rainfall (r = - 0.85, r2 = 0.72, p<0.05, df = 7) and freshwater changes (r = -0.92, r2 = 0.84, p<0.05, df = 7) (figure 6.15). This relationship mirrors the wider SJOS (chapter 6.6.5), with steep trajectories of fishing effort only permissible under shallow gradients of climate change, as both drivers separately degrade resource availability. OM’s core SJOS is therefore dependent upon both social and ecological conditions and specific social-ecological interactions, with the system’s sensitivity to multivariable interactions weakening with stricter governance. However, with only nine core configurations, it is difficult to assess the robustness of these core relationships. The relationships are insignificant across OB’s 28 and OL’s 712 core trajectories, suggesting that increased internal interventions desensitises the system to interactions forming the core SJOS.
This subchapter furthers SJOS operationalisation by identifying interacting social- ecological trajectories leading to the deepest regions of the SJOSs. However, numerous questions involving inter- and intra-generational trade-offs remain before decision-makers might commit to safe and just trajectories over pathways of steeper socioeconomic development. Thus, the next chapter explores wider
194 Chapter 7: Results outcomes and trade-offs, before the safety of the system’s current trajectory is explored.
6.8 Wider social-ecological outcomes
6.8.1 Rationale and indicators
This study highlights two methods to steer systems away from thresholds and towards normatively defined safe and just futures. First, active decision-making seen here through fishing bans and alternative livelihoods may directly limit degradation by widening system resilience. Second, the governance scenarios have multidimensional (core) SJOSs forming explicit signposts towards sustainability, meaning resilient configurations are achievable without transformative governance.
Focusing solely on driver pathways ignores wider social-ecological outcomes and trade-offs influencing whether decision-makers choose to do nothing, follow particular pathways or instigate transformational governance (Janssen and Anderies, 2007). To this end, five social-ecological outcomes, spread across the model’s subsystems (Levin et al., 2015), help to evaluate the safety and justness of the normative futures further:
1. Total fish population: the sum of mature and immature stocks in any year, expressed as a percentage of the average total stock in 2015. The total fish population explores the relationship between catch and fish population (Pauly et al., 2013), plus the level of resource degradation associated with the SJOSs. 2. Total fishery value: the product of annual catch and average fish price, monitored annually by the CDA. 3. Catch per unit effort (CPUE): the conventional metric of fishery efficiency which combines productivity, resource availability and fishing efforts. Combining both fishing fleets, Chilika’s CPUE has averaged 7 kg/boat-days since 2000. According to a database charting fishing efforts from 85 globally distributed lagoons (Pérez-Ruzafa and Marcos, 2012), 7 kg/boat-days is below expected for a system with Chilika’s fisher density. 4. Per capita income: the average financial benefit received by fishers. Beyond the modelled system, income provides access to goods and services underpinning socioeconomic wellbeing, including food, water and education.
195 Chapter 6: Results
5. Differences in resource access has led Pattanaik (2007) and Nayak and Berkes (2010) to argue that Chilika’s traditional fishers are economically marginalised. To this end, the model tracks the ratio of traditional to non-traditional per capita income, assessing the association between marginalisation and different normative futures. This inequality ratio is bound between 0 (disappearance of traditional income) and infinity, where 1 is equality between traditional and non-traditional incomes. The commonly used GINI coefficient (Gini, 1912) cannot be applied here as only two incomes are known.
6.8.2 Future dynamics
The future dynamics of wider social-ecological outcomes have four headlines for the sustainable development of Chilika. First, resource loss is universal across the three governance scenarios, with its magnitude inversely proportional to outcome safety from 2050-2060 (figure 6.16). Resource degradation occurs even when annual catches average MSY from 2050-2060, associated with up to 50% (OM) and 45% (OB) stock losses by mid-century. The persistence of this downwards trajectory implies that safe and just futures may eventually become cautionary, reminding decision-makers that the 2060 horizon is only a checkpoint towards longer-term sustainability. Second, unsafe trajectories of fishery value and per capita income may peak and reverse by 2050, questioning whether cumulative financial benefits are higher under safe or unsafe futures.
Third, safe and just futures are associated with the persistence of today’s income inequality between traditional and non-traditional fishers, creating a lose-lose-lose situation as either: (i) inequality persists as neither fleet dramatically increases in size, meaning current resource access persists, (ii) income inequality dramatically grows after an abrupt increase in non-traditional practices, or (iii) income equality gradually increases as both fleets grow, making the system becomes vulnerable to overexploitation. Moreover, the core SJOS trajectories under OL are associated with growing income inequality, implying that traditional fishers are marginalised even under the most resilient governance scenario.
CPUE dynamics mirror the fish population declines, with the magnitude of efficiency loss proportional to fishery safety. The magnitudes of CPUE loss within the SJOS are similar under OM (-43%) and OB (-44%) by 2050-2060. Therefore, despite different system structures and resilience to the spectrum of plausible futures, restricting fishing efforts under OB produces the same CPUE dynamics as steeper socioeconomic growth and greater resource degradation under OM.
196 Chapter 7: Results
Figure 6.16: Time-series of five wider social-ecological outcomes under OM, OB and OL governance (chapter 5.3). Key: dark green – ‘core safe and just’, green – 'safe and just', orange – 'cautionary', red – ‘dangerous’. See appendix D.1.4 for colour-blind friendly version.
197 Chapter 6: Results
For both OM and OB, safe fishery value follows linear trajectories up to US$130 million by 2060. However, under dangerous futures, fishery value may peak around US$200 million by 2042 before falling to averages of US$34 million (OM) and US$49 million (OB) from 2050-2060. Cautionary futures have characteristics of both safe and dangerous dynamics, with the steepest gradients of change reaching US$250 million, before reversing from the mid-2040s. However, the occurrence of linear trajectories means that cautionary futures average fishery value above both safe and dangerous futures from 2050-2060.
Annual per capita income exhibits a positive linear relationship with fishery value (r2 = 0.92, p < 0.05, df = 998; r2 = 0.92, p < 0.05, df = 998), meaning OM OM OB OB stakeholders depend on the combined effects of catch and unit worth. Consequently, per capita income steadily rises under safe and just trajectories, but reverses under dangerous futures from the mid-2030s. The combination of delayed declines and some linearly rising trends means cautionary futures average income from 2050-2060 equal to US$67,600 (OM) and US$93,000 (OB), as opposed to US$61,100 (OM) and US$71,800 (OB) under safe and just futures, and US$35,800 (OM) and US$49,400 (OB) under dangerous futures.
The outcomes of OL are described separately here due to the dominance of core safe and just futures. Whilst resource availability and CPUE remain positively related to output sustainability, SJOSs may associate with less than 10% resource degradation by 2060 (figure 6.16). Moreover, cautionary resource abundance and CPUE only diverge from safe trajectories from the mid-2030s. Also unlike OM and OB, catch values of US$200 million do not precede collapse. Instead, fishery value may increase ten-fold from today by 2060, equalling a compound annual growth rate (CAGR) of 5.8%. To put this figure in context, CAGR from 1979-2014 equalled 1.1%/year (CDA, 2016), whilst maximum inter-annual growth equalled 5.5%/year (2000-2001). However, despite improved socioeconomic outcomes, the equality ratio under OL remains below unity. In fact, the SJOSs are more unequal under OL because the system is resilient to the abrupt uptake of motorboats and fisher population growth. Therefore, OL decouples the sustainability of the fishery from shifts in resource access, which generally cause unsafe futures under OM and OB.
6.8.3 Outcome trade-offs
Social-ecological trade-offs can hinder sustainable governance as decision-makers consider competing demands, differences between short-term and long-term system performance (e.g. fish catch), and balances between risk and productivity
198 Chapter 7: Results
Figure 6.17: The negative associations between short-term and mid-century catch under (A) OM, (B) OB and (C) OL. The GLMs (black lines) account for the variance of all 1000 points per plot. Coefficients of determination: r2 = 0.33, r2 = 0.34, r2 = 0.10 (all p < 0.05, df = 998). Colours are OM OB OL as per figure 6.16.
199 Chapter 6: Results
(Janssen and Anderies, 2007, Biggs et al., 2015b, Levin et al., 2015). Whilst driver based SJOSs guide sustainable catch, ignoring socioeconomic trade-offs may give an incomplete assessment of the safety and justness of decisions made today.
Trade-offs highlighted so far include the finding that building general resilience requires limits on the number and type of fishing efforts, and the wickedness of present-day resource access and livelihood inequalities. Adding to these intra- generational trade-offs, one potential inter-generational trade-off is the compromise between short-term and long-term system outcomes. For instance, to what extent does allowing catches above MSY advantage short-term benefits? Likewise, does sacrificing short-term fishery value ensure persistence to mid- century?
Regarding inter-generational trade-offs, catch until 2025 is negatively correlated with average catch from 2050-2060 across OM, OB and OL (figure 6.17) (R = - OM 0.57, R = -0.58, R = -0.32). The slope of the linear regression model between OB OL long-term and short-term catch is a proxy for inter-generational trade-off magnitude. Under OM, increasing 2015-2025 average catch by 100 tonnes is associated with -570 tonnes from 2050-2060 (figure 6.17A); the equivalent changes under OB and OL equal -540 tonnes (figure 6.17B) and -120 tonnes (figure 6.17C), respectively. Therefore, the trade-off between short-term productivity and multidecadal sustainability weakens with decommonisation.
Figure 6.18: Time-series charting cumulative annual fishery value under (A) OM and (B) OL governance. Colours are as per figure 6.17.
200 Chapter 7: Results
Moreover, under both OM (figure 6.17A) and OB (figure 6.17B), model outputs indicate that the safe and just future is achievable by averaging up to 13,900 tonnes/year from 2015-2025. This figure aligns with the MSY upper bound of 13,896 tonnes, arguing that safe and just outcomes may only be achieved if catch remains within sustainable bounds over the next decade. Consequently, any catches above 17,000 tonnes/year (chapter 6.2) must emerge after the first decade to avoid establishing a legacy of over-exploitation by mid-century. In contrast, safe and just futures are achievable up to 14,300 tonnes (2015-2025) under OL, suggesting that the path dependency between short-term and mid- century catches can be weakened.
The trade-off between cumulative benefits is comparatively severe. Under OM (figure 6.18A), safe and just futures undergo shallowest growth in cumulative fishery value from the shallowest fish price trajectories. Cautionary futures average cumulative values 51% (OM) and 62% (OB) higher than safe and just futures by 2060, whilst dangerous futures are 49% (OM) and 54% (OB) higher than safe and just. Therefore, even though annual incomes may collapse by mid- century (figure 6.16), following riskier trajectories over the next 45 years may double cumulative revenues. Such trade-offs are absent under OL (figure 6.18B) due to fish catch persistence across the range of fish price trajectories.
This subchapter has shown that normatively defined desirable futures based on system-wide outcomes have trade-offs of short-term versus long-term benefits, cumulative versus annual outcomes and the well-being of traditional versus non- traditional fishers. Trade-off analysis helps to inform system governors of the holistic implications of their decisions, including human and natural elements negatively affected by fishery policies, as well as providing further evidence to evaluate the broader implications of steering SESs towards SJOSs.
6.9 The sustainability of contemporary trajectories
This study has probed divergent exploratory futures forming pathways towards safe and just futures. However, little is currently known about the sustainability of the system’s current direction. To this end, this subchapter extrapolates the trends of the drivers found to have SJOSs (chapter 6.6) to understand the safety of the system’s current course relative to model outputs. Drivers heading for danger would explicate a pressing need for regulatory governance beyond
201 Chapter 6: Results
Figure 6.19: The deterministic trajectory funnels of four critical drivers of fish catch (2015-2050), created by linearly extrapolating their 5 (dot- dash) and 35 year (dashed) trends. business-as-usual, whilst sustainable trajectories may relax calls for interventions as long as contemporary gradients persist.
Fast processes like fish price and motorboat numbers have undergone different periods of stability and growth, meaning single rates of change are unrepresentative of contemporary trajectories. For instance, fish price increased from 12 to 28 INR/kg between 1980 and 1999, before the rate of change increased four-fold in the subsequent nine years, reaching 65 INR/kg by 2009 and 140 INR/kg by 2014 (CDA, 2016). To account for this temporal variability, the historical trends of fish price, active fishers, motorboats and motorboat proportion are extrapolated to 2050 based on the average annual growth rates of: (i) 1981-2015, and (ii) 2010-2015. The resulting deterministic scenario funnels represent historically steep and shallow gradients of driver evolution (figure 6.19).
These funnels subset the spectrum of OM’s plausible trajectories (chapter 6.5) to assess whether the persistence of contemporary trajectories is associated with model derived safe and just, cautionary or dangerous futures. First, it is therefore assumed that model outputs accurately estimate Chilika’s future safety. Second,
202 Chapter 7: Results
Table 6.5 The proportions of different future states within the deterministic funnels of the four socioeconomic drivers (figure 6.19). The risk of a driver exceeding its safe operating space is inversely proportional to the percentage of safe and just outcomes within the funnel.
it is assumed that the system’s direction can be represented by the linear trajectories of four drivers. Third, driver trajectories are assumed to be self- determined (figure 6.19), thus excluding feedbacks between drivers. This independence enables the deterministic funnels to be ranked by the proportion of futures within the scenario funnel that are safe and just:
푁푢푚푏푒푟 표푓 푠푎푓푒 푎푛푑 푗푢푠푡 푓푢푡푢푟푒푠 푤푖푡ℎ푖푛 푠푐푒푛푎푟푖표 푓푢푛푛푒푙 = 100 × ( ) (Eq 6.1) 푇표푡푎푙 푓푢푡푢푟푒푠 푤푖푡ℎ푖푛 푠푐푒푛푎푟푖표 푓푢푛푛푒푙
Overall, the deterministic trajectory funnels imply that Chilika is on course to avoid dangerous futures, as long as the proportion of motorboats within total boats remains below 60%. The motorboat proportion funnel aligns with 95% of all plausible futures due to wide divergence between high and low trajectories (figure 6.19); as such, the continuation of the trend since 1981 may see no traditional boats operating by 2037.
In contrast, the contemporary motorboat funnel is coherent with OM’s core SJOS – located between 4200 and 4600 motorboats by 2049. The discrepancy between the narrow motorboat funnel and wide motorboat proportion funnel is because the proportion of motorboats within the total boats stock has outpaced the absolute rise in the number of motorboats over the past 35 years. The deterministic funnel of the number of motorboats only matches 1.3% of all simulations, arguing that the contemporary rates of motorboat uptake are slow
203 Chapter 6: Results relative to potential nonlinearities triggering under future catch and income trajectories. Therefore, disconnecting fishery value, fisher wealth and adaptive capacity may give a conservative estimate of the number of motorboats in future. Cautionary futures are the modal normative future under the fish price funnel, coherent with the window of opportunity from 250-400 INR/kg (chapter 6.6.2). Therefore, if fish price continues along an evolutionary pathway between that of the previous 5 and 35 years, the resulting internal socioeconomic conditions may initiate unsafe catches by 2050.
Cautionary futures dominate the deterministic funnels of active fishers. However, the meaning of cautionary reflects catches above MSY from 2050-2060, due to the creeping rate of total fishing effort growth relative to the wider plausible futures. On the one hand, avoiding near-term overexploitation associated with sub-MSY catch by 2050 is positive for fishery sustainability. On the other hand, delaying unsustainable catch until mid-century has potentially unknown implications for a slowly degrading fish stock.
The independent trajectories of the key socioeconomic drivers exhibit mixed safeties, as derived from the fully coupled model dynamics. Persistence of the fish price trend since 2010 may cause cautionary or dangerous futures if an abrupt uptake in the number of motorboats occurs. The removal of dependencies and feedbacks captures only linear trajectories during a period when overexploitation was not a concern for decision-makers. Therefore, a key contribution of this work is the uncovering of potential nonlinearities able to force the intensity of fishing efforts towards unsustainable levels. The next subchapter explores such nonlinearities further, assessing the presence of nonlinear driver-response relationships and their association with irreversibility.
6.10 Driver-response relationships
The system dynamics analysed so far have been plotted over time; therefore, little is known about the speeds of driver-outcome relationships, which may mean small changes in driver magnitudes cause disproportional outcome effects (Groffman et al., 2006, Hughes et al., 2013). Finding future driver-response nonlinearities would provide further limits for social and ecological drivers, assess the proximity of normative SJOSs to nonlinearities and evaluate the wider ramifications of normative SJOSs, such as whether linear limits associate with driver-response tipping points and irreversibility.
204 Chapter 7: Results
Driver-response relationships are commonly depicted using phase spaces of state variables (e.g. Dearing, 2008, Zhang, 2016). Features of interest include periods of stability represented by linear bivariate relationships, and thresholds that once exceeded accelerate outcome degradation. Furthermore, hysteretic relationships suggest that recovery only occurs when a driver is reversed back past the point of collapse (Scheffer et al., 2001), with potential implications including the need to forgo productivity and livelihoods to return the system to its desired state.
6.10.1 Driver-response nonlinearities
Whilst catch is the principal social-ecological outcome, Wang et al. (2012) recommend plotting phase spaces of bistable variables controlled by feedback structures. Therefore, total fish population, expressed as a percentage of the average population in 2015, is plotted against a fast (number of motorboats) and a slow driver (active fishers).
Overall, fish population and fishing effort are negatively associated (figure 6.20), although the gradient and linearity of this relationship differs between states, drivers and governance scenarios. Safe futures are relatively resilient to fisher population growth across both governance scenarios, evidenced by linear phase space relationships (figure 6.20A-B). Under dangerous futures, the relationship between the fish population and motorboat numbers is relatively stable until ~6,000 motorboats, suggesting that fishery sustainability is initially insensitive to technological uptake. Beyond ~7,500 motorboats, fish population falls by an average of 82% by 2060 whilst the number of motorboats increases by an average of 48% (figure 6.20C-D) – meaning small changes in driver conditions cause disproportional outcome effects (Groffman et al., 2006). Therefore, the linear relationship between the fisher population and fish stock suggests that nonlinearities may be missed if motorboats numbers continue to be monitored only sporadically. Positively, the nonlinear thresholds of OM and OB are beyond the normative SJOSs of total fishers and motorboats (chapter 6.6.3), sitting within the cautionary zone of risk (Steffen et al., 2015b).
Interestingly, fish population declines of ~50% appear to require 2,000 fewer motorboats under safe futures than under dangerous futures (figure 6.20). This pattern has two possible explanations. First, relatively stable catch over future decades weakens resilience to fast processes (Gunderson and Holling, 2002), reducing the number of motorboats needed to produce the same catch degradation. Alternatively, it is more likely that resource loss under safe and just
205 Chapter 6: Results
Figure 6.20: Phase spaces comparing the driver-response trajectories of safe and dangerous futures under OM and OB. Total fish population is compared with active fishers (top) and motorboat numbers (bottom). futures does not reflect motorboat uptake; instead, the steeper relationships are actually slower in time, reflecting degradation caused by multidecadal population growth. These dynamics explicate the need to monitor multiple drivers over multiple timescales, as only tracking a single (albeit important) driver like the number of motorboats provides an incomplete picture of resilience.
6.10.2 Chilika: a hysteretic system?
Chilika’s cautionary futures are associated with tipping points where increasing the number of motorboats beyond a critical threshold abruptly accelerates resource loss. However, it is currently unclear whether Chilika’s driver-response nonlinearities are smoothly reversible (e.g. Scheffer et al., 2001, fig. 1b) or folded
206 Chapter 7: Results
Figure 6.21: Phase spaces exploring the coupled relationships between fish population and (A) the active fisher population and (B) the number of motorboats. The forward switch covers until 2060; the reverse switch covers 2061-2100. between alternative states (e.g. Scheffer et al., 2001, fig. 1c). Smooth reversals imply that dangerous futures are recoverable by remediating the system by the same stress magnitude that caused the collapse. Discontinuity means that the undesirable state is irreversible unless drivers return past the point of collapse –
207 Chapter 6: Results with potentially catastrophic impacts on fishery value, livelihood opportunities and human wellbeing.
Simulations beyond 2060 are required to explore system recovery. For efficiency, only configurations causing dangerous futures by 2060 are modelled, corresponding to fish price increases between 17 and 30 INR/kg/year, and ‘r’ trajectories producing more than 55,000 fishers by 2050. Diesel price was not found to have a SJOS, so trajectories vary as normal between 0-7 INR/litre/year (figure 5.10). Rather than stationary climate change after 2060, the maximum trajectories of rainfall and temperature change extend to 2100 under RCP8.5 (IPCC, 2013), equalling ~30% rainfall and 4°C temperature increases relative to the 1986-2005 baseline.
Regarding system governance, OM operates until 2060, before decision-makers attempt to reverse degradation by removing 100 non-traditional fishers per month. An alternative method could explore self-organised recovery resulting from livelihood losses under dangerous futures; however, this approach would not give a baseline recovery rate. Recovery is assessed from 360 simulations, matching the number of dangerous futures under OM. This analysis seeks to answer questions surrounding: (i) the possibility and time-scales of recovery; (ii) whether Chilika’s recovery is hysteretic, and if so (iii) the driver conditions determining hysteresis presence and magnitude.
Overall, normative dangerous futures are found to have deeper social-ecological implications beyond the loss of fisher livelihoods. By 2100, 75% of simulations recover to 90% of the initial fish stock, requiring 22 years (±1.2 years) on average from 2060. In systems terms, recovery requires reversing the number of motorboats by an average of 6,900 (±220) and the fisher population by an average of 44,000 (±1400). Therefore, dangerous futures position the system within an alternative state of resource collapse that may not reverse without further and forced livelihood losses (figure 6.21). Moreover, as per figure 6.20, the fish stock versus fisher population relationship is negatively linear until 2060 (figure 6.21A). Finding that hysteretic relationships result from linear forward switches is significant, because hysteretic relationships are conventionally depicted: (i) as symmetrical trends from nonlinear forward shifts (e.g. Scheffer et al., 2001, fig. 1c) and (ii) only for one driver, thus exempting multivariable interactions which may cause linear forwards shifts to appear hysteretic. Additional questions useful for system governance emerge from this evidence of hysteresis, including whether the reversal magnitude is a legacy of pre-collapse
208 Chapter 7: Results
Figure 6.22: Time-series of (A) motorboats, (B) active fishers, (C) catch and (D) fish population (percentage of average 2015 population) classified by the four types of driver-response relationships. The bottom panel shows the phase space portraits of the fish population against (E) motorboats and (F) active fishers. dynamics, and why particular trajectories appear to undergo a second phase of collapse beyond the initial forward switch?
209 Chapter 6: Results
To answer these questions, driver-response relationships are classified into four categories based on the relationship between the number of motorboats and fish stock (figure 6.21B): (i) the forward shift includes all dynamics until 2060; (ii) non-hysteretic relationships happen when fish stock recovery occurs before the critical driver threshold (6000 motorboats); (iii) hysteretic relationships occur when recovery requires the removal of less than 6000 motorboats, and (iv) trapped runs undergo a second phase of collapse by 2100.
Non-hysteretic responses associate with relative shallow trajectories of fisher population and motorboats, producing resource degradation that is reversible with at least 6,000 motorboats operating (figure 6.22E). Therefore, trajectories involving <9,800 motorboats and <62,000 fishers by 2050 do not require driver reversal past the forward tipping point, thus representing the safest region of the normatively dangerous space. Importantly, the economic value of the fishery still supports the livelihood costs of the active fisher population, which involves a relatively high proportion of traditional fishers. Therefore, linear population declines result from the removal of non-traditional fishers, with traditional livelihoods providing resistance against abrupt resource recovery (figure 6.22D).
The pre-2060 trajectories of active fishers and motorboats give little warning of hysteresis, with the steepest fisher population only starting to show signs of reversal due to overcapacity. Due to notions of irreversibility (Dearing et al., 2014, Bohensky et al., 2015), it may be hypothesised that hysteretic relationships require timescales of reversal longer than non-hysteretic relationships. Model outputs reject this hypothesis, as the temporal duration of resource recovery in hysteretic relationships is inversely related to the magnitude of motorboat reversal (figure 6.23). Hysteretic relationships result from the combination of forced livelihood losses and severe overcapacity, where catch disappearance during the 2060s (figure 6.22C) causes the complete loss of fishing as a livelihood source (figure 6.22B). Interestingly, hysteretic relationships appear to have a minimum duration of recovery, exhibited by saturation in the temporal speed of recovery as reversal efforts increase (figure 6.23-inset). The hysteretic system also has a minimum required driver reversal magnitude, here equating to the removal of ~5000 motorboats from the average in 2060 (figure 6.23-inset).
This inverse relationship between reversal duration and hysteresis magnitude presents a social-ecological dilemma for decision-makers, because the system is rapidly recoverable to the least precarious phase space location at the expense of fishery livelihoods (figure 6.23-main). Consequently, this finding argues that a
210 Chapter 7: Results
Figure 6.23: (Main) three-dimensional phase space depicting the number of motorboats, fish population and the year in which the fish population recovers to 90% of its 2015 value; the forward shift has been removed for visual clarity. (Inset) the negative relationship between the duration of recovery and the reversal magnitude (difference between the number of motorboats in the year of recovery and 2060). complete collapse is favourable for systems becoming dangerous in order to encourage rapid self-organisation back to the desirable system state. In contrast, interactions that relatively preserving socioeconomic functions increase the duration of recovery and return the system to a position precariously close to the nonlinear threshold (figure 6.23-main).
Dangerous futures exhibit a third behaviour mode post-2060, as a second phase of decline may follow recovery (figure 6.22A). These trapped futures are found to correspond to a distinct set of sensitive driver trajectories (figure 6.24). Resource degradation is initially shallow relative to the broader spectrum of dangerous futures (figure 6.24E), owing to relatively shallow fish price growth (figure 6.24A) and gradual motorboat uptake over the next two decades (figure 6.24C). However,
211 Chapter 6: Results
Figure 6.24 Driver and outcome time-series of trapped simulations (colour), plotted for reference on top of the wider spectrum of dangerous futures (grey).
212 Chapter 7: Results overcapacity triggers by 2060 due to the steep trajectories of background population (figure 7.24B). Therefore, during the forward switch, trapped runs occupy a region within the dangerous space between high population-high motorboats (hysteretic) and low population-low motorboats (non-hysteretic).
The removal of non-traditional fishers from 2060 triggers resource and catch recovery by 2070 (figure 6.24D-E). The second phase of collapse is then caused by two dynamics that inhibit the full recovery of the fish population. First, an increasingly valuable fishery again accommodates steep fisher population growth. Second, nonstationary fish price and catch restoration causes step-wise motorboat uptake. The resulting effort intensification and extensification outweighs stock regeneration within the modelled time horizon, with the steepest effort trajectories corresponding to the earliest collapse reoccurrences. Interestingly, the second phase of collapse exhibits an improved fish population for a given driver magnitude, suggesting that boom and bust dynamics build robustness against driver intensification (Gunderson and Holling, 2002). However, further into future, following this trapped configuration risks iteratively increasing the number of motorboats needed to be removed in order to recover the fish population.
This subchapter has shown that normative unsafe futures have deeper implications beyond the loss of livelihoods, including nonlinear driver-response relationships, hysteresis which varies in duration and magnitude, and clearly defined alternative states. Recovery from a dangerous future is not straightforward: the steeper the system degradation pre-collapse, the quicker the system may recover, albeit with deeply undesirable social-ecological trade-offs. Special configurations arise where relatively few motorboats but steep fisher population growth may trap the system, causing fishing efforts to fall abruptly in response to catch collapse, but rise again rapidly during catch recovery. Discussed later, these findings have important implications for the future governance of Chilika, regional SJOSs and wider concepts of resilience.
6.11 Defining ‘safe and just’ by catch predictability
6.11.1 Rationale and approach
The conditions used to determine the safety of driver trajectories are normatively defined a priori based on type-I dynamics (Dearing et al., 2014). A proportion of
213 Chapter 6: Results future catches are dangerous based on the perceptions of regional scientists, governors and modellers, meaning that their underlying driver trajectories must also be dangerous. The underlying system dynamics then establish theories about the safety of different trajectories. Type-II dynamics occur when a system exits its historical envelope of variability, whilst type-III and type-IV thresholds signify the presence of outcome nonlinearities, where the latter is marked by EWSs (Dearing et al., 2014). Type-III and type-IV dynamics derive from broader theories and observations from climatic, hydrological, and social-ecological systems (e.g. Alley et al., 2003, Milly et al., 2008, Scheffer, 2009, Scheffer et al., 2012). Therefore, in order to ascertain whether Chilika may exhibit surprising characteristics considered unsafe in other systems, this section creates a further set of guardrails through the novel use of ARIMA forecasting as a measure of ecological surprise and abruptness.
To this end, remaining inside the envelope of variability defined by observed catch from 2002-2015, or exiting but later returning (figure 6.25A), suggests resilience against short-term stresses and the ability to reorganise towards sustainable trajectories in the face of novel driver magnitudes (Holling, 1973, Folke, 2006). Although a longer calibration period is desirable, extending the envelope of variability to pre-2000 would incorporate the undesirable 1990s collapse into the definition of safe and just. Instead, the shortened envelope captures healthy system variability post-ecosystem recovery. Catch moving and remains outside the envelope is indicative of nonstationarity (Milly et al., 2008) and a regime distinct from the historical mean (Dearing et al., 2014).
Here, the distinction between cautionary and dangerous dynamics is the predictability of envelope exit (figure 6.25B-C). Autoregressive integrated moving average (ARIMA) forecasts assess the predictability of future values based on the stationarity, seasonality and autocorrelation of the preceding time-series (Menard, 2007, Wang et al., 2012). Each ARIMA forecast generated in the forecast package (Hyndman and Khandakar, 2008) in R (2008) requires three parameters: (i) autocorrelation at different time lags (p), (ii) degree of differencing making the time-series stationary (d), and (iii) the order of moving average (q). These coefficients are determined by autocorrelation (ACF) and partial autocorrelation functions (PACF) of 50 randomly selected cautionary time-series under OM (appendix D.5), covering 2015 to permanent departure from the envelope of variability. Of this subset, 70% of ACFs exhibited sinusoidal decays, supporting a (p, d, 0) model (Hyndman and Athanasopoulos, 2014); 82% of PACFs were
214 Chapter 7: Results
Figure 6.25: A method to classify the safety of outcomes based on ARIMA forecasting: (A) catch exiting (orange dot) and remaining outside the envelope (green and blue dashed) are unsafe. ARIMA models (grey fill – 95% confidence interval, blue fill – 80% confidence interval) are then projected for each unsafe trajectory from the year after envelope exit. (B) Catch is classified as ‘cautionary’ if the next five records (red) fall inside the 95% confidence intervals of the ARIMA forecast (predictable), but (C) ‘dangerous’ if catch falls outside the forecast at least once (unpredictable).
215 Chapter 6: Results significant at lag-2 without further significances, supporting a (2, d, 0) model (Hyndman and Athanasopoulos, 2014). Finally, the time-series were differenced (d = 1) to induce stationarity in the multidecadal means prior to envelope departure, supporting the interpretation that unsafe catches are nonstationary.
The next section charts the time-series and summary statistics of the catch trajectories, considering whether unpredictable dynamics (dangerous envelope departures) are symptomatic of undesirable catches by mid-century. Thresholds then capture the driver magnitudes at the point of envelope exit, providing additional guardrails to guide sustainable governance.
6.11.2 Outcome trajectories
The alternative SJOS definition has three key implications for the sustainable steering of Chilika’s fishery. First, of the 1000 plausible futures, 8.7% are safe and just, meaning that classifying based on the predictability of outputs leads to a 25% decline in system safety compared to the definition based on MSY catch
Figure 6.26: Future catch under business-as-usual OM governance plotted alongside pre-2015 catch values (black). Whether catch is ‘safe and just’ (green), ‘cautionary’ (orange) or ‘dangerous’ (red) is determined by the predictability of envelope departure. See appendix D.1 for colour-blind friendly version.
216 Chapter 7: Results
Figure 6.27: Temporal distributions of predictable (orange) and unpredictable (red) departures from the historical envelope.
(chapter 6.2). Therefore, 39 of the simulations exit the envelope but still average MSY from 2050-2060, implying that a proportion of the normatively safe futures are beginning to shift from the safe space. In contrast, the proportions of cautionary and dangerous futures are similar to the normative definitions, with 53% (+1.9%) and 38% (+5.6%) of futures exhibiting predictable and unpredictable departures from the envelope, respectively.
Second, both cautionary and dangerous futures average 4,500 tonnes/year from 2050-2060. Therefore, whilst unpredictability is traditionally perceived as undesirable in natural resource management (Holling, 1973), the condition does not determine the degree of social-ecological degradation from 2050-2060. The majority of envelope departures occur during the 2040s, supporting the earlier interpretations of probabilistic resilience, which begins eroding in the 2030s but accelerates during the 2040s (chapter 6.4). Third, cautionary departures from the envelope of variability tend to occur earlier than dangerous (figure 6.27), with the earliest predictable envelope departure in 2034. The dominance of dangerous thresholds post-2050 implies that prolonged system stability leads to comparatively unpredictable envelope departures. In practice, this may be
217 Chapter 6: Results because stable catches parameterise relatively narrow ARIMA forecasts. The bimodal frequencies signify a second stage of transitions from the mid-2050s, corresponding to stressor conditions too shallow to cause collapse during the 2040s, but too steep to keep catch safe until 2060. Thus, reduced threshold frequency from the late-2040s to early-2050s provides a limited window to remediate stressors reaching magnitudes associated with the second phase of thresholds.
Despite offering alternative views of system safety, little is known about the driver trajectories underlying resilience and causing the envelope departures. As with the normative futures (chapter 6.5), the next section relates the predictability of envelope departure to driver magnitude to design a third suite of guardrails towards sustainable futures.
6.11.3 Alternative driver thresholds
Conditional probabilities are inappropriate to visualise these alternative nonlinear thresholds because the frequency of envelope departure changes over time. Therefore, rather than bounding the SJOSs by maximum permissible driver values (chapter 6.6), sustainable pathways are identified from three-dimensional landscapes relating driver magnitude (x-axis), time (y-axis) and threshold frequency (z-axis). The safest pathways avoid cautionary and dangerous threshold dynamics, whilst hazardous pathways are associated with localised peaks of thresholds.
Overall, system sensitivity to the different drivers influences the presence of thresholds. For instance, the absence of cautionary and dangerous thresholds for 5,000 motorboats marks a well-defined safe and just pathway to 2060 (figures 6.28A and 6.29A). This SJOS matches the core SJOS defined in chapter 6.7, with up to 4700 motorboats by 2050 giving high chances of reaching the normative desirable future. The system is safe up to ~7000 motorboats until the mid-2040s and ~8000 motorboats by 2050. Post 2045, the emergence of thresholds at less than 7000 motorboats suggests that futures avoiding unsafe dynamics over the next thirty years become vulnerable at lower stress levels in subsequent decades. This pattern is due to the concurrent growth in total fishing effort over the next four decades, meaning that fewer motorboats are required to trigger threshold dynamics as the system approaches overcapacity.
218 Chapter 7: Results
The flat space centred on ~60,000 fishers in 2045 reflects the absence of plausible population totals (figures 6.28B and 6.29B). Therefore, highlighting the importance of driver interactions, the number of motorboats must remain safe to avoid triggering thresholds associated with between 50,000-55,000 fishers by the mid-2040s. This is supported by the normatively defined SJOS (chapter 6.6), with cautionary futures occurring across all plausible fisher populations from 2050. Such insensitivity to the fisher population trajectory reinforces the argument that Chilika’s sustainability links more closely to the composition of fishing efforts than their combined magnitude, despite regional decision-makers only annually monitoring the latter.
In agreement, distinct peaks of cautionary and dangerous thresholds arise when motorboats represent up to 90% of total boats (figures 6.28E and 6.29E). Therefore, based on departures from the envelope, a multivariable SJOS for the next 30 years is bound by up to 7,000 motorboats with at least 1,000 traditional boats in operation. The critical proportion of motorboats declines over time, as previously safe trajectories may become cautionary with 50-65% motorboats and dangerous with 50-80% motorboats from the 2050s.
Until 2045, cautionary and dangerous thresholds only emerge beyond fish prices ~350 INR/kg (figure 6.28C and figure 6.29C), corresponding to motorboat proportions above 90%. However, the emergence of thresholds post-2050 at less than 300 INR/kg is symptomatic of extensification driven collapses, occurring at prices, total fishers and numbers of motorboats deemed safe during periods of lower fishing effort in earlier years. Consistent with the normatively defined safe and just future (chapter 6.6.3), cautionary and dangerous nonlinear dynamics are absent beneath 2% juvenile catch (figure 6.28D and figure 6.29D). The emergence of thresholds around >3% juvenile catch results from reduced catch selectivity during stock declines, brought about by earlier overexploitation from non- traditional practices. Therefore, monitoring the proximity of juvenile catch to these thresholds may provide an indicator of fish stock health, knowing that the likelihood of envelope departure increases beyond 2% juvenile catch.
How does the system’s contemporary trajectories (chapter 6.9) compare to these alternative thresholds? The deterministic scenarios suggest motorboats will remain <5000 by 2050 (figure 6.19), thus avoiding the rise of dangerous thresholds pre- and post-2050 (figure 6.29A). A fish price of 400 INR/kg by 2050 is on course to meet the ridge of dangerous thresholds post-2050, albeit
219 Chapter 6: Results
Figure 6.28: The landscape of predictable envelope departures for five socioeconomic drivers (A-E) and one ecological driver of catch (F). Two-dimensional density is proportional to the frequency of thresholds for a given driver trajectory at a given time. The salinity threshold landscape reflects the simultaneously occurring human- induced thresholds. avoidable with sustainable motorboat trajectories. Likewise, the fisher population will likely overcome the ridge of nonlinear thresholds if other drivers remain
220 Chapter 7: Results
Figure 6.29: The landscape of dangerous envelope departures for the six drivers depicted in figure 6.28. sustainable – such as less than 7,000 motorboats making up less than 90% of total boats.
In synthesis, this subchapter has operationalised an exploratory method to assess whether a system (un)predictably exits its historical envelope. Overall, simulations remaining inside the envelope for longer tend to exit unpredictably, although the
221 Chapter 6: Results mode of exit does not determine catch magnitude by mid-century. These thresholds build additional multidimensional, time explicit thresholds to persist post-ecorestoration catch variability. To avoid departures from the envelope by 2045, model outputs suggest that average fish price must remain <350 INR/kg, motorboats <7000 (constituting <90% of total boats), and juvenile catch must remain <2%. However, avoiding these pitfalls does not ensure safety, as system dynamics then become vulnerable to thresholds at lower fishing efforts due to longer (albeit shallower) degradation. The narrower SJOSs is arguably counterintuitive given the later emergence of thresholds, yet it indicates an inverse relationship between the length of time that sustainable outcomes are desired and the permissible driver trajectories.
222 Chapter 7: Discussion
Chapter 7: Discussion
7.1 Chapter introduction
The overarching aim of this study is to investigate social-ecological thresholds delineating sustainable pathways for the Chilika lagoon. Developed on a case-study with a legacy of undesirable social-ecological conditions, the modelling framework involves problem conceptualisation, model construction, plausible external trajectories, alternative governance schemes, and ultimately the classification of safe and just, cautionary and dangerous dynamics. Therefore, this study presents novel contributions to three major research areas: (i) the future sustainability and governance of the Chilika lagoon; (ii) the broad concepts of SJOSs, TBM, sustainability and resilience, and (iii) the modelling of SESs. The headline findings with respects to these themes are synthesised below.
This study provides the first medium-term projections of the Chilika lagoon fishery, and therefore represents the first estimations of Chilika’s social-ecological thresholds and tipping points. Periodically dredging the tidal outlet remobilises the fish population to build resilience against the reoccurrence of the 1990s collapse. However, stable resource availability exposes the system to overexploitation by supporting short-term and multidecadal fishery intensification. In parallel, the frequency of sustainable pathways is inversely associated to the gradients of fish price, motorboat use, total fishers and/or boats and juvenile catch, but is generally insensitive to magnitudes of climate change, hydrological change and fishing costs. Managed decommonisation enhances system resilience against exploratory external drivers and internal manifestations; however, the approach is unable to avoid the complete loss of traditional livelihoods, thus questioning the justness of aggregated policies and sustainability classifications.
More widely, this study has operationalised the search for multivariable SJOSs delineating sustainable future pathways (Dearing et al., 2014, Verburg et al., 2016). Therefore, this work builds upon the contemporary outcomes of Dearing et al. (2014) and Cole et al. (2014), and the discrete model scenarios and binary thresholds of Hossain et al. (2017). Consequently, novel driver-based SJOSs are found to be associated with properties of social-ecological resilience responsive to decision-making. Resilience latitude and resistance derive from the complexity of
223 Chapter 7: Discussion human-natural interactions, whereby similar trajectories of fast drivers (e.g. motorboats) produce different states depending on the trajectory of slow drivers (e.g. total fishers). Built from driver ranges found to critically influence Chilika’s normative future, the core SJOSs identifies driver trajectories leading to the least precarious region of safety, thus providing decision-makers with a hierarchy of leverage points to shape sustainability.
This study finds that the implications of normative unsafe spaces extend beyond their linear definition. For example, beyond 6000 motorboats, the fish population response abruptly switches from linear to nonlinear. Moreover, Chilika is found to be a hysteretic system, with resource recovery from overexploitation requiring the number of motorboats and active fishers to be reversed past their forward tipping points. Reversal magnitude (i.e. the change in driver magnitude needed to escape the alternative state) is positively related to the degree of pre-collapse fishery intensification. Subsequently, decision-makers must choose between dangerous futures of complete collapse and rapid renewal, or, relative livelihood preservation and slower ecological recovery.
This study also develops a time-series metric using ARIMA forecasting to assess the predictability of outcome dynamics. Interestingly, the predictability of the system exiting its historical envelope of variability does not foreshadow the magnitude of catch from 2050-2060, questioning the use of only nonlinear dynamics to judge outcome safety and justness. Wider implications of model outputs include the effectiveness of influencing slow and fast drivers (Walker et al., 2012, Biggs et al., 2015a) and the use of exploratory futures in modelling (Kwakkel and Pruyt, 2013, Maier et al., 2016). This study also has various limitations, particularly associated with parametric uncertainties and processes excluded from the modelling framework.
Thus, this chapter discusses this study’s findings in the contexts of: (i) literature on the Chilika lagoon (chapter 7.2), (ii) broader theories of social-ecological governance (chapter 7.3), shifts (chapter 7.4) and modelling (chapter 7.5), and (iii) limitations of study design, modelling and assumptions (chapter 7.6).
224 Chapter 7: Discussion
7.2 Future of the Chilika lagoon
7.2.1 Through the system archetype lens
The quantitative metrics (e.g. values, rates and thresholds) and qualitative modes (e.g. linear and nonlinear growth/decay) inform archetypes of problematic behaviours common to systems of feedback structures (Senge, 1990, Meadows, 2009). System archetypes provide a method to synthesise the future sustainability of Chilika, with the additional benefits of helping to recognise problematic symptoms prior to occurrence, prepare corrective actions and provide insights into the structures driving unsustainability.
Senge (1990) describes 10 archetypes common to political, economic and environmental systems. If judged by mid-century catch alone (figure 6.1), dangerous futures under OM and OB governance might be interpreted as ‘tragedy of the commons’, where fishers overexploit the resource to maximise benefits between 2015 and 2050. In support, the livelihood benefits of continued fishing are seen to shrink as overexploitation is not remediated (Bailey and Jentoft, 1990), whilst further intensification only widens the disparity between resource loss and renewal (Morecroft, 2007). However, Moxnes (2000) and Bueno (2014) argue that basing archetypes purely on aggregated system state can cause misconceptions about the roles of different feedbacks.
Figure 7.1: Causal-loop diagram depicting the ‘success to the successful’ archetype found to drive unsustainable fishery intensification over the next decade; adapted from Senge (1990). Feedbacks: R – reinforcing, B – balancing.
225 Chapter 7: Discussion
An alternative explanation for unsustainable fishery intensification in future would be traditional and non-traditional fleets entering ‘ratchet mode’ to remain competitive (Meadows, 2009). However, rather than adapting to each other’s actions, the modelled fleets intensify through parameterised feedbacks with the environment (e.g. perceived fish abundance) and conditions internal to each fleet (e.g. traditional productivity). The ‘success to the successful’ (Senge, 1990) offers an alternative archetype for the dangerous pathways, characterised by a reinforcing feedback that strengthens non-traditional resource access as more motorboats fish the lagoon (figure 7.1). The success of motorboat fishing under steep socioeconomic gradients accommodates new motorboat entries from (relatively) wealthy traditional fishers. In turn, the poorest traditional fishers become poverty trapped (Carpenter and Brock, 2008) with resource access insufficient to generate profits required to switch practices. Inequality between the two fleets grows until traditional fishing becomes inadequate for livelihood persistence, causing livelihood losses and a shrinking fisher population.
Chilika’s dangerous futures also align with ‘limits to growth’, found in systems with an inherent limiting condition (Meadows et al., 1972). Timely diagnosis of system limits is critical to avoid inappropriate remediation, like the misconception that
Figure 7.2: Causal-loop diagram of the ‘shifting the burden’ archetype, whereby periodic tidal outlet opening promotes fishery intensification until catch surpasses the sustainable limit; adapted from Senge (1990). Feedbacks: R – reinforcing, B – balancing.
226 Chapter 7: Discussion catch is blocked by vegetation or supressed by limited technology. Such shifting of the burden away from unsustainable resource extraction will only increase the possibility that the limit is exceeded in future – failing to address reasons why the limit was approached in the first place (Senge, 1990). Similarly, model outputs suggest that business-as-usual governance is inappropriate when resource availability already supports sustainable catch, as fish stock additions from a sustainable state unintentionally drives the system closer to its environmental ceiling (figure 7.2). Outlet maintenance irrespective of resource condition is a rigidity trap with governance reluctant to adapt from a previously successful strategy, to the detriment of building resilience against nonstationary socioeconomic stresses. In contrast, OM’s and OB’s safe and just futures avoid ‘success to the successful’ because motorised fleets are unable to generate sufficient revenues to monopolise resource access; however, persistence of today’s inequalities means success still remains with the successful.
