Protecting Bhutan's freshwater ecosystem and biodiversity in the context of Nature Needs Half and protected area dynamism
Author Dorji, Tshering
Published 2020-08-10
Thesis Type Thesis (PhD Doctorate)
School School of Environment and Sc
DOI https://doi.org/10.25904/1912/3927
Copyright Statement The author owns the copyright in this thesis, unless stated otherwise.
Downloaded from http://hdl.handle.net/10072/396522
Griffith Research Online https://research-repository.griffith.edu.au Protecting Bhutan’s freshwater ecosystem and biodiversity in
the context of Nature Needs Half and protected area
dynamism
Tshering Dorji
B.Sc., M.Sc. Zoology (Special)
School of Environment and Science
Australian Rivers Institute
Griffith University
Submitted in fulfilment of the requirements of the degree of Doctor of Philosophy
March 2020
Preface
This thesis is dedicated to my paternal granduncle late Lama Kinzang Namgyal for teaching me that there are things beyond mundane in his miraculous passing away in deep meditative pose (Thukdam) within a few days of my departure from Bhutan to embark on this adveanturous journey of learning how to do science.
I also dedicate to the countless organisms that died, are dying and will die to keep my work relevant.
i
Acknowledgements
I am immensely grateful to my supervisors Dr. Simon Linke and Professor Fran Sheldon for seeing me through this journey. Their belief in me never wavered from the time I first requested them to be my supervisors, the time when I was not even sure of securing university admission. Their belief in me remained steadfast when I was going through the trials and tribulations of seeking scholarship to fund my study, which took me two years to get one. In particular, I would like to thank Dr. Simon Linke for allowing me to chart my own journey but being always there to tighten enough the lid of the pressure cooker at right moments so that I found a proper balance between work, work and life. I would also like to thank Pro. Fran Sheldon, in particular, for always remaining the solid bedrock to fall upon when I badly needed additional scholarship from the university. This helped me remain safely on the course of this journey.
I thank the World Wide Fund for Nature (WWF-USA) for the Russel E. Train Faculty
Fellowship that paid my university tuition fee for almost the first three years of my study period. This fellowship set me on this journey at the first place. I would like to thank
Griffith University for the GU Postgraduate Research Scholarship that provided living stipend for almost two years of my study period and for the GU International Postgraduate
Research Scholarship that covered my tuition fee for the remaining part of my study. I am grateful to the management of the College of Natural Resources, Royal University of
Bhutan for the study leave that enabled me to pursue this.
I thank my fellow PhD students working in the HDR-hub 1.1 in the Sir Samuel Griffith
Centre at Nathan campus, GU for the refreshing talks during meal breaks on the few occasions when I worked in the university campus. Sorry, I could not engage more with you all given my difficulty of balancing between my PhD and the almost full-time work beside the PhD. This note also extends to all the great researchers and the faculty members ii in the Australian Rivers Institute. I know I have missed tons of opportunity to learn from you all.
Finally, I would like to thank my father Tenzin Wangchuk and mother Tshumpi for always keeping me in their hearts though often I failed even to make a courtesy telephone call. I would like to thank immensely my father-in-law Nima and mother-in-law Lhamo for bring-up my three young (and understandably difficult) children throughout these years (and may be even few more months in the coming times) despite their old age. I would like to thank my three children, N Namgyel Dorji, Tashi Tshephel and Tsheyang
Tshogyal for bearing with me and my wife’s total absence for these many years. Sorry, we could not come even for a week’s break for these three and half years to physically meet you all. I hope, as you three grow up, you all will understand why we did so. I wold also like to thank my brother-in-law and sisters-in-law for keeping in touch and always concerned about me and my wife’s wellbeing. Now for the final-finally, I would like to thank my wife Choki Wangmo for always being by my side when I was going through the ups and downs of this journey. You kept me sane, and thank you once again for always cooking delicious hot meals for me despite your poor health, and of course also reminding me to eat it when still hot. Thank you all once again for putting up with my vagaries and cheering me on.
iii
Acknowledgement of co-authored papers included in this thesis
Included in this thesis are papers in Chapters 2, 3, 4 and 5 which are co-authored with other researchers. My contribution to each co-authored paper is outlined at the front of the relevant chapter. The bibliographic details for these papers including all authors are:
Chapter 2: Dorji, T., Linke, S., & Sheldon, F. (2019). Half century of protected area dynamism in the country of Gross National Happiness, Bhutan. Conservation Science and Practice, 1, e46. https://doi.org/10.1111/csp2.46
Chapter 3: Dorji, T., Linke, S., & Sheldon, F. (ready to submit). Optimal model selection for Maxent: a case of freshwater species distribution modelling in Bhutan, a data poor country. Ecography.
Chapter 4: Dorji, T., Sheldon, F., & Linke, S. (in review). Fulfilling Nature Needs Half through terrestrial-focused protected areas and their adequacy for freshwater ecosystems and biodiversity protection: a case from Bhutan. Journal for Nature Conservation.
Chapter 5: Dorji, T., Linke, S., & Sheldon, F. (in review). Freshwater conservation planning in the context of nature needs half and protected area dynamism in Bhutan. Biological Conservation.
Appropriate acknowledgements of those who contributed to the research but did not qualify as authors are included in each paper.
(Signed) (Date) 15/03/2020
Tshering Dorji
(Countersigned) (Date) 15/03/2020
Supervisor: Simon Linke
iv
Statement of Originality
This work has not previously been submitted for a degree or diploma in any university.
To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.
(Signed) (Date): 15/03/2020
Tshering Dorji
v
Abstract
Bhutan, a small developing country, is recognised for its pro-environmental and unique development philosophy called Gross National Happiness (GNH). Over the years
Bhutan has continually increased the area coverage under protected area (PA) system and probably it is the first country in the world that has achieved a 50% protection target.
Going beyond the year 2020, there is a growing call to protect at the least 30% of the
World’s terrestrial areas by 2030 with the ultimate goal to protect 50% by year 2050 - a goal commonly termed “Nature Needs Half’ (NNH). However, PAs globally and in
Bhutan have been designated to protect terrestrial ecosystems and biodiversity. Although freshwater ecosystems and biodiversity are most threatened, there are very few freshwater specific protected areas while they are included within existing PAs by coincidence.
The overall objective of this thesis is to assess the potential of existing terrestrial- focussed protected areas of Bhutan in conserving freshwater ecosystems and biodiversity within the context of nature needs half target and protected area dynamism. To achieve this, I have combined a systematic archival literature review with species distribution modelling and systematic conservation planning techniques to address four main issues:
(1) the lack of a country level PA dynamism study and possible link between it and GNH,
(2) lack of adequate freshwater species distribution data in Bhutan, (3) the need to assess potential of existing PA system for freshwater protection in the context of NNH, and (4) the need for systematic freshwater conservation planning in the context of NNH and PA dynamism.
To address the first issue, I conducted systematic archival literature review on PA dynamism in Bhutan for almost the entire modern conservation history of Bhutan (1966-
2016). PA dynamism can be either of gain or loss events; the latter is termed protected area downgrading, downsizing and degazettement (PADDD). The conservation planning vi was the main proximate cause for PA degazettement and downsizing. On the other hand, an increase in infrastructure with hydropower development as the specific cause was the proximate cause for proposed PADDD. Overall PA dynamism reflected Bhutan’s commitment to environmental conservation guided by its GNH philosophy. I recorded differences in the number of PADDD events, PADDD types and PAs affected between the current study and those previously recorded for Bhutan from a regional level study.
This demonstrated the need for country specific systematic archival literature search to better detect PADDD.
To fill in freshwater species data gaps I curated georeferenced occurrence data of fish and odonata from the available literatures till 2017 and also through personal communications. While building SDMs for 10 fish and 28 odonates I compared the performance of four sequential model selection approaches against an expert (EXP) approach in selecting ecologically plausible species models. The sequential approaches using omission rate followed by test AUC (ORTEST) performed better over the sequential approaches that used omission rate followed by difference between training AUC and test
AUC, and test AUC (AUCDIFF). ORTEST approaches could be used as a good first line model screening approach to reduce time taken by the EXP approach.
This thesis contextualized freshwater biodiversity and ecosystems conservation against NNH by categorizing levels of species habitat, river length and lake surface area percentage protection into four progressive targets based on Aichi target 11 and NNH target. Despite covering 50% of Bhutan’s total land area, many freshwater ecosystems and taxa were inadequately represented within the existing PA system of Bhutan. These gaps could be easily filled employing Marxan, a common systematic conservation planning tool. Even when considering hydropower developments, we found solutions that increased the species coverage significantly. That said, high percentages of selected areas
vii are already protected by the existing PA system. This suggests Bhutan may not need to change drastically the existing PA system if PAs are redesigned to better represent both freshwater species and forest types.
This thesis provides the first real world example of inadequacy in protecting freshwater ecosystems and biodiversity by terrestrial-focused PAs even when total area of PA system meets the NNH target. The findings also demonstrate a clear need for freshwater conservation planning. The collated species occurrence data and ecologically plausible SDMs chosen here could be a starting point for freshwater conservation planning in Bhutan. The thesis also showed clear need to conduct country level systematic archival literature review to better detect PADDD. This thesis also linked PA dynamism study with freshwater systematic conservation planning and probably this could be way forward to design robust PA systems.
viii
Table of Contents
1 Introduction...... 1
1.1 Global perspective ...... 1
1.1.1 Declining Biodiversity: varying rates ...... 1
1.1.2 Threats to freshwater biodiversity and ecosystem ...... 2
1.1.3 Protected Areas: the response to declining biodiversity ...... 3
1.1.4 Systematic conservation planning for freshwater ecosystems ...... 7
1.2 National perspective ...... 10
1.2.1 Study area: Bhutan ...... 10
1.2.2 Freshwater ecosystem and biodiversity in Bhutan...... 11
1.2.3 Threats to freshwater biodiversity and ecosystem in Bhutan ...... 12
1.2.4 PAs of Bhutan and Gross National Happiness ...... 13