7.2.2 Implications for users and decision-makers
These first future projections of Chilika’s fishery provide insights into the evolution of social and ecological processes underpinning regional development. Although not directly evaluating aspects of sustainable development like education, sanitation and health (WorldBank, 2001, Kumar et al., 2011), this study contributes to the discussion of contemporary fishing challenges, including the equity and access rights of traditional practices and the impacts of policies on adaptive capacities.
The different system trajectories identified here project the wise use matrix of Kumar et al. (2011) into future. Regarding the system’s current position, a four–fold increase in decadal catch rates and a nine–fold increase in per capita income since 2000 moves the system right of its 1990’s location on the human wellbeing axis (figure 7.3). As for the modelled future trajectories from today, all futures follow downwards trajectories as resource degradation occurs even under the most sustainable future (chapter 6.8.2). Cautionary futures pass through the ‘win-win space’ to a space where social benefits are traded against ecological sustainability, with highest per capita incomes by mid-century associated with steeper-than-safe resource degradation. Dangerous futures follow parabolic trajectories to the ‘no-go’ space of unacceptable ecological decline and wellbeing loss. The resulting ecological conditions may be inferior to the 1990s collapse, because rather than the
227 Chapter 7: Discussion
Figure 7.3: Phase space diagram depicting the trajectories of safe and just (solid green), cautionary (dotted orange) and dangerous (dashed red) futures across the wise use matrix of Kumar et al. (2011). instantaneous recovery of mobility limited stocks, overexploited stocks are found to require decadal rates of recovery even after fishing efforts suddenly cease (chapter 6.10.2).
The application of the wise use matrix here and by Kumar et al. (2011) arguably overlooks Chilika’s poorest stakeholders. Therefore, in future, the matrix could be expanded to a three-dimensional space to factor in poverty alleviation – one of the key objectives of ‘wise use’ (CDA and WI, 2010). If based on absolute outcomes, the safe and just future by 2060 would arguably sit further into the win-win quadrant due to traditional incomes increasing three-fold by 2060 under OM. However, if defined by inter-fleet income and livelihood equalities, the safe and just trajectory would reside within a three-dimensional ‘win-win-lose’ space due to the continued marginalisation of traditional livelihoods.
228 Chapter 7: Discussion
The marginalisation of traditional fishers is critical to the sustainability of practices, the efficacy of system governance and the degree to which outcomes address poverty alleviation. Incentivised by state-wide fishery commercialisation and the booming international prawn trade, non-traditional fishers gained competitive advantage during the 1980s through the use of fine-meshed synthetic nets, motorboats and lease price rises that disproportionately affected the poor (Dujovny, 2009, Nayak and Berkes, 2011). Model outcomes show that without banning motorboat uptake, as advocated by Kadekodi et al. (2000), achieving both fishery sustainability and livelihood equality remains a wicked problem. Fishers desire intensification to increase resource access and personal wellbeing (Nayak, 2014), but these unintentionally advance unsustainable fleet compositions and traditional livelihood marginalisation. Therefore, the uptake of non-traditional practices fits the paradox of innovation (OECD, 2001, Westley et al., 2011), as technological advance benefits a proportion of society before contributing to unsustainability for all.
Therefore, findings suggest that Chilika will remain a contested commons as long as access inequalities persist, thus agreeing with the assertions of Nayak et al. (2015) that Chilika’s social relations of power are key to steering towards sustainable futures. Therefore, this study echoes approaches to rejuvenate stewardship amongst and between stakeholders (Kumar et al., 2011, Nayak and Berkes, 2011), including incentives to forgo intensive practices, such as the subsidy of traditional livelihood costs (Kumar et al., 2011, Nayak, 2014). Likewise, consistent with Kadekodi and Nayampalli (2005), results advocate closer connection between traditional and non-traditional practices through the promotion of joint responsibility for the same resource.
Alternatively, rather than relying on self-organisation and the ‘stewardly nature’ of fishers (Mohanty – personal communication), various regulatory approaches may limit the abilities of users to shape the future. For instance, model outputs support banning fishing within the main tidal outlet (Sahu et al., 2014), limiting motorboat activities (Kadekodi et al., 2000) and limiting population growth to 1%/year (Kadekodi and Nayampalli, 2005). Such increased state control represents decommonisation, causing the loss of common arrangements of excludability and subtractability. Nayak and Berkes (2011) argue the erosion of traditional rights, increasing population density and the individualisation of fishing activities has steadily decommonised Chilika since the 1970s. Whilst increasing total fishers, total boats and motorboats are all associated with unsustainability, juxtaposition exists,
229 Chapter 7: Discussion as the modelled system is most resilient under increased top-down regulation that accommodates steeper and more intense fishing efforts.
The OL governance scenario aligns with the view of Dujovny (2009, p.193) that Chilika is an “organic machine” (White, 1995), with catches becoming increasingly reliant on government interventions that dampen natural variability. Prior to decommonisation, the fishery could be considered a Holocene deltaic subsystem with dynamic interplay between natural and human drivers (Renaud et al., 2013). In turn, periodic outlet dredging tips the system into an Anthropocene state, whereby outcomes are more sensitive to human processes than natural. Chilika may therefore become locked-in to a trajectory of hydraulic interventions (van Staveren and van Tatenhove, 2016), whereby de-engineering would tip the system towards collapse. This study shows that a balance exists between letting nature take its course and introducing regulations that may only detriment the livelihoods of the most vulnerable stakeholders, all the whilst building resilience against highly uncertain and nonstationary external drivers.
To this end, sustainable pathways also require a wider cultural shift away from the high value catches traditionally desired for export and income. At 1000 INR/kg, shellfish like P.monodon and S.serrata are valued approximately 8-times the average daily wage of fishers (CDA, 2015). Assuming prawn is caught at the average CPUE of 2014 and assuming fishers retain all income, a fisher capturing prawn only has to work 43 days to generate the average per capita income of INR 53,000. However, shellfish capture favours fishers travelling to central sector recruitment grounds (Kumar and Pattnaik, 2012), further marginalising relatively immobile traditional fleets which encourages ‘success to the successful’.
Positively, this study has shown that sustainable configurations exist along the continuum of decommonisation (Nayak and Berkes, 2011). Therefore, the exact pathways followed depends on the desired system identity held by decision-makers (Cumming and Collier, 2005). Governors could choose to preserve CPR rights, despite model outputs suggesting that current arrangements increase the likelihood of ‘success to the successful’ and overexploitation. Alternatively, regulating the expansion and operation of fishing efforts enhances aggregated resilience, with the loss of traditional cultures and identity becoming trade-offs under steep development trajectories. Moreover, various other unintended consequences make decommonisation unattractive. Banning motorboats isolates those requiring motors
230 Chapter 7: Discussion to access distant grounds and punishes those who have saved to be able to adapt. Likewise, should alternative livelihoods target non-traditional communities despite their contribution to fishery value, or the more populous traditional communities that have fished for centuries? As a midpoint, model results find that splitting alternative livelihoods evenly triples the frequency of sustainable futures relative to no alternative livelihoods at all.
7.2.3 Implications for the wider system
In addition to direct social, economic and ecological benefits, various secondary functions form part of the wider SES. For instance, service based revenues are gradually increasing across Chilika, with tourist arrivals doubling from 1994 to 2004 (Kumar and Pattnaik, 2012), predominantly in response to the reappearance of Irrawaddy dolphins and the restoration of the Nalabana bird sanctuary. Therefore, mechanisms to enhance fishery sustainability must avoid transferring stresses to other elements of the ecosystem and society, which may in turn feedback to degrade fishery resilience.
Globally vulnerable to habitat degradation and bycatch (IUCN, 2008), Chilika’s Irrawaddy dolphins are found in only two lagoons worldwide (CDA, 2011, Kumar and Pattnaik, 2012). The Irrawaddy population declined to ~20 individuals in 1992 in response to freshwater conditions from the lagoon-sea disconnect (Mohanty and Otta, 2008). In addition to cultural ecosystem services, including mythological connections with traditional fishers, D’Lima et al. (2014) report an association between dolphin presence and fish catch, driven by foraging behaviours chasing fish towards nets. This social-ecological co-dependency on fish availability suggests that dangerous futures found here with >80% fish stock losses threaten the future food security of the Irrawaddy population, with potential feedbacks on fishery productivity during catch declines. D’Lima et al. (2014) also find that non-traditional communities are relatively unsupportive of conservation actions that might limit catch. Thus, pathways dominated by non-traditional practices provide further ecological threats, associated with a weaker sense of co-dependency, increased water and noise pollution (not modelled here), and increased risk of dolphin bycatch (Mohanty and Otta, 2008, D’Lima et al., 2014).
Since 1972, ~300 km2 of dense forest within Chilika’s western and northern catchments has been converted to open forest, whilst ~150 km2 of agricultural land
231 Chapter 7: Discussion has undergone urbanisation (Kumar and Pattnaik, 2012). Lost runoff sinks and increased soil erosion are compounded by networks of flood defences along the Daya and Bhargavi rivers that inhibit sediment deposition upstream of Chilika (Das and Jena, 2008). Positively, this study shows that under decadal outlet maintenance, Chilika’s fishery can remain productive up to at least +30% runoff and sediment influxes by mid-century. Such reliance on engineering is common across the Mahanadi delta, with D'Souza (2006) arguing that rural systems of livelihoods and food production have transitioned from ‘flood-dependent’ to ‘flood-vulnerable’, following the loss of the annual flood pulse. Whilst results here suggest that productivity should be resilient to the hydrological effects of deforestation and urbanisation, the wider impacts of landcover change are excluded. For instance, as Chilika is managed as a ‘pollution-free catchment’ (Pattnaik – personal communication), the model excludes the potential effects of undesirable inputs (e.g. agricultural fertilisers and domestic wastes) on habitat quality. The sensitivity of the fishery to these processes may also increase as the number of agricultural farmers, reservoir aquaculturalists and livestock owners increase through alternative livelihoods.
To this end, as the model does not specify the destination of the alternative livelihoods, it must be recognised that the region’s economic sectors have different capacities to absorb livelihood needs. Discounting natural demographic additions, transferring ~6,000 fishers to agriculture would increase the population of the sector by 6% (Kumar and Pattnaik, 2012). In contrast, the same additions to Chilika’s tourism sector increases the number of members and tourist boats by 600% (Kumar and Pattnaik, 2012) – intensifying disturbances on the same biodiversity tourists pay to visit. Therefore, whilst dampening fishery stresses, undesirable feedbacks and unintended consequences may only emerge once a policy is implemented (Sterman, 2002).
The modelled alternative livelihoods are framed by compliance, as former fishers do not re-enter the lagoon. However, fishers in alternative livelihoods may continue to supplement income, questioning the efficacy of the strategy in reducing fishing efforts (Pollnac et al., 2001, Sievanen et al., 2005, Hill et al., 2012). From the perspective of traditional fishers, the modelled alternative livelihood rate fails as a safety net (Ellis, 1998), as traditional practices are still marginalised or lost under steep socioeconomic trajectories. A complex scenario confuses the targeting scheme, with traditional fishers reluctant to give-up fishing due to their deep-rooted
232 Chapter 7: Discussion identity (Nayak and Berkes, 2010) and non-traditional fishers potentially wanting to supplement income (Ellis, 1998). New fishers may enter the system as others leave (Hill et al., 2012), potentially using more exploitative gear across already overexploited grounds. Although not modelled here, Hoorweg et al. (2006) find a positive association between income diversity and the use of destructive gear across a Kenyan artisanal fishery. Therefore, a double-edged sword exists, whereby targeting traditional fishers threatens replacement by non-traditional practices, but targeting non-traditional fishers may be less effective due to their weaker sense of stewardship and ability to support multiple livelihoods within the surrounding SES.
7.2.4 Subchapter synthesis
This subchapter has discussed how this study builds upon the regional literature focusing on Chilika’s SES and its problematic behaviours. The threat of fishery overexploitation is novel in Chilika’s history, found here to emerge in part from the shift to an Anthropocene deltaic system to overcome historical problems. Despite unique social-ecological settings, patterns and interactions, unsafe pathways can be distilled to a suite of common storylines. At an aggregated level, dangerous futures emulate tragedy of the commons, whereby unchecked intensification overwhelms resource renewal. Yet this work suggests that the causal pathways are more nuanced: short-term success to (relatively) successful traditional fishers can trigger nonlinear fishing effort growth, establishing a legacy of overexploitation and access inequality, which builds vulnerability to fishery overcapacity by mid-century.
Of the normative futures, SJOSs correspond to the system’s wisest use in terms of sustainable development and wider social-ecological outcomes. The sustainability of the different socioeconomic pathways is critically linked to the balance of resource access, with relatively wealthy traditional and non-traditional fishers holding leverage over the system’s future. Consequently, the spectrum of options sits between: (i) committing Chilika to an organic machine through decommonisation by internalising system responses to the spectrum of external trajectories, but potentially diluting traditional rights, or (ii) preserving traditional practices through initiatives of stewardship to anchor socioeconomic trajectories, albeit potentially trading-off individual adaptive capacities. Whichever route is chosen, decision- makers must remember that Chilika is more than a fishery, with potential consequences for the economically and culturally valuable dolphin population, catchment pollution inputs and landuse changes.
233 Chapter 7: Discussion
7.3 Governing regional social-ecological systems
7.3.1 Systems of leverage points
A key consideration emerging from this study is whether governance should continue business-as-usual with a focus on the limited SJOSs, or whether governance and overarching social-ecological functions should transform to widen and deepen the driver-based SJOSs. Defined as internal points of power shaping system dynamics, the theory of leverage points is central to this challenge (Meadows, 1999). Yet, finding where to intervene may be counterintuitive, labour intensive and risk steering a system in an unintended direction.
This study generally supports the leverage point rankings of Meadows (1999, 2009), which increase in influence from parametric adjustments to paradigm shifts, like switching from desiring maximum profits to multidecadal persistence. The deepest leverage point is arguably the realisation that sustainable catch is desirable in future, which promotes safe and just pathways and the targeting of sustainable catch over indefinite intensification. The CDA and CIFRI already incorporate MSY into decision-making (Kumar and Pattnaik, 2012, ICAR-CIFRI, 2016); however, as the fishery intensifies, governance must not interpret returning to MSY after persistent periods above MSY as a drift to lower performance (Meadows, 2009). Therefore, recognising that sustainability is social-ecologically desirable is not a silver bullet, but instead a precursor for further leverage point analysis (Gallopín, 2002), including mechanisms that build resilience against unintended consequences of governance aiming to enhance sustainability (e.g. decadal tidal outlet opening).
The shallowest leverage points may only delay problematic behaviours: linear motorboat uptake may still cause dangerous futures under relatively steep fisher population growth. The importance of parametric adjustments increases when deeper leverage points trigger as a result (Meadows, 2009). Catch declines until ~2030 are buffered by the post-ecorestoration fish stock, which simultaneously facilitates fishing effort growth that eventually causes dangerous outcomes. Therefore, a stabilising, virtuous stock can become destabilising if its size promotes unsustainable dynamics, such as feedbacks driving ‘success to the successful’. Consequently, leverage points should not be viewed in isolation (Abson et al., 2017), but within a system whereby the effectiveness of lever ‘A’ changes with the evolution of leverage points ‘B’, ‘C’ and ‘D’.
234 Chapter 7: Discussion
Moving up the leverage point classification of Meadows (2009), fast drivers are often the main concern for users and decision-makers (Walker et al., 2012), potentially stemming from catastrophes like earthquakes with positively skewed magnitude-frequency relationships (Rudolph and Repenning, 2002, Adger, 2006). In contrast, slow processes are often undetectable or ignored, but may magnify short- term stresses that require stronger remediation (Biggs et al., 2009, Biggs et al., 2015a). Here, the interplay between the system of fast and slow processes is complex. Although nonlinear motorboat increases by 2030 are symptomatic of dangerous futures, the steepest trajectories of population growth are also required to force Chilika beyond the normative cautionary space. Therefore, consistent with Anthropocene deltas, the controlling slow variable shifts from ecological to human, occurring once the ecosystem is stable and motorboat nonlinearities have shocked the system. Chilika is thus becoming a ‘novel’ ecosystem (Collier, 2015) because (i) historically insignificant stresses are growing in importance, and (ii) the system cannot return to its original ecological status without causing dangerous futures (e.g. following NO governance)
Sustainable outputs under OL support keeping socioeconomic stresses as slow as possible. Whilst OM and OB governance scenarios principally adjust stocks and flows to rebalance resource renewal and extraction, OL governance shifts the underlying goals of the system from supporting its dependants to a system that adapts drivers to the resilience of the resource (Berkes, 2011). These deeper leverage points overcome vulnerabilities to fast processes and displace slow processes below dangerous levels to sustain system functions.
This subchapter has built on the discussion of system archetypes to analyse the links between leverage points, governance and sustainability. Ultimately, interacting levers shape progress towards desirable and undesirable consequences over time. Absence of a silver bullet emphasises the usefulness of interacting, multidimensional SJOSs to guide systems towards sustainable futures. Next, the findings uncovered here are compared to other fishery systems to understand the prevalence of challenges found here and whether the system of leverage points fit the general frameworks of fishery sustainability.
235 Chapter 7: Discussion
7.3.2 Sustainable fishery governance
Fishery systems are hotspots of social-ecological challenges, involving appropriate resource use, different user rights and combinations of human and natural stresses (Bailey and Jentoft, 1990, Berkes, 2011). Sustainability principles for fisheries based on these characteristics emphasise the necessity of viewing fisheries as SESs (Berkes, 2011, Kittinger et al., 2013, Lade et al., 2015), differing from the traditional paradigm of biophysical and social disconnection, built around the perception that natural variability influences human uses but not vice versa (Berkes, 2011).
Consistent with the conclusions of Lade et al. (2015) for the Baltic Sea fishery, this study shows that the SES narrative is critical to the sustainability of Chilika. Whilst stable ecosystem dynamics fundamentally open sustainable pathways, only relying on this conventional wisdom would ignore the nonstationary social responses that have contributed to fishery failures across developing nations (FAO, 1984, Bailey and Jentoft, 1990). Consequently, linear forecasts may overestimate the sustainability of contemporary trends in the face of globalisation and environmental change (chapter 7.9). Similarly, externalising fishing efforts by varying initial boat numbers may overestimate immediate stock losses (e.g. Bueno and Basurto, 2009, Martins et al., 2015), destabilising the systems earlier than if efforts had increased from today.
A second key theme in fishery governance is the importance of cross-scale interactions (Kittinger et al., 2013, Nayak and Berkes, 2014). Traditionally, cross- scale interactions have been spatial, such as migrant fisher entries or fleets of highly mobile roving bandits that exploit fish stocks across geographical boundaries (Berkes et al., 2006). This study illustrates how regional fisheries are directly and indirectly vulnerable to temporally cross-scale processes, thus explicating the need for fishery governors to build resilience against creeping rates of fish price and nonstationary annual rainfalls, respectively. Alternatively, modelling fish price by regional supply and demand may have overestimated resilience through a balancing feedback, whereby high productivity causes oversupply relative to domestic demand, which lower prices and so forth. In reality, exports expose Chilika to national and global consumption demands which have tripled and doubled since the 1960s, respectively (FAO, 2015). Exporters demand high prices from burgeoning markets, meaning fishers are triply trapped by export demand, regional food security and livelihood benefits. Elements of this trap
236 Chapter 7: Discussion interact, with exports undermining regional food security and population growth reducing CPUF. Similar dilemmas are found in East African reef fisheries (Cinner, 2011, Cinner et al., 2011), where the poorest are trapped into cheap destructive gears in order to maintain livelihoods.
Coping with cross-scale interactions, feedbacks and surprises is shown to be central to sustainable governance even within the SJOS. Berkes and Seixas (2005) develop a general resilience framework (italicised) from five small-scale fisheries. Learning from past crises, Chilika’s decision-makers have committed to periodic outlet opening to stabilise ecological conditions and resource mobility. Consequently, Chilika must learn to live with change and uncertainty stemming from both inherent (e.g. freshwater inputs) and induced variability (e.g. rejuvenated fish stocks). Decision-makers must thus consider diversifying the portfolio of coping strategies, from official fishery regulations to the stimulation of ecological stewardship. However, this requires a shift in ecological memory from a perceived dependence on resource inputs to the importance of extraction. Likewise, the view that aquatic vegetation degrades catch by blocking fishing grounds (Rajawat et al., 2007) must be replaced by an appreciation of its buffering role against overexploitation.
These social-ecological perspectives also require new social memory (Folke et al., 2002), like suppressing the legacy that humans do not affect resource availability. However, building trust between users may be problematic for systems like Chilika with histories of inequality and conflict (Berkes and Seixas, 2005). To overcome this sticking point, Berkes and Seixas (2005) recommend greater integration between fishing communities through the involvement of fishers in environmental monitoring, the sharing of information between scientists and user groups, and the co-development of case-specific sustainability principles. Decision-makers may resultantly reduce dependence on regulations limiting variability, but may still fail to address wider archetypes of unsustainability, including overcapacity, limits to growth and success to the successful.
In relation, the use of MSY here in the definition of safe and just might be argued to contradict resilient fishery governance (Larkin, 1977, Punt and Smith, 2001). Aside from the data uncertainties in MSY estimation (chapter 7.6), MSY is thought to detrimentally impose stability on an inherently variable resource. However, this study aligns with the ‘new definitions of sustainability’ argument of Andrew et al.
237 Chapter 7: Discussion
(2007, p.234), that catches above MSY are permissible as long as degradation is reversible. Therefore, this study applies MSY differently from conventional fishery management. Rather than a permanent level of productivity to avoid wasting food or efficiency (Larkin, 1977), setting MSY as a future target facilitates short-term variability and opportunities for self-organised resilience. Therefore, sustainability is not measured by the short-term achievement of MSY, but by whether MSY remains an option in future; oppositely, normatively and nonlinearly dangerous futures may not achieve MSY without substantial delays and social-ecological hysteresis. The 2050-2060 horizon is thus a checkpoint towards longer-term sustainability, whereby rerunning the analysis in ten years for 2060-2070 creates a moving window of sustainability.
According to Andrew et al. (2007), setting a desirable outcome operationalises the first principle of sustainable fishery governance. This modelling framework then provides a diagnosis for sustainability by giving decision-makers options to steer the system (second principle - Andrew et al., 2007), akin to a satellite-navigation system guiding a vehicle driver to their desired destination. Evaluating trade-offs associated with future trajectories represents the third principle (Andrew et al., 2007); failing this reflection stage might lead to the perception that sustainable socioeconomic growth is slower than desirable, which in turn might promote the pursuit of unsustainable pathways.
This study supports the growing body of literature calling for fishery systems to be considered as SESs. In turn, cross-scale and interacting social and ecological processes recognise both direct (e.g. fishing effort limits) and indirect (e.g. positive effects of macrophyte coverage) leverage points for sustainability. The next section broadens the scope of discussion to wider SESs thinking; further contextualising sustainability challenges and evaluating the actions needed to operationalise sustainable fishery governance.
7.3.3 The roles of resilience, adaptability and transformability
Chilika’s decision-makers have traditionally focused on resource predictability and productivity, achieved by stabilising resource mobility and the number of resource units. Whilst NO governance shows that resource mobility remains critical in future (chapter 7.2), model outputs also argue that sustainability is linked to numerous second-level variables of the social-ecological systems framework (Ostrom, 2009),
238 Chapter 7: Discussion including the number of users, harvests, and socioeconomic self-organisation. This web of drivers expands once operational rules are introduced under alternative governance scenarios.
Chilika is therefore becoming a more complex system to steer. Positively, this increases the number of entry points for decision-makers, ranging from delaying problematic behaviours through fishing bans to adjusting creeping causes of unsustainability through alternative livelihoods. The ability of resource users to shape sustainability suggests an increased need for participation in decision-making. Chilika therefore stands at a breakpoint in the long-now (Brand, 1999), with decision-making based only on historical legacies threatening to tip the system towards undesirable social-ecological states (Carpenter, 2002).
With these challenges in mind, this study argues that the concepts of specified and general resilience are not mutually exclusive (chapter 2.2.1) (Walker and Abel, 2002). Measures of specified resilience against individual drivers (chapter 6.6) derive from resilient system-wide interactions; thus, pathways such as the core SJOS provide an approach to target individual drivers without disregarding or degrading less important components. Moreover, SJOSs have trade-offs which require resilient support networks; for instance, avoiding unsafe motorboat numbers requires fishers to forgo possible catch improvements, which may only be possible with the trust that others behave similarly. Weak general resilience may fail to generate sustainable self-organisation needed to follow safe trajectories, simultaneously failing marginalised fishers even when outcomes appear safe. More widely, social- ecological resilience requires reorganisation and learning, based on the resilience of experience sharing, knowledge creation and conflict resolution (Folke, 2006, Folke et al., 2010). Thus, this study provides a reminder that general resilience is intrinsically linked to specific trade-offs, inequalities and unintended consequences.
This study also argues that adaptability, defined as the ability of collective actions to shape resilience (Walker et al., 2004), is not necessarily a precursor for transformative sustainability. Adaptability may actually undermine both general and specified system resilience, with relatively wealthy traditional fishers deepening the poverty trap of the poorest. Therefore, building social resilience must utilise strategies that avoid pushing SESs closer to their environmental ceilings, including programmes of public participation, efforts to increase the attractiveness and coping strategies of traditional livelihoods, and methods to increase interest in
239 Chapter 7: Discussion
Figure 7.4: Qualitative scenarios from today’s window of opportunity summarising the adaptability, transformability and stability landscapes of the plausible futures. alternative livelihoods (Lebel et al., 2006, Marshall and Marshall, 2007). It is therefore argued that this discovery of potential problems and solutions opens a window of opportunity for Chilika (figure 7.4). Moreover, the potential trajectories of climate change and international markets encourage transformation whilst the system’s state is currently desirable and resilient – at least relative to the historical collapse and unsafe plausible futures.
To this end, the driver pathways, governance strategies and leverage points can be synthesised into five scenarios relating resilience, adaptability and transformability (figure 7.4). Under OM and OB, stakeholder adaptations may lead to the undesirable basin (figure 7.4A). The system is first forced to the edge of the desirable basin by nonlinear fishing efforts, before the creeping effects of slow parameters like fisher population growth and long-term fish stock depletion develop the undesirable basin. Cautionary futures are similarly forced from the desirable basin (figure 7.4B), but the relative maintenance of traditional livelihoods and slow population growth do not cause sufficient parameter changes to tip the system (Beisner et al., 2003).
240 Chapter 7: Discussion
Safe pathways may be driven by keeping driver levels sustainable from the bottom- up (figure 7.4D). These findings support the need for ‘good governance’ that focuses on social-ecological resilience (Lebel et al., 2006, p.2), including harbouring equality between social groups, social learning and stakeholder empowerment (Lebel et al., 2006, Folke et al., 2010, Lebel et al., 2010). However, system state remains vulnerable to abrupt fishery intensification, hence the shallow desirable basin. The OL scenario combines top-down and bottom-up governance, as managed realignment of fishing effort reduces the precariousness of individual drivers to their thresholds, enhances general resilience to create a deep desirable basin (figure 7.4E) and facilitates sustainable adaptability up to all but the steepest socioeconomic trajectories (figure 7.4C).
7.3.4 Subchapter synthesis
This chapter has evaluated findings with respect to sustainable and resilient governance. In synthesis, remaining within SJOSs requires specific and general resilience, whereby drivers of unsustainability can be targeted without knock-on effects. Therefore, a resilient system accommodates sustainable variability, where users can adapt without undermining resilience elsewhere in the system. Rather than waiting for the first signs of unsustainability, decision-makers should build the resilience of viability loops that may be later overwhelmed by surprising nonlinearities deriving from previously sustainable actions. However, steering Chilika is a complex task even within this model’s structure, requiring a diverse portfolio of leverage points including a combination of traditional engineering strategies and softer approaches promoting traditional livelihoods, self-organisation and human-natural connectedness. These leverage points are found to change in number and type as a function of time and previous interventions, which potentially causes misconceptions and rigidity. Therefore, decision-makers cannot rest on their laurels, because although a system may have recently recovered, understanding of novel stresses, outcomes and leverage points must be continuously updated to avoid sleepwalking off the sustainability cliff.
241 Chapter 7: Discussion
7.4 Undesirable shifts in social-ecological systems
7.4.1 Thresholds, tipping points and regime shifts
Building on the concepts of sustainability, adaptive capacity and transformability, this study uses the paradigm of SJOSs to identify futures that support neither human needs nor ecosystem functions (Raworth, 2012, Dearing et al., 2014). In turn, the SJOS concept develops the planetary boundaries framework (Rockström et al., 2009), which itself developed earlier ideas around ‘safe minimum standards’ where nonlinearities and thresholds were integral (Ciriacy-Wantrup, 1952). Consequently, this study raises questions about the coherence between SJOSs and wider resilience thinking, like whether exiting a normative SJOS can be considered a tipping point and to what extent dangerous futures constitute regime shifts?
Chilika’s shift from safe to dangerous futures has characteristics of parameter- driven and variable-driven alternative stable states (Scheffer et al., 2001, Beisner et al., 2003). Expressed through the canoe metaphor (chapter 2.3.1), Chilika’s canoe has been relatively resilient to small-scale stresses since ecorestoration, including above average monsoonal freshwater inputs and motorboat uptake (figure 7.5A). In future, steep fishery value growth acts as an external shock – like waves hitting the canoe. Internal system responses may further destabilisation, with nonlinear motorboat uptake causing non-analogous catch. The socioeconomic development resulting from mistimed outlet opening is akin to a canoeist shifting position to regain control against external stresses, but unintentionally destabilising the system whilst doing so (figure 7.5B). These fast and slow stresses combine to erode system resilience and cause an alternative state (figure 7.5C), unsupportive of human and ecological functions, and internally uncontrollable without significant reversal efforts. Therefore, consistent with the fundamental functions of the canoe to safely transport its occupants, dangerous futures do not support the livelihood requirements of its dependants.
It is important to reemphasise that the entire trajectories of dangerous futures are classified as such to depict continuous pathways from today. Therefore, despite parallels to the canoe metaphor, the extents to which the different thresholds ([i] conditional probabilities, [ii] variability based and [iii] nonlinear driver-response relationships) constitute social-ecological tipping points vary. Overall, the different
242 Chapter 7: Discussion
Figure 7.5: Shift from a desirable to an undesirable basin under OM and OB governance: (A) a resilient system varies within its desirable basin during unremarkable disturbances (D ); (B) outcome variability norm increases with stress magnitudes, pushing the system into non- analogous regions of the basin; (C) slow parameter changes encourage basin shallowing, reducing outcome resistance to internal variability. thresholds represent critical values that once transgressed associate with a state of degraded social-ecological functions (Renaud et al., 2013) and/or nonlinear driver- response trajectories (Cairns Jr, 2004, Lenton et al., 2008). Thus, under the latter definition, it may be argued that tipping points are absent across the linear phase- spaces (e.g. fish population versus total fishers) and normative futures. Countering this argument, switching from predominantly cautionary to dangerous pathways may only require slightly steepened driver trajectories. For example, steepening the motorboat trajectory by 13% (8000 9000 motorboats) by 2050 under OM collapses the probability of safe and just outcomes from 50% to <5% (figure 6.10). Such disproportionality between driver and outcome state is consistent with the concept of tipping points (Groffman et al., 2006); depicting the SJOSs as conditional probabilities acknowledges the difficulty of defining deterministic futures for
243 Chapter 7: Discussion inherently complex systems. Therefore, only governing with respects to driver- response nonlinearities or unpredictable outcomes might overlook the creeping trajectories undermining resilience against ecological degradation, livelihood losses and human migration.
Normative futures have additional characteristics consistent with regime shifts, including: (i) social-ecologically unsustainable conditions (Cairns Jr, 2004), (ii) threshold values that respond to governance (Walker and Meyers, 2004), and (iii) new social experiences, like the cascading effects of livelihood loss (Kinzig et al., 2006, Christensen and Krogman, 2012). Dangerous futures also align with the criteria of the Regime Shift Database (Biggs et al., 2015c), whereby inadequate ecosystem service provisions and reinforcing feedbacks cause social-ecological reorganisation through impacts on human wellbeing. However, these interpretations are less applicable to cautionary thresholds that do not necessarily reorganise system functions, create new social experiences or degrade human wellbeing.
Consistent with Steffen et al. (2015b), driver-response nonlinearities between the number of motorboats and the fish population are found to sit within the cautionary zone of increased risk en route to dangerous futures. In turn, reversing ecological degradation caused by the steepest driver trajectories is hysteretic – as hypothesised by type-III SJOS boundaries (Dearing et al., 2014). Therefore, it is questionable whether normative cautionary futures represent window of opportunities for remedial governance, as driver-response nonlinearities may accelerate resource degradation before the normative dangerous spaces are reached. Developing on Biggs et al. (2009) who only assess one-at-a-time drivers, it would be interesting to understand how multivariable interactions effect the aversion of a regime shift beyond the forward tipping point. For example, although Biggs et al. (2009) find management can avert shifts from fast drivers up to 10 years into the transition, how might the concurrent creep of population growth over that time resist system reversal?
In relation, driver-responses that enter the cautionary zone as linear trends may exhibit hysteretic reversals due to multivariable interactions and governance affects, thus questioning the conventional linear, nonlinear and hysteretic driver-response relationships (Scheffer et al., 2001, Sterk et al., 2017). This highlights the importance of monitoring multiple and appropriate driver-response relationships, as
244 Chapter 7: Discussion
Figure 7.6: Hysteretic trajectories coloured by a composite metric summing the standardised temporal and driver reversal dimensions (figure 6.23– inset). Red colours represent trajectories where the number of motorboats are reversed relatively rapidly per unit time. The forward shift is removed for visual clarity. only tracking the fish population versus fisher population trajectory might mistakenly lead governors to believe that recovery always follows the same trajectory as collapse (figure 6.20). More widely, Blackwood et al. (2012) asks whether the need to reduce or eliminate fishing efforts to avoid hysteresis can be based on degradation severity. This study affirms that steeper fishing effort growth produces a more resilient (and undesirable) alternative state from 2060; alternatively, relatively shallow resource degradation avoids hysteresis despite normatively dangerous conditions.
These findings also suggest recovery lags stress remediation both temporally and within phase space, building on work that assesses time and driver reversal magnitude in isolation (Blackwood et al., 2012, Nyström et al., 2012, Sugiarto et al., 2015). Critically, the trade-off between the two dimensions questions whether
245 Chapter 7: Discussion decision-makers should prioritise rapid ecological restoration or socioeconomic preservation. A composite metric combining hysteresis duration and reversal magnitude identifies a sweet spot between trajectories that require many years but retain livelihoods, and those that are abrupt but socioeconomically costly (figure 7.6). Interestingly, two trajectories with high efficiency appear to follow trapped-like trajectories (chapter 6.10.2), coherent with the highly responsive dynamics which undergo a second collapse by 2100. In all, these finding show that normative futures share various characteristics with social-ecological thresholds, tipping points and regime shifts. Consequently, this work questions the distinct depictions of the four different types of SJOSs (Dearing et al., 2014), as leaving a normatively defined safe space may be followed by nonlinear dynamics if remedial actions are not implemented in time.
7.4.2 Operationalising threshold-based management through safe and just operating spaces
The previous subchapter discussed various consistencies between normative SJOSs and theories of regime shifts in SESs. Moving from the conceptual aspects, a key contribution of this study is the use of SJOSs to operationalise the setting of target conditions (Kelly et al., 2015), multidimensional safe spaces (Selkoe et al., 2015) and the identification of safe spaces away from thresholds (Biggs et al., 2015b).
Setting appropriate target conditions arguably forms the foundation of TBM, because without a destination, we cannot hope to delineate boundaries. However, for SESs vulnerable to undesirable outcomes, decision-makers may struggle deciding between targeting desirable outcome conditions and precautionary principles keeping the system away from undesirable outcomes (Lempert and Collins, 2007). Dearing et al. (2014) and Levin et al. (2015) agree that targets should be social-ecologically compatible and achievable given today’s norms and baselines, making the definition of ‘desirable’ iterative; for instance, a desirable future for Chilika during the 1990s may have simply overcome the livelihood challenges of fishery collapse. Alternatively, whilst potentially bringing substantial socioeconomic benefits, opting for industrial-scale aquaculture would have been ecologically unsafe. Thus, aiming to average MSY catch from 2050-2060 seeks to balance the lagoon’s ecological and anthropogenic functions, whilst keeping alternative governance strategies open (Rosenhead, 2013). This approach to modelling SJOS includes iteration as a strategic element, in case new information
246 Chapter 7: Discussion makes the previous SJOS definition less desirable. It would be straightforward to recast towards alternative futures in light of new information, such as decision- makers suddenly desiring ‘optimum yield’ following an economic transition away from fishing (Levin, 2014). Abrupt biophysical trends may also be incorporated, such as suddenly accelerated nutrient fluxes.
Directly parallel to regionalising the planetary boundaries, the need for spatiotemporal coherence between governance and thresholds is the second key element of TBM (Foley et al., 2015). Whilst decadal outlet maintenance might be perceived to balance intra-annual resource mobility and multidecadal sedimentation, model outputs suggest mismatch exists with interannual fishery intensification. Therefore, the safe and just pathways here incorporate different temporal horizons to capture the perfect storm of fast and slow drivers (Beddington, 2009), helping to avoid both near-term and longer-term departures from sustainability: akin to combining the short-term safe landing approach (Alcamo and Kreileman, 1996) and the centennial tolerable windows approach (Petschel-Held et al., 1999) when modelling greenhouse gas scenarios.
Debate also exists over whether TBM should remediate drivers most at risk (Liu et al., 2015b) or those associated with the wickedest trade-offs (Kelly et al., 2015). For instance, motorboat numbers may become rapidly unsafe, but outlawing their use might push the practice underground – potentially prolonging ecological degradation and widening social divisions. An alternative view might prioritise cascades where the crossing of one threshold causes knock-on effects, like resource declines leading to the emergence of unsustainable juvenile catch. This also applies to the multidimensional core, where leaving the core motorboat pathway under OM increases the chances that all other drivers will also be unsafe.
Multidimensionality here refers to multiple interacting human and natural drivers (Lade et al., 2013), discouraging decision-makers from focusing on one driver without understanding wider contributions and feedbacks. This study’s SJOSs are additionally multidimensional because outcome sustainability is dependent on both driver trajectories and interactions, producing soft thresholds that build on the conventionally hard and static SJOSs (Cole et al., 2014, Hossain et al., 2017). These non-binary thresholds align the SJOS concept with multidimensional notions of resilience. The range of safe and just trajectories represent resilience latitude, denoting the range of driver trajectories that lead to desirable outcomes (Walker et
247 Chapter 7: Discussion al., 2004). In turn, system resistance is the rate at which safety changes over the driver range. Therefore, a deep driver SJOS indicates attraction to sustainable outcomes over the resilience latitude of the driver (e.g. OB motorboats), but a steep conditional probability gradient means that the system offers little resistance to slight variations in the trajectory of that driver. These interpretations are in line with theories from the Baltic Sea cod fishery, whereby latitude and resistance combine to form the horizontal component of resilience (Vasilakopoulos and Marshall, 2015). Future work may seek to formalise how system precariousness changes over time based on the model defined SJOSs, for instance, the product of the driver’s distance from its SJOS edge and its SJOS conditional probability.
In contrast, traditional binary thresholds only distinguish between trajectories inside and outside the SJOS (Hossain et al., 2017), with little understanding of proximity to the SJOS edge or the presence of a cautionary window. Addressing the challenges of following desirable versus precautionary pathways outlined above, keeping threshold-vulnerable processes within their core SJOS provides a novel method of setting safe limits away from thresholds (Biggs et al., 2015b). Moreover, the core SJOS helps to operationalise the ‘thresholds of potential concern’ used in South Africa’s Kruger Park (Biggs et al., 2015b), where transgressing a ‘worry-level’ prompts meetings between scientists and decision-makers. Thus, core and wider SJOSs facilitate tiered governance (Selkoe et al., 2015), whereby management needs relate to a driver’s position within its SJOS.
Despite the theoretical developments outlined above, time-dependent SJOSs conflict the time-independence philosophy of TBM (Kelly et al., 2015). However, time has been a fundamental component of SJOSs ever since their conception, with retrospective studies basing SJOS definitions on time-series of 10 (Cole et al., 2014) and 100 years (Dearing et al., 2014). This study supports coupling time and system dynamics as dimensions of SJOSs by establishing an explicit future for decision- makers to head towards, and if necessary, remediate trajectories before. System dynamics underpinning progression towards the target are then incrementally measurable (Kelly et al., 2015, Bai et al., 2016), whilst the temporal signature of the outcome can be viewed if the SJOS is missed.
Kelly et al. (2015) argue that TBM increases management efficiency. Whilst this study supports the relative efficiency of SJOSs over managing hysteretic reversals, the broad statement underplays trade-offs associated with sustainability. Business-
248 Chapter 7: Discussion as-usual governance may reap high rewards under cautionary futures, whilst governance either widening SJOSs or keeping drivers within limits is associated with financial trade-offs, decommonisation and issues of equity. Similarly, the degree to which this study constitutes proactive TBM is questionable, given that perceptions of unsustainability might not exist without the 1990s collapse. Countering this, the modelled governance is proactive by recognising that SESs are complex adaptive systems, whereby legacies of path thresholds produce new thresholds. Therefore, whilst Chilika’s future could be modelled without the 1990s collapse ever occurring, the resulting model would potentially exclude the fisher adaptations that emerged due to resource degradation (e.g. dynamic fishing days and catch selectivity).
Overall, this study has operationalised multiple concepts of TBM within a modelling environment, including pathways to inform tiered governance, the use of future targets and concepts of social-ecological resilience. However, all of these contributions are built on the assumption that averaging MSY from 2050-2060 is social-ecologically desirable and symptomatic of a resilient system. Therefore, akin to the suite of social foundations analysed by empirical studies (Raworth, 2012, Cole et al., 2014, Dearing et al., 2014), it is important to discuss the wider implications of these desirable futures to ascertain the extent to which they can be considered safe and just.
7.4.3 How just are safe and just operating spaces?
The term ‘just’ in SJOSs exposes the concept to debates associated with sustainable development, including definitions of successful sustainable development, needs versus rights, and intra- and inter-generational justice (Hopwood et al., 2005, Holden et al., 2014). Therefore, the assertion of this study that the normative future of averaging MSY catch from 2050-2060 is safe and just must be scrutinised to evaluate the degree to which SJOSs facilitate sustainable development and associated challenges.
The notions of justness in the original framework (Raworth, 2012) encompass conditions of well, productive and empowered social priorities. Of the 11 subcategories, this study directly explores the income (well), employment (productive) and resilience (productive) of Chilika’s stakeholders, finding that SJOSs generally produce inter-generational justness. However, Daly (1993) argues that sustainability prioritises quality over quantity, where equity between stakeholders is
249 Chapter 7: Discussion the foundational component. To this end, this work shows that decision-makers must be wary that normatively defined safe and just futures do not necessitate equity in income, livelihood opportunities and social-ecological resilience.
Can such unequal conditions be considered just sustainable development? The question may be answered ‘no’ because the poorest today remain poorest in future, aligning with the argument that resilience concepts exist merely to maintain the status quo (Blühdorn, 2007, Kopnina, 2016). This viewpoint is supported by system resilience under OL, because as per Hardin (1974), traditional fishers must either give up their livelihood or altogether leave the fishery ‘lifeboat’ as relatively wealthy fishers dominate resource consumption. Therefore, relaxed population growth under OL maintains system resistance and latitude, however, improved aggregated resilience masks deeper injustices. These trade-offs emphasise the complex challenges of socially just sustainable development (Dobson, 2000, Raworth, 2012): a more resilient outcome state is associated with the loss of traditional livelihoods; however, the remaining livelihoods are wealthier on average than under less regulatory governance.
Conversely, the earlier question of justness may be answered ‘yes’ if sustainability aims to persist today’s aggregated conditions (Redclift, 2005), or at least not make them worse in the face of accelerating biophysical and socioeconomic stresses. Providing disparities are less than those occurring under different social settings (e.g. fishery collapse), accepting that inequalities are inherent features of sustainable development fits the controversial ‘difference principle’ of Rawls (1999). Moreover, as catch sources are currently unrecorded (i.e. traditional or motorboats), decision-makers might overlook the disappearance of traditional livelihoods as long as the lagoon remains productive, particularly given that the switch to non- traditional practices is positively viewed through the perspectives of social mobility and adaptive capacities.
To address some of these challenges and provide a more holistic assessment of justness, the criteria judging future desirability could be changed. For example, establishing a desirable CPUE would explore whether fishers are receiving just returns for their financial and time outlays. CPUE declines when the volume of active fishing gear per unit time increases, even if catch already greatly exceeds sustainable levels. These interactions between catch and fishing effort in the CPUE calculation are known to overestimate abundance due to a phenomenon known as
250 Chapter 7: Discussion
‘hyperstability’ (Harley et al., 2001), meaning changes in stock abundance may not be noticed until too late. Setting today’s CPUE of ~7 kg/boat-days as the desirable target due to its present association with sustainable catch would reduce the number of futures considered ‘safe and just’ (chapter 6.8.2), with remediation to improve CPUE needing ecologically unsafe catch intensification, or, potentially unjust fisher removals to reduce fishing efforts. Consequently, the transient nature of CPUE makes it difficult to establish a future norm in the face of nonstationary driving conditions, which includes the dependant fisher population.