1.2.5 Freshwater ecosystem and biodiversity conservation in Bhutan ...... 15
1.3 Thesis structure ...... 15
2 Half century of protected area dynamism in the country of Gross National Happiness, Bhutan . 18
2.1 Abstract ...... 19
2.2 Introduction ...... 20
2.3 Methods ...... 23
2.3.1 Study system ...... 23
2.3.2 Gain and PADDD events, count and area affected ...... 25
2.3.3 Proximate causes and count ...... 26
2.3.4 Data collection ...... 26
2.3.5 Data analyses ...... 27
2.4 Results ...... 27
2.4.1 Summary of gain and PADDD events ...... 27
2.4.2 Enacted versus proposed PADDD events ...... 31
2.4.3 Gain and PADDD events temporal trend ...... 31 ix
2.4.4 Gain and PADDD events trend among individual PAs ...... 33
2.4.5 Proximate causes of PADDD ...... 33
2.4.6 PADDD events recorded in www.PADDDtracker.org versus current study ...... 35
2.5 Discussion ...... 37
2.5.1 Overall gain and PADDD events ...... 37
2.5.2 Gain and PADDD events temporal trend ...... 38
2.5.3 Gain and PADDD events trend among individual PAs ...... 42
2.5.4 Proximate cause of PADDD ...... 42
2.5.5 PADDD recorded in www.PADDDtracker.org versus current study ...... 43
3 Optimal model selection for Maxent: a case of freshwater species distribution modelling in
Bhutan, a data poor country ...... 46
3.1 Abstract ...... 47
3.2 Introduction ...... 48
3.3 Methods ...... 52
3.3.1 Species occurrence data ...... 52
3.3.2 Environmental variables ...... 52
3.3.3 Bias files ...... 53
3.3.4 Model setting ...... 53
3.3.5 Selecting optimal models using sequential approaches ...... 55
3.3.6 Selecting optimal models using expert approach ...... 58
3.3.7 Data analysis ...... 59
3.4 Results ...... 60
3.4.1 Model complexity ...... 60
3.4.2 Predicted habitat ...... 66
3.5 Discussion ...... 69
3.5.1 Model complexity ...... 69
3.5.2 Predicted habitat ...... 72 x
4 Fulfilling Nature Needs Half through terrestrial-focused protected areas and their adequacy for freshwater ecosystems and biodiversity protection: a case from Bhutan ...... 74
4.1 Abstract ...... 75
4.2 Introduction ...... 76
4.3 Materials and method ...... 78
4.3.1 Protected areas of Bhutan ...... 78
4.3.2 Freshwater ecosystems representation ...... 79
4.3.3 Freshwater biodiversity representation ...... 80
4.3.4 Percentage protection levels ...... 81
4.3.5 Data analysis ...... 82
4.4 Results ...... 82
4.4.1 Freshwater ecosystem protection ...... 82
4.4.2 Biodiversity protection ...... 88
4.5 Discussion ...... 88
5 Freshwater conservation planning in the context of nature needs half and protected area dynamism in Bhutan ...... 91
5.1 Abstract ...... 92
5.2 Introduction ...... 93
5.3 Methods ...... 96
5.3.1 Conservation features ...... 96
5.3.2 Planning units ...... 96
5.3.3 Reserve design ...... 98
5.3.4 Analysis ...... 100
5.4 Results ...... 100
5.5 Discussion ...... 104
6 General Discussion ...... 107
xi
6.1 Overall Summary ...... 107
6.2 Protected area dynamism ...... 108
6.3 Species distribution modelling...... 110
6.4 Freshwater biodiversity and ecosystem protection against NNH target ...... 112
6.5 Systematic conservation planning in light of NNH and PA dynamism ...... 113
6.6 Significance of this thesis and future research ...... 114
6.7 Conclusion ...... 117
7 References ...... 118
8 Supplementary materials...... 154
8.1 Supplementary materials for Chapter 2 ...... 154
8.2 Supplementary materials for Chapter 3 ...... 158
8.2.1 Appendix 3.1 Detail method of fish and odonata species occurrence data cleaning and
collation for Bhutan ...... 158
8.3 Supplementary materials for Chapter 4 ...... 189
8.4 Supplementary materials for Chapter 5 ...... 191
xii
List of Figures
Figure 1-1. Bhutan and its neighbours. Indian state of Sikkim to west, Indian state of West Bengal to south-
west, Indian state of Assam to south-east, and Tibet Autonomous Region of China the entire north.
(Image adopted from Google Earth @ 2020 Google) ...... 11
Figure 1-2 Existing protected areas of Bhutan along with three Ramsar sites (each PA with unique colour).
BWS: Bumdeling Wildlife Sanctuary, JWS: Jomotshangkha Wildlife Sanctuary, PWS: Phibsoo Wildlife
Sanctuary, SWS: Sakteng Wildlife Sanctuary, JDNP: Jigme Dorji National Park; JSWNP: Jigme Singye
Wangchuck National Park; PNP: Phrumsengla national park, RMNP: Royal Manas National park,
WCNP: Wangchuck Centennial National park, JKSNR: Jigme Khesar Strict Nature Reserve; NC: North
Corridor; JKSNR-JDNP biological corridor, JDNP-JSWNP biological corridor, PWS-JSWNP-RMNP
biological corridor, RMNP-JSWNP-PNP biological corridor, PNP-BWS biological corridor, RMNP-JWS
biological corridor, JWS-SWS biological corridor, BR: Bumdeling Ramsar site, KR: Khotokha Ramsar
site, GPR: Gangtey-Phobji Ramsar site...... 14
Figure 2-1 Protected area systems of Bhutan in (a) 1984 (legend shows PAs degazetted or downsized over
the study period along with other PAs), (b) 1993, (c) 1999, (d) 2008 and (e) 2016. (Figure 2.1 a-c not
to scale; Figure 2.1 a digitized after Thinley et al. 2013; Figure 2.1 b after Seeland (1998); Figure 2.1
c after NCD (2004); Figure 2.1 d & 1e redrawn from shapefiles obtained from Watershed
Management Division, Department of Forest and Park Services, Ministry of Agriculture and Forest).
(BC: Biological Corridor; BWS: Bumdeling Wildlife Sanctuary; DNP: Doga National Park; DWR:
Dungsam Wildlife Reserve; GPR: Gangtey-Phobji Ramsar Site; JDNP: Jigme Dorji National Park;
JDWS: Jigme Dorji Wildlife Sanctuary; JKSNR: Jigme Khesar Strict Nature Reserve; JSWNP: Jigme
Singye Wangchuck National Park; JWS: Jomotshangkha Wildlife Sanctuary; KR: Khotokha Ramsar
Site; KRF: Khaling Reserve Forest; KWS: Khaling Wildlife Sanctuary; MWS: Manas Wildlife Sanctuary;
NaWR: Namgyel Wangchuck Reserve; NWR: Neoli Wildlife Reserve; PNP: Phrumsengla National
Park; PoRF: Pochu Reserved Forest; PRF: Phibsoo Reserve Forest; PWS: Phibsoo Wildlife Sanctuary;
RMNP: Royal Manas National Park; SRF: Sinchula Reserve Forest; SWR: Shumar Wildlife Reserve;
SWS: Sakteng Wildlife Sanctuary; TNP: Thrumshingla National Park; TSNR: Toorsa Strict Nature
Reserve; WCNP: Wangchuck Centennial National Park; ZRF: Zoshing Reserve Forest.) ...... 24
xiii
Figure 2-2 Gain and PADDD events across PA system of Bhutan between 1966 and 2016 (a) number of
gain events; (b) area gained from new protected area, enlarged PA and total; (c) number of enacted
and proposed PADDD events and (d) area lost to downsizing, degazettement and total...... 32
Figure 2-3 Number of proposed and enacted PADDD events due to different proximate causes across PA
system of Bhutan between 1966 and 2016...... 35
Figure 3-1 Schematic diagram of selecting optimal models. Steps with red arrows are unique to AUCDIFF
approaches while all other steps are common to all the four sequential approaches. Green boxes
with “Single model selected” are the optimal models for either ORTEST approaches or AUCDIIF
approaches. If Step 2a is followed then one will derive optimal models for AUCDIFF after Step 2,
otherwise optimal models derived will be for ORTEST approaches. While purple box with “Single
model selected” is the optimal model for only AUCDIFF approaches. Step 5 is the ultimate step
wherein models with lower feature class is chosen as the optimal model when multiple models have
same numbers of parameters and average absolute value...... 58
Figure 3-2 Regularization multiplier and feature class (RM-FC) combinations for fish and odonate of the
optimal models chosen by five model selection approaches...... 61
Figure 3-3 Summary of different features of the optimal models selected through five optimal model
selection approaches for the fish and odonate species of Bhutan. 10 percentile training presence
test omission for the (a) fish and (b) odonate; balance training omission, predicted area and
threshold value test omission for the (c) fish and (d) odonate; AUC difference for the (e) fish and (f)
odonate; average number of parameters for the (g) fish and (h) odonate; test AUC for the (i) fish
and (j) odonate species of Bhutan. EXP: Expert approach based on ecological plausibility of binary
suitable/unsuitable model; ORTEST_PER: sequential approaches using 10 percentile training
presence test omission and test AUC; AUCDIFF_PER: 10 percentile training presence test omission,
AUC difference and test AUC; ORTEST_BAL: balance training omission, predicted area and threshold
value test omission and test AUC; AUCDIFF_BAL: balance training omission, predicted area and
threshold value test omission, AUC difference and test AUC...... 64
Figure 3-4. Area of the predicted habitats of the optimal models selected through the five model selection
approaches for the (a) fish and (b) odonate species of Bhutan...... 67
Figure 3-5. Examples of binary suitable habitat maps derived using 10 percentile training presence Cloglog
threshold (Left panel) and balance training omission, predicted area and threshold value Cloglog xiv
threshold (Right panel) among the optimal models chosen by the Expert and the four sequential
optimal model selection approaches. (a) Cyprinion semiplotum (5 occurrence; EXP, ORTEST_PER,
ORTEST_BAL, AUCDIFF_PER and AUCDIFF_BAL), (b) Neolissochilus hexagonolepis (12 occurrence;
EXP, ORTEST_PER); (c) Neolissochilus hexagonolepis (12 occurrence; ORTEST_BAL, AUCDIFF_PER
and AUCDIFF_BAL); (d) Aristocypha quadrimaculata (8 occurrence; EXP, ORTEST_PER, ORTEST_BAL,
AUCDIFF_PER and AUCDIFF_BAL); (e) Diplacodes trivialis (21 occurrence; EXP); (f) Diplacodes trivialis
(21 occurrence; ORTEST_PER); (g) Diplacodes trivialis (21 occurrence; ORTEST_BAL, AUCDIFF_PER
and AUCDIFF_BAL). 0: Unsuitable habitat; 1: Suitable habitat...... 68
Figure 4-1. Terrestrial protected areas of Bhutan. Drawn from shapefiles obtained from Watershed
Management Division, Department of Forest and Park Services, Ministry of Agriculture and Forestry,
Royal Government of Bhutan. JDNP: Jigme Dorji National Park, JSWNP: Jigme Singye Wangchuck
National Park, PNP: Phrumsengla National Park, RMNP: Royal Manas National Park, WCNP:
Wangchuck Centennial National Park, JKSNR: Jigme Khesar Strict Nature Reserve, BWS: Bumdeling
Wildlife Sanctuary, JWS: Jomotshangkha Wildlife Sanctuary, SWS: Sakteng Wildlife Sanctuary, PWS:
Phibsoo Wildlife Sanctuary. The biological corridors, except for NC (North corridor), are named after
the PAs they connect. For example, JKSNR-JDNP is biological corridor connecting Jigme Khesar Strict
Nature Reserve and Jigme Dorji National Park...... 79
Figure 4-2. Relationship between the percentages protected and surface area of the lakes (a), length of
river reaches (b), habitat area of fish species (c), and habitat area of odonate species (d) of Bhutan.