The equitable share of resources up until the sustainable catch level is a challenge customary to common-pool resources. Additional targets may supplement the use of sustainable catch, akin to the multiple social foundations of Raworth (2012), such as tracking whether traditional income averages below the international poverty line of US$1.90/day. Applying this filter reduces the number of safe and just futures by 60%, 45% and 25% for OM, OB and OL respectively, suggesting that achieving MSY from 2050-2060 may overstate social justness. The wider social- ecological outcomes of chapter 6.8 may form additional filters; their ranking from least to most just helps expose trade-offs of shallow fishing effort growth, for example (figure 7.7). Environmental ceilings like resource availability, mobility and juvenile abundance could also be ranked to avoid anthropocentric views of sustainability (Fischer et al., 2007, Kopnina, 2016). Decision-makers would then be able to choose pathways from the different outcomes or a combined metric, rather than considering all futures achieving MSY as safe and just. However, Holden et al. (2014, 2017) counter this strategy by arguing that multiple targets can equate to no targets at all, especially when underperformance (e.g. traditional income) can be offset against overperformance (e.g. fish population). Therefore, as a first pass, this study portrays pathways to a future with baseline sustainability where intra-state variance is less important than the persistence of desirable social-ecological functions (i.e. inter-state variance).
It is important to note that limited inferences are available into wider aspects of sustainable development like social voice and food availability, whilst issues of education and gender equality are regrettably excluded. Therefore, future modelling efforts or decision-maker engagement may wish to investigate whether sustainable catch would boost regional food availability, given the tendency of governors to intensify exports (Kadekodi et al., 2000); likewise, whether fisher population growth improves the involvement of women in fishing. Positively however, the model
251 Chapter 7: Discussion
Figure 7.7: (Top) annual fish catch and (bottom) motorboat use ranked by the average per capita income (INR/year) associated with each simulation. Please refer to chapter 6 for historical dynamics. recognises futures likely to suppress equality and create voiceless communities. Therefore, if based on the sustainability of a contested natural resource, futures that do not significantly degrade the equality and equity of the poorest might be deemed safe and relatively just, particularly when compared to futures associated with social-ecological collapse for all. This addition confronts the idea that SJOSs are universally just, shown here to be constrained by historically entrenched archetypes, feedbacks and legacies.
252 Chapter 7: Discussion
7.4.4 Subchapter synthesis
This research aimed to understand the extent to which the concept of future SJOSs is useful for governing with respects to thresholds and tipping points, assessing social-ecological trade-offs, and guiding sustainable development. Overall, this study has shown that decision-makers must be aware that transgressing linear SJOSs may lead to temporal and inter-driver nonlinearities, social reorganisation and hysteresis. These occur beyond the normative SJOSs, thus recommending that figure 2 of Dearing et al. (2014) is updated to include multiple types of dynamic behaviours, as opposed to their current isolated arrangement. Similarly, figure 1C of Scheffer et al. (2001) may be updated to include asymmetric hysteresis trajectories resulting from (i) linear driver-response trajectories (figure 6.21B), and (ii) recovery trajectories dependent on the magnitude of resource degradation during the forward shift and system responsiveness and resistance during reverse. Moreover, by appreciating driver multifinality and equifinality, this work has developed the hard boundaries that traditionally classify driver scenarios as either safe or unsafe. Therefore, the SJOSs defined here track safe space latitude, resistance and precariousness to unsafe conditions, and the core SJOSs facilitate tiered governance through the recognition of multidimensional interactions leading to the deepest regions of desirable space.
Moreover, although MSY is inter-generationally just at an aggregated level, intra- generational justness between fleets cannot be guaranteed. Counterintuitively, least just conditions occur when the system is most resilient, as the system becomes robust to inequalities associated with unsustainable pathways under less regulatory governance. Therefore, the SJOSs here locate pathways to avoid aggregated losses, but future efforts are needed to ensure Chilika’s marginalised dependants also remain within the livelihood lifeboat.
7.5 Modelling social-ecological systems
This study has developed a modelling framework based around understanding problematic behaviours, concerns of stakeholders and experts, and interacting pathways reaching different normative futures. Whilst a problem-based focus is common amongst system dynamics studies, this work advances a number of SDM and conventional modelling techniques through the integration of coupled natural and human interactions (Verburg et al., 2016), emergent rather than modeller
253 Chapter 7: Discussion defined problematic behaviours and the use of exploratory external scenarios to uncover threshold locations. This last characteristic is consistent with the rising popularity of exploratory scenarios relative to traditional ‘best-guess’ futures (Maier et al., 2016), which aim to explore highly divergent external drivers, deeply uncertain processes and multiple governance paradigms.
7.5.1 Using models as satellite-navigation systems
Assessing the quality of deeply uncertain Anthropocene futures is a key challenge for decision-makers (Berkhout, 2014, Bai et al., 2016), due in part to the volatile and ambiguous nature of external drivers influenced by global economic forces and climate change (Brugnach et al., 2008, Maier et al., 2016). Consequently, this study adopts an approach similar to a satellite navigation system used to guide drivers towards their desired destination. A safe and just future defined by experts, decision-makers and/or stakeholders forms the destination, whilst the exploratory scenarios are akin to the sat-nav assessing possible routes. In turn, following the core safe and just pathways is analogous to a driver wishing to avoid hazardous routes. Furthermore, decision-makers are forced to think outside conventional reference frames bound by historical dynamics, including leaving behind the perception that rejuvenating resource mobility is sufficient to ensure long-term resource persistence. This quality is akin to a new route potentially beyond the driver’s comfort zone, which would have remained undiscovered if it was not for the sat-nav.
An alternative method could simulate system dynamics under a limited number of predictive scenarios, perhaps relating to low, medium and high rates of socioeconomic development. However, this strategy is akin to using an out of date map that misses newer and potentially less precarious routes. Therefore, a limited number of sustainable routes may not be suitable where cross-scale interactions undermine resilience to local drivers, plus where slight alterations in driver trajectories can cause catastrophic impacts across feedback dependent systems like fisheries and coral reefs (Scheffer et al., 2001). For instance, although Hossain et al. (2017) find scenarios of sea-level rise, temperature change and agricultural subsidy causing dangerous crop production, decision-makers have little understanding of how system precariousness changes between 2°C and 3.5°C warming, whilst 2°C warming is considered equally as safe as 0°C warming. Consequently, decision- makers may interpret 1.5°C warming as always safe, despite co-occurring
254 Chapter 7: Discussion trajectories of agricultural subsidy reductions and water withdrawal potentially undercutting resilience. Therefore, this thesis questions the use of highly divergent storylines when envisioning plausible futures (e.g. Gallopín, 2002, Carpenter et al., 2015a). Whilst stories of landscape abandonment and youth movements are provocative, this study shows that regional thresholds often occur in the spaces between divergent futures, such as 5% differences in the number of active fishers (chapter 6.6).
Moreover, divergent storylines give a level of confidence that decision-makers will agree on interventions despite different knowledge frames and trade-offs (Brugnach et al., 2008), and shape the system as intended despite inherent complexities and deep uncertainties about the likely plausible future. In contrast, exploratory scenarios allow model structures and interactions to define the most frequent futures from a system’s internal structure and interactions. However, it is important to note that thousands of exploratory futures may not suit models with substantial runtimes or significant conceptual components. In these cases such studies will benefit from scenarios accounting for scenario uncertainty (Maier et al., 2016), assessing whether qualitative conclusions persist across slightly different development trajectories, akin to uncertainty propagation in multivariable sensitivity analysis (chapter 4.3.6).
In relation, this study builds on using probabilistic landscapes to represent social- ecological resilience. As per the methodology of Bitterman and Bennett (2016), aggregating all data points for each governance scenario into one landscape creates the illusion that MSY remains the deepest attractor until 2060 (figure 7.8), when actually outputs suggest that only 12% of modelled futures average MSY from 2050- 2060. In line with the findings here, Perz et al. (2013) recognise that stability landscapes shift due to parameter changes. Ignoring the temporal evolution of resilience landscapes may encourage decisions leading away from the desirable future, such as persisting business-as-usual despite steep gradients of non- traditional livelihood uptake, which is shown to force the system from its SJOS and open windows for multidecadal overexploitation (chapter 6.5.2).
This study uses rigid governance policies rather than polices adapting to system conditions. Consequently, this work does not model how undesirable trajectories can be continuously shaped towards sustainability, which would be akin to a sat-nav rerouting around unexpected traffic incidents or road closures. Static governance
255 Chapter 7: Discussion
Figure 7.8: The probabilistic resilience landscape of OM governance involving all outputs from 2015-2060, as per the method of Bitterman and Bennett (2016), rather than the annual method used in chapter 6.4. may therefore underestimate outcome resilience, because adaptive regulations could be introduced when necessary. Regulations across Chilika are thought to have implementation delays of ~10 years (Mohanty – personal communication), meaning planning after the first sign of decline might fail to avert dangerous futures. Therefore, this study focuses on decisions makeable today that enhance persistence, akin to planning the route of a journey before travelling. Therefore, robust policies here produce desirable outcomes which are resistant to driver trajectories (Maier et al., 2016) and do not prevent future interventions in light of new evidence (Schindler and Hilborn, 2015).
7.5.2 Evaluating the role of system dynamics modelling in sustainability science
Social-ecological systems may be modelled through a plethora of dynamical, statistical and conceptual approaches (Letcher et al., 2013, Filatova et al., 2016). An
256 Chapter 7: Discussion
SDM approach was justified on the need to explore feedback-driven sustainability challenges and uncertainty around appropriate system-wide policy scenarios. However, certain aspects of SDMs are seen as inferior to other forms of modelling (Sterman, 2002, Featherston and Doolan, 2012), such as the focus on pattern over prediction and the inability to model heterogeneous frames of reference. Therefore, this subchapter discusses how this study has addressed these criticisms and helped to advance SDM in social-ecological modelling.
A balance ultimately exists between capturing the critical feedbacks, cross-scale interactions and emergence (Verburg et al., 2016), and building ‘Integronsters’ where components become intractably meshed (Voinov and Shugart, 2013). SDMs are reduced complexity models as relatively light data demands and simple feedback structures can produce complex surprises, delays and shifts (Verburg et al., 2016). Hypothetically removing the human-natural feedbacks imagines dynamics that might have occurred under approaches less supportive of feedbacks, like Bayesian networks or integrated assessment models. For instance, spiky fisher populations would be absent if the populations were only driven by birth and death rates. Whilst wealthy traditional fishers would still upgrade, the absence of a shrinking traditional livelihood capacity would fail to produce ‘success to the successful’. Consequently, adaptive capacities would not be a wicked problem – potentially overestimating the resilience of traditional livelihoods as increasing motorboat use would not drive overexploitation and marginalisation. The feedback between fish abundance and days fished also shows that despite its perceived role in limiting degradation during the 1990s (Mohanty – personal communication), the adaptive behaviour does not guard against overexploitation driven collapses. These insights could be further formalised by measuring system sensitivity to feedbacks through formal loop dominance analysis (Ford, 1999, Bueno, 2014); however, absolute driver values are favoured here parallel to the indicators monitored by Chilika’s governors.
Tools like graphical functions and IF-THEN-ELSE statements make it possible for SDMs to explore virtually any problem. However, Letcher et al. (2013) stress feedbacks must be evidenced rather than hypothesised to avoid Integronsters. The collection of qualitative data helped to uncover potentially hidden dynamics across Chilika. For instance, excluding switches from traditional to non-traditional fishing would have missed the fishing effort nonlinearities causing safe space transgressions. Therefore, by expanding the boundaries of our mental models, the
257 Chapter 7: Discussion techniques used in developing and simulating SDMs help to guard against errors of omission where decision-makers fail to foresee undesirable dynamics due to ignorance of underlying processes (Collie et al., 2016). Here, well-intentioned but narrow governance is shown to be associated with policy resistance causing only temporary offset of problematic behaviours (Sterman, 2002). Alternative results might have indicated that business-as-usual overcomes plausible fishing effort trajectories, prioritising climate change resilience over sustainable exploitation. Various other modelling requirements are satisfied and developed, including the effects of biophysical constraints on sustainable development (Letcher et al., 2013, Filatova et al., 2016), the delivery of realistic governance options with respects to complexity and uncertainty (Verburg et al., 2016), and multidimensional evaluation (Jakeman et al., 2006).
A common weakness amongst social-ecological modelling approaches is the capture of emergence (Verburg et al., 2016), conceptualised as the unforeseen appearance of system-wide behaviours driven by local interactions (Dearing et al., 2010). SDMs capture emergence if the key interactions driving problematic processes are known today, for instance, recognising the potential threat posed by non-traditional practices. However, whether this can be considered true emergence is questionable because feedbacks excluded from parameterisation cannot emerge during simulation, such as non-traditional fishers returning to traditional practices or the emergence of heavy industry pollution. Instead, these behaviours result from bottom-up interactions between users and decision-makers captured by techniques like ABM and cellular automata (Dearing et al., 2006, Janssen and Ostrom, 2006), including cooperative behaviours or defections against regulation. Therefore, the SJOSs designed here must be framed as coupled social-ecological pathways of today’s drivers and feedbacks; however, this could be interpreted as rationalising the status quo, discouraging transformative relationships between users, stakeholders and nature.
Similarly, the deterministic nature of SDMs has led to debate over the ability to capture heterogeneous human agency (Featherston and Doolan, 2012), ranging between views that SDM are dehumanising (Jackson, 1991) to the opposite belief that all human decisions are conditioned by their surroundings (Forrester, 1961, Lane, 2000). Here, expressions between model variables are predominantly deterministic, insofar as satisfying a condition triggers an outcome at a higher level (e.g. boat upgrades require a threshold profit and the value of traditional fishing to
258 Chapter 7: Discussion be declining). Probabilistic human processes were included in earlier model versions, such as using Monte Carlo analysis to vary the number of fishers per boat; however, where data permits, deterministic variables offer enhanced output tractability, computational efficiency and sensitivity analysis. Stakeholder and expert interviews led to the inclusion of variables emulating real-life decisions, such as dynamic fishing days and catch selectivity increasing during perceived stock declines. Moreover, distinguishing between the two fishing communities is more humanising than the datasets of aggregate population which overlook social inequalities, adaptive capacities and marginalisation (e.g. Noack et al., 2013, CDA, 2015).
Osgood (2009) develops a SDM of human agency by downscaling variables that are dimensionally consistent across scales. However, additional Monte Carlo simulations to capture the different combinations of decisions would make model runtimes impractical, given the probabilistic techniques already generating future climate conditions and external trajectories. Moreover, the need to disaggregate below the fleet level is questioned by the strong links between social norms and fishing effort, including time spent fishing, group fishing and the sense of community belonging (Nayak and Berkes, 2014).
The pattern versus prediction debate questions whether SDMs should be used to predict the most likely outcome (Ford, 2010). To this end, some system dynamics studies standardize outputs to emphasise output dynamics over absolute values (e.g. Máñez Costa et al., 2011, Chapman and Darby, 2016). However, this study recognises multiple benefits of keeping model outputs in absolute values and units. First, measurement units consistent with those used in decision-making are comparatively tangible, such as catch averaging 1,000 tonnes/year rather than a z- score of -3. Likewise, standardisation can cause misinterpretations when comparing inter-scenario outputs; for example, due to OL’s superior average catch, the scenario appears to undergo divergent catches when scaled by z-scores (figure 7.9). Third, standardization is only suitable once the mean and standard deviation of a complete simulation is known, meaning sustainability assessments cannot be made on the fly. Finally, a neutral z-score may become negative if the output abruptly grows, giving the perception that the pre-growth period was undesirable irrespective of its absolute value.
Proponents of standardising SDM outcomes may question whether the quantitative thresholds defined for Chilika overstretch the capabilities of SDM (chapter 6.6). This
259 Chapter 7: Discussion
Figure 7.9: The 1000 future catch simulations of OM and OL governance scaled 푥−휇 by z-score standardization: 푧 = , where 푥 equals annual catch in 휃 tonnes, 휇 equals the mean of all catches and 휃 equals the standard deviation of all catches. argument would certainly be valid if single ceiling values were used (Rockström et al., 2009, Cole et al., 2014), encouraging decision-makers to drive towards the edge of SJOSs despite ignorance of threshold interactions, latitude and precariousness. Instead, the thresholds are consistent with SDM’s philosophy of goal-seeking behaviour (Forrester, 1961), where the SJOSs are bound by permissible driver trajectories averaging MSY from 2050-2060. This modelling activity has produces numerous insights for decision-makers even if the threshold ranges are considered unreliable, including the identification of the smallest and steepest SJOSs, and the positively skewed SJOSs that encourage shallow fishing effort trajectories to remain within core safe spaces.
7.5.3 Study transferability
This study provides decision-makers with explicit pathways towards normatively sustainable futures, alongside evidence of cross-scale interacting thresholds and potential social-ecological trade-offs. Yet, the usefulness of the modelling framework (figure 3.1) is arguable if it is not transferable to other SESs. Whilst it is not possible to cover all modelling techniques, system types and problematic behaviours, this subchapter generalises the challenges and opportunities of applying this modelling framework to similar fisheries and different SESs with sustainability challenges.
260 Chapter 7: Discussion
Fundamentally, this framework is transferable to SESs where decision-makers wish to surmount contemporary problems by reaching safe and just futures. For fisheries, this target may be futures that persist the ability to catch at MSY, although alternative metrics could include ‘optimum yield’ (Levin, 2014) that adapts MSY as a function of ecological, economic and social constraints. Nonlinear outcomes like historical envelopes of variability, STARS (Rodionov and Overland, 2005) or ARIMA forecasting (chapter 6.11) provide thresholds where normative futures are undefinable. Second, in order to use this SDM framework, problematic behaviours must manifest through temporal dynamics, whilst the modelling strategy must facilitate enough simulations to cover plausible future trajectories and their interactions. Therefore, SDMs ill-suit problems where the spatial flow of ecosystem services is critical to system sustainability, such as surface water dynamics and forest cover structures (Swetnam et al., 2011, Boumans et al., 2015); consequently, these problems tend to rely upon spatially disaggregated datasets unconducive to Monte Carlo simulations.
The modular nature of the model and its individual processes are adaptable to similar fishery systems. The hydroclimatic and ecohydrological modules may be parameterised for any river catchment and lagoon respectively, as long as available datasets are sufficiently long and complete to facilitate parameterisation and evaluation. Importantly, ecological contributions to problematic behaviours should be multidecadal, otherwise specialist hydrological models or ecosystem based models may be favoured for their higher predictive powers. As an example, freshwater and sediment regulation from the Akosombo dam have caused ecoyhdrological disturbances across the Keta and Sangar lagoons of Ghana’s Volta delta (Nairn, 1998, Dankwa et al., 2004). The resulting spread of aquatic vegetation and closures of the lagoon-sea canals are blamed for the collapse of oyster and shrimp fisheries. Due to similar drivers and outcomes, transferring the model to the Volta requires datasets of runoff and sediment downstream of Akosombo, plus estimations of key parameters like eco-K, fishing effort and ecohydrological conditions. In turn, external trajectories may include Akosombo’s water storage, fisher population growth and catch capacities under technological change. Similar ecohydrological stresses effect the Nile delta’s Manzala and Burullus lakes (Negm and Hossen, 2017), whilst unsustainable anthropogenic trajectories and catchment modifications threaten inland fisheries across Southeast Asia (Coates, 2002) and Africa (McIntyre et al., 2016).
261 Chapter 7: Discussion
The approach developed is also transferable to models of non-fishery SESs, particularly those with readily defined external drivers. Developing the eight scenarios of Hossain et al. (2017), MCA could explore interactions between temperature change, water withdrawals and agricultural subsidy keeping rice production within its envelope of variability. Likewise, an interesting experiment would be simulating the World3 model of Meadows et al. (1972) with nonstationary external variables (e.g. land fraction harvested) to explore how interactions between global limits like land fertility and cultivated land area influence the timing of population overshoot.
To illustrate this study’s transferability to regional systems, safe and just trajectories can be defined for California’s Mono Lake basin (figure 7.10), which is facing the challenge of providing just water supplies to surrounding settlements (including Los Angeles) whilst retaining safe lake levels for ecological functions and recreational uses. By projecting spectrums of precipitation, evaporation and runoff diversions with the Lake Mono model of Ford (2010), safe and just outcomes are recognised by water levels above 125 feet (Ford, 2010) after 50 years, whilst dangerous outcomes meet neither just water supply or safe ecological needs (<115 feet). The shallowing lake depths over the first 25 years indicate that the current water balance is unsustainable. Despite being a manmade problem, Lake Mono’s future safety is relatively sensitive to climatic drivers across the range of driver trajectories, evidenced by the distinct SJOS of precipitation change but the absence of a SJOS for water exports. The safest multidimensional pathways correspond to precipitation changes ≥75%, evaporation change ≤-18% and negative export changes. Therefore, this rapid assessment suggests that continuing to govern Lake Mono as business-as-usual (i.e. no export regulations) will likely lead to unsafe futures due to growing water demands in a warming world and annual evaporation across the basin expected to increase by 5% by 2050 (Ficklin et al., 2013).
Unsustainable feedback structures, external processes and deep uncertainty about the future provide a checklist to apply the modelling framework developed here. Adding to the challenges of modelling spatial problems outlined earlier is a suite of problems where this technique does not naturally fit. For problems suited to ABMs, including defector or collective behaviours, the exploratory scenarios could be adapted to investigate SJOSs of individual decisions. For example, varying decision thresholds determining farmland abandonment could explore permissible ranges of collective behaviour (Figueiredo and Pereira, 2011). Moreover, spectrums of
262 Chapter 7: Discussion
Figure 7.10: Time-series of (A) future lake depth under the spectrum of (B-D) precipitation, evaporation and export changes; (E-G) the driver based SJOSs depicting the proportion of future states after 50 years under the external trajectories. Colours are as per figure 6.11.
263 Chapter 7: Discussion biophysical scenarios may also design SJOSs where livelihood decisions, like cropping patterns (Mialhe et al., 2012) or daily fishing efforts (Elliston and Cao, 2006), are influenced by external environmental conditions.
Modelling SJOSs complements rather than replaces optimisation techniques, which are designed to reduce uncertainty about the expected outcomes of a policy (Kwakkel et al., 2016). For example, an optimal scenario of fishery exploitation that balances competing targets of handling times and inter-fleet competition (Agnew, 1979) may be located on the wider spectrum of outcomes. The need to balance competing social and ecological outcomes may position the system towards SJOS edges (chapter 6.8) – fundamentally opposed to the safety and persistence of core trajectories.
Finally, this study also complements backcasting by offering forward-looking pathways towards normative futures. However, where backcasting aims to understand the freedom of policy action and timing (Robinson, 1990), future SJOSs are defined by norms and feasible governance from today. To align with back- casting, the timing and magnitude of governance could be simulated with Monte Carlo analysis, akin to viability theory that seeks all policy scenarios producing outcomes within predetermined constraints (Mathias et al., 2017). However, the method developed here favours policies known to be feasible from today, rather than black-boxed policies in future with little or no causal mechanisms (Biesbroek et al., 2017).
7.5.4 Subchapter synthesis
This study utilises deep uncertainty about the future and inbuilt sensitivity techniques to define continuous and interacting thresholds that bound SJOSs. Consequently, trajectories and interactions beyond what might have been achievable with best-guess scenarios have been identified, thus providing insights for sustainability beyond modelling the continuation of contemporary trends (Börjeson et al., 2006). Such analysis has helped to define the width of the goalposts bounding goal-seeking behaviour, plus whether system structures and governance scenarios are attracted to the goal. In turn, thresholds from SDMs can be defined by trajectories achieving the goal, including qualitative reference behaviours (e.g. linear or nonlinear growth) and/or quantitative metrics (e.g. 6000 motorboats). The resulting multivariable interactions producing drive-response
264 Chapter 7: Discussion equifinality and multifinality are arguably most valuable by encouraging decision- makers to target sustainable interactions forming the core SJOS (‘persistence’) over safe space edges ('efficiency'- Gunderson and Holling, 2002).
Whilst a range of modelling techniques exists, the perceived threats of social- ecological interactions, feedbacks and traps make Chilika’s problematic behaviours particularly suited to an SDM approach. The modelling framework is readily transferable to SESs vulnerable to feedbacks that may intensify under plausible external drivers, leading to problematic behaviours that emerge over time. In contrast, this framework is less suited to spatially explicit problems, short-term forecasting or instances where decision-makers desire optimised outcomes. Moreover, uncertainties of this study are expressed in terms of structural evaluation, behavioural analysis and sensitivity analysis, plus regular descriptions of study limitations during model construction and discussion. Therefore, the next subchapter further discusses how uncertainties in study design, methodological framework and data affect the use of this study to steer Chilika until mid-century.
7.6 Model uncertainties, multiple worldviews and indeterminacy
Uncertainties are inherent to modelling studies (Sterman, 2002). Rather than stunting progress, uncertainty is utilised here to build plausible futures as defined by available data, perceptions of stakeholders and modeller intuitions. Chapter 4 has already detailed uncertainties associated with model structures, parameters and sensitivities, concluding that right quantitative behaviours are produced for the right reasons. However, it is now important to discuss how decision-makers might interpret study outcomes in light of inexact knowledge and potentially more exact methodologies (Refsgaard et al., 2007).
Numerous conceptual frameworks exist to classify uncertainties in modelling studies and environmental sciences (e.g. van Asselt and Rotmans, 2002, Walker et al., 2003, Brugnach et al., 2008). Here, inexact knowledge can be categorised as (van Asselt and Rotmans, 2002): (i) ‘what we roughly know’ – involving uncertainties of natural variability and measurements inaccuracies, (ii) ‘we don’t know what we know’ – including the potential effects of excluded variables and conflicting evidence, and (iii) unknown unknowns due to unidentified events and indeterminacy. These classifications are porous because some uncertainties correspond to multiple
265 Chapter 7: Discussion domains, such as data shortages (type-i) leading an expert to believe that a particular process is unimportant (type-ii) (Zimmermann, 2000).
7.6.1 Uncertainties and limitations that we roughly know
Termed Gaussian uncertainties, every measurement of a physical quantity is effected by some error (van Asselt and Rotmans, 2002), meaning the causes, likelihoods and potential impacts of roughly known uncertainties can be estimated (Pawson et al., 2011). For example, the rainfall-runoff models have uncertainties associated with data, parametric relationships and spatiotemporal aggregation. More specifically, measurement errors derive from the calibration of rainfall gauges, flow meters and thermometers used by the Indian CWC and IMD, including possible systematic biases (Regan et al., 2002).
Different conclusions are reached if variables were to vary across potential uncertainties. From Monte Carlo sensitivity analysis, we roughly know that uncertainties up to ±50% errors in 19 variables do not significantly affect the reproduction of Chilika’s reference mode (table 4.5), including the key variables of the Bueno and Basurto (2009) fishery model, the Ghashghaei et al. (2013) rainfall- runoff model and human population growth model (chapter 4.3.6). Would the model be as useful if these variables were excluded? Freshwater and sediment influxes could be externally generated by sampling from historical distributions. Alternatively, hydrological inputs could be assumed constant, de rigueur for fishery SDMs (Morecroft, 2007, Bueno and Basurto, 2009, Martins et al., 2015). Whilst reducing uncertainties of rainfall-runoff generation, the above techniques would gain little insight into how external natural processes potentially affect future dynamics. Furthermore, lagoon salinity and days fished are indicators used by the CDA to monitor natural and human processes; omitting these variables would be inconsistent with the mental models of decision-makers. Moreover, catch is insensitive to survival rate and fecundity, but the two variables are required to drive fish regeneration within fishery SDMs (Bueno and Basurto, 2009, Duer-Balkind et al., 2013) .
In contrast, six variables have uncertainties where errors ≥10% tend to produce outputs different from the historical reference. Chapter 4.3.6 provides a detailed evaluation of these parametric uncertainties, arguing that rather than introducing errors, the sources and magnitudes of uncertainty are consistent with qualitative
266 Chapter 7: Discussion and quantitative system knowledge. Therefore, despite uncertainties inherent to the different datasets, the model’s relatively simple structure and equations are able to replicate the timing, magnitude and relative contributions of past problematic behaviours.
A shortcoming of the model evaluation conducted here is the lack of quantitative sensitivity analysis around changes in the form of graphical functions (e.g. how would monthly salinity variations change if the non-monsoon graphical function took linear form?). It is however possible to have a qualitative discussion based on knowledge gained from the alternative pasts (chapter 4.3.4) and sensitivity analyses (chapters 4.3.5 and 4.3.6). Arguably, the most significant impacts on model outputs would derive from changes in the rainfall-runoff and sediment generation functions. For example, giving LMC rainfall-runoff a steep linear form would overestimate non- monsoonal sediment fluxes, causing earlier than expected outlet sedimentation. Thus, we can have confidence in the parameterised form given its consistency with observed magnitudes of freshwater and sediment delivery (chapter 4.3.2), plus its relative contribution to the historical reference mode of fish catch (chapter 4.3.4). Likewise, linear functions between Chilika’s ecohydrological conditions and fish survival (chapter 3.4.2.5) would likely undermine fish stock resilience to fresher, warmer and deoxygenated waters, overlooking the deleterious effects that accelerate with ecological degradation. In contrast, the alternative pasts and sensitivity analyses suggest that altering the graphical form of the number of days fished would have little impact on the resulting fish catch.
A specific shortcoming of chapter 4.3.6 is the inability to assess the effects of disaggregating key model structures. For instance, spatially lumped parameters aggregate runoff-generating processes to the catchment scale. Such models are less exact than spatially distributed models that simulate the effects of local surface and subsurface characteristics (Beven, 2002). To this end, Ahmed and Simonovic (2004) develop a spatially disaggregated SDM of the Red River basin in Canada. However, assigning one runoff stock per grid cell (4km x 4km) is only computationally feasible for small catchments, over sub-annual timescales and limited future scenarios. Similarly, disaggregating Chilika’s fish stock into its 317 species (Mohanty et al., 2015) is infeasible on the grounds of data availability, model capacity and inexact knowledge surrounding how different species interact with each other, the environment and catch efforts. A comparatively plausible approach would be to parameterise individual stocks for the 17 most commercially important
267 Chapter 7: Discussion finfish and shellfish (Kumar and Pattnaik, 2012), exploring whether disaggregated stocks and target species build resilience against external pressures, internal nonlinearities and inequalities.
Rather than a SDM, specialist hydrological models like the Integrated Catchment Model (INCA) (Whitehead et al., 2016) and Soil and Water Assessment Tool (SWAT) (Arnold et al., 1998) could be coupled with an ecosystem based fishery model (for review see Plagányi, 2007) to explore the dynamics of individual stocks, trophic interactions and ecosystem health. However, an ecosystem-based framework has limitations that would hamper the search for social-ecological thresholds and SJOSs. Models like SWAT require run-times up to tens of hours (Sloboda and Swayne, 2011), meaning exactness comes with the cost of limited simulations, which potentially miss threshold regions within phase space. Second, model resolutions and dimensions must be cross-compatible, but inconsistencies often arise between integrated models developed within different research spheres (Voinov and Shugart, 2013). Furthermore, complex models with process disaggregation may be perceived as black-boxes by decision-makers unable to understand their ‘mathematical wizardry’ (Mulligan, 2013, p.339). Usability and scrutiny are traded-off, potentially disengaging the stakeholders and decision-makers the models aim to serve.
Therefore, we know that reduced complexity models like SDMs must prioritise medium-term patterns over short-term forecasts requiring finer precision, such as hydrological flows in a given day/month. Thus, model outputs must be interpreted as what-if scenarios: what if fishers continue to adapt from traditional to motorboats when financially viable? And, what if today’s attitudes persist (e.g. MSY and business-as-usual governance)? Rather than ignoring parametric uncertainties, these interpretations help to bound insights within limits of what we roughly know.
7.6.2 We do not know if we know the effects of excluded processes
The multiple worldviews informing this model’s structure makes knowledge of system processes fundamentally uncertain, as the representation of dynamics is open to questioning from different mental models (Brugnach et al., 2008). Indicating the lagoon’s social-ecological health and safety of plausible futures, perhaps the most fundamental uncertainties are associated with the catch record. Six days per month, catch weights of every fourth boat entering Chilika’s 18 landing sites are recorded (Kumar and Pattnaik, 2012). Small boats are assessed by eye and
268 Chapter 7: Discussion catches are reported without confidence intervals, adding to the uncertainties associated with interpolating between samples. Moreover, stakeholders and officials give conflicting accounts of productivity (Dujovny, 2009), including ICAR-CIFRI (2016, p.25) simultaneously quoting catch in 2015 as 9,184 tonnes and 12,000 tonnes! Thus, it is critical models can reproduce relatively certain reference modes, like the 1990s catch decline that is agreed upon by both decision-makers and stakeholders (Nayak, 2014).
Perhaps most noticeable to readers familiar with Chilika is the exclusion of prawn aquaculture. Although Kadekodi and Nayampalli (2005) promote aquaculture to benefit catch and livelihoods, it is argued to have contributed to the 1990s collapse through the spread of disease, pollution and encroachment on traditional grounds (Dujovny, 2009, Nayak, 2014). Aquaculture across Chilika was outlawed in 2001, although Kumar and Pattnaik (2012) map remnants around the lagoon’s periphery. Therefore, we do not know whether the model overestimates fishery persistence if prawn aquaculture is to proliferate by mid-century. Exclusion from the model is first due to the lack of information linking prawn aquaculture to habitat effects – making the potential feedbacks speculative at best. Second, decision-makers prioritised the catch of juveniles, tidal outlet fishing and overcapacity when asked to consider future threats. Model outputs are therefore framed by the assumption that aquaculture does not impede MSY, thus encouraging decision-makers to uphold its outlaw.
Although the system exhibits resilience to monthly and annual hydroclimatic inputs, future resilience to cyclones and storm surges is unknown. From 1975-2013, 14 cyclones affected Odisha (Yadav and Barve, 2017), including the 1999 Super Cyclone which killed ~10,000 people. The 2013 Cyclone Phailin disrupted Chilika’s salinity, dissolved oxygen and temperature cycles, with Barik et al. (2017) finding the normal annual salinity regime supressed for 20 months post-Phailin. Meteorological extremes also affect fishing operations by damaging boats, landing jetties and roads (Iwasaki and Shaw, 2010b). The natural effects of cyclones could be modelled by pulsing additional rainfall to the LMC and WC, assessing whether sub-monthly processes undermine resilience against slower stresses. However, exploratory analysis attempting to screen cyclone rainfalls from the monthly data found that the signal of the most powerful cyclone to date is hidden, with rainfall in October 1999 (Super Cyclone) equalling the median October rainfalls of both Khordha and Puri districts from 1973-2009. Therefore, this model instead aims to
269 Chapter 7: Discussion find system configurations that are resilient to multidecadal trends, under the assumption that this increases the likelihood of overcoming extreme events.
We also do not know whether growth in the number of motorboats and the active fisher population will cause toxic conditions for fish due to waste and petroleum hydrocarbon emissions. Clusters of relatively high hydrocarbon concentrations around jetty stations suggest spatial modelling techniques are required (Mohanty et al., 2016). Mohanty et al. (2016) find hydrocarbon concentrations around jetties to be ~15 times below the 100 µg L− 1 threshold of the Indian Central Pollution Control Board (CPCB, 1993), whilst Baliarsingh et al. (2014) note that the main source of pollution is the unsafe handling of fuels, rather than exhaust emissions. Therefore, rather than erroneously modelling pollution dynamics and potentially causing unjustified concern, this study raises awareness of plausible causal trajectories to encourage further study of Chilika’s pollution dynamics and its effects on habitat quality.
Despite the existence of predictive models (Malek et al., 2011), zooplankton and phytoplankton dynamics are excluded in favour of directly linking water quality indicators to explicit drivers of catch (e.g. macrophyte coverage, survival and migration). Such parsimony inhibits the exploration of how primary productivity and nutrient fluxes under climate change may buffer catch losses (Cloern et al., 2014). The excluded processes explicate a limitation of finding the sweet-spot (Collie et al., 2016) for data driven models between including all processes, which increases the degrees of freedom for errors (Jakeman et al., 2006), and only including processes perceived to be key to past, present and future dynamics (Verburg et al., 2016). This study adopts a relatively conservative stance, favouring the exclusion of processes and acknowledgement of uncertainties over erroneous parameterisation that may undermine the tractability of relatively certain processes.
Uncertainties also derive from human behaviour. In terms of information gained from the interviews, fishers may display cognitive dissonance between perceived and actual fishing habits, potentially caused by the want to appear environmentally conscious. All fishers interviewed expressed negative views towards juvenile catch and outlet fishing; therefore, including fishers involved in destructive practices might have increased understanding into the motivations behind such behaviours. Similarly, the model excludes relatively non-rational behaviours like the acts of fishing despite bans, the poorest fishers upgrading boats and the maintenance of
270 Chapter 7: Discussion fishing days despite stock declines. Dissonance may also exist between the plans and implementations of governors (e.g. policing fishing bans) when faced with the real-life resistance of international consumers and fishery cooperatives absent from the model.
Hitherto, this chapter has discussed factors questioning the extent to which we know the model accurately captures holistic social-ecological functions. Given the various exclusions, driver pathways are framed by an effective absence of pollution, prawn aquaculture and defector behaviour, which become implicit dimensions of the SJOSs. However, the SJOSs do not build reciprocal robustness against the excluded processes, meaning further study of these uncertain processes must be encouraged to widen the scope of future models.
7.6.3 A brief word on indeterminacy
The third major suite of uncertainties in modelling studies are processes not determinable by humans (van Asselt and Rotmans, 2002). Difficult to discuss due to their very nature beyond the mental models of observers, the definition of ‘unknown unknowns’ (after Donald Rumsfeld’s famous press briefing as US Secretary of State, 2002) is relaxed to include processes with undeterminable risks or impacts over the horizon of study (Cleden, 2009).
The fish price funnel rising from today assumes that domestic and international demand for fish will continue to rise, thus implying that demand declines are inconceivable. Carpenter et al. (2015a) provide an alternative approach by including a plausible future of weakened consumerism in their model of agricultural landuse change. Here, decision-makers might diverge from desiring MSY, perhaps following unexpected fish population declines that make MSY suddenly unsafe. Additional indeterminable risks include the industrialisation of aquaculture, akin to the failed joint ventures of Tata and the Government of Odisha in 1991 (Ray and Garada, 2017), the introduction of marine trawlers, invasive species and fish disease. These stresses all stem from natural events or social tipping points with unknown chances and impacts (Pawson et al., 2011), meaning that the SJOSs assume that these unknowns also remain absent. Prioritising particular known processes is a fundamental limitation of problem-based modelling (Wintle et al., 2010); however, failing to build resilience against known processes may mean that the system transgresses a tipping point before unknowns even emerge.
271 Chapter 7: Discussion
7.6.4 Subchapter synthesis
The aphorism all models are wrong (Box, 1976, 1979, Sterman, 2002) emerged from the inherent nature of uncertainty in modelling studies. However, the saying is qualified by a number of caveats, which make models useful as tools for decision support and explorers of system dynamics.
The three types of uncertainty discussed above are fundamental to interpret the future sustainability of Chilika. Parametric uncertainties are effectively universal, as the quantitative and qualitative variables are sampled from populations. Therefore, it is not important whether uncertainties exist, but whether changing parameter estimations over their plausible range produce fundamentally different conclusions about outcome behaviours. Critically, this study has defined permissible error ranges so that we roughly know that conclusions are consistent up to ±10% errors across the most sensitive system drivers. Additional data does not necessarily reduce uncertainties if the modelling strategy is also inexact; for instance, the sub- district rainfall data must be aggregated for use in spatially implicit rainfall-runoff models. To this end, specialist hydrological, economic and fishery models offer alternative approaches, but often exclude social-ecological feedbacks, performance evaluation and external scenarios achievable with SDMs. Consequently, model projections here should not be viewed as deterministic forecasts but plausible patterns of behaviours emerging from different external trajectories, key internal feedbacks and decisions.
Limited insights are gained into the sustainability of processes excluded due to parsimony, lack of observations or conflicting evidence. Ideally, the model would capture all human and natural processes, but speculative relationships have been avoided where possible to improve the tractability of processes and feedbacks important in the eyes of stakeholders and experts. Ultimately, within the bounds of limited knowledge, this model has uncovered insights into the sustainability of Chilika’s internal feedbacks, adaptive behaviours and decommonisation. Crucially, this study also shows that decision-makers may degrade resilience through actions designed to be sustainable, which Sterman (2002) promotes as the crucial indicator of model usefulness.
272 Chapter 8: Conclusions
Chapter 8: Conclusions and future research
This study is grounded in the concepts of SESs, sustainability, resilience and system dynamics. In all, this project details the conceptualisation, construction and simulation of an ambitious model designed to investigate the future of a deltaic natural resource system, focusing on pathways bounding sustainability, system responses to decision-making and resulting social-ecological trade-offs. Chapter 8.1 presents five conclusions to summarise the contributions of this study to the key themes, before chapter 8.2 describes future research directions to begin addressing unanswered and emerging questions.
8.1 Five-point synthesis
Conclusion 1: The Chilika lagoon fishery is an Anthropocene deltaic system: the persistence of social-ecological functions is dependent on but also vulnerable to human actions
The Chilika lagoon of India’s Mahanadi delta has a legacy of resource collapse. Productivity during the 1990s reached a 50-year low, principally blamed on the unchecked sedimentation of the Magarmukh tidal outlet channel that degraded the lagoon’s brackish environment, inhibited fish migration to and from the Bay of Bengal, and promoted the spread of freshwater macrophytes across the lagoon. Following the dredging of a shorter tidal channel in 2000, reported catch recovered from an average of 3,100 tonnes/year in the 1990s, to 12,100 tonnes/year since 2005. However, uncertainty surrounds the efficacy of this policy to overcome cross- scale interactions, unsustainable fishing efforts and future habitat changes. Therefore, this research builds on the plethora of historical analysis by providing the first horizon scan of problematic behaviours.
Synthesising the relationship between governance and future system persistence, managed decommonisation is found to build resilience against abrupt fishing effort intensification and creeping gradients of total efforts. As a worst-case scenario, ceasing outlet maintenance under do nothing governance retriggers the historical cause and rate of collapse, found to be proportional to the rate of decadal rainfall change and minimum annual salinity. Decadal tidal outlet maintenance tips the fishery from a Holocene SES to an Anthropocene SES, where persistence of
273 Chapter 8: Conclusions ecological and human functions are dependent on and vulnerable to human actions. In turn, stable resource mobility under business-as-usual builds climate resilience, but rejuvenating the fish stock every ten year facilitates the growth of unsustainable fishing efforts under steep trajectories of fish price and fisher population growth. Fishing bans across an area equivalent to the new outlet almost doubles the frequency of sustainable catch (OB), whilst dangerous futures disappear once OB is combined with an exploratory alternative livelihood policy removing 1/2000 fishers/fleet/month (OL).
In terms of future pathways, sustainability is also found to be inversely proportional to the magnitudes of socioeconomic drivers by 2049, including externally driven fish price, active fishers, total boats, motorboats and juvenile catch. Managed decommonisation builds specified resilience against these components of fishing effort, with the latitude of permissible trajectories increasing from OM to OL. Despite this, the SJOSs of juvenile catch and the number of motorboats remain distinct even under OL, suggesting that Chilika is inherently vulnerable to effort and catch compositions. Furthermore, motorboat numbers and juvenile catch have the most distinct SJOSs relative to their dangerous spaces – signifying key levers to steer system sustainability. In terms of temporal dynamics, stepwise motorboat proliferation over the next decade is symptomatic of unsafe futures, deriving from relatively wealthy traditional fishers adapting during falling traditional productivity. A spiked fisher population is indicative of success to the successful, whereby non- traditional resource access dominance causes the loss of traditional livelihoods. However, whether an unsafe trajectory ultimately becomes dangerous is positively linked to the gradient of fishing effort beyond 2040.
This research shows that the persistence of Chilika’s social-ecological functions is dependent on short-term fishing intensities, multidecadal effort extents and the ability of governance to build internal resilience against external trajectories. Moreover, whilst hydroclimatic processes are found to have statistically indistinct SJOSs, negative correlations exist with the permissible gradients of fishing effort within the SJOS. In particular, the sustainable number of motorboats shallows as freshwater and sediment deliveries steepen, highlighting the antagonistic relationship between degraded resource mobility and exploitation. Therefore, cross- scale interactions undermine resilience to regional processes, meaning decision- makers must also consider external trajectories of climate change and globalisation when steering internal dynamics.
274 Chapter 8: Conclusions
Chilika’s future fish population, CPUE, per capita income and income equality display complexities relating to driver pathways. Generally, fish population and CPUE decline linearly over time, with the gradient of change inversely related to safety. Sustainable futures have linear trends of income and fishery value, which abruptly rise then fall from 2040 under dangerous futures. This inflection is delayed under cautionary futures, producing highest average income and fishery value from 2050-2060. This dilemma adds to a wider suite of trade-offs for decision-makers to consider going forward, including whether to follow shallow socioeconomic trajectories that might require stakeholders to forego adaptive capacities – effectively trapping traditional fishers into less productive practices. Second, whether Chilika’s historical CPR status should be waived in order to build resilience against the spectrum of plausible external stresses. Whilst missing a definitive answer, this study ultimately empowers Chilika’s decision-makers to make an informed decision about desired system identity, based on the dynamics of social- ecological pathways, unintended consequences and cross-scale interactions that underpin system persistence.
Conclusion 2: Future regional SJOSs can be operationalised as multidimensional pathways of interacting drivers and thresholds
The SJOSs paradigm emerged from the planetary boundaries concept to find sustainable development that meets social foundations but avoids environmental ceilings. Studies to date have principally compared contemporary observations to predefined thresholds at the international (Raworth, 2012), national (Cole et al., 2014) and regional (Dearing et al., 2014) scales. A model of the Ganges- Brahmaputra-Meghna delta shows that rice production might be forced from its envelope of variability under particular climatic and economic scenarios (Hossain et al., 2017); however, these scenarios are limited in number and diversity, whilst the model only involves statistical couplings and inter-driver interactions are absent. Consequently, little is known about how SJOSs may evolve in future, the driver interactions needed to achieve safety, and how the pathways respond to governance and cross-scale processes like climate change and globalisation.
Transcending the importance of the specific pathways for Chilika, this study shows that SJOSs can be operationalised by driver trajectories reaching a desirable future defined a priori that balances socially just (catch) and ecologically safe (resource consumption) conditions. Therefore, depicting causal dynamics from today
275 Chapter 8: Conclusions achieving sustainable catch by mid-century converts the SJOS concept from a purely outcome-based assessment of today’s sustainability to a driver based tool to guide future evolutions. Whilst only one outcome is used as a first pass, multiple social foundations and environmental ceilings may be added in future to constrain the pathways further. Moreover, through the concept of conditional probabilities, SJOSs communicate how sustainability changes over the plausible range of driver trajectories. Rather than static/single value thresholds, driver based SJOSs are better thought of as soft thresholds where safety is partly conditioned by multivariable interactions. In turn, multivariable interactions operationalise the concept of core safe spaces (Steffen et al., 2015b) by identifying the drivers that critically underpin sustainability. This study takes the concept one-step further by delineating the permissible trajectories giving at least a 75% chance of safe and just outcomes.
Conclusion 3: Linear SJOSs have various characteristics of nonlinear concepts of social-ecological resilience
Normative thresholds (type-I: Dearing et al., 2014) are used here to operationalise SJOSs within a modelling environment. Despite linear definitions, the SJOSs of social-ecological drivers are found to exhibit nonlinear properties of social- ecological resilience and regime shifts. When viewed as conditional probabilities between a given driver trajectory and outcome state, the SJOSs resemble stability landscapes commonly used to describe general resilience and system attraction. In actuality, the SJOSs depict resilience to specific drivers, with the range of driver trajectories producing safe and just outcomes akin to resilience latitude. Resistance, ‘the ease or difficulty of changing the system’ (Walker et al., 2004, p.2), is communicated by the rate that safety changes across the latitude. The precariousness of the system to a given driver is a function of latitude and resistance, where the least precarious position would be within a wide, deep and shallowly changing SJOS. However, this study highlights a trade-off where the deepest and most distinct SJOSs (e.g. motorboats) also have the steepest boundaries, meaning outcomes become rapidly unsafe as the edge of the SJOS approaches. This represents a key difference to conventional outcome-based stability landscapes, where steep sides are desirable for outcome recovery. Instead, the SJOSs found here align with the idea that resilient social-ecological functions persist across driver magnitudes (Folke, 2006, Carpenter et al., 2015b).
276 Chapter 8: Conclusions
Developing the nonlinear aspects of the SJOSs concept, driver-response nonlinearities and hysteresis are found to be associated with normative cautionary spaces. Therefore, driver-response nonlinearities within cautionary spaces questions whether unsustainable trajectories can be restored once within the normative window of opportunity. Beyond the window of opportunity, hysteresis trajectories are found to have added dimensions beyond their theoretical depiction (e.g. Scheffer et al., 2001), with the temporal duration of resource recovery positively associated with the magnitude of resource loss. Moreover, the reverse trajectory through phase space is dependent on the priorities of decision-makers. Chilika’s overexploited fish stock may be recovered within ten years if ecological functions are prioritised at the expense of socioeconomic functions; alternatively, recovery may require 30 years if decision-makers wish to retain livelihood generation.
This conclusion builds on the operationalisation of driver based SJOSs to note their numerous similarities with wider resilience thinking. Rather than binary attributions of sustainability, this study shows that driver-based SJOSs exhibit latitude, resistance and precariousness all dependent on multivariable interactions. Therefore, system safety cannot be gauged by one driver alone, promoting the core SJOS to find simultaneously sustainable trajectories.
Conclusion 4: Normative safe and just futures are associated with trade-offs effecting holistic sustainability
Previous studies overlooked the creation of winners and losers within SJOSs, with the default position being that remaining between the environmental ceilings and social foundations is safe and just for all stakeholders. This study finds that SJOSs are associated with heterogeneous outcomes, particularly when feedbacks drive poverty traps with imperceptible effects at the system level. Importantly for decision-makers designing Chilika’s identity, building general resilience through managed decommonisation accommodates levels of income inequality associated with collapse under OM and OB. Consequently, general resilience converts the ‘success to the successful’ archetype from a problematic behaviour to a trade-off of fishery persistence. These results also have important implications for Chilika’s governors, who do not currently differentiate between traditional and non- traditional catch, income or livelihood opportunities.
Similarly, safe and just futures under business-as-usual are partly dependent on the continued marginalisation of traditional fishers, as stepwise uptake of non-
277 Chapter 8: Conclusions traditional practices risks overexploitation. Consequently, a delicate balance exists between promoting adaptation to further livelihood opportunities and shifting fishing effort intensities towards unsustainable levels. The suggestion that the adaptive capacities of certain users should be limited to preserve the wider system is itself unjust. Therefore, this work calls for future studies to assess how deeper structural leverage points (e.g. feedbacks and self-organisation) can bolster the social resilience of Chilika’s most marginalised without edging the system closer to its environmental ceilings.
Conclusion 5: Combining divergent external trajectories with reduced complexity modelling probes interactions causing safe space transitions, tipping points and regime shifts.
Alternative plausible futures are gaining traction within SESs thinking as a means of envisaging divergent and transformative futures (Carpenter et al., 2015a, Bai et al., 2016). Modelling highly divergent external conditions builds upon previous research that uses expert appraisal to investigate how internal interactions and feedbacks attract a region towards different divergent futures. To this end, societal goals (e.g. MSY average from 2050-2060) involve both equifinality and multifinality. Equifinality emerges when the same normative future results from different driver trajectories; for instance, cautionary futures achieved from linear and nonlinear motorboat use. Multifinality argues that similar trajectories can lead to different normative futures – discouraging rigid decision-making or the focus on one driver whilst others become unsustainable.
Consequently, the location of thresholds may be brought forward or pushed back due to social-ecological interactions. Interestingly, drivers with insignificant connections to outcome state are found to influence the size of statistically distinct SJOSs. Drivers like climate change and livelihood costs may not directly cause dangerous futures, but their disregard may promote unsafe outcomes when combined with anthropogenic trajectories approaching SJOS limits. Thus, when modelling social-ecological interactions and thresholds, predictive scenarios must be at least combined with a buffer zone of uncertainty (e.g. Maier et al., 2016, figure 1d) to begin to understand how resilience latitude, resistance and precariousness changes over plausible interacting driver trajectories.
278 Chapter 8: Conclusions
8.2 Future research directions
This thesis has developed and critically evaluated a novel framework to project plausible futures and subset safe and just driver limits. Various future research directions have emerged from this study, forming additional requirements to operationalise policies across Chilika, expand model breadth and depth, and undertake investigations to further or verify a number of implications for SESs and their governance.
Following the commitment of the CDA to reopen the tidal outlet every decade, the realisation that sustainability is also intrinsically linked to future fisher dynamics is arguably the key message for Chilika. Therefore, the main future research demand for Chilika is to focus on the pathways underpinning sustainability, because within the headline lies the respective roles of traditional and non-traditional practices, with the persistence of the former critical to avoid overexploitation from the latter. This study showing that a future with <9,000 total boats and <5,000 motorboats by 2050 is safe, whilst <9,000 total boats but 8,000 motorboats is unsafe. However, routine surveys presently aggregate both fleets (Noack et al., 2013, CDA, 2015), whilst any datasets with differentiation are sporadic at best (Kadekodi et al., 2000, Mohapatra et al., 2007b, Mohanty et al., 2016). Therefore, this study highlights a need to routinely monitor the different fleets, productivity contributions and adaptive behaviours, potentially foreseeing transitions towards non-traditional dominance and the marginalisation of traditional livelihoods.
Through only a single feedback, this study has shown that system-wide sustainability is sensitive to the timing and number of switches between traditional and non-traditional fishing. Therefore, adding to the quantitative data demand is the need for qualitative insights into the decisions causing traditional fishers to intensify livelihood practices. Building knowledge of the wider push and pull factors will improve the exactness of model parameterisation, plus highlight additional levers to steer socioeconomic trajectories and address root causes of unsustainability (Kadekodi and Gulati, 1999, Abson et al., 2017). For instance, subsidising costs of traditional practices may be ineffective if younger generations continue to disconnect with the traditional identity (Ray and Garada, 2017), favouring efficient practices over ecosystem stewardship (Nayak, 2014).
279 Chapter 8: Conclusions
Whilst formal regulations forcibly anchor driver trajectories, the resulting trade-offs effect holistic sustainability through decommonisation and the marginalisation of traditional livelihoods. Consequently, research may wish to focus on the role of shadow networks (Olsson et al., 2006) and informal institutions (Pelling et al., 2007) operating outside formal regulations. Cross-livelihood cooperatives and education programmes may foster sustainable self-organisation from the bottom-up, helping to limit divergence between socioeconomic pathways and the emergence of inequitable archetypes. The CDA recognise the absence of a social scientist as an institutional weakness (Kumar and Pattnaik, 2012); this study explicates the necessity to continue expanding research into the regional impacts of humans upon Chilika’s nature, rather than only the legacy of nature upon humans.
This refocus may be aided by developing the usefulness and exactness of this model. The former is currently on-going, with study insights being fed-back to decision-makers via informal discussions, formal meetings and scientific publications (e.g. Cooper and Dearing, in prep.). Regarding model exactness, a number of plausible expansions have been discussed in chapter 7.6, including the modelling of defector behaviours and pulse events. An alternative direction would be to use the horizon scans conducted here to guide more specialised modelling efforts. Whilst deep regions of dangerous spaces may be relatively certain, ecosystem-based models could target scenarios at the edge of the SJOSs, assessing whether simulating all trophic interactions shifts safe spaces and associated trade- offs (Daw et al., 2015). Likewise, the spatial disaggregation of multispecies migration, fleet dynamics and ecological conditions may explore how these heterogeneities influence the speed and cause of collapse.
Another logical direction would be to model adaptive governance, whereby decisions are only made in direct response to undesirable conditions. Adaptive governance could be operationalised through the use of model switches (e.g. IF catch < MSY THEN dredge outlet ELSE do nothing), exploring whether small yet frequent interventions help to dampen the variability found to occur under scenarios of decadal outlet openings and steep socioeconomic trajectories. The engineering resilience of constantly returning the system to equilibrium could be compared with the results found here, exploring the successes of persistent steering against implementing major decisions today.
280 Chapter 8: Conclusions
In reality, the model and framework can be expanded in almost every direction. EWSs have been overlooked due to the linear SJOSs definitions and the linear fish population declines over time. Therefore, future research could explore whether Chilika’s resource stock exhibits the variance and autocorrelation based EWSs found in the toy fishery model of Biggs et al. (2009), and if so, how their appearance relates to the space between dangerous normative and dangerous nonlinear thresholds. Ideally, EWSs will appear even before the normative dangerous space, prompting action based on time-series properties as well as linear thresholds. Disappearance of the fish population without EWSs would be the worst-case scenario (Hastings and Wysham, 2010), entering a hysteretic regime as a result of long-term albeit temporally linear resource degradation.
To further investigate the conclusion that normative SJOS are not always safe and just for everyone, a future study might enhance the number of constraints defining the safe and just future. Exploring the interplay between numerous environmental ceilings and social foundations helps scrutinise the trade-off between simply aiming for safe and just persistence and desiring relatively precise outcomes with narrower thresholds.
This research has developed and tested a dynamic social-ecological model to produce the first projections of the Chilika lagoon fishery system and operationalise interacting pathways to safe and just futures. This study may be ultimately expanded in breadth (e.g. process inclusion), depth (e.g. SJOS constraints) and application (e.g. different SESs); the transferability of the model and wider approaches makes these directions possible in the presence of new data, emergent knowledge or research demands.
8.3 Closing thoughts
The development of the model and associated framework has been an incredibly interesting and fulfilling experience, stemming from Professor John Dearing’s desire to ‘build a model and crash it into the ground’. I initially aimed to model an entire deltaic SES, focusing on the wider water-energy-food nexus to capture interactions between deltaic natural fluxes and social functions like agriculture, fisheries and migration. I soon realised that the Chilika lagoon provided a deltaic subsystem with clearly defined boundaries, external processes and a problematic history. The final version of this model has vastly improved early iterations which barely captured the
281 Chapter 8: Conclusions
1990s’ collapse (appendix C.4); access to qualitative and quantitative data during the fieldtrip to Odisha in spring 2016 was invaluable in this respect, and I am exceptionally grateful to my supervisors, DECCMA and the wider Indian team for this opportunity. The discussions of uncertainties concede there are undoubtedly aspects of this study that could be improved with further research. However, this work paves the way for future studies to develop models that seek to uncover safe and just dimensions of regional SESs. Likewise, the SJOSs designed here are intended to add to the suite of tools building resilience across SESs, particularly where the plausible future driver trajectories, outcome responses and trade-offs represent novel concerns for system governors and stakeholders alike.
282 Appendices
Appendices
Appendix A – Interview information sheet and question matrices
A.1 Participant information sheet
A.2 Key informant
A.3 Stakeholder
Appendix B – Model parameters
Appendix C – Model evaluation
C.1 Qualitative parameter assessment
C.2 Quantitative extreme conditions
C.3 External trajectory combinations
C.4 Early model evaluation
Appendix D – Model outputs
D.1 Colour-blind friendly graphs
D.1.1 Normative catch
D.1.2 Catch predictability
D.1.3 Driver pathways
D.1.4 Wider social-ecological outcomes
D.2 Mean and variability of catch pathways
D.3 Safe and just spaces correlation analysis
D.4 Wider core spaces
D.5 Autocorrelation and partial autocorrelation functions
283 Appendices
Appendix A – Interview information sheet and question matrices
A.1 Participant information sheet
Study Title: Social-ecological tipping points in world deltas: designing a safe and just operating space for the Chilika lagoon fishery, India
Researcher: Gregory Cooper Ethics number: 17721
What is the research about?
The purpose of this research is to discover the key natural and human factors affecting past, present and future fishery productivity across the Chilika Lagoon. This study will learn from the experiences and expectations of local and regional stakeholders and experts, to help model future changes in fishery production across the Chilika system under different biophysical and human conditions.
Who is the researcher?
The researcher is a postgraduate student at the University of Southampton in the United Kingdom (UK). This project forms part of the training which will lead to a Doctorate of Philosophy (PhD) in the subject of Geography. Specific research interests include modelling complex social-ecological and natural resource extraction systems of river deltas.
Why have I been chosen?
The participants in this project are all adults from different stakeholder backgrounds within the Chilika Lagoon region. The researcher is looking to speak to and learn from people with a variety of different experiences of working, living and governing in the Chilika Lagoon environment.
What does the interview involve?
If you agree to take part, the interview can either happen right away or at a later arranged time. The location is up to you, but it is best to choose somewhere quiet and without disturbance, and where you will feel comfortable. The interview will take no longer than 45 minutes. There is no financial cost to take part. If you
284 Appendices consent, the questions will focus on your past experiences fishing in the Chilika Lagoon and the expectations you have about your future abilities to fish. An interpreter will be part of our conversation, so that we can all effectively communicate. The conversation will be recorded for later analysis; only the researcher has access to the recordings.
Are there any benefits in my taking part?
I hope that you will find taking part an interesting experience. The results of this research may be useful to fishers, scientists and regional managers, looking to sustain future fishery production across the Chilika Lagoon by controlling the relevant natural and human influences. Although there is much research on the health of the Chilika Lagoon, there are few studies which project the future sustainability of the fishery, given the likely evolution of the lagoon’s natural and human conditions.
Are there any risks involved?
There are no risks in terms of your safety and you are not obliged to talk about any experiences you feel uncomfortable discussing or find distressing.
Will my participation be confidential?
Confidentiality is very important in this project. The recording and any documents will be stored on a computer and they will be password protected so that they cannot be accessed by anyone other than the researcher. In any written documents your name and the names of anyone else you mention (such as family members or work colleagues) will be changed, as will any other details by which you could be identified. Information will be kept safe in line with UK laws (the Data Protection Act) and University of Southampton policy. The interpreter in the interview has signed a confidentiality agreement.
What happens if I change my mind?
You have the right to withdraw at any time and this will not have an effect on any of your rights. Please contact the people below to withdraw from the study.
What happens if something goes wrong?
If you have any concerns or complaints about how this research is conducted, you may contact:
285 Appendices
Research and Governance Office
University of Southampton, UK
+44 2380 594684 [email protected]
If you would prefer to speak with someone directly in India in order to raise your concerns or find out more information about the research, you can contact:
Dr Sugata Hazra,
Jadavpur University,
India.
Email: [email protected].
Tel.: +913324146242.
Where can I get more information?
If you have any further questions once you have read this information sheet, please get in touch with me using the following details:
Gregory Cooper
University of Southampton, UK
286 Appendices
A.2 Key informant
The matrix below details the range of questions asked to the key informants interviewed (chapter 3.3.1) from the Chilika Development Authority (CDA), Odisha Integrated Coastal Zone Management Project (ICZMP), Odisha Government Directorate of Fisheries and Chilika Fisherman’s Central Cooperative Society. Due to the semi-structured nature of the interviews, the exact combination of questions differed between interviews as different issues were raised.
Topic Questions
Understanding the What is your occupation? background of the What role does your occupation have in the key informant governance/study of Chilika? How long have you been in your role?
Chilika’s past, Which do you consider to be the more important present and future determinant of Chilika’s fish catch, social-economic or biophysical processes? Which do you consider to have a more significant impact upon fish catch levels across Chilika: resource availability or resource accessibility? Over the past fifty years, what have been the main drivers of Chilika’s fish catch? What is the lagoon’s current ecological status? If you could ensure that one stress never becomes unsustainable or reaches a dangerous level, what would it be? If I am to visit Chilika in 20 years, how do you think the system will look different? Is there the potential for a similar shift from a high to a low productivity regime to happen again in future? How do you expect the natural conditions of the
287 Appendices
lagoon to evolve in future?
Plausible Are there any difficulties implementing regulation governance across Chilika? options If yes, do you see the difficulties relaxing in future? Talk to me about the 1990’s fishery collapse: for example, what were the main reasons for the decline and what were the main management strategies ‘x’ implemented? Do you believe the lagoon’s ecorestoration will have long- lasting positive effects? What influence over catchment processes does Chilika governance have? How do you envisage governance to react to your expected future changes?
288 Appendices
A.3 Stakeholder
The matrix below details the questions asked to the stakeholders interviewed (chapter 3.3.2) from the four fishing villages. The original matrix included more questions that were relatively complex (e.g. what do you believe were the main causes of the 1990s collapse?); however questions were revised after consultation with local informants to instead focus on the everyday knowledge of fishers.
Topic Questions
Background What is your occupation? How long have you fished in Chilika? questions Where do you fish? What is your monthly income? How many people in your family fish? How many people in your family are educated?
Fishery perceptions What are the commercially important species? What are the prices of the species? What are the greatest threats to Chilika’s fish? What do you think about the actions of the government (e.g. new tidal outlet)?
Future plans What are your future plans? Are there any other livelihood options? Do your children plan to fish?
289 Appendices
Appendix B – Model equations and constants
Future trajectories
DIESEL PRICE PER LITREt (Indian Rupees[INR]/litres) = 퐈퐅 퐓퐈퐌퐄 < 504 퐓퐇퐄퐍 OBSERVED DIESEL PRICEt 퐄퐋퐒퐄 DIESEL PRICEt−1 + DIESEL PRICE INCREASEt × DT
DIESEL PRICEt=504 = 60
DIESEL PRICE INCREASEt (INR) = 퐏퐔퐋퐒퐄(퐑퐀퐍퐃퐎퐌(퐌퐈퐍 = 0, 퐌퐀퐗 = 7), 퐈퐍퐈퐓퐈퐀퐋 퐓퐈퐌퐄 = 516, 퐅퐑퐄퐐퐔퐄퐍퐂퐘 = 12)
OBSERVED DIESEL PRICEt (INR) = Verman and Fernandez (2010) dataset
Where t equals time step in months.
FISH PRICE PER KGt (INR/KG) = 퐈퐅 퐓퐈퐌퐄 < 504 퐓퐇퐄퐍 OBSERVED FISH PRICEt 퐄퐋퐒퐄 FISH PRICEt−1 + FISH PRICE INCREASEt × DT
FISH PRICEt=504 (INR) = 133
FISH PRICE INCREASEt (INR) = 퐏퐔퐋퐒퐄(퐑퐀퐍퐃퐎퐌(0, 30), 516, 12)
OBSERVED FISH PRICEt (INR) = CDA (2016) dataset
FUTURE HUMAN POPULATION GROWTH RATE [FUTURE R]t (DMNSL) = 퐈퐅 퐓퐈퐌퐄 < 504 퐓퐇퐄퐍 OBSERVED Rt 퐄퐋퐒퐄 Rt−1 + R INCREASEt × DT
Rt=504 (Dimensionless[DMNSL]) = 14
R INCREASEt (DMNSL) = 퐏퐔퐋퐒퐄(퐑퐀퐍퐃퐎퐌(0, 0.022), 516, 12)
OBSERVED Rt (DMNSL) = OBSERVED BIRTH RATEt − OBSERVED DEATH RATEt
OBSERVED BIRTH RATEt (DMNSL) = OBSERVED BIRTH RATEt − OBSERVED DEATH RATEt = ‘India Development Indicators Revised, 2012’ dataset, published by the United Nations StatisticsDivision, and The 2013 − 14 Odisha Economic Survey
OBSERVED DEATH RATEt (DMNSL) = see OBSERVED BIRTH RATE
TEMPERATURE CHANGEt (°C) = 1 + TEMPERATURE CLIMATE CHANGE × TIME GRAPHICALt
TIME GRAPHICALt (months) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍 (DT) {(1,0), (504, 0), (1524, 1)}
TEMPERATURE CLIMATE CHANGE (DMNSL) = 퐑퐀퐍퐃퐎퐌(0, 0.2)
RAINFALL CHANGEt (%) = 1 + RAINFALL CLIMATE CHANGE × TIME GRAPHICALt
290 Appendices
RAINFALL CLIMATE CHANGE (DMNSL) = 퐑퐀퐍퐃퐎퐌(0, 0.22)
Governance
PERSISTENT OUTLET OPENINGt (tonnes) = 퐈퐅 퐓퐈퐌퐄 = 576, 696, 816, 936, 1056) 퐓퐇퐄퐍 0.3 × CHILIKA SEDIMENT STOCKt 퐄퐋퐒퐄 0
LAGOON AREA (km2) = 1000
OUTLET AREA (km2) = 70
LAGOON AREA FISHING BAN (DMNSL) = 1 − (( ) × IMPLEMENTATION PROCESS × BAN ACTIVE SCENARIO) t ANNUAL AVERAGE AREA t
IMPLEMENTATION PROCESSt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(DT) {(1,0), (516,0), (636,1), (1056,1)}
BAN ACTIVE SCENARIO (DMNSL) = can equal 0 or 1, as determined by modeller.
ALTERNATIVE LIVELIHOOD UPTAKE (DMNSL) = 퐈퐅 ALTERNATIVE LIVELIHOOD ACTIVE = 1 퐓퐇퐄퐍 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍 (DT) {(1,0), (515,0), (516,0.001), (1056,0.001)}퐄퐋퐒퐄 0
ALTERNATIVE LIVELIHOOD ACTIVE (DMNSL) = can equal 0 or 1, as determined by modeller.
Fish population
Population renewal
IMMATURE POPULATION STOCKt (fish) = 퐈퐍퐓퐄퐆 (IMMATURE POPULATION STOCKt−1 + (REGENERATION RATEt − MATURATION RATEt − JUVENILE CATCHi,t) 퐝퐭
Where i is the boat type (i.e. traditional or non-traditional/motorised fleets).
IMMATURE POPULATION STOCKt=0(fish) = 67,500,000
REGENERATION RATEt = (BIRTHSt × MODIFIED SURVIVAL RATEt) − (MIGRATIONAL FRACTION × BIRTHSt × MODIFIED SURVIVAL RATEt × OUTLET CLOSUREt)
MIGRATIONAL FRACTION (DMNSL) = 0.7
MATURATION RATEt (fish) = 퐃퐄퐋퐀퐘(BIRTHSt, ퟏퟐ, ퟏퟐ) − (IMMATURE CATCHi,t + BIRTHS LOSTt)
IMMATURE CATCHi,t (fish) VEGETATION AREA = IMMATURE FISHING EFFORT × (1 − VEGETATION COEFFICIENT × t) i,t LAGOON AREA
VEGETATION COEFFICIENT (DMNSL) = 0.4
291 Appendices
FECUNDITY BIRTHS (fish) = MATURE POPULATION × FEMALE PERCENTAGE × ( ) t t REPRODUCTIVE LIFE
BIRTHS LOSTt(fish) = (MIGRATIONAL FRACTION × BIRTHS × MODIFIED SURVIVAL RATE × OUTLET CLOSUREt)
FECUNDITY (DMNSL) = 105
FEMALE PERCENTAGE (DMNSL) = 0.5
REPRODUCTIVE LIFE (months) = 108
SEDIMENT STOCK OUTLET CLOSURE (DMNSL) = t t SEDIMENT CAPACITY
SEDIMENT CAPACITY (tonnes) = 44E6
NATURAL SURVIVAL RATEt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(FISH DENSITYt)
{(0.00, 0.50), (0.10, 0.49), (0.20, 0.48), (0.40, 0.46), (0.60, 0.42), (0.80, 0.34),
(0.90, 0.25), (0.95, 0.15), (1.00, 0.00), (1.50, 0.00), (2.00, 0.00)}
TOTAL FISH POPULATION FISH DENSITY (DMNSL) = ( t ) 퐭 ECOLOGICAL CARRYING CAPACITY
ECOLOGICAL CARRYING CAPACITYt (fish) = 675,000,000
MODIFIED SURVIVAL RATE t (DMNSL) = NATURAL SURVIVAL RATEt × HABITAT QUALITY IMPACTt
Mature fish dynamics
MATURE POPULATION STOCK t(fish) = 퐈퐍퐓퐄퐆 (MATURE POPULATION STOCKt−1 + (MATURATION RATEt + MATURE INMIGRATIONt − MATURE CATCHi,t − NATURAL DEATH
− MATURE OUTMIGRATIONt) 풅풕
MATURE POPULATION STOCKt=0(fish) = 608,000,000
MATURE INMIGRATIONt (fish) = MATURE OUTMIGRATIONt × (1 − OUTLET CLOSUREt)
(MIGRATIONAL FRACTION × MATURE POPULATION ) MATURE OUTMIGRATION (fish) = t t 12
MATURARE POPULATION STOCK NATURAL DEATHS (fish) = t LIFE LENGTH
LIFE LENGTH (months) = 120
Fish catch
292 Appendices
TOTAL FISH POPULATIONt (fish) = MATURE POPULATION STOCKt + IMMATURE POPULATION STOCKt
VEGETATION AREA MATURE CATCH (fish) = MATURE FISHING EFFORT × (1 − t) i,t i,t LAGOON AREA
MATURE FISH PER KG (fish/kg) = 25
IMMATURE FISH PER KG (fish/kg) = 50
MATURE CATCH MATURE FISH LANDING (kg) = i,t i,t 1000 × MATURE FISH PER KG
IMMATURE CATCH IMMATURE FISH LANDING (kg) = i,t i,t 1000 × IMMATURE FISH PER KG
Fish landings then converted to tonnes (i.e. 1000 kg).
TOTAL LANDINGi,t (tonnes) = MATURE CATCHi,t + IMMATURE CATCHi,t
Socioeconomic
Fishing efforts
TRADITIONAL CATCH CAPACITY (fish/month) = 6500
MOTORISED CATCH CAPACITY (fish/month) = 8000
MATURE FISHING EFFORTi,t (fish/boat/month) = FISH DENSITYt × MATURE HARVEST FORCEi,t
MATURE HARVEST FORCEi,t(fish/boat/month) = DISCOUNT RATEi,t × CATCH CAPACITYi,t × BOATSi,t × FISHING BANt
MOTORISED DISCOUNT RATEt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(MOTORISED PERCEIVED FISH DENSITYt)
{(0.10, 0.64), (0.20, 0.78), (0.30, 0.48), (0.40, 0.88), (0.50, 0.90), (0.60, 0.92),
(0.70, 0.93), (0.80, 0.94), (0.90, 0.95), (1.00, 0.95)}
TRADITIONAL DISCOUNT RATEt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(TRADITIONAL PERCEIVED FISH DENSITYt)
{(0.10, 0.66), (0.20, 0.68), (0.30, 0.70), (0.40, 0.70), (0.50, 0.71),
(0.60, 0.71), (0.70, 0.71), (0.80, 071), (0.90, 0.71), (1.00, 0.71)}
PERCEIVED FISH DENSITYi,t (DMNSL) = 퐃퐄퐋퐀퐘(FISH DENSITYt, 1)
IMMATURE FISHING EFFORTi,t (DMNSL) = FISH DENSITYt × IMMATURE HARVEST FORCEi,t
TRADITIONAL IMMATURE HARVEST FORCEt (fish/boat/month) = (0.71 − TRADITIONAL DISCOUNT RATEt) × TRADITIONAL CATCH CAPACITYt × TRADITIONAL BOATSt)
293 Appendices
MOTORISED IMMATURE HARVEST FORCEt(fish/boat/month) = (0.95 − MOTORISED DISCOUNT RATEt) × MOTORISED CATCH CAPACITYt × MOTORISED BOATSt)
(BOAT OWNERHIP RATE × FISHER POPULATION ) BOATS (boats) = i,t i,t FISHERS PER BOATS
FISHERS PER BOATS (fishers) = 6
BOAT OWNERHIP RATE (DMNSL) = 0.95
TRADITIONAL FISHERSt (fishers) = 퐈퐍퐓퐄퐆 (TRADITIONAL FISHERSt−1 + (TRADITIONAL CARRYING CAPACITY CHANGEt − TRADITIONAL UPGRADEt − (ALTERNATIVE LIVELIHOOD UPTAKE × TRADITIONAL FISHERSt−1))) 퐝퐭
TRADITIONAL FISHERSt=0 (fishers) = 19,000
TRADITIONAL CARRYING CAPACITY CHANGEt (fishers) = 퐈퐅 TRADITIONAL FISHERSt
> TRADITIONAL CARRYING CAPACITYt 퐓퐇퐄퐍 퐒퐌퐎퐎퐓퐇 (FISHER Rt × TRADITIONAL FISHERSt
1 − TRADITIONAL FISHERSt × ( ) , ퟐퟏퟔ) 퐄퐋퐒퐄 (FISHER Rt × TRADITIONAL FISHERSt) TRADITIONAL CARRYING CAPACITYt
TOTAL FLEET REVENUEi,t LIVEHOOD CARRYING CAPACITYi,t(fishers) = LIVELIHOOD COST PER FISHER i,t
FISHER Rt (DMNSL) = 퐈퐅 퐓퐈퐌퐄 (OBSERVED BIRTH RATE − OBSERVED DEATH RATE ) FUTURE R < 506 퐓퐇퐄퐍 t t 퐄퐋퐒퐄 t 12000 12000
MOTORISED FISHERSt(fishers) = 퐈퐍퐓퐄퐆 (MOTORISED FISHERSt−1 + (MOTORISED CARRYING CAPACITY CHANGEt + TRADITIONAL UPGRADEt − (ALTERNATIVE LIVELIHOOD UPTAKE × MOTORISED FISHERSt−1))) 퐝퐭
MOTORISED FISHERSt=0(fishers) = 0
MOTORISED CARRYING CAPACITY CHANGEt (fishers) = 퐈퐅 MOTORISED FISHERSt
> MOTORISED CARRYING CAPACITYt 퐓퐇퐄퐍 퐒퐌퐎퐎퐓퐇 (FISHER Rt × MOTORISED FISHERSt
1 − MOTORISED FISHERSt × ( ) , ퟐퟏퟔ) 퐄퐋퐒퐄 (FISHER Rt × MOTORISED FISHERSt) MOTORISED CARRYING CAPACITYt
TOTAL FISHERSt (fishers) = MOTORISED FISHERSt + TRADITIONAL FISHERSt
TOTAL LIVELIHOOD CARRYING CAPACITYt (fishers) = TRADITIONAL CARRYING CAPACITYt +
MOTORISED CARRYING CAPACITYt
Fisher incomes
TOTAL FLEET REVENUEi,t(INR) = TOTAL LANDINGi,t × FISH PRICE PER KG t
TOTAL FLEET REVENUEi,t PER CAPITA INCOMEi,t (INR/fisher) = FISHER POPULATIONi,t
294 Appendices
PER CAPITA PROFITSi,t (INR/fisher) = PER CAPITA INCOMEi,t − PER CAPITA COSTSi,t
Fisher costs
MOTORISED LIVELIHOOD COSTSt (INR/fisher/month) = (MOTORISED MONTHLY COSTSt + MOTORISED ONE OFF COSTSt + DIESEL USE PRICEt) × MOTORISED DISCOUNT RATEt
(DIESEL PRICE PER LITRE × LITRES USED PER DAY × 30) DIESEL USE PRICE (INR) = t t FISHERS PER BOAT
LITRES USED PER DAY (litres/day) = 3.7
MOTORISED MONTHLY COSTSt (INR/month) (MOTORISED LICENCE COST + NET REPAIR COST + BOAT REPAIR COST + ENGINE REPAIR COST) = t t 12 × FISHERS PER BOAT
BOAT REPAIR COSTt (INR) = 3970 × COST INCREASEt
NET REPAIR COSTt (INR) = NET REPAIR COST PER UNITt × NET USAGE
NET REPAIR COST PER UNITt (INR) = 130 × COST INCREASEt
ENGINE REPAIR COSTt (INR) = 5480 × COST INCREASEt
(LIFETIME ENGINE COSTt + LIFETIME BOAT COSTt + LIFETIME NETS COSTt +LIFETIME MOTORISED LICENCE COST ) MOTORISED LIFETIME COSTS (INR) = t t FISHERS PER BOAT
COST INCREASEt (INR) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍 (TIME) {(1,0.34), (492,1), (1056,1.86)}
ONE OFF ENGINE COST LIFETIME ENGINE COST (INR) = t t ENGINE LIFETIME
ONE OFF ENGINE COSTt (INR) = 18000 × COST INCREASEt
ENGINE LIFETIME (months) = 98.4
ONE OFF BOAT COST LIFETIME BOAT COST (INR) = t t BOAT LIFETIME
ONE OFF BOAT COSTt (INR) = 38000 × COST INCREASEt
BOAT LIFETIME = 127.2 (months)
ONE OFF NET COST LIFETIME NETS COST (INR) = t t NET LIFETIME
NET LIFETIME (months) = 70.8
ONE OFF NET COSTt (INR) = NET USAGE × NET USAGE COSTt
NET USAGE (kg) = 63
NET USAGE COSTt (INR) = 680 × COST INCREASEt
295 Appendices
MOTORISED LICENCE COST LIFETIME MOTORISED LICENCE COST (INR) = t t LICENCE LIFETIME
MOTORISED LICENCE COSTt (INR) = 235
LICENCE LIFETIME (month) = 127.2
MONTHLY COST PER TRADITIONAL FISHERt (INR/fisher/month) = (TRADITIONAL MONTHLY COSTSt + TRADITIONAL ONE OFF COSTSt) × TRADITIONAL DISCOUNT RATEt
TRADITIONAL LIVELIHOOD COSTSt (INR/fisher/month) (TRADITIONAL LICENCE COST + NET REPAIR COST + BOAT REPAIR COST ) = t t 12 × FISHERS PER BOAT
TRADITIONAL LICENCE COSTt (INR) = 210
Switch from traditional to motorised boats (upgrades)
TRADITIONAL UPGRADEt(fishers) = TRADITIONAL FISHERSt × PROPORTION UPGRADINGt
PROPORTION UPGRADINGt(DMNSL) = 퐈퐅 퐓퐈퐌퐄 < 120 퐓퐇퐄퐍 0 퐄퐋퐒퐄 WILLINGNESS TO UPGRADEt × UPGRADE PROBABILITYt
STANDARD DEVIATION PROFITSt (INR) 260 = ( ) × DAYS IN MONTH × TRADITIONAL DISCOUNT RATE FISHERS PER BOAT t
PER CAPITA UPGRADE COSTt (INR/fishers) (ONE OFF ENGINE COSTt + (MOTORISED LICENCE COSTt − TRADITIONAL LICENCE COSTt) +BOAT REPAIR COST ) = t FISHERS PER BOAT + DIESEL USE PRICEt
Z − SCORE TRADITIONAL PROFITSt (DMNSL) (TRADITIONAL PER CAPITA PROFITS − PER CAPITA UPGRADE COST ) = t t STANDARD DEVIATION PROFITSt
UPGRADE PROBABILITYt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(Z − SCORE TRADITIONAL PROFITSt) {(-3.40, 0.0003), ,(-3.20, 0.0007), (-3.00, 0.0013), (-2.80, 0.0026), (-2.60, 0.0047), (-2.40, 0.0082), (-2.20, 0.0139), (-2.00, 0.0228), (-1.80, 0.0359), (-1.60, 0.0548), (-1.40, 0.0808), (-1.20, 0.1151), (-1.00, 0.1687), (-0.80, 0.2119), (-0.60, 0.2743), (-0.40, 0.3446), (-0.20, 0.4207), (0.00, 0.500), (0.20, 0.5793), (0.40, 0.6554), (0.60, 0.7257), (0.80, 0.7881), (1.00, 0.8413), (1.20, 0.8849), (1.40, 0.9192), (1.60, 0.94452), (1.80, 0.9641), (2.00, 0.9772), (2.20, 0.9861), (2.40, 0.9918), (2.60, 0.9953), (2.80, 0.9974), (3.00, 0.9987), (3.20, 0.9993), (3.40, 0.9997)}
WILLINGNESS TO UPGRADE (DMNSL) = 퐈퐅 퐓퐈퐌퐄 < 132 퐎퐑 ∆ TRADITIONAL PER CAPITA PROFITS > 0 퐓퐇퐄퐍 0 퐄퐋퐒퐄 1
∆ TRADITIONAL PROFITS (INR) = TRADITIONAL PER CAPITA PROFITSt − TRADITIONAL PER CAPITA PROFITSt−1
Eco-hydrological
Aquatic vegetation
296 Appendices
2 AQUATIC VEGETATION STOCKt (km ) = 퐈퐍퐓퐄퐆 (AQUATIC VEGETATION STOCKt−1 + VEG INCREASEt − VEG DECREASEt) 퐝퐭
2 AQUATIC VEGETATION STOCKt=0 (km ) = 20
2 VEG INCREASEt (km ) = 퐈퐅 SALINITY < 18 퐓퐇퐄퐍 (18 − SALINITY) × DO × OUTLET CLOSUREt × 0.039 퐄퐋퐒퐄 0
2 1 ퟏ VEG DECREASEt (km ) = 퐈퐅 SALINITY > 14 퐓퐇퐄퐍 (2.5 × SALINITY − 14) × ( ) × ( ) 퐄퐋퐒퐄 0 DO OUTLET CLOSUREt
Dissolved oxygen (DO)
BASELINE DO (ppm) = 7.4
DOt (ppm) = 퐃퐄퐋퐀퐘(퐒퐌퐎퐎퐓퐇 ((BASELINE DO D + POSTMONSOON DO DEVIATION AQUATIC VEGETATION STOCK − NONMONSOON DO DEVIATION) × (1 − ) , ퟑ) , ퟒ) 15
3 MONSOON FRESHWATER FLOW (m ) = 퐈퐅 FRESHWATER INFLOWt ≥ 800E6 퐓퐇퐄퐍 FRESHWATER INFLOWt 퐄퐋퐒퐄 0
NORMALISED MONSOON FRESHWATER FLOW (DMNSL) = 퐈퐅 MONSOON FRESHWATER INFLOWt (MONSOON FRESHWATER INFLOW − 1000E6) > 1 퐓퐇퐄퐍 t 퐄퐋퐒퐄 0 1000E6
MONSOON DO DEVIATION (ppm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(NORMALISED MONSOON FRESHWATER INFLOWt){(0.0, 1.95), (0.40, 2.27), (0.70, 2.53), (1.10, 2.75), (1.40, 2.95), (1.80, 3.11), (2.10, 3.20), (2.50, 3.33), (2.85, 3.42), (3.20, 3.49), (3.60, 3.56), (3.90, 3.56), (4.30, 3.62), (4.65, 3.64), (5.00, 3.64)}
3 NONMONSOON FRESHWATER FLOW(m ) = 퐈퐅 FRESHWATER INFLOWt < 800E6 퐓퐇퐄퐍 FRESHWATER INFLOWt 퐄퐋퐒퐄 0
NORMALISED NONMONSOON FRESHWATER INFLOWt (DMNSL) = 퐈퐅 NONMONSOON FRESHWATER INFLOWt (NONMONSOON FRESHWATER INFLOW − 113E6) > 1 퐓퐇퐄퐍 t 퐄퐋퐒퐄 0 113E6
NONMONSOON DO DEVIATION (ppm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(NORMALISED NONMONSOON FRESHWATER INFLOWt){(0.0, 0.50), (0.80, 0.49), (1.60, 0.46), (2.40, 0.43), (3.20, 0.39), (4.00, 0.35), (4.80, 0.30), (5.60, 0.25), (6.40, 0.19), (7.20, 0.13), (8.00, 0.00)}
Salinity
OUTLET CLOSURE SALINITY EFFECTt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐒퐌퐎퐎퐓퐇(OUTLET CLOSUREt), ퟒ){(0.00, 0.286), (0.11, 0.38), (0.22, 0.46), (0.33, 0.54), (0.44, 0.63), (0.56, 0.71), (0.67, 0.79), (0.78, 0.86), (0.89, 0.93), (1.00, 1.00)}
3 PREMONSOON FRESHWATER INFLOWt (m ) = 퐈퐅 FRESHWATER INFLOWt ≤ 350E6 퐓퐇퐄퐍 FRESHWATER INFLOWt 퐄퐋퐒퐄 ퟎ
297 Appendices
NORMALISED PREMONSOON FRESHWATER INFLOWt (DMNSL) (PREMONSOON FRESHWATER INFLOW − 74.0E6) = t 74.0E6
PREMONSOON SALINITY GRAPHt (ppt) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(NORMALISED PREMONSOON FRESHWATER INFLOWt){(-1.00, 0.99), (-0.94, 0.99), (-0.885, 0.99), (-0.81, 0.98), (-0.75, 0.95), (-0.69, 0.93), (-0.63, 0.91), (-0.56, 0.88), (-0.50, 0.82), (- 0.44, 0.72), (-0.38, 0.63), (-0.31, 0.55), (-0.25, 0.45), (-0.19, 0.36), (-0.13, 0.25), (-0.06, 0.12), (0.00, 0.01)}
PREMONSOON SALINITY DEVIATIONt (ppt) = 퐈퐅 NORMALISED PREMONSOON FRESHWATER INFLOWt = −ퟏ 퐀퐍퐃 SEASON = 2 퐓퐇퐄퐍 0 퐄퐋퐒퐄 PREMONSOON SALINITY GRAPHt
3 MONSOON AND POST MONSOON (MAPM) FRESHWATER INFLOW (m ) = 퐈퐅 FRESHWATER INFLOWt > 350E6 퐓퐇퐄퐍 FRESHWATER INFLOWt 퐄퐋퐒퐄 ퟎ
FRESHWATER SEASONt (DMNSL) = 퐈퐅 MAPM FRESHWATER INFLOWt > 350E6 퐓퐇퐄퐍 1 퐎퐑 3 ELSE 2
(MAPM FRESHWATER INFLOW − 576E6) NORMALISED MAPM FRESHWATER INFLOW (DMNSL) = t t 576E6