...... 85
Figure 5-1. Planning units (PUs) with different status for the two Marxan scenarios. Available: PUs
available for both scenarios; Locked in: PUs with three Ramsar sites locked in for both scenarios;
Locked out: PUs with hydropower sites (of existing, under construction and detailed project report
stages) and one preceding and one following the PUs with hydropower sites along the river flow
direction were locked out for Scenario 2. We used PUs outside Bhutan’s international boundary to
emphasis importance of longitudinal connectivity in identifying freshwater reserve design...... 97
Figure 5-2. Connectivity penalty (CP) calibration. We set a medium uniform species target of 650 km2 and
ran Marxan with 10 repeat runs for 13 different CPs (0, 0.01, 0.05, 0.1, 1, 5, 10, 50, 100, 500, 1000,
5000 and 10000) using fixed SPF of 100. CP 50 has the highest increase in connectivity with the
lowest increase in cost...... 101 xv
Figure 5-3. Percentage of modelled species distribution and forest type distribution areas protected within
the two reserves and existing protected area system of Bhutan...... 102
Figure 5-4. Reserve 1 and Reserve 2 with the location of hydropower sites and the common and unique
PUs selected. Reserve 1(top panel): all PUs available; Reserve 2 (bottom panel): PUs with
hydropower sites locked out. Note we locked in three Ramsar sites for both the reserves...... 102
Figure 5-5. Percentage protection level of PUs of Reserve 1 (top panel) and Reserve 2 (bottom panel).
...... 104
xvi
List of Tables
Table 2-1 Summary of changes to protected area system of Bhutan between 1966 and 2016...... 28
Table 2-2 Protected areas of Bhutan affected by gain events between 1966 and 2016 with gain event types
(enlarged PA and policy upgrade), current, primary and after gain event names of protected areas,
year of event and area affected...... 30
Table 2-3 Protected areas of Bhutan affected by PADDD events between 1966 and 2016 with PADDD
events type, area lost, status, year of event, and proximate and specific causes...... 34
Table 2-4 Difference between PADDD events recorded in www.PADDDtracker.org and current study for
PA system of Bhutan...... 36
Table 3-1 Number of fish and odonate species of Bhutan with same model setting (Regularization
Multiplier-Feature Class (RM-FC) combination) of their optimal models as selected by all five model
selection approaches (ALL) and by the different pairs of the selection approaches...... 61
Table 4-1. Distribution of number of lakes within six agro-ecological zones of Bhutan and their number
with different percentage surface area protection...... 83
Table 4-2. Distribution of number of river reaches within twenty-two river reach types found in Bhutan
and their number with different percentage length protection...... 87
Supplementary Materials
Table S2.1 Summary of changes to protected area system of Bhutan between 1966 and 2016. (aProtected Area Downgrading, Downsizing and Degazettement; bProtected areas; cArea affected unknown for 3 policy upgrade events; dArea affected by all policy downgrade events unknown; e14 were proposed and 6 enacted policy downgrade events.) Avaialble from https://conbio.onlinelibrary.wiley.com/doi/abs/10.1111/csp2.46 as Supporting Information csp246-sup-0001-AppendixS1.xlsx
Table S2.2 Protected areas of Bhutan affected by gain events between 1966 and 2016 with gain event types (enlarged PA and policy upgrade), current, primary and after gain event names of protected areas, year of event and area affected. Available from https://conbio.onlinelibrary.wiley.com/doi/abs/10.1111/csp2.46 as Supporting Information csp246-sup-0002-AppendixS2.xlsx
Table S2.3. Protected areas from different data sources and consolidated area.
xvii
Table S2.4 Percentage distribution of number of PA gain and PADDD events and area affected across Bhutan’s PA system between 1966 and 2016.
Table S2.5. Number and percentage of proposed and enacted PADDD events across PA system of Bhutan between 1966 and 2016.
Table S2.6. Distribution of gain events (A) and PADDD events (B) and percentage composition across PA system of Bhutan between 1966 and 2016.
Table S2.7 Change in area and percentage across the PA system of Bhutan between 1966 and 2016 due to (A) new and enlarged PAs and (B) degazettement and downsizing events.
Table S2.8. Number and percentage of different proximate causes for different PADDD types across Bhutan’s PA system between 1966 and 2016.
Table S2.9. Summary of PADDD recorded in www.PADDDtracker.org across Bhutan’s PA system
Table S3.1. List of cleaned and collated (a) fish and (b) odonata species of Bhutan with georeferenced coordinates and remarks on identification and occurrence. Grey portions contain the species with adequate species identification and georeferenced coordinates and were considered for model building in this study.
Table S3.2. Shapiro Wilk test for normality of different model features for five optimal model selection approaches and the number of occurrence record for the fish and odonate species of Bhutan. Percentile OR: 10 percentile training presence test omission; Balance OR: Balance training omission, predicted area and threshold value test omission; No. of parameters: Number of parameters with non-zero lambda coefficients; Habitat area: predicted habitat area; Occurrence: Number of occurrence records of the species used for model building.
Table S3.3 Regularization multiplier and feature class (RMxFC) combinations of the optimal models chosen by five model selection approaches for the fish and odonate species of Bhutan. Number of occurrence points used for model building is also given.
Table S3.4 Spearman’s correlation coefficient (rs) and two-tailed significance (p) value among the optimal models chosen through different model selection approaches for the different model features for the fish and odonate species of Bhutan. Features: 10 percentile OR: 10 percentile training presence test omission; Balance OR: Balance training omission,
predicted area and threshold value test omission; AUCDIFF: Difference between training xviii
AUC and test AUC; Parameters: number of parameters with non-zero lambda coefficient values; Test AUC; Area: predicted habitat area.
Table S4.1. Shapiro-Wilk normality test results for the lake surface area, river reach length, fish and odonata species habitat area and the percentage protected.
Table S4.2. Habitat area, protected habitat area and percentage protected of the habitat of (a) fish and (b) odonate species of Bhutan.
Table S5.1. Conservation feature targets calibrated to obtain Marxan best solutions with reserve area of 50% Bhutan’s total land area.
Table S5.2. Conservation feature category, total area of conservation features and their percentage protected within best solutions of Marxan scenario 1 and 2, and the existing protected area system (PA) of Bhutan.