MAPM SALINITY GRAPHt (ppt) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(NORMALISED MAPM FRESHWATER INFLOWt){(-1.00, 0.00), (-0.50, 0.18), (0.00, 0.36), (0.50, 0.48), (1.00, 0.58), (1.50, 0.70), (2.00, 0.78), (2.50, 0.84), (3.00, 0.89), (3.50, 0.92), (4.00, 0.95), (4.50, 0.96), (5.00, 0.98), (5.50, 0.98), (6.00, 0.98)}
MAPM SALINITY DEVIATIONt (ppt) = 퐈퐅 FRESHWATER SEASONt ≠ 2 퐓퐇퐄퐍 11 × MAPM SALINITY GRAPHt 퐄퐋퐒퐄 0
SALINITYt (ppt) = (12.6 + (PREMONSOON SALINITY DEVIATIONt − MAPM SALINITY DEVIATIONt) × (1 − OUTLET CLOSURE SALINITY EFFECTt)
WATER TEMPERATUREt (°C) = (0.80 × AIR TEMPERATUREt + 7.94) − MONSOON BURSTt
FRESHWATER INFLOW MONSOON BURST (°C) = 퐈퐅 FRESHWATER INFLOW > 1000E6 퐓퐇퐄퐍 0.5 × ( t) 퐄퐋퐒퐄 0 t t 1000E6
Habitat quality modifications on fish survival
DO SURVIVAL MODIFICATION (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(DOt){(0.00, 0.000), (3.00, 0.700), (5.00, 0.900), (7.00, 1.000), (10.00, 1.000)}
TEMPERATURE SURVIVAL MODIFICATION (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(WATER TEMPERATUREt){(25.00, 1.000), (31.00, 1.000), (35.00, 0.800), (45.00, 0.000)}
SALINITY SURVIVAL MODIFICATION (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(SALINITY)(0.00, 0.200), (5.00, 0.700), (8.00, 0.900), (15.00, 1.000), (20.50, 1.000)}
HABITAT QUALITY IMPACTt (DMNSL) = DO SURVIVAL MODIFICATION × TEMPERATURE SURVIVAL MODIFICATION × SALINITY SURVIVAL MODIFICATION
Hydroclimatic
Temperatures and potential evapotranspiration (PET)
298 Appendices
Observed PET across the districts of Chilika’s Western Catchments (WC)
KHORDHA OBSERVED PETt (mm) = IMD monthly PET data 1973 − 2002
PURI OBSERVED PETt (mm) = IMD monthly PET data 1973 − 2002
Observed PET across the districts of the Lower Mahanadi Catchment (LMC)
ANGUL OBSERVED PETt (mm) = IMD monthly PET data 1973 − 2002
BAUDH OBSERVED PETt (mm) = IMD monthly PET data 1973 − 2002
SAMBALPUR OBSERVED PETt (mm) = IMD monthly PET data 1973 − 2002
SONEPUR OBSERVED PETt (mm) = IMD monthly PET data 1973 − 2002
Generating future temperatures across Chilika’s WC
CLIMATE MONTHt (DMNSL) = Vector from 1 → 12 that repeats over the length of each simulation
DAYS IN MONTHt (DMSNL) = Vector of 12 numbers with the number of days per month (no leap year) repeated over the simulation
JAN TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 1 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 18.500), (0.200, 19.300), (0.400, 19.700), (0.600, 20.700), (0.800, 21.000), (1.000, 22.300)}
FEB TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 2 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 20.500), (0.200, 21.700), (0.400, 22.300), (0.600, 22.500), (0.800, 23.000), (1.000, 24.000)}
MAR TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 3 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 23.200), (0.200, 24.400), (0.400, 24.800), (0.600, 25.250), (0.800, 15.800), (1.000, 26.200)}
APR TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 24.700), (0.200, 26.000), (0.400, 26.700), (0.600, 26.900), (0.800, 27.000), (1.000, 28.300)}
MAY TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 5 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 25.600), (0.200, 26.500), (0.400, 27.500), (0.600, 28.000), (0.800, 28.500), (1.000, 29.000)}
JUN TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 6 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 25.100), (0.200, 26.700), (0.400, 27.300), (0.600, 27.700), (0.800, 27.900), (1.000, 29.000)}
JUL TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 7 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 24.600), (0.200, 25.300), (0.400, 26.000), (0.600, 26.500), (0.800, 26.700), (1.000, 28.000)}
299 Appendices
AUG TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 8 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 24.500), (0.200, 25.700), (0.400, 25.950), (0.600, 26.400), (0.800, 26.700), (1.000, 27.000)}
SEP TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 9 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 25.600), (0.200, 26.200), (0.400, 26.400), (0.600, 26.500), (0.800, 26.700), (1.000, 27.100)}
OCT TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 10 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 23.200), (0.200, 24.600), (0.400, 25.400), (0.600, 25.700), (0.800, 25.900), (1.000, 26.600)}
NOV TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 11 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 21.200), (0.200, 21.600), (0.400, 22.700), (0.600, 23.000), (0.800, 23.400), (1.000, 24.400)}
DEC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 12 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 18.500), (0.200, 19.600), (0.400, 20.000), (0.600, 20.300), (0.800, 20.600), (1.000, 21.700)}
Generating future temperatures across the LMC
JAN LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 1 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 18.900), (0.200, 19.500), (0.400, 19.800), (0.600, 20.700), (0.800, 21.000), (1.000, 22.200)}
FEB LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 2 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 21.800), (0.200, 22.700), (0.400, 23.200), (0.600, 23.500), (0.800, 23.900), (1.000, 24.800)}
MAR LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 3 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 25.600), (0.200, 26.200), (0.400, 26.750), (0.600, 27.300), (0.800, 27.600), (1.000, 27.800)}
APR LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 28.500), (0.200, 29.100), (0.400, 28.800), (0.600, 30.000), (0.800, 30.600), (1.000, 32.000)}
MAY LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 20.250), (0.200, 30.250), (0.400, 20.800), (0.600, 31.250), (0.800, 31.500), (1.000, 32.400)}
JUN LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 27.200), (0.200, 28.100), (0.400, 28.700), (0.600, 29.300), (0.800, 29.700), (1.000, 30.400)}
JUL LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 25.200), (0.200, 26.200), (0.400, 26.400), (0.600, 26.800), (0.800, 27.200), (1.000, 28.300)}
AUG LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 8 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 24.900), (0.200, 26.200), (0.400, 26.400), (0.600, 26.700), (0.800, 26.900), (1.000, 27.300)}
300 Appendices
SEP LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 9 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 25.600), (0.200, 26.250), (0.400, 26.400), (0.600, 25.600), (0.800, 26.900), (1.000, 27.400)}
OCT LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 10 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 22.800), (0.200, 24.300), (0.400, 24.600), (0.600, 25.000), (0.800, 25.200), (1.000, 26.000)}
NOV LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 11 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 19.800), (0.200, 21.100), (0.400, 21.300), (0.600, 22.000), (0.800, 22.300), (1.000, 24.000)}
DEC LMC TEMPERATURE ECDF (°C) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 12 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 18.200), (0.200, 18.900), (0.400, 19.200), (0.600, 19.800), (0.800, 20.300), (1.000, 21.100)}
Calculating future PET across Chilika’s WC
OBSERVED WC PET FUTURE WC PET WC PET (mm) = 퐈퐅 퐓퐈퐌퐄 < 324 퐓퐇퐄퐍 t 퐄퐋퐒퐄 t t 1000 1000
OBSERVED WC PETt(mm) = 퐌퐄퐀퐍(KHORDHA OBSERVED PETt, KHORDHA OBSERVED PETt)
(700 × FUTURE WC Tm) ( + (15 × FUTURE WC )) (100 − WC LATITUDE) t−td FUTURE WC PETt (mm) = (80 − FUTURE WC TEMPERATUREt)
FUTURE WC TEMPERATUREt (°C) = 퐒퐔퐌(JAN WC TEMPERATURE ECDF, FEB WC TEMPERATURE ECDF, … , NOV WC TEMPERATURE ECDF, DEC WC TEMPERATURE ECDF) × TEMPERATURE CHANGEt
FUTURE WC Tmt (°C) = FUTURE WC TEMPERATUREt + 0.006 × WC ELEVATION
FUTURE WCt−td (°C) = 0.0023 × 3 + 0.37 × FUTURE WC TEMPERATUREt + 0.53 × WC DAILY TEMP RANGE + 0.35 × WC MONTHLY TEMP RANGE − 10.9
WC ELEVATION (m) = 5
WC LATITUDE (°) = 19.45
WC DAILY TEMP RANGE (°C) = 8.7
WC MONTHLY TEMP RANGE (°C) = 7.4
Calculating future PET across the LMC
OBSERVED LMC PET FUTURE LMC PET LMC PET (m) = 퐈퐅 퐓퐈퐌퐄 < 324 퐓퐇퐄퐍 t 퐄퐋퐒퐄 t t 1000 1000
OBSERVED LMC PETt (mm) = 퐌퐄퐀퐍(ANGUL OBSERVED PETt, BAUDH OBSERVED PETt,
SAMBALPUR OBSERVED PETt, SONEPUR OBSERVED PETt)
301 Appendices
(700 × FUTURE LMC Tm) ( + (15 × FUTURE LMC )) (100 − LMC LATITUDE) t−td FUTURE LMC PETt (mm) = (80 − FUTURE LMC TEMPERATUREt)
FUTURE LMC TEMPERATUREt (°C) = 퐒퐔퐌(JAN LMC TEMPERATURE ECDF, FEB LMC TEMPERATURE ECDF, … , NOV LMC TEMPERATURE ECDF, DEC LMC TEMPERATURE ECDF) × TEMPERATURE CHANGEt
FUTURE LMC Tmt (°C) = FUTURE LMC TEMPERATUREt + 0.006 × LMC ELEVATION
FUTURE LMCt−td (°C) = 0.0023 × LMC ELEVATION + 0.37 × FUTURE LMC TEMPERATUREt + 0.53 × LMC DAILY TEMP RANGE + 0.35 × LMC MONTHLY TEMP RANGE − 10.9
LMC ELEVATION (m) = 192
LMC LATITUDE (°) = 20.75
LMC DAILY TEMP RANGE (°C) = 10.25
LMC MONTHLY TEMP RANGE (°C) = 12.03
Rainfall
Observed rainfall across the districts of Chilika’s WC
KHORDHA OBSERVED RAINFALLt (mm) = IMD monthly rainfall data 1973 − 2009
PURI OBSERVED RAINFALLt (mm) = IMD monthly rainfall data 1973 − 2009
Observed rainfall across the districts of the LMC
ANGUL OBSERVED RAINFALLt (mm) = IMD monthly rainfall data 1973 − 2009
BAUDH OBSERVED RAINFALLt (mm) = IMD monthly rainfall data 1973 − 2009
SAMBALPUR OBSERVED RAINFALLt (mm) = IMD monthly rainfall data 1973 − 2009
SONEPUR OBSERVED RAINFALLt (mm) = IMD monthly rainfall data 1973 − 2009
Generating future rainfall across Chilika’s WC
JAN WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 1 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.000), (0.200, 6.000), (0.400, 10.000), (0.600, 20.000), (0.800, 30.000), (1.000, 60.000)}
FEB WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 2 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.000), (0.200, 5.000), (0.400, 20.000), (0.600, 24.000), (0.800, 38.000), (1.000, 90.000)}
MAR WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 3 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.000), (0.200, 10.000), (0.400, 23.000), (0.600, 30.000), (0.800, 50.000), (1.000, 110.000)}
302 Appendices
APR WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 5.000), (0.200, 20.000), (0.400, 27.000), (0.600, 38.000), (0.800, 57.000), (1.000, 90.000)}
MAY WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 25.000), (0.200, 55.000), (0.400, 75.000), (0.600, 100.000), (0.800, 120.000), (1.000, 275.000)}
JUN WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 130.000), (0.200, 220.000), (0.400, 260.000), (0.600, 300.000), (0.800, 370.000), (1.000, 470.000)}
JUL WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 250.000), (0.200, 410.000), (0.400, 480.000), (0.600, 520.000), (0.800, 550.000), (1.000, 660.000)}
AUG WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 8 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 340.000), (0.200, 410.000), (0.400, 450.000), (0.600, 520.000), (0.800, 550.000), (1.000, 640.000)}
SEP WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 9 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 180.000), (0.200, 270.000), (0.400, 320.000), (0.600, 380.000), (0.800, 460.000), (1.000, 500.000)}
OCT WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 10 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 150.000), (0.200, 210.000), (0.400, 250.000), (0.600, 400.000), (0.800, 420.000), (1.000, 750.000)}
NOV WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 11 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 10.000), (0.200, 35.000), (0.400, 50.000), (0.600, 110.000), (0.800, 140.000), (1.000, 210.000)}
DEC WC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 12 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(0.000, 0.000), (0.200, 3.000), (0.400, 8.000), (0.600, 15.000), (0.800, 24.000), (1.000, 55.000)}
WC RAINFALLt (mm) = 퐈퐅 퐓퐈퐌퐄 < ퟒퟒퟒ 퐓퐇퐄퐍 OBSERVED WC RAINFALLt 퐄퐋퐒퐄 FUTURE WC RAINFALLt
OBSERVED WC RAINFALLt (mm) = 0.001 × 퐒퐔퐌(KHORDHA OBSERVED RAINFALLt, PURI OBSERVED RAINFALLt)
FUTURE WC RAINFALLt (mm) = 0.001 × (퐒퐔퐌(JAN WC RAINFALL ECDF, FEB WC RAINFALL ECDF, … , NOV WC RAINFALL ECDF, DEC WC RAINFALL ECDF) × RAINFALL CHANGEt)
3 DIRECT RAINFALL INFLOWt (mm ) = WC RAINFALLt × LAGOON AREA × 1E6
Generating future rainfall across the LMC
JAN LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 1 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.0), (0.200, 5.0), (0.400, 10.0), (0.600, 25.0), (0.800, 40.0), (1.000, 117.0)}
FEB LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 2 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.0), (0.200, 10.0), (0.400, 25.0), (0.600, 50.0), (0.800, 70.0), (1.000, 299.0)}
303 Appendices
MAR LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 3 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.000), (0.200, 10.000), (0.400, 30.000), (0.600, 120.000), (0.800, 150.000), (1.000, 290.000)}
APR LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.000), (0.200, 20.000), (0.400, 45.000), (0.600, 70.000), (0.800, 110.000), (1.000, 365.000)}
MAY LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(0.000, 0.000), (0.200, 75.000), (0.400, 90.000), (0.600, 130.000), (0.800, 190.000), (1.000, 340.000)}
JUN LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 300.000), (0.200, 500.000), (0.400, 640.000), (0.600, 740.000), (0.800, 900.000), (1.000, 1420.000)}
JUL LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 4 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌(0,1) 퐄퐋퐒퐄 − 1){(-1.000, 0.000), (0.000, 875.000), (0.200, 1100.000), (0.400, 1200.000), (0.600, 1450.000), (0.800, 1600.000), (1.000, 2100.000)}
AUG LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 8 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 700.000), (0.200, 1100.000), (0.400, 1300.000), (0.600, 1400.000), (0.800, 1500.000), (1.000, 2256.000)}
SEP LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 9 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 350.000), (0.200, 540.000), (0.400, 650.000), (0.600, 750.000), (0.800, 800.000), (1.000, 1220.000)}
OCT LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 10 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(-1.000, 0.000), (0.000, 60.000), (0.200, 120.000), (0.400, 200.000), (0.600, 220.000), (0.800, 480.000), (1.000, 641.000)}
NOV LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 11 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(0.000, 0.000), (0.200, 10.000), (0.400, 30.000), (0.600, 50.000), (0.800, 75.000), (1.000, 238.000)}
DEC LMC RAINFALL ECDF (mm) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(퐈퐅 CLIMATE MONTHt = 12 퐓퐇퐄퐍 퐑퐀퐍퐃퐎퐌 (ퟎ, ퟏ) 퐄퐋퐒퐄 − ퟏ){(0.000, 0.000), (0.200, 0.000), (0.400, 0.000), (0.600, 5.000), (0.800, 15.000), (1.000, 80.000)}
LMC RAINFALLt (mm) = 퐈퐅 퐓퐈퐌퐄 < ퟒퟒퟒ 퐓퐇퐄퐍 OBSERVED LMC RAINFALLt 퐄퐋퐒퐄 FUTURE LMC RAINFALLt
OBSERVED LMC RAINFALLt (mm) = 0.001 × 퐒퐔퐌(ANGUL OBSERVED RAINFALLt, BAUDH OBSERVED RAINFALLt, SONEPUR
OBSERVED RAINFALLt, SAMPULPUR OBSERVED RAINFALLt)
FUTURE LMC RAINFALLt (mm) = 0.001 × (퐒퐔퐌(JAN LMC RAINFALL ECDF, FEB LMC RAINFALL ECDF, … , NOV LMC RAINFALL ECDF, DEC LMC RAINFALL ECDF) × RAINFALL CHANGEt)
Monthly freshwater flows
304 Appendices
LMC runoff generation
3 LMC RUNOFF VOLUMEt (m ) = 퐈퐍퐓퐄퐆 (LMC RUNOFF VOLUMEt−1 + LMC OVERLAND FLOWt + LMC BASEFLOWt − LMC DOWNSTREAM FLOWt − LMC ABSTRACTIONt − LMC EVAPORATIONt ) 퐝퐭
3 LMC RUNOFF VOLUMEt=0 (m ) = 0
3 LMC OVERLAND FLOWt (m ) = LMC AREA × LMC RAINFALLt × LMC RUNOFF COEFFICIENTt
3 LMC BASEFLOWt (m ) = GROUNDWATER OUTt × LMC BASEFLOW COEFFICIENT
LMC BASEFLOW COEFFICIENT (m3) = 0.6
LMC AREA (m2) = 1.4E10
3 LMC DOWNSTREAM FLOWt (m ) = (1 − LMC ABSTRACTION COEFFICIENTt) × LMC RUNOFF VOLUMEt
3 LMC ABSTRACTIONt (m ) = LMC RUNOFF VOLUMEt × LMC ABSTRACTION COEFFICIENTt
3 LMC GROUNDWATER VOLUMEt (m ) = 퐈퐍퐓퐄퐆 (LMC GROUNDWATER VOLUMEt−1 + LMC GROUNDWATER INFLOWt − LMC GROUNDWATER OUTFLOWt) 풅풕
3 LMC GROUNDWATER VOLUMEt=0 (m ) = 0
3 LMC GROUNDWATER INFLOWt (m ) = (1 − LMC RUNOFF COEFFICIENTt) × LMC OVERLAND FLOWt
LMC GROUNDWATER VOLUME LMC GROUNDWATER OUTFLOW (m3) = t t LMC GROUNDWATER COEFFICIENT
LMC GROUNDWATER COEFFICIENT (m3) = 12
3 LMC EVAPORATIONt (m ) = 퐈퐅 LMC AREA × LMC PETt < LMC RUNOFF EVAPORATIONt 퐓퐇퐄퐍 LMC AREA × LMC RUNOFF VOLUME LMC PET 퐄퐋퐒퐄 t t LMC EVAPORATION COEFFICIENT
LMC RUNOFF VOLUME LMC RUNOFF EVAPORATION (m3) = ( t ) − LMC DOWNSTREAM FLOW t LMC EVAPORATION COEFFICIENT t
LMC ABSTRACTION COEFFICIENTt (DMNSL) = 0.001
LMC RUNOFF COEFFICIENTt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(LMC RAINFALLt){(0.00, 0.00), (0.01, 0.01), (0.05, 0.15), (0.17, 0.20), (0.40, 0.25), (0.65, 0.27), (1.50, 0.99)}
LMC EVAPORATION COEFFICIENT (DMNSL) = 0.1
3 OBSERVED LMC DOWNSTREAM FLOWt (m ) = Indian Central Water Commission Non − classified monthly hydro − observation dataset from Tikarapara gauging station: 1973 − 1999
305 Appendices
WC runoff generation
3 WC RUNOFF VOLUMEt (m ) = 퐈퐍퐓퐄퐆 (WC RUNOFF VOLUMEt−1 + WC OVERLAND FLOWt + WC BASEFLOWt − WC DOWNSTREAM FLOWt − WC ABSTRACTIONt − WC EVAPORATIONt) 퐝퐭
3 WC RUNOFF VOLUMEt=0(m ) = 0
3 WC OVERLAND FLOWt (m ) = WC AREA × WC RAINFALLt × WC RUNOFF COEFFICIENTt
3 WC BASEFLOWt (m ) = GROUNDWATER OUTt × WC BASEFLOW COEFFICIENT
WC BASEFLOW COEFFICIENT (DMNSL) = 0.4
WC AREA (m2) = 4.8E9
3 WC DOWNSTREAM FLOWt (m ) = (1 − WC ABSTRACTION COEFFICIENTt) × WC RUNOFF VOLUMEt
3 WC ABSTRACTIONt (m ) = WC RUNOFF VOLUMEt × WC ABSTRACTION COEFFICIENTt
3 WC GROUNDWATER VOLUMEt (m ) = 퐈퐍퐓퐄퐆(WC GROUNDWATER VOLUMEt−1 + WC GROUNDWATER INFLOWt − WC GROUNDWATER OUTFLOWt) 퐝퐭
3 WC GROUNDWATER VOLUMEt=0(m ) = 0
3 WC GROUNDWATER INFLOWt (m ) = (1 − LMC RUNOFF COEFFICIENTt) × LMC OVERLAND FLOWt
WC GROUNDWATER VOLUME WC GROUNDWATER OUTFLOW (m3) = t t WC GROUNDWATER COEFFICIENT
WC GROUNDWATER COEFFICIENT (DMNSL) = 4
3 WC EVAPORATIONt (m ) = 퐈퐅 WC AREA × WC PETt < WC RUNOFF EVAPORATIONt 퐓퐇퐄퐍 WC AREA × WC RUNOFF VOLUME WC PET 퐄퐋퐒퐄 t t WC EVAPORATION COEFFICIENT
WC RUNOFF VOLUME WC RUNOFF EVAPORATION (m3) = ( t ) − WC DOWNSTREAM FLOW t WC EVAPORATION COEFFICIENT t
WC ABSTRACTION COEFFICIENTt (DMNSL) = 0.002
WC RUNOFF COEFFICIENTt (DMNSL) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(WC RAINFALLt){(0.0000, 0.000), (0.0100, 0.008), (0.0200, 0.012), (0.1000, 0.032), (0.2000, 0.057), (0.3400, 0.095), (0.7150, 0.230)}
WC EVAPORATION COEFFICIENT (DMNSL) = 5
Monthly sediment flows
LMC sediment generation
306 Appendices
LMC SEDIMENT YIELDt(grams) = LMC SEDIMENT RATEt × LMC DAILY FRESHWATER FLOW × DAYS IN MONTHt
LMC SEDIMENT YIELD (grams) LMC SEDIMENT YIELD (tonnes) = t t 1E6
OBSERVED LMC SEDIMENT FLOW RATEt (grams/litre) = Indian Central Water Commission Non − classified monthly hydro − observation dataset from Tikarapara gauging station: 1973 − 1999
LMC DOWNSTREAM FLOW DAILY LMC SEDIMENT GENERATION (grams/litre) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍( t ){(0, 0), t DAYS IN MONTH×60×60×24 (2500, 100), (5000, 300), (7500, 500), (10000, 750), (12500, 1100), (15000, 1350), (17500, 1600), (20000, 1900), (22500, 2190)}
LMC SEDIMENT RATEt(grams) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(DAILY LMC SEDIMENT GENERATIONt){(0, 0.00), (180, 0.29), (360, 0.40), (540, 0.45), (710, 0.48), (890, 0.50), (1070, 0.51), (1250, 0.52), (1430, 0.52), (1610, 0.53), (1790, 0.54), (1960, 0.54), (2140, 0.54), (2320, 0.54), (2500, 0.54)}
LMC DOWNSTREAM FLOW DAILY LMC FRESHWATER FLOW RATE (m3) = t t DAYS IN MONTH
3 DAILY LMC FRESHWATER FLOW RATEt (litres) = DAILY LMC FRESHWATER FLOW RATEt(m ) × 1000
LMC SEDIMENT TO CHILIKAt (tonnes) = 0.025 × LMC SEDIMENT YIELDt
WC sediment generation
WC SEDIMENT YIELDt (grams) = WC SEDIMENT RATEt × WC DAILY FRESHWATER FLOW × DAYS IN MONTHt
WC SEDIMENT YIELD (grams) WC SEDIMENT YIELD (tonnes) = t t 1E6
WC DOWNSTREAM FLOW DAILY WC SEDIMENT GENERATION (grams/litre) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍( t ){(0, 0), (100, t DAYS IN MONTH×60×60×24 186), (200, 330), (300, 410), (450, 500), (650, 560), (1000, 650)}
WC SEDIMENT RATEt(grams) = 퐆퐑퐀퐏퐇퐈퐂퐀퐋 퐅퐔퐍퐂퐓퐈퐎퐍(DAILY WC SEDIMENT GENERATIONt){(0.0, 0.0), (27.3, 0.14), (54.6, 0.21), (81.8, 0.25), (110, 0.27), (140, 0.28), (165, 0.30), (190, 0.30), (220, 0.30), (250, 0.31), (270, 0.32), (300, 0.32)}
WC DOWNSTREAM FLOW DAILY WC FRESHWATER FLOW RATE (m3) = t t DAYS IN MONTH
DAILY WC FRESHWATER FLOW RATEt(litres) = DAILY WC FRESHWATER FLOW RATEt(cubic metres) × 1000
Freshwater and sediment deliveries to Chilika
Freshwater flows to Chilika
3 TOTAL FRESHWATER INFLOWt (m ) = TOTAL LMC FRESHWATER DELIVERYt + DIRECT RAINFALL INFLOWt + TOTAL WC FRESHWATER DELIVERYt
307 Appendices
3 TOTAL LMC FRESHWATER DELIVERYt (m ) = LMC DOWNSTREAM FLOW ARRIVINGt + WC PROPORTIONING × WC DOWNSTREAM FLOWt
3 TOTAL WC FRESHWATER DELIVERYt (m ) = WC DOWNSTREAM FLOWt × (1 − WC PROPORTIONING)
WC PROPORTIONING (DMNSL) = 0.4
3 LMC DOWNSTREAM FLOW ARRIVINGt (m ) = 퐈퐅 LMC DOWNSTREAM FLOWt > 1E10 퐀퐍퐃 LMC FLOW (cumecs)t > 2830 퐓퐇퐄퐍 LMC DOWNSTREAM FLOWt × 0.12 퐄퐋퐒퐄 0
LMC DOWNSTREAM FLOWt LMC FLOW t (cubic metres per second) = DAYS IN MONTHt × 60 × 60 × 24
Chilika sediment stock
CHILIKA SEDIMENT STOCKt (tonnes) = 퐈퐍퐓퐄퐆(CHILIKA SEDIMENT STOCKt−1 + WC SEDIMENT INFLOWt + LMC SEDIMENT INFLOWt + LITTORAL SEDIMENTt − OUTLFLOW TO BAYt − PERSISTANT OUTLET OPENINGt − OBSERVED OPENING) 퐝퐭
CHILIKA SEDIMENT STOCKt=0(tonnes) = 17E6
LMC SEDIMENT INFLOWt (tonnes) = LMC SEDIMENT YIELDt + WC PROPORTIONING × WC SEDIMENT YIELDt
WC SEDIMENT INFLOWt (tonnes) = WC SEDIMENT YIELDt × (1 − WC PROPORTIONING)
LITTORAL SEDIMENTt (tonnes) = 8300
OUTLFLOW TO BAYt (tonnes) = 퐈퐅 TOTAL FRESHWATER INFLOWt < 200E6 퐓퐇퐄퐍 ((0.9 × ANNUAL SEDIMENT FLUSHED × REMOVAL PROPORTIONt) × (1 − AQUATIC VEGETATION STOCKt)) 퐄퐋퐒퐄 (1.1 × ANNUAL SEDIMENT FLUSHED × REMOVAL PROPORTIONt) × (1 − AQUATIC VEGETATION STOCKt))
TOTAL FRESHWATER INFLOW REMOVAL PROPORTION (DMNSL) = t t 5.1E9
OBSERVED OPENING (tonnes) = 퐈퐅 퐓퐈퐌퐄 = 324 퐓퐇퐄퐍 0.8 × CHILIKA SEDIMENT STOCKt 퐄퐋퐒퐄 0
308 Appendices
Appendix C – Model evaluation
C.1 Qualitative parameter assessment
Module Model variable Source(s) Information type Score
Hydroclimatic Lower Mahanadi rainfall-runoff Ghashghaei et al (2013)/personal intuition Modelled/personal intuition - coefficients
Western catchments rainfall-runoff Ghashghaei et al (2013)/personal intuition Modelled/personal intuition - coefficients Lower Mahanadi Mishra and Jena (2015) Statistical/predictive 67% sediment rating Western catchments Mishra and Jena (2015); Santra and Das Statistical 58% sediment rating (2013) Lower Mahanadi Asokan and Dutta (2008); CWC Statistical 67% area
Naraj barrage flow Das and Jena (2008); Kumar and Pattnaik Statistical 67% proportioning (2012)
Littoral sediment Rao (n.d) Statistical 67% inputs Chilika sediment capacity/outlet Personal Intuition Personal Intuition - closure
309 Appendices
Ecohydrological Chilika salinity CDA; Kumar and Pattnaik (2012) statistical 75% Chilika dissolved CDA; Kumar and Pattnaik (2012) statistical 75% oxygen salinity effect on personal intuition/statistical personal intuition/statistical - macrophyte growth dissolved oxygen effect on personal intuition/statistical personal intuition/statistical - macrophyte growth sedimentation effect on macrophyte personal intuition/statistical personal intuition/statistical - growth Habitat quality effects on fish various (see chapter 3.4.2.5) statistical/controlled 67% survival Core fishery Default fish survival Ricker (1975); Bueno and Basurto (2009) statistical/controlled 50% Ecosystem carrying Mohanty- personal communication expert knowledge 83% capacity Average Bueno and Basurto (2009); Duer-Balkind et reproductive life statistical 58% al (2013) length Natural death rate Bueno and Basurto (2009); Duer-Balkind et statistical 58% constant al (2013) Bueno and Basurto (2009); Duer-Balkind et Fecundity constant statistical 58% al (2013) Proportion of stock Kumar and Pattnaik (2012); Kumar et al Uncontrolled experiment 66% migrating (2011); Mohanty et al (2015) Socioeconomic Traditional boats CDA; Kumar and Pattnaik (2012) statistical 67% catch capacity Motor boats catch CDA; Kumar and Pattnaik (2012) statistical 67% capacity Traditional boats Jones and Sujansingani (1954); Holling statistical/controlled 75% days fished (1959b)
310 Appendices
Motor boats days Kumar Sethi- personal communication; expert knowledge/controlled 75% fished Holling (1959b)
Fish per kg see section 4.3.3 expert/stakeholder knowledge 75% Standard deviation Noack et al (2013) Uncontrolled experiment 67% of traditional profits Fish price at first sale CDA Statistical 67%
Diesel price per litre Verma and Fernandez (2010) Statistical 67%
311 Appendices
C.2 Quantitative extreme scenarios
Figure A 1: Quantitative extreme conditions associated with (top 3) the fish survival rate and (bottom 3) number of new fishers added to the lagoon each month.
312 Appendices
C.3 External trajectory combinations
Figure A 2: Two-dimensional densities of external trajectory combinations; (top) combinations of future temperature and rainfall changes, measured as the average percentage changes (2050-2060) relative to their 1986-2005 baselines; (bottom) combinations of diesel and fish price trajectories, measured as the average gradients between 2015-2060.
313 Appendices
C.4 Early model evaluation
Figure A 3: Comparison of modelled catch (red) relative to the observed record (black) from the SDM as of (a) September 2015 and (b) January 2017.
314 Appendices
Appendix D – Model outputs
D.1 Colour-blind friendly graphs
D.1.1 Normative catch
Figure A 4: Future catch time-series generated under OM, OB and OL governance; futures are classified as: green – safe and just, yellow – cautionary, and red – dangerous. NO governance is excluded here as all runs collapse (figure 6.1).
315 Appendices
D.1.2 Catch predictability
Figure A 5: Future catch under OM governance plotted alongside pre-2015 catch values (black). Whether catch is ‘safe and just’, ‘cautionary’ or ‘dangerous’ is determined by the predictability of envelope departure (chapter 6.11). Time-series colours are as above.
316 Appendices
D.1.3 Driver pathways
Figure A 6: Exploratory external trajectories under OM, OB and OL governance, classified as safe and just, cautionary or dangerous depending on their normative future from 2050-2060 (figure 6.1). Climate trajectories are plotted as decadal anomalies from the 1986-2005 baseline (blue dashed). Historical – black. Time-series colours are as above.
317 Appendices
Figure A 7: Internal fishing effort trajectories under OM, OB and OL governance. Time-series colours are the same as above.
318 Appendices
Figure A 8: Internal ecohydrological trajectories classified as safe and just, cautionary or dangerous, depending on their respective fishery outcomes. Time-series colours are as above.
319 Appendices
D.1.4 Wider social-ecological outcomes
Figure A 9: Time-series of five wider social-ecological outcomes under OM, OB and OL governance. Colours are as above.
320 Appendices
D.2 Mean and variability of catch pathways
Table A 1: Summary statistics describing the catch characteristics along the safe and just trajectories up to 2049 and within the regional SJOS from 2050 onwards.
Governance scenario
Catch (tonnes) NO OM OB OL
Mean pathway: - 13,800 13,700 13,700 2015-2049
Standard deviation of - 1,100 1,000 1,060 pathways: 2015-2049
Mean: 2050-2060 - 12,400 12,800 13,000
Standard deviation: - 1,660 1,530 1,350 2050-2060
321 Appendices
D.3 Safe and just spaces correlation analysis
Table A 2: Correlation between twelve OM drivers within the SJOS. Relationships marked as significant (*) (p < 0.05) are cross- scale social-ecological interactions only
322 Appendices
Table A 3: Correlation between twelve OB drivers within the SJOS: Relationships marked as significant (*) (p < 0.05) are cross-scale social-ecological interactions only.
Table A 4: Correlation between twelve OB
323
D.4 Wider core spaces
Figure A 10: The core SJOS trajectories of the five system drivers found to have indistinct sustainable and dangerous configurations (chapter 6.6). Colours are as per figure 6.14.
325 Appendices
D.5 Autocorrelation and partial autocorrelation functions
Figure A 11: The autocorrelation functions (ACF) of the first 25 randomly selected cautionary futures under OM (chapter 6.11.1). Blue dashed line represents the 95% confidence interval.
326
Figure A 12: The second 25 ACFs of the randomly selected cautionary futures under OM.
327 Appendices
Figure A 13: The partial autocorrelation functions (PACF) of the first 25 randomly selected cautionary futures under OM (chapter 6.11.1). Blue dashed line represents the 95% confidence interval.
328
Figure A 14: The second 25 PACFs of the randomly selected cautionary futures under OM.
329
List of References
Abson, D. J., J. Fischer, J. Leventon, J. Newig, T. Schomerus, U. Vilsmaier, H. von Wehrden, P. Abernethy, C. D. Ives, N. W. Jager & D. J. Lang (2017) Leverage points for sustainability transformation. Ambio, 46, 1, 30-39. 10.1007/s13280-016-0800-y
Adduci, M. (2009) Neoliberal Wave Rocks Chilika Lake, India: Conflict over Intensive Aquaculture from a Class Perspective. Journal of Agrarian Change, 9, 4, 484-511. 10.1111/j.1471-0366.2009.00229.x
Adger, W. N. (2006) Vulnerability. Global Environmental Change, 16, 3, 268-281. http://dx.doi.org/10.1016/j.gloenvcha.2006.02.006
Adger, W. N., K. Brown, D. R. Nelson, F. Berkes, H. Eakin, C. Folke, K. Galvin, L. Gunderson, M. Goulden & K. O'Brien (2011) Resilience implications of policy responses to climate change. Wiley Interdisciplinary Reviews: Climate Change, 2, 5, 757-766
Agnew, T. T. (1979) Optimal exploitation of a fishery employing a non-linear harvesting function. Ecological Modelling, 6, 1, 47-57. http://dx.doi.org/10.1016/0304-3800(79)90057-7
Ahmad, S. & S. P. Simonovic (2004) Spatial system dynamics: new approach for simulation of water resources systems. Journal of Computing in Civil Engineering, 18, 4, 331-340
Ahmed, S. & S. P. Simonovic (2004) Spatial system dynamics: new approach for simulation of water resources systems. Journal of Computing in Civil Engineering, 18, 4, 331-340
Alabaster, J. S. & R. S. Lloyd. 1980. Water Quality Criteria for Freshwater Fish London, U.K: Butterworths.
Albert, R., H. Jeong & A.-L. Barabasi (2000) Error and attack tolerance of complex networks. Nature, 406, 6794, 378-382
Alcamo, J. & E. Kreileman (1996) Emission scenarios and global climate protection. Global Environmental Change, 6, 4, 305
Allen, M. R. & W. J. Ingram (2002) Constraints on future changes in climate and the hydrologic cycle. Nature, 419, 6903, 224-232
Alley, R. B., J. Marotzke, W. D. Nordhaus, J. T. Overpeck, D. M. Peteet, R. A. Pielke, R. T. Pierrehumbert, P. B. Rhines, T. F. Stocker, L. D. Talley & J. M. Wallace (2003) Abrupt Climate Change. Science, 299, 5615, 2005-2010. 10.1126/science.1081056
Alverson, D. L., M. H. Freeburg, S. A. Murawski & J. G. Pope. 1996. Estimates of fisheries bycatch and discards. In A Global Assessment of Fisheries Bycatch and Discards, ed. FAO. Rome, Italy: Food and Agriculture Organization of the United Nations.
Anderies, J. M. (2015) Managing variance: Key policy challenges for the Anthropocene. Proceedings of the National Academy of Sciences, 112, 47, 14402-14403. 10.1073/pnas.1519071112
331 Appendices
Anderies, J. M., M. A. Janssen & E. Ostrom (2004) A Framework to Analyze the Robustness of Social-ecological Systems from an Institutional Perspective. Ecology and Society, 9, 1, 18. 10.5751/ES-00610-090118
Andersen, T., J. Carstensen, E. Hernández-García & C. M. Duarte (2009) Ecological thresholds and regime shifts: approaches to identification. Trends in Ecology & Evolution, 24, 1, 49-57. http://dx.doi.org/10.1016/j.tree.2008.07.014
Andrew, N. L., C. Béné, S. J. Hall, E. H. Allison, S. Heck & B. D. Ratner (2007) Diagnosis and management of small-scale fisheries in developing countries. Fish and Fisheries, 8, 3, 227-240. 10.1111/j.1467-2679.2007.00252.x
Anthoff, D. & R. S. J. Tol (2013) The uncertainty about the social cost of carbon: A decomposition analysis using fund. Climatic Change, 117, 3, 515-530. 10.1007/s10584-013-0706-7
Anyadike, R. N. C. (1987) The Linacre evaporation formula tested and compared to others in various climates over West Africa. Agricultural and Forest Meteorology, 39, 2–3, 111-119. http://dx.doi.org/10.1016/0168- 1923(87)90030-X
Arnold, J. G., R. Srinivasan, R. S. Muttiah & J. R. Williams (1998) Large area hydrologic modeling and assessment part I: model development. Journal of the American Water Resources Association, 34, 1, 73-89
Arthur, W. B. 2013. Complexity Economics: A Different Framework for Economic Thought. In SFI Working Paper: 2013-04-012, 24. New Mexico, USA: Santa Fe Institute.
Ashkenazy, A., T. Calvão Chebach, K. Knickel, S. Peter, B. Horowitz & R. Offenbach (2017) Operationalising resilience in farms and rural regions – Findings from fourteen case studies. Journal of Rural Studies. https://doi.org/10.1016/j.jrurstud.2017.07.008
Asokan, S. M. & D. Dutta (2008) Analysis of water resources in the Mahanadi River Basin, India under projected climate conditions. Hydrological Processes, 22, 18, 3589-3603. 10.1002/hyp.6962
Audouin, M., R. Preiser, S. Nienaber, L. Downsborough, J. Lanz & S. Mavengahama (2013) Exploring the implications of critical complexity for the study of social-ecological systems. Ecology and Society, 18, 3. 10.5751/ES-05434- 180312
Azar, C., J. Holmberg & K. Lindgren (1996) Socio-ecological indicators for sustainability. Ecological Economics, 18, 2, 89-112. http://dx.doi.org/10.1016/0921-8009(96)00028-6
Bagstad, K. J., G. W. Johnson, B. Voigt & F. Villa (2013) Spatial dynamics of ecosystem service flows: A comprehensive approach to quantifying actual services. Ecosystem Services, 4, 0, 117-125. http://dx.doi.org/10.1016/j.ecoser.2012.07.012
Bai, J. (1994) Least squares estimation of a shift in linear processes. Journal of Time Series Analysis, 15, 5, 453-472
Bai, X., S. van der Leeuw, K. O’Brien, F. Berkhout, F. Biermann, E. S. Brondizio, C. Cudennec, J. Dearing, A. Duraiappah, M. Glaser, A. Revkin, W. Steffen & J. Syvitski (2016) Plausible and desirable futures in the Anthropocene: A new
332
research agenda. Global Environmental Change, 39, 351–362. http://dx.doi.org/10.1016/j.gloenvcha.2015.09.017
Bailey, C. & S. Jentoft (1990) Hard choices in fisheries development. Marine Policy, 14, 4, 333-344. http://dx.doi.org/10.1016/0308-597X(90)90055-V
Bak, P. & K. Chen (1991) Self-organized criticality. Journal Name: Scientific American; (United States); Journal Volume: 264:1, Medium: X; Size: Pages: 46-53
Bak, P., C. Tang & K. Wiesenfeld (1987) Self-organized criticality. Physical Review A, 38, 1, 364-374
Baland, J. & J. Platteau. 1996. Halting Degradation of Natural Resources: Is there A Role for Rural Communities? Rome, Italy: Food and Agriculture Organization of the United Nations.
Baliarsingh, S. K., S. Sahoo, A. Acharya, H. B. Dalabehera, K. C. Sahu & A. A. Lotliker (2014) Oil pollution in Chilika lagoon: An anthropogenic threat to biodiversity. Current Science, 104, 4, 516-517
Balke, T., P. M. J. Herman & T. J. Bouma (2014) Critical transitions in disturbance- driven ecosystems: identifying Windows of Opportunity for recovery. Journal of Ecology, 102, 3, 700-708. 10.1111/1365-2745.12241
Barbier, E. B., E. W. Koch, B. R. Silliman, S. D. Hacker, E. Wolanski, J. Primavera, E. F. Granek, S. Polasky, S. Aswani, L. A. Cramer, D. M. Stoms, C. J. Kennedy, D. Bael, C. V. Kappel, G. M. E. Perillo & D. J. Reed (2008) Coastal Ecosystem-Based Management with Nonlinear Ecological Functions and Values. Science, 319, 5861, 321-323. 10.1126/science.1150349
Barik, S. K., P. R. Muduli, B. Mohanty, A. T. Behera, S. Mallick, A. Das, R. N. Samal, G. Rastogi & A. K. Pattnaik (2017) Spatio-temporal variability and the impact of Phailin on water quality of Chilika lagoon. Continental Shelf Research, 136, 39-56. http://dx.doi.org/10.1016/j.csr.2017.01.019
Barko, J. W., D. Gunnison & S. R. Carpenter (1991) Ecology of Submersed Aquatic Macrophytes Sediment interactions with submersed macrophyte growth and community dynamics. Aquatic Botany, 41, 1, 41-65. http://dx.doi.org/10.1016/0304-3770(91)90038-7
Barlas, Y. (1989) Multiple tests for validation of system dynamics type of simulation models. European Journal of Operational Research, 42, 1, 59- 87. http://dx.doi.org/10.1016/0377-2217(89)90059-3
Barlas, Y. (1996) Formal aspects of model validity and validation in system dynamics. System Dynamics Review, 12, 3, 183-210. 10.1002/(SICI)1099- 1727(199623)12:3<183::AID-SDR103>3.0.CO;2-4
Barlas, Y. & K. Kanar. 2000. Structure-oriented behaviour tests in model validation. In International System Dynamics Conference, 19. Bergen, Norway: System Dynamics Society.
Barnes, R. S. K. (1999) The conversation of brackish-water systems: priorities for the 21st century. Aquatic Conservation: Marine and Freshwater Ecosystems, 9, 523-527
Barnett, J. (2003) Security and climate change. Global Environmental Change, 13, 1, 7-17. http://dx.doi.org/10.1016/S0959-3780(02)00080-8
333 Appendices
Barnosky, A. D., E. A. Hadly, J. Bascompte, E. L. Berlow, J. H. Brown, M. Fortelius, W. M. Getz, J. Harte, A. Hastings, P. A. Marquet, N. D. Martinez, A. Mooers, P. Roopnarine, G. Vermeij, J. W. Williams, R. Gillespie, J. Kitzes, C. Marshall, N. Matzke, D. P. Mindell, E. Revilla & A. B. Smith (2012) Approaching a state shift in Earth's biosphere. Nature, 486, 7401, 52-58
Beals, M., L. Gross & S. Harrell. 1999. Type II Functional Response: Holling's Disk Equation. ed. T. I. o. E. Modeling. University of Tennessee.
Beck, M. & T. Krueger (2016) The epistemic, ethical, and political dimensions of uncertainty in integrated assessment modeling. Wiley Interdisciplinary Reviews: Climate Change, 7, 5, 627-645. 10.1002/wcc.415
Beck, M. B. & J. Chen. 2000. Assuring the quality of models designed for predictive tasks. In Sensitivity Analysis, eds. A. Saltelli, K. Chan & E. M. Scott, 401-420. Chichester, U.K: John Wiley & Sons Ltd.
Beddington, J. 2009. Presentation given to the Sustainable Development UK Annual Conference, 19 March 2009. In http://webarchive.nationalarchives.gov.uk/?/http://www.dius.gov.uk/news _and_speeches/speeches/john_beddington/perfect-storm. London, UK.
Beisner, B. E., D. T. Haydon & K. Cuddington (2003) Alternative stable states in ecology. Frontiers in Ecology and the Environment, 1, 7, 376-382
BenDor, T., J. Scheffran & B. Hannon (2009) Ecological and economic sustainability in fishery management: A multi-agent model for understanding competition and cooperation. Ecological Economics, 68, 4, 1061-1073. http://dx.doi.org/10.1016/j.ecolecon.2008.07.014
Bennett, E. M., G. S. Cumming & G. D. Peterson (2005) A Systems Model Approach to Determining Resilience Surrogates for Case Studies. Ecosystems, 8, 8, 945-957. 10.1007/s10021-005-0141-3
Bennett, E. M., G. D. Peterson & L. J. Gordon (2009) Understanding relationships among multiple ecosystem services. Ecology Letters, 12, 12, 1394-1404. 10.1111/j.1461-0248.2009.01387.x
Berkes, F. 2011. Restoring unity: the concept of marine social-ecological systems. In World Fisheries: A Social Ecological Analysis, eds. R. E. Ommer, R. I. Perry, K. Cochrane & P. Cury, 9-28. Wiley-Blackwell.
Berkes, F. & C. Folke. 1998. Linking Social and Ecological Systems: Management Practices and Social Mechanisms for Building Resilience. Cambridge, UK: Cambridge University Press.
Berkes, F., T. P. Hughes, R. S. Steneck, J. A. Wilson, D. R. Bellwood, B. Crona, C. Folke, L. H. Gunderson, H. M. Leslie, J. Norberg, M. Nyström, P. Olsson, H. Österblom, M. Scheffer & B. Worm (2006) Globalization, Roving Bandits, and Marine Resources. Science, 311, 5767, 1557. 10.1126/science.1122804
Berkes, F. & C. S. Seixas (2005) Building Resilience in Lagoon Social–Ecological Systems: A Local-level Perspective. Ecosystems, 8, 8, 967-974. 10.1007/s10021-005-0140-4
Berkhout, F. (2014) Anthropocene Futures. The Anthropocene Review, 1, 2, 154- 159. 10.1177/2053019614531217
334
Bestelmeyer, B. T., A. M. Ellison, W. R. Fraser, K. B. Gorman, S. J. Holbrook, C. M. Laney, M. D. Ohman, D. P. C. Peters, F. C. Pillsbury, A. Rassweiler, R. J. Schmitt & S. Sharma (2011) Analysis of abrupt transitions in ecological systems. Ecosphere, 2, 12, art129. 10.1890/ES11-00216.1
Beven, K. J. 2002. Rainfall-Runoff Modelling: The Primer. West Sussex, U.K: John Wiley & Sons, LTD.
Bhaumik, U. (2015) Migration of Hilsha Shad in the Indo-Pacific region- a review. International Journal of Current Research and Academic Review, 3, 11, 139-155
Biermann, F. (2007) ‘Earth system governance’ as a crosscutting theme of global change research. Global Environmental Change, 17, 3–4, 326-337. http://dx.doi.org/10.1016/j.gloenvcha.2006.11.010
Biermann, F. (2012) Planetary boundaries and earth system governance: Exploring the links. Ecological Economics, 81, 4-9. http://dx.doi.org/10.1016/j.ecolecon.2012.02.016
Biesbroek, R., J. Dupuis & A. Wellstead (2017) Explaining through causal mechanisms: resilience and governance of social–ecological systems. Current Opinion in Environmental Sustainability, 28, Supplement C, 64-70. https://doi.org/10.1016/j.cosust.2017.08.007
Biggs, R., S. R. Carpenter & W. A. Brock (2009) Turning back from the brink: Detecting an impending regime shift in time to avert it. Proceedings of the National Academy of Sciences, 106, 3, 826-831. 10.1073/pnas.0811729106
Biggs, R., L. Gordon, C. Raudsepp-Hearne, M. Schlüter & B. Walker. 2015a. Principle 3- manage slow variables and feedbacks. In Principles of Building Resilience, eds. R. Biggs, M. Schlüter & M. L. Schoon, 105-141. Cambridge, U.K: Cambridge University Press.
Biggs, R., C. Rhode, S. Archibald, L. M. Kunene, S. S. Mutanga, N. Nkuna, P. O. Ocholla & L. J. Phadima (2015b) Strategies for managing complex social- ecological systems in the face of uncertainty: examples from South Africa and beyond. Ecology and Society, 20, 1. 10.5751/ES-07380-200152
Biggs, R. O., G. D. Peterson & J. C. C. Rocha. 2015c. The regime shifts database: a framework for analyzing regime shifts in social-ecological systems.
Binder, C. R., J. Hinkel, P. W. G. Bots & C. Pahl-Wostl (2013) Comparison of Frameworks for Analyzing Social-ecological Systems. Ecology and Society, 18, 4. 10.5751/ES-05551-180426
Biswas, K. P. 1995. Ecological and Fisheries Development in Wetlands: A Study of Chilika Lagoon. Delhi, India: Daya Publishing House.
Bitterman, P. & D. A. Bennett (2016) Constructing stability landscapes to identify alternative states in coupled social-ecological agent-based models. Ecology and Society, 21, 3. 10.5751/ES-08677-210321
Blackwood, J. C., A. Hastings & P. J. Mumby (2012) The effect of fishing on hysteresis in Caribbean coral reefs. Theoretical Ecology, 5, 1, 105-114. 10.1007/s12080-010-0102-0
335 Appendices
Blühdorn, I. (2007) Sustaining the unsustainable: Symbolic politics and the politics of simulation. Environmental Politics, 16, 2, 251-275. 10.1080/09644010701211759
Bodin, Ö. & M. Tengö (2012) Disentangling intangible social–ecological systems. Global Environmental Change, 22, 2, 430-439. http://dx.doi.org/10.1016/j.gloenvcha.2012.01.005
Boettiger, C. & A. Hastings (2013) Tipping points: From patterns to predictions. Nature, 493, 7431, 157-158
Bohensky, E. L., L. A. Evans, J. M. Anderies, D. Biggs & F. C. 2015. Principle 4- foster complex adaptive systems thinking. In Principles for Building Resilience: Sustaining Ecosystem Services in Social-Ecological Systems, eds. R. Biggs, M. Schlüter & M. L. Schoon, 142-173. Cambridge, U.K: Cambridge University Press.
Börjeson, L., M. Höjer, K.-H. Dreborg, T. Ekvall & G. Finnveden (2006) Scenario types and techniques: Towards a user's guide. Futures, 38, 7, 723-739. http://dx.doi.org/10.1016/j.futures.2005.12.002
Boulding, K. E. 1966. The Economics of the Coming Spaceship Earth. [online] http://dieoff.org/page160.htm.
Boumans, R., J. Roman, I. Altman & L. Kaufman (2015) The Multiscale Integrated Model of Ecosystem Services (MIMES): Simulating the interactions of coupled human and natural systems. Ecosystem Services, 12, 0, 30-41. http://dx.doi.org/10.1016/j.ecoser.2015.01.004
Box, G. E. P. (1976) Science and Statistics. Journal of the American Statistical Association, 71, 356, 791-799. 10.1080/01621459.1976.10480949
Box, G. E. P. 1979. Robustness in the strategy of scientific model building. New York, U.S.A: Academic Press.
Boyd, C. E. 1982. Water Quality Management for Pond Fish Culture (Developments in Aquaculture and Fisheries Science). Amsterdam, The Netherlands: Elsevier Science Publishing.
Brand, F. S. & K. Jax (2007) Focusing the Meaning(s) of Resilience: Resilience as a Descriptive Concept and a Boundary Object. Ecology and Society, 12, 1,
Brand, S. 1999. The Clock of the Long Now. New York, USA: Basic Books.
Brandt, G. & A. Merico (2013) Tipping points and user-resource system collapse in a simple model of evolutionary dynamics. Ecological Complexity, 13, 46-52. http://dx.doi.org/10.1016/j.ecocom.2012.12.003
Brock, W. A. & D. Starrett (2003) Managing Systems with Non-convex Positive Feedback. Environmental and Resource Economics, 26, 4, 575-602. 10.1023/B:EARE.0000007350.11040.e3
Brondizio, E. S., N. D. Vogt, A. V. Mansur, E. J. Anthony, S. Costa & S. Hetrick (2016) A conceptual framework for analyzing deltas as coupled social– ecological systems: an example from the Amazon River Delta. Sustainability Science, 11, 4, 591-609. 10.1007/s11625-016-0368-2
336
Brook, B. W., E. C. Ellis, M. P. Perring, A. W. Mackay & L. Blomqvist (2013) Does the terrestrial biosphere have planetary tipping points? Trends in Ecology & Evolution, 28, 7, 396-401. http://dx.doi.org/10.1016/j.tree.2013.01.016
Brucet, S., S. Poikane, A. Lyche-Solheim & S. Birk (2013) Biological assessment of European lakes: ecological rationale and human impacts. Freshwater Biology, 58, 6, 1106-1115. 10.1111/fwb.12111
Bruckner, T., G. Petschel-Held, F. L. Tóth, H. M. Füssel, C. Helm, M. Leimbach & H. J. Schellnhuber (1999) Climate change decision-support and the tolerable windows approach. Environmental Modeling & Assessment, 4, 4, 217-234. 10.1023/A:1019076603956
Brugnach, M., A. Dewulf, C. Pahl-Wostl & T. Taillieu (2008) Toward a Relational Concept of Uncertainty: about Knowing Too Little, Knowing Too Differently, and Accepting Not to Know. Ecology and Society, 13, 2. 10.5751/ES-02616- 130230
Bueno, N. P. (2014) Misperceptions of feedbacks and the resilience of common- pool resource systems: a discussion for irrigation systems based on loop dominance analysis. International Journal of the Commons, 8, 1, 79-106. http://doi.org/10.18352/ijc.388
Bueno, N. P. & X. Basurto (2009) Resilience and collapse of artisanal fisheries: a system dynamics analysis of a shellfish fishery in the Gulf of California, Mexico. Sustainability Science, 4, 139-149. DOI 10.1007/s11625-009- 0087-z
Bunting, S. W. 2013. Principles of Sustainable Aquaculture: Promoting Social, Economic and Environmental Resilience. Abingdon, UK: Routledge.
Burmaster, D. E. & P. D. Anderson (1994) Principles of good practice for the use of monte carlo techniques in human health and ecological risk assessments. Risk Analysis, 14, 4, 477-481
Burt, T. P. 1996. The hydrology of headwater catchments. In River Flows and Channel Forms, eds. G. Petts & P. Calow, 6-31. Oxford, U.K: Blackwell Science.
Cairns Jr, J. (2004) Ecological tipping points: a major challenge for experimental sciences. Asian Journal of Experimental Sciences, 18, 1-2, 1-16
Campbell, G. S. & J. M. Norman. 1998. An Introduction to Environmental Biophysics. New York, NY, USA: Springer Science & Business Media, Inc. .
Campolongo, F., A. Saltelli, T. Sørense & S. Tarantola. 2000. Hitchhiker's guide to sensitivity analysis. In Sensitivity Analysis, eds. A. Saltelli, K. Chan & E. M. Scott, 15-50. Chichester, U.K: John Wiley & Sons Ltd.
Carpenter, S., B. Walker, J. M. Anderies & N. Abel (2001) From Metaphor to Measurement: Resilience of What to What? Ecosystems, 4, 8, 765-781. 10.1007/s10021-001-0045-9
Carpenter, S. R. (2002) Ecological futures: building an ecology of the long now. Ecology, 83, 2069-2083
Carpenter, S. R., E. G. Booth, S. Gillon, C. J. Kucharik, S. Loheide, A. S. Mase, M. Motew, J. Qiu, A. R. Rissman, J. Seifert, E. Soylu, M. Turner & C. B. Wardropper (2015a) Plausible futures of a social-ecological system: Yahara
337 Appendices
watershed, Wisconsin, USA. Ecology and Society, 20, 2. 10.5751/ES-07433- 200210
Carpenter, S. R., W. Brock & P. Hanson (1999) Ecological and Social Dynamics in Simple Models of Ecosystem Management. Ecology and Society, 3, 2,
Carpenter, S. R. & W. A. Brock (2008) Adaptive Capacity and Traps. Ecology and Society, 13, 2, 16. 10.5751/ES-02716-130240
Carpenter, S. R., W. A. Brock, C. Folke, E. H. van Nes & M. Scheffer (2015b) Allowing variance may enlarge the safe operating space for exploited ecosystems. Proceedings of the National Academy of Sciences, 112, 46, 14384-14389. 10.1073/pnas.1511804112
CDA. 2011. Chilika Development Authority Annual Report 2010-2011. ed. A. K. Pattanaik, 36. Orissa, India: Chilika Development Authority.
CDA. 2015. Fish landings in Chilika lake during 1999-00 to 2013-14: catch and effort analysis. Bhubaneswar, India: Chilika Development Authority, Department of Water Resources (Orissa).
CDA. 2016. Year-wise fisheries output and value (revenue generation) during the period 1978-1979 to 2014-2015. Bhubaneswar, India: Chilika Development Authority, Department of Water Resources (Orissa).
CDA. n.d-a. CDA ecological monitoring records. Bhubaneswar, Odisha: Chilika Development Authority, Department of Water Resources (Orissa).
CDA. n.d-b. Total volume of fresh water flow in (M.Cu.M) into Chilika lagoon during 1999-00 - 2009-10. Bhubaneswar, India: Chilika Development Authority, Department of Water Resources (Orissa).
CDA. n.d-c. Variation of meteorological parameters at Chandraput for 2006-2009. Bhubaneswar, India: Chilika Development Authority, Department of Water Resources (Orissa).
CDA & WI. 2010. Chilika Newsletter. eds. R. Kumar & A. K. Pattnaik, 28. Bhubaneswar, India: Chilika Development Authority & Wetlands International- South Asia.
Chapman, A. & S. Darby (2016) Evaluating sustainable adaptation strategies for vulnerable mega-deltas using system dynamics modelling: Rice agriculture in the Mekong Delta's An Giang Province, Vietnam. Science of The Total Environment, 559, 326-338. http://dx.doi.org/10.1016/j.scitotenv.2016.02.162
Christensen, L. & N. Krogman (2012) Social Thresholds and their Translation into Social-ecological Management Practices. Ecology and Society, 17, 1. 10.5751/ES-04499-170105
CIFRI. 2006. Assessment of fish yield potential of Chilika Lake. 1-25. Barrackpore, Kolkata, India: Central Inland Fisheries Research Institute.
Cilliers, P. 1998. Complexity and postmodernism: understanding complex systems. London, UK: Routledge.
Cilliers, P., H. C. Biggs, S. Blignaut, A. G. Choles, J.-H. S. Hofmeyr, G. P. W. Jewitt & D. J. Roux (2013) Complexity, Modeling, and Natural Resource Management. Ecology and Society, 18, 3, 12. 10.5751/ES-05382-180301
338
Cinner, J. E. (2011) Social-ecological traps in reef fisheries. Global Environmental Change, 21, 3, 835-839
Cinner, J. E., C. Folke, T. Daw & C. C. Hicks (2011) Responding to change: Using scenarios to understand how socioeconomic factors may influence amplifying or dampening exploitation feedbacks among Tanzanian fishers. Global Environmental Change, 21, 1, 7-12. http://dx.doi.org/10.1016/j.gloenvcha.2010.09.001
Ciriacy-Wantrup, S. V. 1952. Resource Conservation: Economics and Policies. California, U.S.A: University of California Press.
Clark, W. C. & N. M. Dickson (2003) Sustainability science: the emerging research program. Proceedings of the National Academy of Sciences, 100, 14, 8059- 8061
Cleden, D. 2009. Managing Project Uncertainty. Farnham, U.K: Gower Publishing Limited.
Cloern, J. E., S. Q. Foster & A. E. Kleckner (2014) Phytoplankton primary production in the world's estuarine-coastal ecosystems. Biogeosciences, 11, 9, 2477-2501. 10.5194/bg-11-2477-2014
Coates, D. 2002. Inland capture fishery statistics of Southeast Asia: current status and information needs. In RAP publication 2002/11 ed. A.-P. F. Commission, 129. Regional Office for Asia and the Pacific, Bangkok, Thailand: Food and Agriculture Organization of the United Nations.
Cole, M. J., R. M. Bailey & M. G. New (2014) Tracking sustainable development with a national barometer for South Africa using a downscaled “safe and just space” framework. Proceedings of the National Academy of Sciences, 111, 42, E4399-E4408. 10.1073/pnas.1400985111
Collie, J. S., L. W. Botsford, A. Hastings, I. C. Kaplan, J. L. Largier, P. A. Livingston, É. Plagányi, K. A. Rose, B. K. Wells & F. E. Werner (2016) Ecosystem models for fisheries management: finding the sweet spot. Fish and Fisheries, 17, 1, 101-125. 10.1111/faf.12093
Collie, J. S., K. Richardson & J. H. Steele (2004) Regime shifts: Can ecological theory illuminate the mechanisms? Progress in Oceanography, 60, 2, 281- 302. http://dx.doi.org/10.1016/j.pocean.2004.02.013
Collier, M. J. (2015) Novel ecosystems and social-ecological resilience. Landscape Ecology, 30, 8, 1363-1369. 10.1007/s10980-015-0243-z
Contamin, R. & A. M. Ellison (2009) Indicators of regime shifts in ecological systems: what do we need to know and when do we need to know it? Ecological Applications, 19, 3, 799-816. 10.1890/08-0109.1
Costanza, R., R. D'Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S. Naeem, R. O'Neill, J. Paruelo, R. G. Raskin, P. Sutton & M. van den Belt (1997) The value of the world's ecosystem services and natural capital. Nature, 387, 253-260
Costanza, R., R. de Groot, P. Sutton, S. van der Ploeg, S. J. Anderson, I. Kubiszewski, S. Farber & R. K. Turner (2014) Changes in the global value of ecosystem services. Global Environmental Change, 26, 152-158. http://dx.doi.org/10.1016/j.gloenvcha.2014.04.002
339 Appendices
Costanza, R., I. Kubiszewski, D. Ervin, R. Bluffstone, J. Boyd, D. Brown, H. Chang, V. Dujon, E. Granek, S. Polasky, V. Shandas & A. Yeakley (2011) Valuing ecological systems and services. F1000 Biology Reports, 3, 14. 10.3410/B3-14
Coulthard, S. (2008) Adapting to environmental change in artisanal fisheries— Insights from a South Indian Lagoon. Global Environmental Change, 18, 3, 479-489. http://dx.doi.org/10.1016/j.gloenvcha.2008.04.003
Cox, M. (2014) Applying a Social-Ecological System Framework to the Study of the Taos Valley Irrigation System. Human Ecology, 42, 2, 311-324. 10.1007/s10745-014-9651-y
CPCB. 1993. Criteria For Classification and Zoning Of Coastal Waters, Coastal Pollution Control Series. 12. Central Pollution Control Board.
Crépin, A.-S., R. Biggs, S. Polasky, M. Troell & A. de Zeeuw (2012) Regime shifts and management. Ecological Economics, 84, 0, 15-22. http://dx.doi.org/10.1016/j.ecolecon.2012.09.003
Cretney, R. (2014) Resilience for Whom? Emerging Critical Geographies of Socio- ecological Resilience. Geography Compass, 8, 9, 627-640. 10.1111/gec3.12154
Crutzen, P. J. & E. F. Stoermer (2000) The 'Anthropocene'. IGBP Newsletter, 41, 17- 18
Cuddington, K., M. J. Fortin, L. R. Gerber, A. Hastings, A. Liebhold, M. O'Connor & C. Ray (2013) Process-based models are required to manage ecological systems in a changing world. Ecosphere, 4, 2, art20. 10.1890/ES12- 00178.1
Cumming, G. S. & J. Collier (2005) Change and Identity in Complex Systems. Ecology and Society, 10,
Cumming, G. S. & G. D. Peterson (2017) Unifying Research on Social–Ecological Resilience and Collapse. Trends in Ecology & Evolution, 32, 9, 695-713. https://doi.org/10.1016/j.tree.2017.06.014
CWC. n.d. Disharge data at Tikarapara gauging station, Odisha: 1972-2012. [online]: http://www.india-wris.nrsc.gov.in/wris.html: Central Water Commission.
D'Souza, R. 2006. Drowned and Dammed: Colonial Capitalism and Flood Control in Eastern India. Delhi, India: Oxford University Press India.
D’Lima, C., H. Marsh, M. Hamann, A. Sinha & R. Arthur (2014) Positive Interactions Between Irrawaddy Dolphins and Artisanal Fishers in the Chilika Lagoon of Eastern India are Driven by Ecology, Socioeconomics, and Culture. Ambio, 43, 5, 614-624. 10.1007/s13280-013-0440-4
Daily, G. 1997. Nature's Services: Societal Dependence on Natural Ecosystems. Washington DC, USA: Island Press.
Daily, G. C. & P. R. Ehrlich (1992) Population, Sustainability, and Earth's Carrying Capacity. BioScience, 42, 10, 761-771. 10.2307/1311995
Dakos, V., M. Scheffer, E. H. van Nes, V. Brovkin, V. Petoukhov & H. Held (2008) Slowing down as an early warning signal for abrupt climate change.
340
Proceedings of the National Academy of Sciences of the United States of America, 105, 38, 14308-14312. 10.1073/pnas.0802430105
Dalin, C., M. Konar, N. Hanasaki, A. Rinaldo & I. Rodriguez-Iturbe (2012) Evolution of the global virtual water trade network. Proceedings of the National Academy of Sciences of the United States of America, 109, 16, 5989-5994. 10.1073/pnas.1203176109
Daly, H. 1993. Sustainable growth: an impossibility theorem. In Valuing the Earth: Economics, Ecology Ethics, eds. H. Daly & K. Townsend, 267-274. Cambridge, Massachusetts, U.S.A: MIT Press.
Dammers, E., A. K. Bregt, J. Edelenbos, H. Meyer & B. Pel (2014) Urbanized Deltas as Complex Adaptive Systems: Implications for Planning and Design. Built Environment, 40, 2, 156-168
Dankwa, H. R., J. M. Shenker, J. Lin, P. K. Ofori-Danson & Y. Ntiamoa-Baidu (2004) Fisheries of two tropical lagoons in Ghana, West Africa. Fisheries Management and Ecology, 11, 6, 379-386. 10.1111/j.1365- 2400.2004.00406.x
Das, B. P. & J. Jena. 2008. Impact of Mahanadi Basin development on eco- hydrology of Chilika. In Taal 2007: the 12th World Lake Conference, eds. M. Sengupta & R. Dalwani, 697-702. Jaipur, India.
Daw, T. M., S. Coulthard, W. W. L. Cheung, K. Brown, C. Abunge, D. Galafassi, G. D. Peterson, T. R. McClanahan, J. O. Omukoto & L. Munyi (2015) Evaluating taboo trade-offs in ecosystems services and human well-being. Proceedings of the National Academy of Sciences, 112, 22, 6949-6954. 10.1073/pnas.1414900112
Day, J. W., J. F. Martin, L. Cardoch & P. H. Templet (1997) System functioning as a basis for sustainable management of deltaic ecosystems. Coastal Management, 25, 2, 115-153. 10.1080/08920759709362315
De Young, B. 1999. Canadian Marine Fisheries in a Changing and Uncertain World: A Report Prepared for the Canadian Global Change Program of the Royal Society of Canada. Ottawa, Ontario, Canada: National Research Council of Canada.
Dearing, J., B. Acma, S. Bub, F. Chambers, X. Chen, J. Cooper, D. Crook, X. Dong, M. Dotterweich, M. Edwards, T. Foster, M.-J. Gaillard, D. Galop, P. Gell, A. Gil, E. Jeffers, R. Jones, A. Krishnamurthy, P. Langdon, R. Marchant, F. Mazier, C. McLean, L. Nunes, S. Raman, I. Suryaprakash, M. Umer, X. Yang, R. Wang & K. Zhang (2015) Social-ecological systems in the Anthropocene: the need for integrating social and biophysical records at regional scales. The Anthropocene Review, 1-27. 10.1177/2053019615579128
Dearing, J. A. (2008) Landscape change and resilience theory: a palaeoenvironmental assessment from Yunnan, SW China. Holocene, 18, 1, 117-227. 10.1177/0959683607085601
Dearing, J. A., A. K. Braimoh, A. Reenberg, B. L. Turner & S. van der Leeuw (2010) Complex Land Systems: the Need for Long Time Perspectives to Assess their Future. Ecology and Society, 15, 4, 19. 10.5751/ES-03645-150421
Dearing, J. A., N. Richmond, A. J. Plater, J. Wolf, D. Prandle & T. J. Coulthard (2006) Modelling approaches for coastal simulation based on cellular automata:
341 Appendices
the need and potential. Philosophical Transactions of the Royal Society A: Mathematical, Physical, and Engineering Sciences, 364, 1841, 1051-71. 10.1098/rsta.2006.1753
Dearing, J. A., R. Wang, K. Zhang, J. G. Dyke, H. Haberl, M. S. Hossain, P. G. Langdon, T. M. Lenton, K. Raworth, S. Brown, J. Carstensen, M. J. Cole, S. E. Cornell, T. P. Dawson, C. P. Doncaster, F. Eigenbrod, M. Flörke, E. Jeffers, A. W. Mackay, B. Nykvist & G. M. Poppy (2014) Safe and just operating spaces for regional social-ecological systems. Global Environmental Change, 28, 227-238. http://dx.doi.org/10.1016/j.gloenvcha.2014.06.012
Dearing, J. A., X. Yang, X. Dong, E. Zhang, X. Chen, P. G. Langdon, K. Zhang, W. Zhang & T. P. Dawson (2012) Extending the timescale and range of ecosystem services through paleoenvironmental analyses, exemplified in the lower Yangtze basin. Proceedings of the National Academy of Sciences, 109, 18, E1111–E1120. 10.1073/pnas.1118263109
Denscombe, M. 2007. The Good Research Guide: For Small Social Research Projects. Maidenhead, U.K: Open University Press.
Dobson, A. 2000. Sustainable development and the defence of the natural world. In Global ustainable Development in the 21st Century, eds. K. Lee, A. Holland & D. McNeill, 49-60. Edinburgh, Scotland: Edinburgh University Press.
Doncaster, C. P., V. Alonso Chávez, C. Viguier, R. Wang, E. Zhang, X. Dong, J. A. Dearing, P. G. Langdon & J. G. Dyke (2016) Early warning of critical transitions in biodiversity from compositional disorder. Ecology, 97, 11, 3079-3090. 10.1002/ecy.1558
Doubilet, P., C. B. Begg, M. C. Weinstein, P. Braun & B. J. McNeil (1985) Probabilistic Sensitivity Analysis Using Monte Carlo Simulation: A Practical Approach. Medical Decision Making, 5, 2, 157-177. 10.1177/0272989x8500500205
Drijfhout, S., S. Bathiany, C. Beaulieu, V. Brovkin, M. Claussen, C. Huntingford, M. Scheffer, G. Sgubin & D. Swingedouw (2015) Catalogue of abrupt shifts in Intergovernmental Panel on Climate Change climate models. Proceedings of the National Academy of Sciences, 112, 43, E5777-E5786. 10.1073/pnas.1511451112
Dube, A. D., G. Jayaraman, A. D. Rao & P. K. Mohanty. 2008. Numerical simulation of salinity structure in Chilika Lake. In Monitoring and Modelling Lakes and Coastal Environments, ed. P. K. Mohanty, 136-150. Dordrecht, The Netherlands: Springer.
Duer-Balkind, M., K. R. Jacobs, B. Güneralp & X. Basurto (2013) Resilience, Social- Ecological Rules, and Environmental Variability in a Two-Species Artisanal Fishery. Ecology and Society, 18, 4. 10.5751/ES-05751-180450
Dujovny, E. (2009) The deepest cut: political ecology in the dredging of the new sea mouth in Chilika Lake, Orissa, India. Conservation & Society, 7, 3, 192- 204
Eakin, H. & A. L. Luers (2006) Assessing he vulnerability of social-environmental systems. Annual Review of Environment and Resources, 31, 365-394
342
Eder, D. & L. Fingerson. 2001. Interviewing children and adolescents. In Handbook of Interview Research: Context and Method, eds. J. F. Gubrium & J. A. Holstein. California, U.S.A: Sage Publications, Inc.
Egelston, A. E. 2013. Sustainable Development: A History. Dordrecht, Netherlands: Springer.
Ellis, F. (1998) Household strategies and rural livelihood diversification. The Journal of Development Studies, 35, 1, 1-38. 10.1080/00220389808422553
Elliston, L. & L. Cao (2006) An agent-based bioeconomic model of a fishery with input controls. Mathematical and Computer Modelling, 44, 5, 565-575. https://doi.org/10.1016/j.mcm.2006.01.010
Engel, S. 1990. Ecosystem responses to growth and control of submerged macrophytes: a literature review. In Technical Bulletin No. 170, 24. Madison, Wisconsin, U.S.A: Deparment of Natural Resources.
Ercsey-Ravasz, M., Z. Toroczkai, Z. Lakner & J. Baranyi (2012) Complexity of the International Agro-Food Trade Network and Its Impact on Food Safety. PLOS ONE, 7, 5, e37810. 10.1371/journal.pone.0037810
Ericson, J. P., C. J. Vörösmarty, S. L. Dingman, L. G. Ward & M. Meybeck (2006) Effective sea-level rise and deltas: Causes of change and human dimension implications. Global and Planetary Change, 50, 1–2, 63-82. http://dx.doi.org/10.1016/j.gloplacha.2005.07.004
Eurostat. 1999. Towards environmental pressure indicators for the EU. In Panorama of the European Union, Theme 8, Environment and Energy. Luxembourg.
Fagherazzi, S. (2008) Self-organization of tidal deltas. Proceedings of the National Academy of Sciences, 105, 48, 18692-18695. 10.1073/pnas.0806668105
FAO. 1984. Report of the FAO World Conference on Fisheries Management and Development: Rome, 27 June to 6 July 1984. 60. Rome, Italy: Food and Agriculture Organization of the United Nations.
FAO. 1990. The tools and how to use them: semi-structured interviews. In The community's toolbox: The idea, methods and tools for participatory assessment, monitoring and evaluation in community forestry, ed. D. D. Case. Food and Agriculture Organization of the United Nations.
FAO. 2002. Section N: Fishing Effort. In CWP Handbook of Fishery Statistical Standards. Rome, Italy: Food and Agriculture Organization of the United Nations: Fisheries and Aquaculture Department
FAO. 2008. FAO Methodology for the Measurement of Food Deprivation: Updating the Minimum Dietry Energy Requirements. ed. F. S. Division, 16. Rome, Italy: Food and Agricultural Organisation of the United Nations.
FAO. 2015. Food and Agriculture Organization of the United Nations Statistics Division
Fath, B. D., C. A. Dean & H. Katzmair (2015) Navigating the adaptive cycle: an approach to managing the resilience of social systems. Ecology and Society, 20, 2. 10.5751/ES-07467-200224
343 Appendices
Featherston, C. R. & M. Doolan. 2012. A Critical Review of the Criticisms of System Dynamics In Proceedings of the 30th International Conference of the System Dynamics Society, 13. St.Gallen, Switzerland: International System Dynamics Society.
Fedra, K. 1983. Environmental Modeling Under Uncertainty: Monte Carlo Simulation. In IIASA Research Report, 86. Laxenburg, Austria: International Institute for Applied Systems Analysis.
Fennessey, N. M. & R. M. Vogel (1996) Regional models of potential evaporation and reference evapotranspiration for the northeast USA. Journal of Hydrology, 184, 337-354
Ficklin, D. L., I. T. Stewart & E. P. Maurer (2013) Effects of projected climate change on the hydrology in the Mono Lake Basin, California. Climatic Change, 116, 1, 111-131. 10.1007/s10584-012-0566-6
Figueiredo, J. & H. Pereira (2011) Regime shifts in a socio-ecological model of farmland abandonment. Landscape Ecology, 26, 5, 737-749. 10.1007/s10980-011-9605-3
Filatova, T., J. G. Polhill & S. van Ewijk (2016) Regime shifts in coupled socio- environmental systems: Review of modelling challenges and approaches. Environmental Modelling & Software, 75, 333-347. http://dx.doi.org/10.1016/j.envsoft.2015.04.003
Finnigan, J. J. (2017) Society as a complex system: Can we find a safe and just operating space for humanity? Journal and Proceedings of the Royal Society of New South Wales, 150, 463/464, 31-47
Fischer, J., A. D. Manning, W. Steffen, D. B. Rose, K. Daniell, A. Felton, S. Garnett, B. Gilna, R. Heinsohn, D. B. Lindenmayer, B. MacDonald, F. Mills, B. Newell, J. Reid, L. Robin, K. Sherren & A. Wade (2007) Mind the sustainability gap. Trends in Ecology & Evolution, 22, 12, 621-624. http://dx.doi.org/10.1016/j.tree.2007.08.016
Foley, M. M., R. G. Martone, M. D. Fox, C. V. Kappel, L. A. Mease, A. L. Erickson, B. S. Halpern, K. A. Selkoe, P. Taylor & C. Scarborough (2015) Using Ecological Thresholds to Inform Resource Management: Current Options and Future Possibilities. Frontiers in Marine Science, 2, 95
Folke, C. (2006) Resilience: The emergence of a perspective for social–ecological systems analyses. Global Environmental Change, 16, 3, 253-267. http://dx.doi.org/10.1016/j.gloenvcha.2006.04.002
Folke, C., R. Biggs, A. V. Norström, B. Reyers & J. Rockström (2016) Social- ecological resilience and biosphere-based sustainability science. Ecology and Society, 21, 3. 10.5751/ES-08748-210341
Folke, C., S. Carpenter, T. Elmqvist, L. Gunderson, C. S. Holling & B. Walker (2002) Resilience and Sustainable Development: Building Adaptive Capacity in a World of Transformations. AMBIO: A Journal of the Human Environment, 31, 5, 437-440. 10.1579/0044-7447-31.5.437
Folke, C., S. R. Carpenter, B. Walker, M. Scheffer, T. Chapin & J. Rockström (2010) Resilience Thinking: Integrating Resilience, Adaptability and Transformability. Ecology and Society, 15, 4,
344
Forbes, A. T. & D. P. Cyrus. 1998. Status and future of the St. Lucia Lake system, a large estuary of the Southwestern Indian Ocean. In Large Marine Ecosystems of the Indian Ocean: Assessment, Sustainability, and Management, eds. K. Sherman, E. N. Okemwa & M. J. Ntiba, 283-296. Oxford, U.K: Blackwell Science.
Ford, A. 2010. Modeling the Environment. Washington DC, USA: Island Press.
Ford, D. N. (1999) A behavioural approach to feedback loop dominance analysis. Systems Dynamics Review, 15, 1, 3-36
Forrester, J. W. 1961. Industrial Dynamics. Cambridge, Massachusetts, U.S.A: Massachusetts Institute of Technology Press.
Forrester, J. W. & P. M. Senge. 1980. Tests for building confidence in system dynamics models. ed. 1980, 40. Cambridge, Massachesetts: Massachesetts Insitute of Technology.
Franco, D. 2007. Fifity years of table functions. In International System Dynamics Conference, 9. Boston, USA.
Friedenburg, J. 2009. Dynamical Psychology: Complexity, Self-organization and Mind. Mansfield, USA: ISCE Publishing.
Fulton, E. A., A. D. M. Smith, D. C. Smith & P. Johnson (2014) An Integrated Approach Is Needed for Ecosystem Based Fisheries Management: Insights from Ecosystem-Level Management Strategy Evaluation. PLOS ONE, 9, 1, e84242. 10.1371/journal.pone.0084242
Gallopín, G. C. (1997) Branching futures and energy projections. Renewable Energy for Development, 10, 3, 4-5
Gallopín, G. C. 2002. Planning for resilience: scenarios, surprises, and branch points. In Panarchy: Understanding Transformations in Human and Natural Systems, eds. L. H. Gunderson & C. S. Holling, 361-394. Washington DC, U.S.A Island Press.
Gallopín, G. C., S. Funtowicz, M. O'Connor & J. Ravetz (2001) Science for the Twenty-First Century: From Social Contract to the Scientific Core. International Social Science Journal, 53, 168, 219-229. 10.1111/1468- 2451.00311
Gallopín, G. C., P. Gutman & P. Maletta (1989) Global impoverishment, sustainable development and the environment: a conceptual approach. International Social Science Journal, 121, 375-397
Gallopín, G. C. & P. Raskin (1998) Windows on the Future: Global Scenarios & Sustainability. Environment: Science and Policy for Sustainable Development, 40, 3, 6-11. 10.1080/00139159809603187
Gehrke, P. C., M. J. Sheaves, D. Boseto, B. S. Figa & J. Wani. 2011. Vulnerability of freshwater and estuarine fisheries in the tropical Pacific to climate change. In Vulnerability of Tropical Pacific Fisheries and Aquaculture to Climate Change, eds. J. D. Bell, J. E. Johnson & A. J. Hobday, 577-546. Noumea, New Caledonia: Secretariat of the Pacific Community
Gell-Mann, M. 2010. Transformations of the twenty-first century: transitions to greater sustainability. In Global Sustainability—A Nobel Cause, eds. H. J.
345 Appendices
Schellnhuber, M. Molina, N. Stern, V. Huber & S. Kadner, 1-8. Cambridge, UK: Cambridge University Press.
Ghashghaei, M., A. Bagheri & S. Morid (2013) Rainfall-runoff Modeling in a Watershed Scale Using an Object Oriented Approach Based on the Concepts of System Dynamics. Water Resources Management, 27, 15, 5119-5141. 10.1007/s11269-013-0457-2
Ghosh, A. & A. K. Pattnaik. 2003. Experience Brief for Chilika Lake Orissa, India. 37. Indian Center for Environment and Development
Ghosh, A., A. K. Pattnaik & T. J. Ballatore (2006) Chilika Lagoon: restoring ecological balance and livelihoods through re-salinization. Lakes & Reservoirs: Research and Managment 11, 239-255
Ghosh, S., D. Raje & P. P. Mujumdar (2010) Mahanadi streamflow: climate change impact assessment and adaptive strategies. Current Science, 98, 8, 1084- 1091
Gini, C. (1912) Variabilità e Mutabilità, contributo allo studio delle distribuzioni e relazioni statistiche (Variability and Instability, a contribution to the study of statistical distributions and relationships). Studi Economico-Giuridici dell' Univ. di Cagliari, 3, 2, 1-158
Giri, C., J. Long, S. Abbas, R. M. Murali, F. M. Qamer, B. Pengra & D. Thau (2014) Distribution and dynamics of mangrove forests of South Asia. Journal of Environmental Management, 0. http://dx.doi.org/10.1016/j.jenvman.2014.01.020
Glaser, M. & B. Glaeser (2014) Towards a framework for cross-scale and multi- level analysis of coastal and marine social-ecological systems dynamics. Regional Environmental Change, 14, 6, 2039-2052. 10.1007/s10113-014- 0637-5
Gosain, A. K., S. Rao & D. Basuray (2006) Climate change impact assessment on hydrology of Indian river basins. Current Science 90, 3, 346-353
Graham, A. K. 1976. Parameter formulation and estimation in system dynamics models. In International System Dynamics Conference, 541-580. Geilo, Norway.
Green, A. J., P. Alcorlo, E. T. H. M. Peeters, E. P. Morris, J. L. Espinar, M. A. Bravo- Utrera, J. Bustamante, R. Díaz-Delgado, A. A. Koelmans, R. Mateo, W. M. Mooij, M. Rodríguez-Rodríguez, E. H. van Nes & M. Scheffer (2017) Creating a safe operating space for wetlands in a changing climate. Frontiers in Ecology and the Environment, 15, 2, 99-107. 10.1002/fee.1459
Groffman, P., J. Baron, T. Blett, A. Gold, I. Goodman, L. Gunderson, B. Levinson, M. Palmer, H. Paerl, G. Peterson, N. L. Poff, D. Rejeski, J. Reynolds, M. Turner, K. Weathers & J. Wiens (2006) Ecological Thresholds: The Key to Successful Environmental Management or an Important Concept with No Practical Application? Ecosystems, 9, 1, 1-13. 10.1007/s10021-003-0142-z
Grossman, G., M & A. B. Krueger (1995) Economic growth and the environment. Quaterly Journal of Economics, 110, 353-377
Gunderson, L. & C. S. Holling. 2002. Panarchy: Understanding Transformations in Human and Natural Systems. Washington DC, USA: Island Press.
346
Guo, H. C., L. Liu, G. H. Huang, G. A. Fuller, R. Zou & Y. Y. Yin (2001) A system dynamics approach for regional environmental planning and management: A study for the Lake Erhai Basin. Journal of Environmental Management, 61, 93-101
Gupta, H., S.-J. Kao & M. Dai (2012) The role of mega dams in reducing sediment fluxes: A case study of large Asian rivers. Journal of Hydrology, 464–465, 0, 447-458. http://dx.doi.org/10.1016/j.jhydrol.2012.07.038
Guttal, V. & C. Jayaprakash (2008) Changing skewness: an early warning signal of regime shifts in ecosystems. Ecology Letters, 11, 5, 450-460. 10.1111/j.1461-0248.2008.01160.x
Hamilton, L. C. & M. J. Butler (2001) Outport adaptations: social indicators through Newfoundland's Cod crisis. Research in Human Ecology, 8, 2, 1-11
Hammond, R. & P. S. McCullagh. 1974. Quantitative Techniques in Geography: An Introduction. Oxford, UK: Clarendon Press.
Hannon, B. & M. Ruth. 2009. Dynamic Modeling of Diseases and Pests. New York, New York, USA: Springer.
Hanspach, J., T. Hartel, A. I. Milcu, F. Mikulcak, I. Dorresteijn, J. Loos, H. von Wehrden, T. Kuemmerle, D. Abson, A. Kovács-Hostyánszki, A. Báldi & J. Fischer (2014) A holistic approach to studying social-ecological systems and its application to southern Transylvania. Ecology and Society, 19, 4. 10.5751/ES-06915-190432
Hardin, G. (1974) Living on a lifeboat. Bioscience, 24, 10, 561-8
Harley, S. J., R. A. Myers & A. Dunn (2001) Is catch-per-unit-effort proportional to abundance? Canadian Journal of Fisheries and Aquatic Sciences, 58, 9, 1760-1772
Harris, J. M. & P. Vinobaba (2013) Influence of hydrochemistry on biotic components of the Batticaloa lagoon, Sri Lanka. International Journal of Environmental Sciences, 3, 5, 1603-1613
Harrison, R. L. (2010) Introduction To Monte Carlo Simulation. AIP conference proceedings, 1204, 17-21. 10.1063/1.3295638
Hastings, A. & D. B. Wysham (2010) Regime shifts in ecological systems can occur with no warning. Ecology Letters, 13, 4, 464-472. 10.1111/j.1461- 0248.2010.01439.x
Häyhä, T., P. L. Lucas, D. P. van Vuuren, S. E. Cornell & H. Hoff (2016) From Planetary Boundaries to national fair shares of the global safe operating space — How can the scales be bridged? Global Environmental Change, 40, 60-72. http://dx.doi.org/10.1016/j.gloenvcha.2016.06.008
Heath, M. R. 1992. Field Investigations of the Early Life Stages of Marine Fish. In Advances in Marine Biology, eds. J. H. S. Blaxter & A. J. Southward, 1-174. Academic Press.
Heckbert, S. (2013) MayaSim: An Agent-Based Model of the Ancient Maya Social- Ecological System. Journal of Artificial Societies and Social Simulation, 16, 4, 11
347 Appendices
Hekimoglu, M. & Y. Barlas. 2010. Sensitivity analysis of system dynamics models by behaviour pattern measures. In International Conference of the System Dynamics Society, 31. Seoul, Korea.
Helton, J. C. & F. J. Davis. 2000. Sampling-based methods. In Sensitivity Analysis, eds. A. Saltelli, K. Chan & E. M. Scott, 101-154. Chichester, U.K: John Wiley & Sons Ltd.
Helton, J. C., J. D. Johnson, C. J. Sallaberry & C. B. Storlie. 2006. Survey of Sampling-Based Methods for Uncertainty and Sensitivity Analysis. In Sandia Report, 84. Albuquerque, New Mexico, USA: Sandia National Laboratories.
Heylighen, F., P. Cilliers & C. Gershenson. 2007. Complexity and philosophy. In Complexity, Science and Society, eds. J. Bogg & R. Geyer, 117-134. Oxford, UK: Radcliffe.
Hill, N. A. O., J. M. Rowcliffe, H. J. Koldewey & E. J. Milner-Gulland (2012) The Interaction between Seaweed Farming as an Alternative Occupation and Fisher Numbers in the Central Philippines. Conservation Biology, 26, 2, 324-334. 10.1111/j.1523-1739.2011.01796.x
Hinkel, J., M. E. Cox, M. Schlüter, C. R. Binder & T. Falk (2015) A diagnostic procedure for applying the social-ecological systems framework in diverse cases. Ecology and Society, 20, 1. 10.5751/ES-07023-200132
Hjorth, P. & A. Bagheri (2006) Navigating towards sustainable development: A system dynamics approach. Futures, 38, 1, 74-92. https://doi.org/10.1016/j.futures.2005.04.005
Holden, E., K. Linnerud & D. Banister (2014) Sustainable development: Our Common Future revisited. Global Environmental Change, 26, 130-139. http://dx.doi.org/10.1016/j.gloenvcha.2014.04.006
Holden, E., K. Linnerud & D. Banister (2017) The Imperatives of Sustainable Development. Sustainable Development, 25, 3, 213-226. 10.1002/sd.1647
Holdschlag, A. & B. M. W. Ratter (2016) Caribbean island states in a social- ecological panarchy? Complexity theory, adaptability and environmental knowledge systems. Anthropocene, 13, Supplement C, 80-93. https://doi.org/10.1016/j.ancene.2016.03.002
Holland, J. H. 1995. Hidden Order: How Adaptation Builds Complexity. New York, USA: Helix Books.
Holling, C. S. (1959a) The components of predation as revealed by a study of small mammal predation of the European pine sawfly. Canadian Entomologist, 91, 293-320. https://doi.org/10.4039/Ent91293-5
Holling, C. S. (1959b) Some characteristics of simple types of predation and parasitism. Canadian Entomologist, 91, 385-398. https://doi.org/10.4039/Ent91385-7
Holling, C. S. (1973) Resilience and Stability of Ecological Systems. Annual Review of Ecology and Systematics, 4, 1, 1-23. doi:10.1146/annurev.es.04.110173.000245
Holling, C. S. 1986. The resilience of terrestrial ecosystems: local surprise and global change. In Sustainable Development of the Biosphere: Interactions
348
Between the World Economy and the Global Environment, eds. W. C. Clark & R. E. Munn, 292-317. Cambridge, UK: Cambridge University Press.
Holling, C. S. (2001) Understanding the Complexity of Economic, Ecological, and Social Systems. Ecosystems, 4, 5, 390-405. 10.1007/s10021-001-0101-5
Holling, C. S., G. Peterson, P. Marples, J. Sendzimir, K. Redford, L. Gunderson & D. Lambert. 1996. Self-organization in ecosystems: lumpy geometries, periodicities and morphologies. In Global Change and Terrestrial Ecosystems, eds. B. Walker & W. Steffen, 622. Cambridge, UK: Cambridge University Press.
Holmberg, J., K. H. Robèrt & K. E. Eriksson. 1996. Socio-ecological principles for a sustainable society- scientific background and Swedish experience. In Getting Down to Earth: Practical Application for Ecological Economics, ed. R. Costanza, 17-48. Washington DC, USA.: Island Press.