Table S5.3. Number of planning units and their percentage falling within different percentage protected range by the existing protected area system for the Marxan scenario 1 and scenario 2.
xix
1 Introduction
1.1 Global perspective
1.1.1 Declining Biodiversity: varying rates
Despite some local success stories (Butchart, 2010), global biodiversity continues to decline (Butchart, 2010; World Wide Fund for Nature (WWF), 2016; WWF, 2018; He et al., 2019; Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem
Services (IPBES), 2019) across a broad range of taxonomic groups, including freshwater crabs (Cumberlidge et al., 2008), odonates (Clausnitzer et al., 2009), mammals (Ceballos
& Ehrlich, 2006) and amphibians (Stuart et al., 2008). The current extinction rate has been estimated to be 1000 times the likely background rate (Pimm et al., 2014; IPBES,
2019). The persistent downward trend in the abundance of 3,706 vertebrate species monitored since 1970 (WWF, 2016) and the likely chance that 121-219 vertebrate species could become threatened by 2030 if forest loss continues at its present rate (Betts et al.,
2017) further supports the predicted increase in species extinction rate in coming times
(Pimm et al., 2014; IPBES, 2019).
The rate of biodiversity loss among the terrestrial, marine and freshwater ecosystems differs (WWF, 2016; WWF, 2018). The rate for freshwater biodiversity loss far exceeds that of terrestrial (WWF, 2016; Harrison et al., 2018; WWF 2018; Tickner et al, 2019) and marine counterparts (WWF, 2016; Carrizo et al., 2017; WWF, 2018). For example, among the vertebrate populations monitored from 1970 to 2014 the abundance of freshwater biodiversity dropped by 81% whereas the abundance of terrestrial and marine vertebrates dropped only by 38% and 36% respectively (WWF, 2016). Further, the threat of extinction rates differs among freshwater organisms (Collen et al., 2014).
For instance, freshwater megafaunas were found the most threatened among the freshwater vertebrates monitored from 1970 to 2014, - among these fish were the most
1 threatened followed by reptiles and mammals (He et al., 2017). Among higher taxonomic groups amphibians are more threatened (Clausnitzer et al., 2009; Pimm et al., 2014) compared to crabs (Cumberlidge et al., 2009), while odonates were the least threatened
(Clausnitzer et al., 2009). While these generalities across taxonomic groups exist, within group extinction risk differed from region to region, place to place and ecosystem types as found for odonates (Clausnitzer et al., 2009) and crabs (Cumberlidge et al., 2009).
These differences reflect the different levels of threat posed to habitats and to the uniqueness of the habitats (Clausnitzer et al., 2009) and conservation efforts (Betts et al.,
2017).
1.1.2 Threats to freshwater biodiversity and ecosystem
Threats to biodiversity in general include habitat degradation and loss, species overexploitation, pollution, invasive species and diseases, and climate change (WWF,
2016; WWF, 2018; IPBES, 2019). Global freshwater biodiversity in particular is threatened by overexploitation, water pollution, flow modification, habitat destruction or degradation, invasion by exotic species (Dudgeon et al., 2006; Dudgeon, 2010) and climate change (Dudgeon et al., 2006; WWF, 2016; WWF, 2018; Reid et al., 2019).
Among these overexploitations, habitat degradation and loss are the most common threats to declining freshwater biodiversity (WWF, 2018; IPBES, 2019).
Freshwater systems tend to be hotspots of human activities and this increased the pressure on freshwater ecosystems and their biodiversity, and this pressure is likely to increase overtime (Dudgeon et al., 2006; Strayer & Dudgeon, 2010) with recent studies suggesting a rise in the threat to freshwater biodiversity (e.g. WWF, 2016; Dudgeon,
2019; Reid et al., 2019). In recent times previously recognised threats to freshwater biodiversity have increased and new and novel threats have emerged, particularly in the decade since the influential work of Dudgeon et al., (2006) on the threats to freshwater
2 biodiversity (Reid et al., 2019). Our understanding of emergent threats like engineered nanomaterials, microplastics and emerging contaminants are at an early stage, but the impact of these threats to freshwater systems are expected to increase (Reid et al., 2019).
Among many previously recognised threats that are increasing, the hydropower development has increased manyfold over these years and may continue to do so (Reid et al., 2019). Hydropower dams mainly possess great threat to migrating fishes, for example to already threatened megafauna fishes (He eta al., 2017; Carrizo et al., 2019) and even to freshwater eels which are otherwise very adaptive (Drouineau et al., 2018). Already there are very few rivers over 1000 km length which are without dams globally and these are located in remote regions (Grill et al., 2019).
1.1.3 Protected Areas: the response to declining biodiversity
Modern protected areas (PAs) were initially aimed to conserve scenic areas for public recreation, and they originated in the nineteenth century new nations of North
America, Australia, New Zealand and South Africa (Phillips, 2004). Over time PAs have been designated to serve varying purposes (Watson, Dudley, Segan, & Hockings, 2014), however, now they are considered the main tool to conserve biodiversity and halt its decline (Bernard, Penna, & Araújo, 2014; Hermoso, Abell, Linke, & Boon, 2016) among numerous other methods used and proposed (Butchart et al., 2010; WWF, 2016).
The number and area of PAs in all realms have been increasing in recent times
(Deguignet et al. 2014; Watson et al., 2014; Hermoso et al., 2016; Venter et al., 2017;
Acreman et al, 2020). As of August, 2014, there were 209,429 PAs with a total surface area of 32,868,673 km2 which made up 14% of the world’s terrestrial area inclusive of
Antarctica (Deguignet et al. 2014). The increase in PAs was driven by the Aichi target 11 of the Convention on Biological Diversity (CBD) which envisaged to designate at least
3
17 per cent of terrestrial and inland water areas and 10 per cent of coastal and marine areas as PA, or other effective area-based conservation measure, by year 2020 (CBD,
2010). Going beyond 2020 the need for setting more ambitious targets for PAs called
Nature Need Half (NNH) has been proposed (e.g. Locke 2014, Dinerstein et al., 2019;
Visconti et al., 2019).