Holmes, D. L. (1993) Rise of the Nile Delta. Nature, 363, 6428, 402-403
Homer-Dixon, T. F. 2006. The Upside of Down: Catastrophe, Creativity and the Renewal of Civilisation. London, United Kingdom: Souvenir Press.
Hoorweg, J., N. Versleijen, B. Wangila & A. Degen. 2006. Income diversification and fishing practices among artisanal fishers on the Malindi-Kalifi coast. In Coastal Ecology Conference IV, 18. Mombasa, Kenya.
Hopfenburg, R. (2003) Human carrying capacity is determined by food availability. Population and Environment, 25, 2, 109-117. 10.1023/B:POEN.0000015560.69479.c1
Hopwood, B., M. Mellor & G. O'Brien (2005) Sustainable development: mapping different approaches. Sustainable Development, 13, 1, 38-52. 10.1002/sd.244
Hoshino, E., I. van Putten, W. Girsang, B. P. Resosudarmo & S. Yamazaki (2016) A Bayesian belief network model for community-based coastal resource management in the Kei Islands, Indonesia. Ecology and Society, 21, 2. 10.5751/ES-08285-210216
Hossain, M. S., J. A. Dearing, F. Eigenbrod & F. A. Johnson (2017) Operationalizing safe operating space for regional social-ecological systems. Science of The Total Environment, 584-585, 673-682. http://dx.doi.org/10.1016/j.scitotenv.2017.01.095
Houghton, J. 2009. Global Warming: The Complete Briefing. Cambridge, U.K: Cambridge University Press.
Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell & C. A. Johnson. 2001. Climate Change 2001: The Scientific Basis. In Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, ed. IPCC, 94. Cambridge, U.K.
Hugget, R. J. 1985. Earth Surface Systems. Berlin, Germany: Springer-Verlag.
Huggett, R. J. 2004. Fundamentals of Biogeography. Abindgon, U.K: Routledge.
Hughes, T. P., M. L. Barnes, D. R. Bellwood, J. E. Cinner, G. S. Cumming, J. B. C. Jackson, J. Kleypas, I. A. van de Leemput, J. M. Lough, T. H. Morrison, S. R.
349 Appendices
Palumbi, E. H. van Nes & M. Scheffer (2017) Coral reefs in the Anthropocene. Nature, 546, 7656, 82-90. 10.1038/nature22901 http://www.nature.com/nature/journal/v546/n7656/abs/nature22901.html#sup plementary-information
Hughes, T. P., C. Linares, V. Dakos, I. A. van de Leemput & E. H. van Nes (2013) Living dangerously on borrowed time during slow, unrecognized regime shifts. Trends in Ecology & Evolution, 28, 3, 149-155. http://dx.doi.org/10.1016/j.tree.2012.08.022
Hutchings, J. A. (2000) Collapse and recovery of marine fishes. Nature, 406, 6798, 882-885
Hyndman, R. & G. Athanasopoulos. 2014. Forecasting: principles and practice. [online] accessed 10th January 2017: Otexts.
Hyndman, R. & Y. Khandakar (2008) Automatic time series forecasting: the forecast package for R. Journal of Statistical Software, 27, 2, 22
ICAR-CIFRI. 2016. Annual Report: 2015-16. 168. Barrackpore, Kolkata: ICAR- Central Inalnd Fisheries Research Institute.
IPCC. 2013. Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. eds. T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex & P. M. Midgley, 1535 pp. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
IPCC. 2014a. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change In IPCC Core Writing Team, eds. R. K. Pachauri & L. A. Meyer, 151. Geneva, Switzerland.
IPCC. 2014b. Summary for Policy Makers. In Climate Change 2014: Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change eds. O. Edenhofer, Y. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seybold, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel & J. C. Minx, 30. Cambridge: Intergovernmental Panel on Climate Change.
ISEE. 2015. STELLA: Systems Thinking for Education and Research. In http://www.iseesystems.com/softwares/Education/StellaSoftware.aspx. ISEE systems.
IUCN. 2008. Orcaella brevirostris The IUCN Red List of Threatened Species: International Union for Conservation of Nature.
Ives, A. R. (1995) Measuring Resilience in Stochastic Systems. Ecological Monographs, 65, 2, 217-233. 10.2307/2937138
Iwasaki, S., B. Razafindrabe & R. Shaw (2009) Fishery livelihoods and adaptation to climate change: a case study of Chilika lagoon, India. Mitigation and Adaptation Strategies for Global Change, 14, 4, 339-355. 10.1007/s11027-009-9167-8
350
Iwasaki, S. & R. Shaw (2009) Linking human security to natural resources: perspective from a fishery resource allocation system in Chilika lagoon, India. Sustainability Science, 4, 2, 281-292. 10.1007/s11625-009-0084-2
Iwasaki, S. & R. Shaw. 2010a. Integrated climate change adaptation in Chilika Lagoon Fisheries, India. In Climate Change Adaptation and Disaster Risk Reduction: An Asian Perspective, eds. R. Shaw, J. M. Pulhin & J. J. Pereira, 81-103. Bingley, UK: Emerald Group Publishing Limited.
Iwasaki, S. & R. Shaw. 2010b. Integrated lagoon fisheries management: resource dynamics and adaptation. Bingley, UK: Emeral Group Publishing Limited.
Jackson, M. C. 1991. Systems Methodology for the Management Sciences. New York City, New York, U.S.A: Springer Science & Business Media.
Jaegar, C., P. Jansson, S. van der Leeuw, M. Resch & J. D. Tàbara. 2013. GSS: Towards a Research Program for Global Systems Science. In Second Open Global Systems Science Conference, 28. Brussels.
Jakeman, A. J., R. A. Letcher & J. P. Norton (2006) Ten iterative steps in development and evaluation of environmental models. Environmental Modelling & Software, 21, 5, 602-614. http://dx.doi.org/10.1016/j.envsoft.2006.01.004
Janssen, M. A. & J. M. Anderies (2007) Robustness trade-offs in social-ecological systems. International Journal of the Commons, 1, 1, 43-66. http://doi.org/10.18352/ijc.12
Janssen, M. A., Ö. Bodin, J. M. Anderies, T. Elmqvist, H. Ernstson, R. R. J. McAllister, P. Olsson & P. Ryan (2006) Toward a Network Perspective of the Study of Resilience in Social-Ecological Systems. Ecology and Society, 11, 1,
Janssen, M. A. & E. Ostrom. 2006. Chapter 30 Governing Social-Ecological Systems. In Handbook of Computational Economics, eds. L. Tesfatsion & K. L. Judd, 1465-1509. Elsevier.
Jayaraman, G., A. D. Rao, A. D. Dube & P. K. Mohanty (2007) Numerical simulation of circulation and salinity structure in Chilika Lagoon. Journal of Coastal Research, 23, 4, 861-877
Jena, P. P., C. Chatterjee, G. Pradhan & A. Mishra (2014) Are recent frequent high floods in Mahanadi basin in eastern India due to increase in extreme rainfalls? Journal of Hydrology, 517, 0, 847-862. http://dx.doi.org/10.1016/j.jhydrol.2014.06.021
Johnson, C. J. (2013) Identifying ecological thresholds for regulating human activity: Effective conservation or wishful thinking? Biological Conservation, 168, 0, 57-65. http://dx.doi.org/10.1016/j.biocon.2013.09.012
Jolliffe, D. & E. B. Prydz. 2016. Estimating International Poverty Lines from Comparable National Thresholds. In Policy Research Working Paper 7606, ed. D. R. G. Poverty and Inequality Team, 36.
Jones, S. & K. H. Sujansingani (1954) Fish and fisheries of the Chilika Lake with statistics of fish catches for the years 1948-1950. Indian Journal of Fisheries, 1, 1, 256-347
Jul-Larsen, E., J. Kolding, R. Overå, J. R. Nielsen & P. A. M. van Zwieten. 2003. Management, co-management or no management? Major dilemmas in
351 Appendices
southern African freshwater fisheries. In FAO Fisheries Technical Paper, ed. FAO, 9. Rome, Italy: Food and Agriculture Organization of the United Nations.
Kadekodi, G. K. & S. C. Gulati. 1999. Root causes of biodiversity losses in Chilika Lake. In CMDR Monograp Series No.-26, ed. WWF-India, 29.
Kadekodi, G. K., S. C. Gulati, S. K. Adhikari, R. Ranjan & D. Kar. 2000. India: Chilika Lake. In The Root Causes of Biodiversity Loss, eds. A. Wood, P. Stedman-Edwards & J. Mang, 213-230. London, U.K: Earthscan.
Kadekodi, G. K. & A. Nayampalli. 2005. Reversing biodiversity degradation: a case of Chilika Lake. In Dimensions of Environmental and Ecological Economics, eds. N. C. Sahu & A. K. Choudhury, 527-545. Hyderabad, India: Universities Press (India) Private Limited.
Kasprzyk, J. R., S. Nataraj, P. M. Reed & R. J. Lempert (2013) Many objective robust decision making for complex environmental systems undergoing change. Environmental Modelling & Software, 42, 55-71. http://dx.doi.org/10.1016/j.envsoft.2012.12.007
Kates, R. W., W. C. Clark, R. Corell, J. M. Hall, C. C. Jaeger, I. Lowe, J. J. McCarthy, H. J. Schellnhuber, B. Bolin, N. M. Dickson, S. Faucheux, G. C. Gallopin, A. Grübler, B. Huntley, J. Jäger, N. S. Jodha, R. E. Kasperson, A. Mabogunje, P. Matson, H. Mooney, B. Moore, T. O'Riordan & U. Svedin (2001) Sustainability Science. Science, 292, 5517, 641-642. 10.1126/science.1059386
Kauffman, S. A. 1991. The Sciences of Complexity and Origins of Order. In SFI Working Paper: 1991-04-021, 25. New Mexico, USA: Santa Fe Institute.
Keck, B. P., Z. H. Marion, D. J. Martin, J. C. Kaufman, C. P. Harden, J. S. Schwartz & R. J. Strange (2014) Fish Functional Traits Correlated with Environmental Variables in a Temperate Biodiversity Hotspot. PLoS ONE, 9, 3, e93237. 10.1371/journal.pone.0093237
Kellie-Smith, O. & P. M. Cox (2011) Emergent dynamics of the climate-economy system in the Anthropocene Philosophical Transactions of the Royal Society A, 369, 868-886
Kelly, R. P., A. L. Erickson, L. A. Mease, W. Battista, J. N. Kittinger & R. Fujita (2015) Embracing thresholds for better environmental management. Philosophical Transactions of the Royal Society B, 370, 1659. 10.1098/rstb.2013.0276
Kemker, C. 2013. Dissolved Oxygen. In Fundamentals of Environmental Measurements, ed. F. E. Inc. online.
Keyl, F. & M. Wolff (2008) Environmental variability and fisheries: what can models do? Reviews in Fish Biology and Fisheries, 18, 3, 273-299. 10.1007/s11160-007-9075-5
Khan, N., A. Lucas & K. Linard. 2004. Development of a reference mode for characterisation of salinity problem in the Murray Darling Basin. In International System Dynamics Conference, 29. Oxford, UK.
Kim, D. H. (1991) Guidelines for drawing causal loop diagrams. The Systems Thinker, 3, 1, 5-6
352
King, A. (1995) Avoiding Ecological Surprise: Lessons from Long-Standing Communities. The Academy of Management Review, 20, 4, 961-985. 10.2307/258962
Kingdon, J. W. 1995. Agendas, Alternatives, and Public Policies. New York, USA: Harper Collins.
Kinzig, A. P., P. Ryan, M. Etienne, H. Allison, T. Elmqvist & B. H. Walker (2006) Resilience and Regime Shifts: Assessing Cascading Effects. Ecology and Society, 11, 1, 23
Kittinger, J. N., E. M. Finkbeiner, N. C. Ban, K. Broad, M. H. Carr, J. E. Cinner, S. Gelcich, M. L. Cornwell, J. Z. Koehn, X. Basurto, R. Fujita, M. R. Caldwell & L. B. Crowder (2013) Emerging frontiers in social-ecological systems research for sustainability of small-scale fisheries. Current Opinion in Environmental Sustainability, 5, 3, 352-357. http://dx.doi.org/10.1016/j.cosust.2013.06.008
Kocurek, G. & R. C. Ewing (2005) Aeolian dune field self-organization – implications for the formation of simple versus complex dune-field patterns. Geomorphology, 72, 1–4, 94-105. http://dx.doi.org/10.1016/j.geomorph.2005.05.005
Kopnina, H. (2016) The victims of unsustainability: a challenge to sustainable development goals. International Journal of Sustainable Development & World Ecology, 23, 2, 113-121. 10.1080/13504509.2015.1111269
Koshel, P. & K. McAllister. 2008. Transitioning to Sustainability Through Research and Developmnet on Ecosystem Services and Biofuels. In Science and Technology for Sustainability Program, 130. Washington D.C., USA: National Research Council of the National Academies.
Kramer, K.-L. 2012. User Experience in the Age of Sustainability: A Practitioner’s Blueprint. Waltham, USA: Morgan Kaufmann.
Kuikka, S., M. Hildén, H. Gislason, S. Hansson, H. Sparholt & V. Olli. 1999. Modeling environmentally driven uncertainties in Baltic cod (Gadus morhua) management by Bayesian influence diagrams.
Kumar, K. K., A. Kulkarni, S. Patwardhan & K. K. Rao. 2012. Indo-UK Collaborative Research Programme on the Impacts of Climate Change on India. 30. Pune, India: Indian Institute of Tropical Meteorology.
Kumar, R., P. Horwitz, G. R. Milton, S. S. Sellamuttu, S. T. Buckton, N. C. Davidson, A. K. Pattnaik, M. Zavagli & C. Baker (2011) Assessing wetland ecosystem services and poverty interlinkages: a general framework and case study. Hydrological Sciences Journal, 56, 8, 1602-1621
Kumar, R. & A. K. Pattnaik. 2012. Chilika: An Integrated Management Planning Framework for Conservation and Wise Use. 178. New Delhi, India and Bhubaneswar, India: Wetlands International South Asia and Chilika Development Authority.
Kummu, M. & O. Varis (2007) Sediment-related impacts due to upstream reservoir trapping, the Lower Mekong River. Geomorphology, 85, 3–4, 275-293. http://dx.doi.org/10.1016/j.geomorph.2006.03.024
Kwakkel, J. H., S. Eker & E. Pruyt. 2016. How Robust is a Robust Policy? Comparing Alternative Robustness Metrics for Robust Decision-Making. In
353 Appendices
Robustness Analysis in Decision Aiding, Optimization, and Analytics, eds. M. Doumpos, C. Zopounidis & E. Grigoroudis, 221-237. Cham: Springer International Publishing.
Kwakkel, J. H. & E. Pruyt (2013) Exploratory Modeling and Analysis, an approach for model-based foresight under deep uncertainty. Technological Forecasting and Social Change, 80, 3, 419-431. https://doi.org/10.1016/j.techfore.2012.10.005
Kwakkel, J. H. & J. S. Timmermans. 2012. Safe operating spaces for human water use: applying exploratory modeling and patient rule induction to ANEMI. In 2012 International Congress on Environmental Modelling and Software Managing Resources of a Limited Plane, 10. Leipzig, Germany: International Environmental Modelling and Software Society (iEMSs)
Kwakkel, J. H., W. E. Walker & V. A. W. J. Marchau (2010) Classifying and communicating uncertainties in model-based policy analysis. International Journal of Technology, Policy and Management, 10, 4, 299-315
Lade, S., A. Tavoni, S. Levin & M. Schlüter (2013) Regime shifts in a social- ecological system. Theoretical Ecology, 6, 3, 359-372. 10.1007/s12080- 013-0187-3
Lade, S. J., S. Niiranen, J. Hentati-Sundberg, T. Blenckner, W. J. Boonstra, K. Orach, M. F. Quaas, H. Österblom & M. Schlüter (2015) An empirical model of the Baltic Sea reveals the importance of social dynamics for ecological regime shifts. Proceedings of the National Academy of Sciences, 112, 35, 11120- 11125. 10.1073/pnas.1504954112
Ladyman, J., J. Lambert & K. Wiesner (2013) What is a complex system? European Journal for Philosophy of Science, 3, 1, 33-67. 10.1007/s13194-012-0056- 8
Lambin, E. F. & H. J. Geist. 2006. Land-use and land-cover change: local processes and global impacts. Berlin, Germany: Springer.
Lane, D. C. (2000) Should system dynamics be described as a ‘hard’ or ‘deterministic’ systems approach? Systems Research and Behavioural Science, 17, 1, 3-22
Lang, D., A. Wiek, M. Bergmann, M. Stauffacher, P. Martens, P. Moll, M. Swilling & C. Thomas (2012) Transdisciplinary research in sustainability science: practice, principles, and challenges. Sustainability Science, 7, 1, 25-43. 10.1007/s11625-011-0149-x
Lansing, J. S. (2003) Complex Adaptive Systems. Annual Review of Anthropology, 32, 183-204
Lansing, J. S. & J. N. Kremer (1993) Emergent properties of Balinese water temples. American Anthropologist, 95, 1, 97-114
Larkin, P. A. (1977) An Epitaph for the Concept of Maximum Sustained Yield. Transactions of the American Fisheries Society, 106, 1, 1-11. 10.1577/1548-8659(1977)106<1:AEFTCO>2.0.CO;2
Lawrence, D. P. 2003. Environmental Impact Assessment: Practical Solutions to Recurrent Problems. New Jersey, USA: Wiley-Interscience.
354
Lazar, A. N., R. J. Nicholls, A. Payo, H. Adams, C. Mortreux, N. Suckall, C. Vincent, H. Sugata, B. A. Amisigo, M. Rahman, A. Haque, N. Adger & C. Hill. 2015. A method to assess migration and adaptation in deltas: A preliminary fast track assessment. In DECCMA Working Paper. Deltas, Vulnerability and Climate Change: Migration and Adaptation.
Leal Neto, A. C., L. L. Legey, M. González-Araya & S. Jablonski (2006) A System Dynamics Model for the Environmental Management of the Sepetiba Bay Watershed, Brazil. Environmental Management, 38, 5, 879-888. 10.1007/s00267-005-0211-5
Lebel, L., J. M. Anderies, B. Campbell, C. Folke, S. Hatfield-Dodds, T. P. Hughes & J. Wilson (2006) Governance and the Capacity to Manage Resilience in Regional Social-Ecological Systems. Ecology and Society, 11, 1. 10.5751/ES- 01606-110119
Lebel, L., T. Grothmann & B. Siebenhüner (2010) The role of social learning in adaptiveness: insights from water management. International Environmental Agreements: Politics, Law and Economics, 10, 4, 333-353. 10.1007/s10784-010-9142-6
Lee, C.-S. & B. Menu (1981) Effects of salinity on egg development and hatching in grey mullet Mugil cephalus L. Journal of Fish Biology, 19, 2, 179-188. 10.1111/j.1095-8649.1981.tb05822.x
Lélé, S. (1991) Sustainable development: a critical review. World Development, 19, 607-621
Lempert, R. J. & M. T. Collins (2007) Managing the Risk of Uncertain Threshold Responses: Comparison of Robust, Optimum, and Precautionary Approaches. Risk Analysis, 27, 4, 1009-1026. 10.1111/j.1539- 6924.2007.00940.x
Lempert, R. J., S. W. Popper & S. C. Bankes. 2003. Shaping the Next One Hundred Years: New Methods for Quantitative Long-term Stategy Analysis. 209. Santa Monica, California, U.S.A: RAND Corperation.
Lenton, T. M., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf & H. J. Schellnhuber (2008) Tipping elements in the Earth's climate system. Proceedings of the National Academy of Sciences, 105, 6, 1786-1793. 10.1073/pnas.0705414105
Letcher, R. A., A. J. Jakeman, O. Barreteau, M. E. Borsuk, S. ElSawah, S. H. Hamilton, H. J. Henriksen, S. Kuikka, H. R. Maier, A. E. Rizzoli, H. van Delden & A. A. Voinov (2013) Selecting among five common modelling approaches for integrated environmental assessment and management. Environmental Modelling & Software, 47, 0, 159-181. http://dx.doi.org/10.1016/j.envsoft.2013.05.005
Leung, F.-H. & R. Savithiri (2009) Spotlight on focus groups. Canadian Family Physician, 55, 2, 218-219
Lévêque, C. 1997. Biodiversity Dynamics and Conservation: the Freshwater Fish of Tropical Africa. Cambrdige, U.K: Cambridge University Press.
Levin, P. S. (2014) New Conservation for the Anthropocene Ocean. Conservation Letters, 7, 4, 339-340. 10.1111/conl.12108
355 Appendices
Levin, P. S., G. D. Williams, A. Rehr, K. C. Norman & C. J. Harvey (2015) Developing conservation targets in social-ecological systems. Ecology and Society, 20, 4. 10.5751/ES-07866-200406
Levin, S., T. Xepapadeas, A.-S. Crépin, J. Norberg, A. de Zeeuw, C. Folke, T. Hughes, K. Arrow, S. Barrett, G. Daily, P. Ehrlich, N. Kautsky, K.-G. Mäler, S. Polasky, M. Troell, J. R. Vincent & B. Walker (2013) Social-ecological systems as complex adaptive systems: modeling and policy implications. Environment and Development Economics, 18, 02, 111-132. doi:10.1017/S1355770X12000460
Levontin, P., S. Kulmala, P. Haapasaari & S. Kuikka (2011) Integration of biological, economic, and sociological knowledge by Bayesian belief networks: the interdisciplinary evaluation of potential management plans for Baltic salmon. ICES Journal of Marine Science, 68, 3, 632-638. 10.1093/icesjms/fsr004
Lewis-Beck, M. S., A. Bryman & T. F. Lao. 2004. The SAGE Encyclopedia of Social Science Research Methods. online: SAGE Publications, Inc. .
Lewontin, R. C. (1969) The meaning of stability. Brookhaven Symposia in Biology, 22, 13-24
Linacre, E. T. (1977) A simple formula for estimating evaporation rates in various climates, using temperature data alone. Agricultural Meteorology, 18, 6, 409-424. http://dx.doi.org/10.1016/0002-1571(77)90007-3
Lindegren, M., V. Dakos, J. P. Gröger, A. Gårdmark, G. Kornilovs, S. A. Otto & C. Möllmann (2012) Early Detection of Ecosystem Regime Shifts: A Multiple Method Evaluation for Management Application. PLoS ONE, 7, 7, e38410. 10.1371/journal.pone.0038410
Lister, G. S., D. M. Livingstone, B. Ammann, D. Arizlegui, W. Haeberli, A. F. Lotter, C. Ohlendorf, C. Pfister, Schwander, F. Schweingruber, B. Stauffer & F. Sturm. 1998. Alpine paleoclimatology. In Views from the Alps: Regional Perspectives on Climate Change, eds. P. Cebon, U. Dahinden, H. C. Davies, D. Imboden & C. C. Jaegar, 73-170. Cambridge, Massachesetts, U.S.A: The MIT Press.
Liu, H., G. Benoit, T. Liu, Y. Liu & H. Guo (2015a) An integrated system dynamics model developed for managing lake water quality at the watershed scale. Journal of Environmental Management, 155, 0, 11-23. http://dx.doi.org/10.1016/j.jenvman.2015.02.046
Liu, J., T. Dietz, S. R. Carpenter, M. Alberti, C. Folke, E. Moran, A. N. Pell, P. Deadman, T. Kratz, J. Lubchenco, E. Ostrom, Z. Ouyang, W. Provencher, C. L. Redman, S. H. Schneider & W. W. Taylor (2007) Complexity of coupled human and natural systems. Science, 317, 5844, 1513-1516. 10.1126/science.1144004
Liu, J., G. Kattel, H. P. H. Arp & H. Yang (2015b) Towards threshold-based management of freshwater ecosystems in the context of climate change. Ecological Modelling, 318, 265-274. http://dx.doi.org/10.1016/j.ecolmodel.2014.09.010
Loo, Y. Y., L. Billa & A. Singh (2015) Effect of climate change on seasonal monsoon in Asia and its impact on the variability of monsoon rainfall in
356
Southeast Asia. Geoscience Frontiers, 6, 6, 817-823. http://dx.doi.org/10.1016/j.gsf.2014.02.009
Lotka, A. J. 1925. Elements of Physical Biology. Baltimore, USA: Williams and Wilkins Company.
Loucks, D. P., E. van Beek, J. R. Stedinger, J. P. M. Dijkman & M. T. Villars. 2005. Water resources systems planning and management: an introduction to methods, models and applications. In UNESCO Studies and Reports in Hydrology, eds. D. P. Loucks & E. van Beek, 690. Paris, France: United Nations Educational, Scientific and Cultural Organization.
Luna-Reyes, L. F. & D. L. Andersen (2003) Collecting and analyzing qualitative data for system dynamics: methods and models. System Dynamics Review, 19, 4, 271-296. 10.1002/sdr.280
Lynam, T., W. de Jong, D. Sheil, T. Kusumanto & K. Evans (2007) A Review of Tools for Incorporating Community Knowledge, Preferences, and Values into Decision Making in Natural Resources Management. Ecology and Society, 12, 1,
Mace, G., H. Masundire & J. Baillie. 2005. Biodiversity. In Ecosystems and Human Wellbeing: Current State and Trends, eds. H. Hassan, R. J. Scholes & N. J. Ash, 79-115. Washington DC, U.S.A: Island Press.
Mace, P. M. (2001) A new role for MSY in single-species and ecosystem approaches to fisheries stock assessment and management. Fish and Fisheries, 2, 1, 2-32. 10.1046/j.1467-2979.2001.00033.x
Maier, H. R., J. H. A. Guillaume, H. van Delden, G. A. Riddell, M. Haasnoot & J. H. Kwakkel (2016) An uncertain future, deep uncertainty, scenarios, robustness and adaptation: How do they fit together? Environmental Modelling & Software, 81, 154-164. http://dx.doi.org/10.1016/j.envsoft.2016.03.014
Malek, S., S. M. Syed Ahmad, S. K. K. Singh, P. Milow & A. Salleh (2011) Assessment of predictive models for chlorophyll-a concentration of a tropical lake. BMC Bioinformatics, 12, 13, S12. 10.1186/1471-2105-12- S13-S12
Máñez Costa, M. A., E. J. Moors & E. D. G. Fraser (2011) Socioeconomics, Policy, or Climate Change: What is Driving Vulnerability in Southern Portugal? Ecology and Society, 16, 1. 10.5751/ES-03703-160128
Mannion, A. M. 1997. Global Environmental Change: A Natural and Cultural Environmental History. Abingdon, UK: Routledge.
Marshall, N. A. & P. A. Marshall (2007) Conceptualizing and Operationalizing Social Resilience within Commercial Fisheries in Northern Australia. Ecology and Society, 12, 1. 10.5751/ES-01940-120101
Martell, S. & R. Froese (2013) A simple method for estimating MSY from catch and resilience. Fish and Fisheries, 14, 4, 504-514. 10.1111/j.1467- 2979.2012.00485.x
Martin, S. B., A. T. Hitch, K. M. Purcell, P. L. Klerks & P. L. Leberg (2009) Life history variation along a salinity gradient in coastal marshes. Aquatic Biology, 8, 15-28. 10.3354/ab00203
357 Appendices
Martins, J. H., A. S. Camanho, M. M. Oliveira & M. B. Gaspar (2015) A system dynamics model to support the management of artisanal dredge fisheries in the south coast of Portugal. International Transactions in Operational Research, 22, 4, 611-634. 10.1111/itor.12090
Massey, F. J. (1951) The Kolmogorov-Smirnov Test for Goodness of Fit. Journal of the American Statistical Association, 46, 253, 68-78. 10.2307/2280095
Mathias, J.-D., J. M. Anderies & M. A. Janssen (2017) On our rapidly shrinking capacity to comply with the planetary boundaries on climate change. Nature Scientific Reports, 7, 9. DOI: 10.1038/srep42061
Matyas, D. & M. Pelling (2015) Positioning resilience for 2015: the role of resistance, incremental adjustment and transformation in disaster risk management policy. Disasters, 39, s1, s1-s18. 10.1111/disa.12107
Maunder, M. N. 2008. Maximum Sustainable Yield In Encyclopedia of Ecology, ed. B. D. Fath, 2292-2296. Oxford: Academic Press.
May, R. M. (1977) Thresholds and breakpoints in ecosystems with a multiplicity of stable states. Nature, 269, 5628, 471-477
McCartney, M., G. Forkuor, A. Sood, B. Amisigo, F. Hattermann & L. Muthuwatta. 2012. The water resource implications of changing climate in the Volta River Basin. In IWMI Research Report 146, 46. Colombo, Sri Lanka: IWMI.
McIntyre, P. B., C. A. Reidy Liermann & C. Revenga (2016) Linking freshwater fishery management to global food security and biodiversity conservation. Proceedings of the National Academy of Sciences, 113, 45, 12880-12885
McKinney, M. L., R. M. Schoch & L. Yonavjak. 2007. Environmental Science: Systems and Solutions. Massachusetts, USA: Jones and Bartlett Publishers, Inc.
MEA. 2005. Millennium Ecosystem Assessment Synthesis Report. 155. Washington DC, USA: World Resources Institute.
Meadows, D. H. 1999. Leverage Points: Places to Intervene in a System. 15. Leverage Points: Places to Intervene in a System.
Meadows, D. H. 2009. Thinking in Systems: A Primer. Abingdon, United Kingdom: Routledge.
Meadows, D. H., D. L. Meadows, J. Randers & W. W. Behrens III. 1972. The Limits to Growth: a Report for the Club of Rome's Project on the Predicament of Mankind. New York, USA: Universe Books.
Meehl , G. A., T. F. Stocker, W. D. Collins, P. Friedlingstein, A. T. Gaye, J. M. Gregory, A. Kitoh, R. Knutti, J. M. Murphy, A. Noda, S. C. B. Raper, I. G. Watterson, A. J. Weaver & Z.-C. Zhao. 2007. Global Climate Projections. In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, eds. S. Solomon, D. Quin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor & H. L. Miller, 100. Cambridge, U.K.
Menard, S. 2007. A brief introduction to time series analysis. Burlington, MA, USA: Academic Press.
358
Mendoza, G. A. & R. Prabhu (2006) Participatory modeling and analysis for sustainable forest management: overview of soft system dynamics models and applications. Forest Policy and Economics, 9, 2, 179-196. http://dx.doi.org/10.1016/j.forpol.2005.06.006
Meyerson, L. A., K. Saltonstall & R. M. Chambers. 2009. Phragmites australis, in eastern North America: a historical and ecological perspective. In Salt Marshes: Global Siege, eds. B. R. Silliman, E. Grosholz & M. D. Bertness, 57- 82. Berkeley, California, U.S.A: University of California Press.
Mialhe, F., N. Becu & Y. Gunnell (2012) An agent-based model for analyzing land use dynamics in response to farmer behaviour and environmental change in the Pampanga delta (Philippines). Agriculture, Ecosystems & Environment, 161, 0, 55-69. http://dx.doi.org/10.1016/j.agee.2012.07.016
Mileti, D. S. 1999. Disasters by Design: a Reassessment of Natural Hazards in the United States. Washington DC, USA: Joseph Henry Press.
Milly, P. C. D., J. Betancourt, M. Falkenmark, R. M. Hirsch, Z. W. Kundzewicz, D. P. Lettenmaier & R. J. Stouffer (2008) Stationarity Is Dead: Whither Water Management? Science, 319, 5863, 573-574. 10.1126/science.1151915
Minoiu, C. & J. A. Reyes (2013) A network analysis of global banking: 1978–2010. Journal of Financial Stability, 9, 2, 168-184. https://doi.org/10.1016/j.jfs.2013.03.001
Mishra, S. P. & J. Jena (2014) Migration of tidal inlets of Chilika Lagoon, Odisha, India- a critical case study. International Journal of Engineering and Technology, 6, 5, 2453-2464
Mishra, S. P. & J. Jena (2015) Intervention of Naraj Barrage: Its effects on sediment inflows into Chilika Lake, India. Lakes & Reservoirs: Research & Management, 20, 2, 77-84. 10.1111/lre.12088
Mishra, S. R. & A. L. Griffin (2010) Encroachment: A threat to resource sustainability in Chilika Lake, India. Applied Geography, 30, 3, 448-459. http://dx.doi.org/10.1016/j.apgeog.2009.12.001
Mitchell, B. (1971) A Comparison of Chi-Square and Kolmogorov-Smirnov Tests. Area, 3, 4, 237-241
Mitchell, M. & M. Newman. 2002. Complex systems theory and evolution. In Encyclopedia of Evolution, ed. M. Pagel, 1-5. New York, USA: Oxford University Press.
Mitra, K., R. Bharadwaj, S. Kaur & A. Khan. 2012. State Level Vulnerability Assessment and Adaptation Strategies- A Case Study of Orissa. In Indo-UK Collaborative Research Programme on the Impacts of Climate Change on India, 137. New Delhi, India.
Mohanty, B., P. R. Muduli, A. T. Behera, D. Mahapatro, S. K. Barik, S. K. Nag, R. N. Samal & A. K. Pattnaik (2016) Assessment of petroleum hydrocarbon in a tropical brackish water lagoon: Chilika, India. Chemistry and Ecology, 32, 7, 653-668. 10.1080/02757540.2016.1177521
Mohanty, P. K. & S. N. Otta (2008) Population of Irrawaddy Dolphin in Chilika Lagoon. Odisha Review, June-2008, 21-26
359 Appendices
Mohanty, P. K., S. R. Pal & P. K. Mishra (2001) Monitoring Ecological Conditions of a Coastal Lagoon using IRS Data: A Case Study in Chilka, East Coast of India. Journal of Coastal Research, 459-469. 10.2307/25736312
Mohanty, S. K., K. S. Bhatta, R. K. Mohanty, S. Mishra, A. Mohapatra & A. K. Pattnaik. 2008. Eco-restoration impact on fishery biodiversity and population structure in Chilika Lake. In Monitoring and Modelling Lakes and Coastal Environments, ed. P. K. Mohanty, 1-32. Heidelberg, Germany: Springer.
Mohanty, S. K., S. S. Mishra, M. Khan, R. K. Mohanty, A. Mohapatra & A. K. Pattnaik (2015) Ichthyofaunal diversity of Chilika Lake, Odisha, India: an inventory, assessment of biodiversity status and comprehensive systematic checklist (1916-2014). Check List, 11, 6. 10.15560/11.6.1817
Mohapatra, A., R. Mohanty, S. K. Mohanty, K. S. Bhatta & N. R. Das (2007a) Fisheries enhancement and biodiversity assessment of fish, prawn and mud crab in Chilika lagoon through hydrological intervention. Wetlands Ecology and Management, 15, 3, 229-251. 10.1007/s11273-006-9025-3
Mohapatra, A., R. K. Mohanty, S. K. Mohanty & N. R. Das (2007b) Fisheries enhancement and biodiversity assessment of fish, prawn and mud crab in Chilika lagoon through hydrological intervention. Wetlands Ecology and Management, 15, 3, 229-251. 10.1007/s11273-006-9025-3
Mohapatra, G. N., U. S. Panda & P. K. Mohanty (2007c) Annual cycle of surface meteorological and solar energy parameters of Orissa. Indian Journal of Radio & Space Physics, 36, 128-144
Mohapatra, M., U. C. Mohanty & S. Behera (2003) Spatial variability of daily rainfall over Orissa, India, during the southwest summer monsoon season. International Journal of Climatology, 23, 15, 1867-1887. 10.1002/joc.974
Morecroft, J. 2007. Strategic Modelling and Business Dynamics: A Feedback Systems Approach. Chichester, U.K: John Wiley and Sons, Ltd.
Motesharrei, S., J. Rivas & E. Kalnay (2014) Human and nature dynamics (HANDY): Modeling inequality and use of resources in the collapse or sustainability of societies. Ecological Economics, 101, 0, 90-102. http://dx.doi.org/10.1016/j.ecolecon.2014.02.014
Moxnes, E. (2000) Not only the tragedy of the commons: misperceptions of feedback and policies for sustainable development. System Dynamics Review, 16, 4, 325-348. 10.1002/sdr.201
Mulligan, M. 2013. Models Supporting Decision-Making and Policy Evaluation. In Environmental Modelling, 333-348. John Wiley & Sons, Ltd.
Nagendra, H. & E. Ostrom (2014) Applying the social-ecological system framework to the diagnosis of urban lake commons in Bangalore, India. Ecology and Society, 19, 2. 10.5751/ES-06582-190267
Nairn, R. B., MacIntosh, K. J., Hayes, M. O., Nai, G., Anthonio, S. L. and Valley, W. S. . 1998. Coastal erosion at Keta Lagoon, Ghana- large scale solution to a large scale problem. In Coastal Engineering (1998), 3192-3205.
Nayak, A. K., S. K. Mukhi, B. Kumar Behera, C. K. Misra, B. R. Samantaray & M. P. Singh Kohli (2001) A survey of various gears, crafts and fishing methods
360
adopted by the fisherman of Chilika Lake. Applied Fisheries and Aquaculture, 1, 1, 91-93
Nayak, B. K., B. C. Acharya, U. C. Panda, B. B. Nayak & S. K. Acharya (2004) Variation of water quality in Chilika lake, Orissa. Indian Journal of Marine Sciences, 33, 2, 164-169
Nayak, P., D. Armitage & M. Andrachuk (2015) Power and politics of social– ecological regime shifts in the Chilika lagoon, India and Tam Giang lagoon, Vietnam. Regional Environmental Change, 1-15. 10.1007/s10113-015- 0775-4
Nayak, P. K. 2011. Change and Marginalisation: Livelihoods, Commons Institutions and Environmental Justice in Chilika Lagoon, India. In Faculty of Graduate Studies 359. University of Manitoba, Winnipeg, Manitoba: The University of Manitoba.
Nayak, P. K. (2014) The Chilika Lagoon Social-Ecological System: An Historical Analysis. Ecology and Society, 19, 1. 10.5751/ES-05978-190101
Nayak, P. K. & F. Berkes (2010) Whose marginalisation? Politics around environmental injustices in India's Chilika lagoon. Local Environment, 15, 6, 553-567. 10.1080/13549839.2010.487527
Nayak, P. K. & F. Berkes (2011) Commonisation and decommonisation: Understanding the processes of change in the Chilika Lagoon, India. Conservation & Society, 9, 2, 132-145
Nayak, P. K. & F. Berkes (2014) Linking global drivers with local and regional change: a social-ecological system approach in Chilika Lagoon, Bay of Bengal. Regional Environmental Change, 14, 6, 2067-2078. 10.1007/s10113-012-0369-3
Negm, A. M. & H. Hossen. 2017. Sustainability of Water Bodies of Northern Egyptian Lakes: Case Studies, Burrulus and Manzalla Lakes. In The Nile Delta, ed. A. M. Negm, 455-467. Cham: Springer International Publishing.
Neuwirth, C., A. Peck & S. P. Simonović (2015) Modeling structural change in spatial system dynamics: A Daisyworld example. Environmental Modelling & Software, 65, 30-40. http://dx.doi.org/10.1016/j.envsoft.2014.11.026
Nicholls, R. J., P. P. Wong, V. R. Burkett, J. O. Codignotto, J. E. Hay, R. F. McLean, S. Ragoonaden & C. D. Woodroffer. 2007. Coastal systems and low-lying areas. In Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, eds. M. L. Parry, O. F. Canziani, J. P. Palutikof, P. J. ven der Linden & C. E. Hanson, 315-356. Cambridge, U.K: IPCC.
Noack, F., M.-C. Riekhof & M. F. Quaas. 2013. Redistribution and Development in Resource Abundant Economies: Theory and Evidence from Fishing Communities in India. 51. Kiel, Olsausenstrasse, Germany: Department of Economics, Christian Albrechts University zu Kiel.
Noack, F., M.-C. Riekhof & M. F. Quaas. 2015. Use Rights for Common Pool Resources and Economic Development. In Economic Working Paper 15/219, 55. Centre of Economic Research, ETH Zurich.
361 Appendices
Nordhaus, W. D. (1993) Rolling the ‘DICE’: an optimal transition path for controlling greenhouse gases. Resource and Energy Economics, 15, 1, 27- 50. https://doi.org/10.1016/0928-7655(93)90017-O
Nordhaus, W. D. & Z. Yang (1996) A Regional Dynamic General-Equilibrium Model of Alternative Climate-Change Strategies. The American Economic Review, 86, 4, 741-765. 10.2307/2118303
Nuno, A., N. Bunnefeld & E. J. Milner-Gulland (2014) Managing social-ecological systems under uncertainty: implementation in the real world. Ecology and Society, 19, 2. 10.5751/ES-06490-190252
Nyström, M., A. V. Norström, T. Blenckner, M. de la Torre-Castro, J. S. Eklöf, C. Folke, H. Österblom, R. S. Steneck, M. Thyresson & M. Troell (2012) Confronting Feedbacks of Degraded Marine Ecosystems. Ecosystems, 15, 5, 695-710. 10.1007/s10021-012-9530-6
OECD. 2001. Sustainable Development: Critical Issues. Paris, France: Organisation for Economic Co-Operation and Development.
OECD. 2010. Globalisation in Fisheries and Aquaculture. OECD Publishing.
OED. 2015. "complex, adj.". Oxford University Press.
Olson, M. 1982. The Rise and Decline of Nations. New Haven, U.S.A: Yale University Press.
Olsson, P., C. Folke & T. Hahn (2004) Social-Ecological Transformation for Ecosystem Management: the Development of Adaptive Co-management of a Wetland Landscape in Southern Sweden. Ecology and Society, 9, 4, 26
Olsson, P., L. H. Gunderson, S. R. Carpenter, P. Ryan, L. Lebel, C. Folke & C. S. Holling (2006) Shooting the Rapids: Navigating Transitions to Adaptive Governance of Social-Ecological Systems. Ecology and Society, 11, 1,
Oreskes, N. 2003. The role of quantitative models in science. In Models in Ecosystem Science, eds. C. D. Canham, J. J. Cole & W. K. Lauenroth, 13-31. Princeton, New Jersey, U.S.A: Princeton University Press.
Osgood, N. (2009) Lightening the performance burden of individual-based models through dimensional analysis and scale modeling. System Dynamics Review, 25, 2, 101-134. 10.1002/sdr.417
Österblom, H., A. Merrie, M. Metian, W. J. Boonstra, T. Blenckner, J. R. Watson, R. R. Rykaczewski, Y. Ota, J. L. Sarmiento, V. Christensen, M. Schlüter, S. Birnbaum, B. G. Gustafsson, C. Humborg, C.-M. Mörth, B. Müller-Karulis, M. T. Tomczak, M. Troell & C. Folke (2013) Modeling Social—Ecological Scenarios in Marine Systems. BioScience, 63, 9, 735-744. 10.1093/bioscience/63.9.735
Ostrom, E. (2007) A diagnostic approach for going beyond panaceas. Proceedings of the National Academy of Sciences, 104, 39, 15181-15187. 10.1073/pnas.0702288104
Ostrom, E. (2009) A General Framework for Analyzing Sustainability of Social- Ecological Systems. Science, 325, 5939, 419-422. 10.1126/science.1172133
362
Overeem, I. & J. P. M. Syvitski. 2009. Dynamics and vulnerbility of delta systems. In Land-Ocean Interactions in the Coastal Zone (LOICZ) eds. I. Overeem & J. P. M. Syvitski, 56. Germany: Core Project of the International Geosphere- Biosphere Programme (IGBP) and the International Human Dimensions Programme on Global Environmental Change (IHDP).
Pal, S. R. & P. K. Mohanty (2002) Use of IRS-1B data for change detection in water quality and vegetation of Chilka lagoon, east coast of India. International Journal of Remote Sensing, 23, 6, 1027-1042. 10.1080/01431160110076243
Palmer, M. A., C. A. Reidy Liermann, C. Nilsson, M. Flörke, J. Alcamo, P. S. Lake & N. Bond (2008) Climate change and the world's river basins: anticipating management options. Frontiers in Ecology and the Environment, 6, 2, 81- 89. 10.1890/060148
Panda, D. K., A. Kumar, S. Ghosh & R. K. Mohanty (2013) Streamflow trends in the Mahanadi River basin (India): Linkages to tropical climate variability. Journal of Hydrology, 495, 0, 135-149. http://dx.doi.org/10.1016/j.jhydrol.2013.04.054
Panigrahi, S., B. Acharya, R. Panigrahy, B. Nayak, K. Banarjee & S. Sarkar (2007) Anthropogenic impact on water quality of Chilika lagoon RAMSAR site: a statistical approach. Wetlands Ecology and Management, 15, 2, 113-126. 10.1007/s11273-006-9017-3
Paredes, C., C. Godoy & R. Castedo (2015) Soft-cliff retreat, self-organizaed critical phenomena in the limit of predictability? Fractals, 23, 01, 25. doi:10.1142/S0218348X15400095
Parida, S., K. S. Bhatta & B. C. Guru (2013) Study of sectoral variation of some commercially important fish groups in relation to environmental variables of Chilika Lake, east coast of India. International Journal of Bioassays, 2, 8, 1118-1123
Parsons, M. & M. C. Thoms (2017) From academic to applied: Operationalising resilience in river systems. Geomorphology. https://doi.org/10.1016/j.geomorph.2017.08.040
Pattanaik, S. (2007) Conservation of environment and protection of marginalized fishing communities of Lake Chilika in Orissa, India. Journal of Human Ecology, 22, 4, 291-302
Pattnaik, A. K. & S. Kobayashi. 2009. Data book of socioeconomic survey of fishers in Chilika Lagoon. Bhubaneswar, India: Japan International Cooperation Agency and Chilika Development Authority.
Pattnaik, A. K. & R. Kumar. 2013. Chilika Newsletter. 32. Chilika Development Authority, Wetlands International.