NNH had its early beganning in early 1990’s from the growing scientific evidences which showed most regions studied would need on average 50% to be protected to maintain full biodiversity (Kopnina, 2016). This was further enhanced by Edward O.
Wilson’s opinion blog called Half Earth (Kopnina 2016) and formation of Nature Needs
Half coalition (https://natureneedshalf.org/) among scientists, conservationists, nonprofits, and public officials in 2009 during 9th World Wilderness Congress in Mexico
(https://www.wild.org/who-we-are/history/). The Proponents of Nature Needs Half
(NNH) movement has argued for the need to protect at least 30% of the earth by 2030 and then increase to 50% by 2050 (Dinerstein et al., 2017; Dinerstein et al., 2019).
However, increasing only the area placed under protection alone will not guarantee biodiversity conservation (Pimm et al., 2018; Visconti et al., 2019). In order for PAs to be successful it depends on where they are placed and how well they are managed (Pimm et al., 2018; Visconti et al., 2019), highlighting the need for both efficient and effective
PA systems.
1.1.3.1 Protected area downgrading, downsizing and degazettement (PADDD)
PAs, though designated to conserve biodiversity for perpetuity (Golden Kroner et al, 2019), are dynamic entities that keep changing overtime (Mascia & Pailler, 2011;
Cook, Valkan, Mascia, & Mc Geoch, 2017; Golden Kroner et al., 2019). PAs are particularly under threat from protected area downgrading, downsizing and degazettement (PADDD) events that not only reduce their efficiency but are sometimes
4 altogether removed from the system (Mascia & Pailler, 2011; Cook et al., 2017; Golden
Kroner et al., 2019). Global level analysis found infrastructure development as one of the major proximate causes for PADDD, and hydropower development was the major specific cause under infrastructure development (Mascia & Pailler, 2011; Mascia et al.,
2014; Golden Kroner et al., 2019). However, proximate causes for PADDD also include conservation planning and hence PADDD events are not always negative for conservation
(Mascia & Pailler, 2011; Symes, Rao, Mascia, & Carrasco, 2016; Golden Kroner et al.,
2019). Understanding what the drivers and proximate causes are for PADDD would not only help prevent PADDDs from impacting existing PA systems but can also help design robust PA systems (Symes et al., 2016; Tesfaw et al., 2018). However, study of PADDD as a conservation phenomenon is at its early stage and there are needs for further study to understand its full scope (Mascia & Pailler, 2011; Symes et al., 2016; Golden Kroner et al., 2019).
1.1.3.2 PAs are terrestrial-focussed
Protected areas have been historically used for the protection of terrestrial biodiversity and habitat (Pittock et al., 2015) with a recent increase in marine biodiversity conservation (Worm, 2017). PAs have played a minimal role in freshwater biodiversity and ecosystem conservation (Pittock et al., 2015; Hermoso et al., 2016; Acreman et al.,
2020). Policymakers, water resource managers and even conservationists have considered conservation of freshwater biodiversity of secondary importance (Abell et al., 2008;
Pittock et al., 2015). However, recognizing that freshwater biodiversity and ecosystems are the most globally threatened ecosystems, some freshwater protected areas (FPAs) equivalent to terrestrial and marine PAs have been used to manage freshwater environments (Acreman et al., 2020). Further, more recent studies have recognized the need for FPA designation (e.g. Chessman, 2013; Raghavan et al., 2016). However, the
5 number of actual FPAs declared has been very low (Abell et al., 2008; Acreman et al.,
2020) and conservation of freshwater ecosystems like rivers (Abell, Lehner, Thieme, &
Linke, 2017) and seasonal inland wetlands (Reis et al., 2017) remains below the Aichi
2020 target. This is despite recent methodological suggestions for FPA creation and management (Kingsford, Biggs, & Pollard, 2011; Linke, Turak, & Nel, 2011; Pittock et al., 2015; Hermoso et al., 2016; Linke et al., 2019; Reis et al., 2019). These could be due to unique feature of freshwater ecosystems (Pittock et al., 2015; Hermoso et al., 2016;
Acreman et al., 2020) which requires management within and beyond PA boundary
(Abell et al., 2017).
1.1.3.3 Terrestrial PAs and potential for freshwater conservation
Freshwater ecosystems are present within many terrestrial PAs (Acreman et al.,
2020) and many have been retroactively designated as FPAs (Bower, Lennox, & Cooke,
2014). Integration of terrestrial and freshwater ecological management requirements have been proposed against a backdrop of reluctance in FPA creation (Dudgeon et al., 2006;
Linke, 2011; Pittock et al., 2015; Raghavan, Das, Nameere, Bijkumar, & Dahanukar,
2016). Recently, studies that assessed the potential of terrestrial-focussed PAs in conserving freshwater ecosystems (e.g. Poppe, Biffard, & Stevens, 2016; Abell et al.,
2017; Reis et al., 2017) and species (e.g. Januchowski-Hartley, Pearson, Puschendorf, &
Rayner, 2011; Raghavan et al., 2016) have increased but found mixed results with some studies suggesting even negative role of PAs on freshwater conservation (see Acreman et al., 2020). This was expected since freshwater ecosystems and biodiversity were actually only included by chance and had not been considered while designing the PAs (Pittock et al., 2015). For a PA to be effective as FPA it may require whole-catchment management, maintenance of natural flow and exclude non-native species (Saunders, Meeuwig, &
Vincent, 2002). However, the studies assessed the potential or gap for freshwater
6 conservation against the Aichi target 11 of 17% protection (e.g. Abell et al., 2017; Reis et al., 2017). Moving beyond 2020 may require assessments to be made against the NNH target. Already representation of terrestrial ecosystems (e.g. Dinerstein et al., 2017) and terrestrial mammals, terrestrial birds and amphibians (Pimm et al., 2018) within existing
PAs had been assessed against NNH target.