Paul, J. F. & M. E. McDonald (2005) Development of empirical, geographically specific water quality criteria: a conditional probability analysis approach. Journal of the American Water Resources Association, 41, 5, 1211-1223. 10.1111/j.1752-1688.2005.tb03795.x
Pauly, D., V. Christensen, J. Dalsgaard, R. Froese & F. Torres (1998) Fishing Down Marine Food Webs. Science, 279, 5352, 860-863
363 Appendices
Pauly, D., R. Hilborn & T. A. Branch (2013) Fisheries: Does catch reflect abundance? Nature, 494, 7437, 303-306. 10.1038/494303a
Pawson, R., G. Wong & L. Owen (2011) Known knowns, known unknowns, unknown unknowns. American Journal of Evaluation, 32, 4, 518-546
PCD. 2011. Economic Survey 2010-11: Orissa. ed. D. o. E. S. Planning & Coordination Department, Government of Orissa, 395. Bhubaneswar: Planning & Coordination Department, Government of Orissa.
PCD. 2014. Odisha Economic Survey. 402. Bhubaneswar, Odisha, India: Planning & Coordination Department, Directorate of Economics & Statistics, Government of Orissa.
Peetabas, N. & R. P. Panda (2015) Conservation and management of bioresources of Chilika Lake, Odisha, India. International Journal of Scientific and Research Publications, 5, 7, 4
Pelling, M., C. High, J. A. Dearing & D. Smith (2007) Shadow spaces for social learning: a relational understanding of adaptive capacity to climate change within organisations. Environment & Planning, 40, 4, 867-884
Pelling, M. & D. Manuel-Navarrete (2011) From Resilience to Transformation: the Adaptive Cycle in Two Mexican Urban Centers. Ecology and Society, 16, 2,
Pérez-Ruzafa, A. & C. Marcos (2012) Fisheries in coastal lagoons: An assumed but poorly researched aspect of the ecology and functioning of coastal lagoons. Estuarine, Coastal and Shelf Science, 110, 0, 15-31. http://dx.doi.org/10.1016/j.ecss.2012.05.025
Perretti, C. T. & S. B. Munch (2012) Regime shift indicators fail under noise levels commonly observed in ecological systems. Ecological Applications, 22, 6, 1772-1779. 10.1890/11-0161.1
Pervez, M. S. & G. M. Henebry (2014) Projections of the Ganges–Brahmaputra precipitation—Downscaled from GCM predictors. Journal of Hydrology, 517, 0, 120-134. http://dx.doi.org/10.1016/j.jhydrol.2014.05.016
Perz, S. G., R. Muñoz-Carpena, G. Kiker & R. D. Holt (2013) Evaluating ecological resilience with global sensitivity and uncertainty analysis. Ecological Modelling, 263, 174-186. http://dx.doi.org/10.1016/j.ecolmodel.2013.04.024
Peters, D. P. C., R. A. Pielke, B. T. Bestelmeyer, C. D. Allen, S. Munson-McGee & K. M. Havstad (2004) Cross-scale interactions, nonlinearities, and forecasting catastrophic events. Proceedings of the National Academy of Sciences of the United States of America, 101, 42, 15130-15135. 10.1073/pnas.0403822101
Peterson, G. D. (2002) Estimating Resilience Across Landscapes. Ecology and Society, 6, 1. 10.5751/ES-00375-060117
Petr, T. 2000. Interactions between fish and aquatic macrophytes in inland waters. In FAO Fisheries Technical Paper 396, ed. FAO. Rome, Italy: Food and Agriculture Organization of the United Nations.
Petschel-Held, G., H.-J. Schellnhuber, T. Bruckner, F. L. Tóth & K. Hasselmann (1999) The Tolerable Windows Approach: Theoretical and Methodological
364
Foundations. Climatic Change, 41, 3, 303-331. 10.1023/A:1005487123751
Phillips, W. 1976. Monte Carlo tests of conclusion robustness. In Proceedings of the International System Dynamics Conference, 19. Geilo, Norway.
Pianosi, F., K. Beven, J. Freer, J. W. Hall, J. Rougier, D. B. Stephenson & T. Wagener (2016) Sensitivity analysis of environmental models: A systematic review with practical workflow. Environmental Modelling & Software, 79, 214-232. http://dx.doi.org/10.1016/j.envsoft.2016.02.008
Pindyck, R. S. (2013) Climate Change Policy: What Do the Models Tell Us? Journal of Economic Literature, 51, 3, 860-72
Pindyck, R. S. 2015. The use and misuse of models for climate policy. In Review of Environmental Policy and Economics Symposium, 17.
Plagányi, É. 2007. Models for an ecosystem approach to fisheries. In FAO Fisheries Technical Paper 477, ed. FAO, 126. Rome, Italy: Food and Agricultural Organization of the United Nations.
Pollnac, R. B., R. S. Pomeroy & I. H. T. Harkes (2001) Fishery policy and job satisfaction in three southeast Asian fisheries. Ocean & Coastal Management, 44, 7–8, 531-544. https://doi.org/10.1016/S0964- 5691(01)00064-3
Pretty, J. N. (1995) Participatory learning for sustainable agriculture. World Development, 23, 8, 1247-1263. http://dx.doi.org/10.1016/0305- 750X(95)00046-F
Primavera, J. H. (1997) Socio-economic impacts of shrimp culture. Aquaculture Research, 28, 815-827
Punt, A. E. & A. D. M. Smith. 2001. The gospel of maximum sustainable yield in fisheries management: birth, crucifixion and reincarnation. In Conservation of Exploited Species, eds. J. D. Reynolds, G. M. Mace, K. H. Redford & J. G. Robinson, 41-66. Cambridge, U.K: Cambridge University Press.
Purnomo, H. & G. A. Mendoza (2011) A system dynamics model for evaluating collaborative forest management: a case study in Indonesia. The International Journal of Sustainable Development & World Ecology, 18, 2, 164
Quinn, T. J. & J. S. Collie (2005) Sustainability in single-species population models. Philosophical Transactions of the Royal Society B: Biological Sciences, 360, 1453, 147-162. 10.1098/rstb.2004.1577
R. 2008. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria: R Development Core Team.
Rajawat, A. S., M. Gupta, B. C. Acharya & S. Nayak (2007) Impact of new mouth opening on morphology and water quality of the Chilika Lagoon – a study based on Resourcesat‐1 LISS‐III and AWiFS and IRS‐1D LISS‐III data. International Journal of Remote Sensing, 28, 5, 905-923. 10.1080/01431160600904949
Rao, R. J. n.d. Environmental Impact of Dredging of Outer Channel and Creation of a New Mouth Opening to Chilika Lake with the Bay of Bengal, Odisha, India. 8. online.
365 Appendices
Raven, P. H., L. R. Berg & D. M. Hassenzahl. 2012. Environment. New Jersey, U.S.A: John Wiley & Sons.
Rawls, J. 1999. A Theory of Justice. Cambridge, Massachusetts, U.S.A: Harvard University Press.
Raworth, K. 2012. A safe and just space for humanity: can we live within the doughnut. In Oxfam Policy and Practice: Climate Change and Resilience, 1- 26.
Ray, S. & R. Garada (2017) Boat automation and fishery livelihood: a case of Chilika Lake in Odisha. Environment, Development and Sustainability. 10.1007/s10668-017-9995-8
Redclift, M. (2005) Sustainable development (1987–2005): an oxymoron comes of age. Sustainable Development, 13, 4, 212-227. 10.1002/sd.281
Refsgaard, J. C., J. P. van der Sluijs, A. L. Højberg & P. A. Vanrolleghem (2007) Uncertainty in the environmental modelling process – A framework and guidance. Environmental Modelling & Software, 22, 11, 1543-1556. http://dx.doi.org/10.1016/j.envsoft.2007.02.004
Regan, H. M., M. Colyvan & M. A. Burgman (2002) A taxonomy and treatment of uncertainty for ecology and conservation biology. Ecological Applications, 12, 2, 618-628
Renaud, F., J. Birkmann, M. Damm & G. Gallopín (2010) Understanding multiple thresholds of coupled social–ecological systems exposed to natural hazards as external shocks. Natural Hazards, 55, 3, 749-763. 10.1007/s11069-010-9505-x
Renaud, F., J. Friedrich, Z. Sebesvari & L. Giosan (2014) Tipping points for delta social-ecological systems. LOICZ Inprint 1, 5-13
Renaud, F. G., T. T. H. Le, C. Lindener, V. T. Guong & Z. Sebesvari (2015) Resilience and shifts in agro-ecosystems facing increasing sea-level rise and salinity intrusion in Ben Tre Province, Mekong Delta. Climatic Change, 133, 1, 69-84. 10.1007/s10584-014-1113-4
Renaud, F. G., J. P. M. Syvitski, Z. Sebesvari, S. E. Werners, H. Kremer, C. Kuenzer, R. Ramesh, A. Jeuken & J. Friedrich (2013) Tipping from the Holocene to the Anthropocene: How threatened are major world deltas? Current Opinion in Environmental Sustainability, 5, 6, 644-654. http://dx.doi.org/10.1016/j.cosust.2013.11.007
Renaud, F. G., S. Szabo & Z. Matthews (2016) Sustainable deltas: livelihoods, ecosystem services, and policy implications. Sustainability Science, 11, 4, 519-523. 10.1007/s11625-016-0380-6
Ricker, W. E. 1975. Computation and interpretation of biological statistics of fish populations. In Bulletin 191, 382. Department of Environment, Fisheries and Marine Service, Ottawa.
Riley, S., W. Schouten & S. Cahill (2003) Exploring the Dynamics of Subjectivity and Power Between Researcher and Researched. 2003, 4, 2,
Robert, C. & G. Casella. 2011. A short history of MCMC: subjective recollections from incomplete data. In Handbook of Markov Chain Monte Carlo, eds. S.
366
Brooks, A. Gelman, G. Jones & X. Meng, 49-66. Boca Raton, Floria, USA: Chapman & Hall.
Robinson, J., S. Burch, S. Talwar, M. O'Shea & M. Walsh (2011) Envisioning sustainability: Recent progress in the use of participatory backcasting approaches for sustainability research. Technological Forecasting and Social Change, 78, 5, 756-768. https://doi.org/10.1016/j.techfore.2010.12.006
Robinson, J. B. (1990) Futures under glass: A recipe for people who hate to predict. Futures, 22, 8, 820-842. https://doi.org/10.1016/0016- 3287(90)90018-D
Robson, J. P. & P. K. Nayak (2010) Rural out-migration and resource-dependent communities in Mexico and India. Population and Environment, 32, 2-3, 263-284. 10.1007/s11111-010-0121-1
Rocha, J. C., G. D. Peterson & R. Biggs (2015) Regime Shifts in the Anthropocene: Drivers, Risks, and Resilience. PLoS ONE, 10, 8, e0134639. 10.1371/journal.pone.0134639
Rockström, J., W. Steffen, K. Noone, Å. Persson, F. S. Chapin, III, E. Lambin, T. M. Lenton, M. Scheffer, C. Folke, H. J. Schellnhuber, B. Nykvist, C. A. de Wit, T. Hughes, S. van der Leeuw, H. Rodhe, S. Sörlin, P. K. Snyder, R. Costanza, U. Svedin, M. Falkenmark, L. Karlberg, R. W. Corell, V. J. Fabry, J. Hansen, B. Walker, D. Liverman, K. Richardson, P. Crutzen & J. Foley (2009) Planetary Boundaries: Exploring the Safe Operating Space for Humanity. Ecology and Society, 14, 2, 33. 10.5751/ES-03180-140232
Rodionov, S. & J. E. Overland (2005) Application of a sequential regime shift detection method to the Bering Sea ecosystem. ICES Journal of Marine Science: Journal du Conseil, 62, 3, 328-332. 10.1016/j.icesjms.2005.01.013
Rodionov, S. N. (2004) A sequential algorithm for testing climate regime shifts. Geophysical Research Letters, 31, 9, n/a-n/a. 10.1029/2004GL019448
Roe, E. & M. van Eeten (2001) Threshold-based resource management: a framework for comprehensive ecosystem management. Environmental Management, 27, 2, 195-214. 10.1007/s002670010143
Rose, K. A., J. H. Cowan, K. O. Winemiller, R. A. Myers & R. Hilborn (2001) Compensatory density dependence in fish populations: importance, controversy, understanding and prognosis. Fish and Fisheries, 2, 4, 293- 327. 10.1046/j.1467-2960.2001.00056.x
Rosenhead, J. 2013. Robustness Analysis. In Encyclopedia of Operations Research and Management Science, eds. S. I. Gass & M. C. Fu, 1346-1347. Boston, MA: Springer US.
Roughgarden, J. & F. Smith (1996) Why fisheries collapse and what to do about it. Proceedings of the National Academy of Sciences, 93, 10, 5078-5083
Rudolph, J. W. & N. P. Repenning (2002) Disaster Dynamics: Understanding the Role of Quantity in Organizational Collapse. Administrative Science Quarterly, 47, 1, 1-30. 10.2307/3094889
Saeed, K. 1998. Defining a problem or constructing a reference mode. In International System Dynamics Conference, 29. Quebec City, Canada.
367 Appendices
Sahu, B., P. Pati & R. C. Panigrahy (2014) Environmental conditions of Chilika Lake during pre and post hydrological intervention: an overview. Journal of Coastal Conservation, 18, 3, 285-297. 10.1007/s11852-014-0318-z
Sala, S., F. Farioli & A. Zamagni (2013) Progress in sustainability science: lessons learnt from current methodologies for sustainability assessment: Part 1. The International Journal of Life Cycle Assessment, 18, 9, 1653-1672. 10.1007/s11367-012-0508-6
Salagrama, V. 2006. Trends in Poverty and Livelihoods in Coastal Fishing Communities of Orissa State, India. Rome, Italy: Food and Agriculture Organization of the United Nations.
Saltelli, A. (1999) Sensitivity analysis: Could better methods be used? Journal of Geophysical Research: Atmospheres, 104, D3, 3789-3793. 10.1029/1998JD100042
Saltelli, A. 2000a. Fortune and future of sensitivity analysis. In Sensitivity Analysis, eds. A. Saltelli, K. Chan & E. M. Scott, 421-426. Chichester, U.K: John Wiley & Sons Ltd.
Saltelli, A. 2000b. What is sensitivity analysis. In Sensitivity Analysis, eds. A. Saltelli, K. Chan & E. M. Scott, 3-14. Chichester, U.K: John Wiley & Sons Ltd.
Samal, K. C. & S. Meher (2003) Fishing Communities on Chilika Lake: Comparative Socio-Economic Study. Economic and Political Weekly, 38, 31, 3319-3325
Santra, P. & B. S. Das (2013) Modeling runoff from an agricultural watershed of western catchment of Chilika lake through ArcSWAT. Journal of Hydro- environment Research, 7, 4, 261-269. http://dx.doi.org/10.1016/j.jher.2013.04.005
Saraswathy, R., M. Muralidhar, J. K. Sundaray, N. Lalitha & P. Kumararaja. 2015. Water quality managment in fish hatchery and grow-out systems. In Advances in Marine and Brackishwater Aquaculture, eds. S. Perumal, A. R. Thirunavukkarasu & P. Pachiappan, 217-226. New Delhi, India: Springer.
Sauriau, P.-G., J.-P. Robin & J. Marchand. 1994. Effects of the excessive organic enrichment of the Loire Estuary on the downstream migratory patterns of the amphihaline greg mullet Liza ramada (Pisces: Mugilidae). In Changes in Fluxes in Estuaries: Implications from Science to Management, eds. K. R. Dyer & R. J. Orth, 349-356. Fredensborg, Denmark: Olsen & Olsen.
Scheffer, M. 2009. Critical Transitions in Nature and Society. New Jersey, USA: Princeton University Press.
Scheffer, M., S. Barrett, S. R. Carpenter, C. Folke, A. J. Green, M. Holmgren, T. P. Hughes, S. Kosten, I. A. van de Leemput, D. C. Nepstad, E. H. van Nes, E. T. H. M. Peeters & B. Walker (2015) Creating a safe operating space for iconic ecosystems. Science, 347, 6228, 1317-1319. 10.1126/science.aaa3769
Scheffer, M., J. Bascompte, W. A. Brock, V. Brovkin, S. R. Carpenter, V. Dakos, H. Held, E. H. van Nes, M. Rietkerk & G. Sugihara (2009) Early-warning signals for critical transitions. Nature, 461, 7260, 53-59. 10.1038/nature08227
Scheffer, M., S. Carpenter, J. A. Foley, C. Folke & B. Walker (2001) Catastrophic shifts in ecosystems. Nature, 413, 6856, 591-596
368
Scheffer, M., S. R. Carpenter, T. M. Lenton, J. Bascompte, W. Brock, V. Dakos, J. van de Koppel, I. A. van de Leemput, S. A. Levin, E. H. van Nes, M. Pascual & J. Vandermeer (2012) Anticipating Critical Transitions. Science, 338, 6105, 344-348. 10.1126/science.1225244
Scheffer, M., F. Westley & W. Brock (2003) Slow Response of Societies to New Problems: Causes and Costs. Ecosystems, 6, 5, 493-502. 10.1007/PL00021504
Schindler, D. E. & R. Hilborn (2015) Prediction, precaution, and policy under global change. Science, 347, 6225, 953
Schlüter, M., R. Biggs, M. L. Schoon, M. D. Robards & J. M. Anderies. 2015. Reflections on building resilience- interactions among principles and implications for governance. In Principles for Building Resilience, eds. R. Biggs, M. Schlüter & M. L. Schoon, 251-282. Cambridge, U.K: Cambridge University Press.
Schlüter, M., H. Leslie & S. Levin (2009) Managing water-use trade-offs in a semi- arid river delta to sustain multiple ecosystem services: a modeling approach. Ecological Research, 24, 3, 491-503. 10.1007/s11284-008- 0576-z
Schlüter, M., R. R. J. McAllister, R. Arlinghaus, N. Bunnefeld, K. Eisenack, F. Hölker, E. J. Milner-Gulland, B. Müller, E. Nicholson, M. Quaas & M. Stövn (2012) New horizons for managing the environment: a review of coupled social- ecological systems modeling Natural Resource Modeling, 25, 1, 219-272. 10.1111/j.1939-7445.2011.00108.x
Schmid, M., S. Hunziker & A. Wüest (2014) Lake surface temperatures in a changing climate: a global sensitivity analysis. Climatic Change, 124, 1, 301-315. 10.1007/s10584-014-1087-2
Schulz, M., H.-P. Kozerski, T. Pluntke & K. Rinke (2003) The influence of macrophytes on sedimentation and nutrient retention in the lower River Spree (Germany). Water Research, 37, 3, 569-578. http://dx.doi.org/10.1016/S0043-1354(02)00276-2
Scoones, I. 1998. Sustainable rural livelihoods: a framework for analysis. In IDS Working Paper 72, 22. Institute of Development Studies, University of Sussex, Brighton.
Scott, G. & R. Garner. 2013. Doing Qualitative Research: Designs, Methods and Techniques. New Jersey, USA: Pearson Education Inc.
Sebesvari, Z., F. G. Renaud, S. Haas, Z. Tessler, M. Hagenlocher, J. Kloos, S. Szabo, A. Tejedor & C. Kuenzer (2016) A review of vulnerability indicators for deltaic social–ecological systems. Sustainability Science, 11, 4, 575-590. 10.1007/s11625-016-0366-4
Seijo, J. C., O. Defeo & S. Salas. 1998. Fisheries Bioeconomics: Theory, Modelling and Management. Rome, Italy: Food and Agriculture Organization of the United Nations.
Sekhar, N. U. (2004) Fisheries in Chilika lake: how community access and control impacts their management. Journal of Environmental Management, 73, 3, 257-266. http://dx.doi.org/10.1016/j.jenvman.2004.07.006
369 Appendices
Selkoe, K. A., T. Blenckner, M. R. Caldwell, L. B. Crowder, A. L. Erickson, T. E. Essington, J. A. Estes, R. M. Fujita, B. S. Halpern, M. E. Hunsicker, C. V. Kappel, R. P. Kelly, J. N. Kittinger, P. S. Levin, J. M. Lynham, M. E. Mach, R. G. Martone, L. A. Mease, A. K. Salomon, J. F. Samhouri, C. Scarborough, A. C. Stier, C. White & J. Zedler (2015) Principles for managing marine ecosystems prone to tipping points. Ecosystem Health and Sustainability, 1, 5, 1-18. 10.1890/EHS14-0024.1
Senge, P. M. 1990. The fifth discipline: the art and practice of the learning organization. New York, USA: Doubleday/Currency.
Seybold, H., J. S. Andrade & H. J. Herrmann (2007) Modeling river delta formation. Proceedings of the National Academy of Sciences, 104, 43, 16804-16809. 10.1073/pnas.0705265104
Sharp, J. A. 1976. Sensitivity analysis methods for system dynamics models. In Proceedings of the International System Dynamics Conference, 761-804. Geilo, Norway.
Shaw, B. P., N. P. Rout, B. C. Barman, S. B. Choudhury & K. H. Rao (2000) Distribution of macrophytic vegetation in relation to salinity in the Chilka lake, a lagoon along east coast of India. Indian Journal of Marine Sciences, 29, 2, 144-148
Sievanen, L., B. Crawford, R. Pollnac & C. Lowe (2005) Weeding through assumptions of livelihood approaches in ICM: Seaweed farming in the Philippines and Indonesia. Ocean & Coastal Management, 48, 3–6, 297-313. https://doi.org/10.1016/j.ocecoaman.2005.04.015
Simonovic, S. P. 2010. Systems Approach to Management of Disasters: Methods and Applications. Hoboken, New Jersey, U.S.A: John Wiley & Sons.
Sloboda, M. & D. Swayne. 2011. Autocalibration of environmental process modelis using a PAC learning hypothesis. In Environmental Software Systems: Frameworks of eEnvironment, eds. J. Hřebíček, G. Schimak & R. Denzer, 529-534. Brno, Czech Republic: 9th IFIP WG 5.11 International Symposium.
Smit, B. & J. Wandel (2006) Adaptation, adaptive capacity and vulnerability. Global Environmental Change, 16, 282-292
Sneddon, C., R. B. Howarth & R. B. Norgaard (2006) Sustainable development in a post-Brundtland world. Ecological Economics, 57, 253-268
Soulié, J.-C. & O. Thébaud (2006) Modeling fleet response in regulated fisheries: An agent-based approach. Mathematical and Computer Modelling, 44, 5, 553-564. http://dx.doi.org/10.1016/j.mcm.2005.02.011
Spear, R. C. & G. M. Hornberger (1980) Eutrophication in Peel Inlet-II. Identification of critical uncertainties via generalized sensitivity analysis. Water Research, 14, 1, 43-49. https://doi.org/10.1016/0043- 1354(80)90040-8
Srichandan, S., J. Y. Kim, P. Bhadury, S. K. Barik, P. R. Muduli, R. N. Samal, A. K. Pattnaik & G. Rastogi (2015) Spatiotemporal distribution and composition of phytoplankton assemblages in a coastal tropical lagoon: Chilika, India. Environmental Monitoring and Assessment, 187, 2, 17
370
Stanton, E., F. Ackerman & S. Kartha (2009) Inside the integrated assessment models: Four issues in climate economics. Climate and Development, 1, 2, 166-184. 10.3763/cdev.2009.0015
Steffen, W., W. Broadgate, L. Deutsch, O. Gaffney & C. Ludwig (2015a) The trajectory of the Anthropocene: The Great Acceleration. The Anthropocene Review, 2, 1, 81-98. 10.1177/2053019614564785
Steffen, W., J. Grinevald, P. Crutzen & J. McNeill (2011) The Anthropocene: conceptual and historical perspectives. Philosophical Transactions of the Royal Society A, 369, 842-867
Steffen, W., K. Richardson, J. Rockström, S. E. Cornell, I. Fetzer, E. M. Bennett, R. Biggs, S. R. Carpenter, W. de Vries, C. A. de Wit, C. Folke, D. Gerten, J. Heinke, G. M. Mace, L. M. Persson, V. Ramanathan, B. Reyers & S. Sörlin (2015b) Planetary boundaries: Guiding human development on a changing planet. Science, 347, 6223, 736-747. 10.1126/science.1259855
Steffen, W., R. A. Sanderson, P. D. Tyson, J. Jäger, P. A. Matson, B. Moore III, F. Oldfield, K. Richardson, H.-J. Schellnhuber, B. L. Turner & R. J. Wasson. 2004. Global Change and the Earth System: A Planet under Pressure.
Sterk, M., I. A. van de Leemput & E. T. H. M. Peeters (2017) How to conceptualize and operationalize resilience in socio-ecological systems? Current Opinion in Environmental Sustainability, 28, 108-113. https://doi.org/10.1016/j.cosust.2017.09.003
Sterman, J. D. 1983. Appropriate summary statistics for evaluating the historical fit of system dynamics models. In Internatonal System Dynamics Conference, 203-260. Massachusetts, USA.
Sterman, J. D. 2000. Business Dynamics: Systems Thinking and Modeling for a Complex World. New York City, New York, U.S.A: McGraw-Hill Education.
Sterman, J. D. (2002) All models are wrong: reflections on becoming a systems scientist. System Dynamics Review, 18, 4, 501-531. 10.1002/sdr.261
Stern, D. I. (2004) The rise and fall of the environmental Kuznets curve. World Development, 32, 8, 1419-1439
Sterner, T., M. Troell, J. Vincent, S. Aniyar, S. Barrett, W. Brock, S. Carpenter, K. Chopra, P. Ehrlich, M. Hoel, S. Levin, K.-G. Mäler, J. Norberg, L. Pihl, T. Söderqvist, J. Wilen & A. Xepapadeas (2006) Quick Fixes for the Environment: Part of the Solution or Part of the Problem? Environment: Science and Policy for Sustainable Development, 48, 10, 20-27. 10.3200/ENVT.48.10.20-27
Stirling, A. 1846. Orissa: its Geography, Statistics, History, Religion and Antiquities. London: John Snow.
Sugiarto, H. S., N. N. Chung, C. H. Lai & L. Y. Chew (2015) Socioecological regime shifts in the setting of complex social interactions. Physica Review E, 91, 6, 11
Summers, D. M., B. A. Bryan, W. S. Meyer, G. Lyle, S. Wells, J. McLean, T. Moon, G. van Gaans & M. Siebentritt (2015) Simple models for managing complex social–ecological systems: The Landscape Futures Analysis Tool (LFAT). Environmental Modelling & Software, 63, 0, 217-229. http://dx.doi.org/10.1016/j.envsoft.2014.10.002
371 Appendices
Sundar, V., S. A. Sannasiraj, K. Murali, V. Sriram, U. S. Panda, M. M. Mahanty, V. R. Rao, S. Patra & P. Mishra (2015) Hydrodynamics and Water Quality in Chilika Lagoon-A Modelling Approach. Procedia Engineering, 116, 639-646. http://dx.doi.org/10.1016/j.proeng.2015.08.337
Sušnik, J., L. S. Vamvakeridou-Lyroudia, D. A. Savić & Z. Kapelan (2012) Integrated System Dynamics Modelling for water scarcity assessment: Case study of the Kairouan region. Science of The Total Environment, 440, 0, 290-306. http://dx.doi.org/10.1016/j.scitotenv.2012.05.085
Suweis, S. & P. D'Odorico (2014) Early Warning Signs in Social-Ecological Networks. PLOS ONE, 9, 7, e101851. 10.1371/journal.pone.0101851
Swart, R. J., P. Raskin & J. Robinson (2004) The problem of the future: sustainability science and scenario analysis. Global Environmental Change, 14, 2, 137-146. http://dx.doi.org/10.1016/j.gloenvcha.2003.10.002
Sweeny, K. & F. Griffiths. 2002. Complexity and Healthcare: An Introduction. Abingdon, UK: Radcliffe Publishing Ltd.
Swetnam, R. D., B. Fisher, B. P. Mbilinyi, P. K. T. Munishi, S. Willcock, T. Ricketts, S. Mwakalila, A. Balmford, N. D. Burgess, A. R. Marshall & S. L. Lewis (2011) Mapping socio-economic scenarios of land cover change: A GIS method to enable ecosystem service modelling. Journal of Environmental Management, 92, 3, 563-574. http://dx.doi.org/10.1016/j.jenvman.2010.09.007
Sylvester, J. R., C. E. Nash & C. R. Emberson (1975) Salinity and oxygen tolerances of eggs and larvae of Hawaiian striped mullet, Mugil cephalus L.*. Journal of Fish Biology, 7, 5, 621-629. 10.1111/j.1095-8649.1975.tb04635.x
Syvitski, J. P. M., A. J. Kettner, I. Overeem, E. W. H. Hutton, M. T. Hannon, G. R. Brakenridge, J. Day, C. Vorosmarty, Y. Saito, L. Giosan & R. J. Nicholls (2009) Sinking deltas due to human activities. Nature Geosci, 2, 10, 681- 686. http://www.nature.com/ngeo/journal/v2/n10/suppinfo/ngeo629_S1.html
Szabo, S., E. Brondizio, F. G. Renaud, S. Hetrick, R. J. Nicholls, Z. Matthews, Z. Tessler, A. Tejedor, Z. Sebesvari, E. Foufoula-Georgiou, S. da Costa & J. A. Dearing (2016) Population dynamics, delta vulnerability and environmental change: comparison of the Mekong, Ganges–Brahmaputra and Amazon delta regions. Sustainability Science, 11, 4, 539-554. 10.1007/s11625- 016-0372-6
Tainter, J. 1990. The Collapse of Complex Societies. Cambridge, UK: Cambridge University Press.
Tainter, J. A. & C. Crumley. 2005. Climate, complexity, and problem solving in the Roman Empire. In 96th Dahlem Workshop Integrated History and the Future of People on Earth (IHOPE). Freie Universitat, Berlin.
Tansley, A. G. (1935) The use and abuse of vegetational concepts and terms. Ecology, 16, 284-307
Tarantola, S., J. Jesinghaus & M. Paulamaa. 2000. Global sensitivity analysis: a quality assurance tool in environmnetal policy modelling. In Sensitivity Analysis, eds. A. Saltelli, K. Chan & E. M. Scott, 385-400. Chichester, U.K: John Wiley & Sons Ltd.
372
Thépot, V. & D. R. Jerry (2015) The effect of temperature on the embryonic development of barramundi, the Australian strain of Lates calcarifer (Bloch) using current hatchery practices. Aquaculture Reports, 2, 132-138. http://dx.doi.org/10.1016/j.aqrep.2015.09.002
Thrush, S. F., N. Lewis, R. Le Heron, K. T. Fisher, C. J. Lundquist & J. Hewitt (2016) Addressing surprise and uncertain futures in marine science, marine governance, and society. Ecology and Society, 21, 2. 10.5751/ES-08574- 210244
Timmerman, P. & R. White (1997) Megahydropolis: coastal cities in the context of global environmental change. Global Environmental Change, 7, 3, 205-234. http://dx.doi.org/10.1016/S0959-3780(97)00009-5
Tracey, S. R. & J. M. Lyle (2011) Linking scallop distribution and abundance with fisher behaviour: implication for management to avoid repeated stock collapse in a recreational fishery. Fisheries Management and Ecology, 18, 3, 221-232. 10.1111/j.1365-2400.2010.00775.x
Tremblay, M.-A. (1957) The Key Informant Technique: A Nonethnographic Application. American Anthropologist, 59, 4, 688-701. 10.1525/aa.1957.59.4.02a00100
Troell, M., L. Pihl, P. Rönnbäck, H. Wennhage, T. Söderqvist & N. Kautsky (2005) Regime Shifts and Ecosystem Service Generation in Swedish Coastal Soft Bottom Habitats: When Resilience is Undesirable. Ecology and Society, 10, 1,
Tucker Jr, J. W., D. J. Russel & M. A. Rimmer (2002) Barramundi culture: a success story for aquaculture in Asia and Australisa. World Aquaculture Magazine, 33, 3, 67-72
Turner II, B. L., R. E. Kasperson, P. A. Matson, J. J. McCarthy, R. W. Corell, L. Christensen, N. Eckley, J. X. Kasperson, A. Luers, M. L. Martello, C. Polsky, A. Pulsipher & A. Schiller (2003a) A framework for vulnerability analysis in sustainability science. Proceedings of the National Academy of Sciences, 100, 14, 8074-8079. 10.1073/pnas.1231335100
Turner II, B. L., P. A. Matson, J. J. McCarthy, R. W. Corell, L. Christensen, N. Eckley, G. K. Hovelsrud-Broda, J. X. Kasperson, R. E. Kasperson, A. Luers, M. L. Martello, S. Mathiesen, R. Naylor, C. Polsky, A. Pulsipher, A. Schiller, H. Selin & N. Tyler (2003b) Illustrating the coupled human–environment system for vulnerability analysis: Three case studies. Proceedings of the National Academy of Sciences, 100, 14, 8080-8085. 10.1073/pnas.1231334100
Underwood, F. M., G. Parkes & J. H. Swasey (2016) Building Bayesian Belief Networks to investigate how fishery performance responds to management interventions. Fisheries Research, 182, Supplement C, 28-38. https://doi.org/10.1016/j.fishres.2015.12.005
UNWCED. 1987. Our Common Future: Report of the World Commission on Environment and Development. 400. Switzerland: United Nations World Commission on Environment and Development.
Uusitalo, L. (2007) Advantages and challenges of Bayesian networks in environmental modelling. Ecological Modelling, 203, 3, 312-318. https://doi.org/10.1016/j.ecolmodel.2006.11.033
373 Appendices van Asselt, M. B. A. & J. Rotmans (2002) Uncertainty in Integrated Assessment Modelling. Climatic Change, 54, 1, 75-105. 10.1023/A:1015783803445 van Bilsen, A., Bekebrede, G. and Meyer, I. (2010) Understanding complex adaptive systems by playing games. Infromatics in Education, 9, 1, 1-18 van der Leeuw, S., R. Costanza, S. Aulenbach, S. Brewer, M. Burek, S. Cornell, C. Crumley, J. A. Dearing, C. Downy, L. J. Graumlich, S. Heckbert, M. Hegmon, K. Hibbard, S. T. Jackson, I. Kubiszewski, P. Sinclair, S. Sörlin & W. Steffen (2011) Toward an Integrated History to Guide the Future. Ecology and Society, 16, 4. 10.5751/ES-04341-160402 van Staveren, M. F. & J. P. M. van Tatenhove (2016) Hydraulic engineering in the social-ecological delta: understanding the interplay between social, ecological, and technological systems in the Dutch delta by means of delta trajectories. Ecology and Society, 21, 1. 10.5751/ES-08168-210108
Vasilakopoulos, P. & C. T. Marshall (2015) Resilience and tipping points of an exploited fish population over six decades. Global Change Biology, 21, 5, 1834-1847. 10.1111/gcb.12845
Venkatesan, A. K., S. Ahmad, W. Johnson & J. R. Batista (2011) Systems dynamic model to forecast salinity load to the Colorado River due to urbanization within the Las Vegas Valley. Science of The Total Environment, 409, 13, 2616-2625. http://dx.doi.org/10.1016/j.scitotenv.2011.03.018
Verbesselt, J., R. Hyndman, G. Newnham & D. Culvenor (2010) Detecting trend and seasonal changes in satellite image time series. Remote Sensing of Environment, 114, 1, 106-115. http://dx.doi.org/10.1016/j.rse.2009.08.014
Verburg, P. H., J. A. Dearing, J. G. Dyke, S. v. d. Leeuw, S. Seitzinger, W. Steffen & J. Syvitski (2016) Methods and approaches to modelling the Anthropocene. Global Environmental Change, 39, 328-340. http://dx.doi.org/10.1016/j.gloenvcha.2015.08.007
Verkerk, P. J., M. Lindner, M. Pérez-Soba, J. S. Paterson, J. Helming, P. H. Verburg, T. Kuemmerle, H. Lotze-Campen, A. Moiseyev, D. Müller, A. Popp, C. J. E. Schulp, J. Stürck, A. Tabeau, B. Wolfslehner & E. H. van der Zanden (2016) Identifying pathways to visions of future land use in Europe. Regional Environmental Change. 10.1007/s10113-016-1055-7
Verma, N. & C. Fernandez. 2010. India's petrol, diesel, kerosene and LPG prices. Online: Reuters.
Vitousek, P. M., H. A. Mooney, J. Lubchenco & J. M. Melillo (1997) Human Domination of Earth's Ecosystems. Science, 277, 5325, 494-499. 10.1126/science.277.5325.494
Voinov, A. & F. Bousquet (2010) Modelling with stakeholders. Environmental Modelling & Software, 25, 11, 1268-1281. http://dx.doi.org/10.1016/j.envsoft.2010.03.007
Voinov, A. & H. H. Shugart (2013) ‘Integronsters’, integral and integrated modeling. Environmental Modelling & Software, 39, 149-158. http://dx.doi.org/10.1016/j.envsoft.2012.05.014
Wade, A. J., G. M. Hornberger, P. G. Whitehead, H. P. Jarvie & N. Flynn (2001) On modeling the mechanisms that control in-stream phosphorus, macrophyte,
374
and epiphyte dynamics: an assessment of a new model using general sensitivity analysis. Water Resources Research, 37, 11, 2777-2792
Walker, B. & N. Abel. 2002. Resilient rangelands- adaptation in complex systems. In Panarchy: Understanding Transformations in Human and Natural Systems, eds. L. H. Gunderson & C. S. Holling, 293-314. Washington DC, U.S.A: Island press.
Walker, B., C. S. Holling, S. R. Carpenter & A. P. Kinzig (2004) Resilience, Adaptability and Transformability in Social–ecological Systems. Ecology and Society, 9, 2. 10.5751/ES-00650-090205
Walker, B. & J. A. Meyers (2004) Thresholds in Ecological and Social-Ecological Systems: a Developing Database. Ecology and Society, 9, 2. 10.5751/ES- 00664-090203
Walker, B. H., S. R. Carpenter, J. Rockstrom, A.-S. Crépin & G. D. Peterson (2012) Drivers, "Slow" Variables, "Fast" Variables, Shocks, and Resilience. Ecology and Society, 17, 3. 10.5751/ES-05063-170330
Walker, G. (2006) Climate change: The tipping point of the iceberg. Nature, 441, 7095, 802-805
Walker, W. E., P. Harremoës, J. Rotmans, J. P. van der Sluijs, M. B. A. van Asselt, P. Janssen & M. P. Krayer von Krauss (2003) Defining Uncertainty: A Conceptual Basis for Uncertainty Management in Model-Based Decision Support. Integrated Assessment, 4, 1, 5-17. 10.1076/iaij.4.1.5.16466
Walsh, W. A., C. Swanson & C.-S. Lee (1991) Combined effects of temperature and salinity on embryonic development and hatching of striped mullet, Mugil Cephalus. Aquaculture, 97, 281-290
Walters, C., V. Christensen & D. Pauly (1997) Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Reviews in Fish Biology and Fisheries, 7, 2, 139-172. 10.1023/A:1018479526149
Wang, C. & J. M. Blackmore (2009) Resilience concepts for water resource systems. Journal of Water Resources Planning and Management, 135, 6, 528-536
Wang, R., J. A. Dearing, P. G. Langdon, E. Zhang, X. Yang, V. Dakos & M. Scheffer (2012) Flickering gives early warning signals of a critical transition to a eutrophic lake state. Nature, 492, 7429, 419-422. http://www.nature.com/nature/journal/v492/n7429/abs/nature11655.ht ml#supplementary-information
Warner, K., M. Hamza, A. Oliver-Smith, F. Renaud & A. Julca (2010) Climate change, environmental degradation and migration. Natural Hazards, 55, 3, 689-715. 10.1007/s11069-009-9419-7
Wei, S., H. Yang, J. Song, K. C. Abbaspour & Z. Xu (2012) System dynamics simulation model for assessing socio-economic impacts of different levels of environmental flow allocation in the Weihe River Basin, China. European Journal of Operational Research, 221, 1, 248-262. http://dx.doi.org/10.1016/j.ejor.2012.03.014
Werners, S. E., S. Pfenninger, E. van Slobbe, M. Haasnoot, J. H. Kwakkel & R. J. Swart (2013) Thresholds, tipping and turning points for sustainability under climate change. Current Opinion in Environmental Sustainability, 5, 3–4, 334-340. http://dx.doi.org/10.1016/j.cosust.2013.06.005
375 Appendices
Westley, F., P. Olsson, C. Folke, T. Homer-Dixon, H. Vredenburg, D. Loorbach, J. Thompson, M. Nilsson, E. Lambin, J. Sendzimir, B. Banerjee, V. Galaz & S. van der Leeuw (2011) Tipping Toward Sustainability: Emerging Pathways of Transformation. Ambio, 40, 7, 762-780. 10.1007/s13280-011-0186-9
Weyant, J. (2017) Some Contributions of Integrated Assessment Models of Global Climate Change. Review of Environmental Economics and Policy, 11, 1, 115-137. 10.1093/reep/rew018
White, R. 1995. The Organic Machine. New York City, USA: Hill and Wang.
Whitehead, P. G., H. Leckie, K. Rankinen, D. Butterfield, M. N. Futter & G. Bussi (2016) An INCA model for pathogens in rivers and catchments: Model structure, sensitivity analysis and application to the River Thames catchment, UK. Science of The Total Environment, 572, 1601-1610. http://dx.doi.org/10.1016/j.scitotenv.2016.01.128
Wilson, J., L. Yan & C. Wilson (2007) The precursors of governance in the Maine lobster fishery. Proceedings of the National Academy of Sciences, 104, 39, 15212-15217. 10.1073/pnas.0702241104
Wintle, B. A., M. C. Runge & S. A. Bekessy (2010) Allocating monitoring effort in the face of unknown unknowns. Ecology Letters, 13, 11, 1325-1337. 10.1111/j.1461-0248.2010.01514.x
Wisker, G. 2008. The Postgraduate Research Handbook. Basingstoke, U.K: Palgrave Macmillan.
WorldBank. 2001. Poverty Trends and Voices of the Poor. In World Development Report. Washington DC, USA: World Bank.
Xu, D., Y. Cai, X. Wu, H. Jiang, X. Leng & S. An (2015) Regime Shifts and Resilience of the Lake Taihu Social-Ecological System Under Long-term External Disturbance (1960s–2000s). CLEAN – Soil, Air, Water, 43, 4, 561-568. 10.1002/clen.201300870
Yadav, D. K. & A. Barve (2017) Analysis of socioeconomic vulnerability for cyclone-affected communities in coastal Odisha, India. International Journal of Disaster Risk Reduction, 22, 387-396. http://dx.doi.org/10.1016/j.ijdrr.2017.02.003
Yang, Q., Y. Ding, B. de Vries, Q. Han & H. Ma (2014) Assessing Regional Sustainability Using a Model of Coordinated Development Index: A Case Study of Mainland China. Sustainability, 6, 12, 9282-9304
Yeo, A., R. Legard, J. Keegan, K. Ward, C. McNaughton Nicholls & J. Lewis. 2014. In-depth interviews. In Qualitative Research Practice: A Guide for Social Science Students and Researchers, 177-210. London, UK: Sage.
Young, O. R. (2010) Institutional dynamics: Resilience, vulnerability and adaptation in environmental and resource regimes. Global Environmental Change, 20, 3, 378-385. http://dx.doi.org/10.1016/j.gloenvcha.2009.10.001
Young, O. R., F. Berkhout, G. C. Gallopín, M. A. Janssen, E. Ostrom & S. van der Leeuw (2006) The globalization of socio-ecological systems: An agenda for scientific research. Global Environmental Change, 16, 3, 304-316. http://dx.doi.org/10.1016/j.gloenvcha.2006.03.004
376
Zarghami, M. & S. Akbariyeh (2012) System dynamics modeling for complex urban water systems: Application to the city of Tabriz, Iran. Resources, Conservation and Recycling, 60, Supplement C, 99-106. https://doi.org/10.1016/j.resconrec.2011.11.008
Zeileis, A. n.d. Conditional Density Plots. RDocumentation.
Zeileis, A., F. Leisch, K. Hornik, C. Kleiber, B. Hansen & E. C. Merkle. 2015. Package 'strucchange'. ed. A. Zeileis, 69.
Zeleny, M. 2009. Knowledge and Self-Production Processes in Social Systems. In Systems Science and Cybernetics, ed. F. Parra-Luna, 20. United Nations Educational, Scientific and Cultural Organization.
Zhang, K. (2016) Regime shifts and resilience in China’s coastal ecosystems. Ambio, 1-10. 10.1007/s13280-015-0692-2
Zhang, K., J. A. Dearing, T. P. Dawson, X. Dong, X. Yang & W. Zhang (2015) Poverty alleviation strategies in eastern China lead to critical ecological dynamics. Science of The Total Environment, 506–507, 164-181. http://dx.doi.org/10.1016/j.scitotenv.2014.10.096
Zhang, X. H., H. W. Zhang, B. Chen, G. Q. Chen & X. H. Zhao (2008) Water resources planning based on complex system dynamics: A case study of Tianjin city. Communications in Nonlinear Science and Numerical Simulation, 13, 10, 2328-2336. http://dx.doi.org/10.1016/j.cnsns.2007.05.031
Zhu, J., X. Wang, L. Zhang, H. Cheng & Z. Yang (2015) System dynamics modeling of the influence of the TN/TP concentrations in socioeconomic water on NDVI in shallow lakes. Ecological Engineering, 76, 0, 27-35. http://dx.doi.org/10.1016/j.ecoleng.2014.06.030
Zimmermann, H. J. (2000) An application-oriented view of modeling uncertainty. European Journal of Operational Research, 122, 2, 190-198. http://dx.doi.org/10.1016/S0377-2217(99)00228-3
Zurlini, G., K. B. Jones, K. H. Riitters, B.-L. Li & I. Petrosillo (2014) Early warning signals of regime shifts from cross-scale connectivity of land-cover patterns. Ecological Indicators, 45, 549-560. http://dx.doi.org/10.1016/j.ecolind.2014.05.018
377