1.1.4 Systematic conservation planning for freshwater ecosystems
The need for systematic conservation planning for freshwater has been recognised
(Linke et al., 2011). There are many decision support tools that can be used for systematic conservation planning (Janßen, Göke, & Luttmann, 2019). Marxan is one very popular conservation planning tool and had traditionally been used for planning terrestrial and marine conservation areas (Linke et al., 2011). However, recent growth in both theoretical and methodological improvements (e.g. Linke et al., 2008; Linke et al., 2011; Hermoso et al., 2012; Hermoso, Cattarino, Kennard, & Linke, 2015; Linke et al., 2019; Reis,
Hermoso, Hamilton, Bunn, & Linke, 2019) on the use of Marxan for freshwater conservation planning has increased its use for freshwater conservation (Linke et al.,
2012; Flitcroft, Cooperman, Harrison, Juffe-Bignoli, & Boon, 2019; Linke, Hermoso, &
Januchowski-Hartley, 2019). The notable recent methodological improvements in freshwater conservation are inclusion of groundwater within freshwater planning framework (e.g. Linke et al., 2019) and integration of river-floodplain and upstream catchment connectivity (Reis et al., 2019).
Studies (e.g. Amis et al., 2009; Nel et al., 2009) that employed Marxan for freshwater conservation planning found the potential of existing PAs to conserve freshwater biodiversity can increase with minor modification and/or expansion of existing
PAs (Amis et al., 2009; Nel et al., 2009). Going beyond 2020 the NNH target will provide further scope for freshwater conservation planning in the countries which have PA size
7 smaller than the NNH target (Tickner et al., 2020). Even countries that have met NNH targets can also benefit from the conservation planning exercise when considered against the dynamic nature of PAs. The freshwater priority areas identified within existing PAs could be targeted for better management to help freshwater conservation (Pittock et al.,
2015; Raghavan et al., 2016). However, while planning for freshwater conservation there is a need to account for threats (Linke et al., 2019; Hermoso, Filipe, Segurado, & Beja,
2017) in order to identify robust priority areas that can be designated as PAs.
1.1.4.1 Data for freshwater conservation planning
Systematic conservation planning and management (Linke et al., 2011), and the identification of conservation planning gaps (Raghavan et al., 2016; Pittock et al., 2015) require data on freshwater biodiversity and ecosystems. However, data on freshwater biodiversity is generally scarce (Darwall et al., 20111; Brooks et al., 2016; Darwall et al.,
2019) and the available data are biased towards vertebrates (Gatti, 2016). Biodiversity data is even scarcer in parts of the biodiverse developing countries (Esselman & Allan,
2011) where threats to biodiversity are also higher (Esselman & Allan, 2011; Darwall et al., 2019). These can be overcome by ever increasing open access biodiversity data sources like Global Biodiversity Information Facility (gbif.org) and iNaturalis
(inaturalis.org) (Pearson, 2018), and strengthening of biodiversity monitoring
(Lindenmeyer et al., 2012).
Further, in many developing countries there are no adequate data on freshwater ecosystems for better planning, though some studies have developed new data and this has been used for freshwater conservation in some countries (see Linke et al., 2011).
Encouragingly, the availability of open access data on both biodiversity and other ecosystem components required for freshwater conservation planning are becoming
8 available. For example, the global river classification (GloRiC) dataset can be used to designate river ecosystem types (Quellet-Dallaire et al., 2018), hydroshed datasets
(Lehner & Grill, 2013) can be used to derive planning units for systematic conservation planning, and global human footprint data (Venter et al., 2016; Venter et al., 2018) can be used to derive planning unit cost.
1.1.4.2 Species distribution modelling to fill biodiversity data gap
Species distribution modelling has been used increasingly for conservation planning in data poor study areas (Guisan et al., 2016; Araújo et al., 2019). Freshwater conservation planning in recent times has also employed species distribution modelling tools to fill in patchy data (e.g. Hermosa et al., 2012; Domisch, Jähnig, Simaika,
Kuemmerlen, & Stoll, 2015; Domisch, Wilson, & Jetz, 2016; Balaguera-Reina, 2017;
Domisch et al., 2019), with many SDM tools able to be used depending on data availability. Maxent has been popular among many SDM tools that require only species occurrence data due to its better performance with very poor data and also due to its user friendly graphical user interface (Galante et al., 2018). However, Maxent users have to consider many key factors while building species distribution models (Merow et al., 2102;
Araújo et al., 2019).
Model tuning and optimal model selection are key issues with Maxent modelling
(Galante et al., 2018) once other issues are considered to build sound models that can be used in conservation decision making (Araújo et al., 2019). While building multiple models using different combinations of regularization multiplier and feature class can help overcome model tuning issue, but the issue of selecting the optimal model from the set of models still remains (Warren & Seifert, 2011; Shcheglovitova & Anderson, 2013;
Muscarella et al., 2014; Radosavljevic & Anderson, 2014; Galante et al., 2018). Out of the two main approaches used for model selection, namely information criteria and
9 withheld data (Galante et al., 2018) the users of latter approach have multiple model evaluation criteria available like the ‘area under the curve of the receiver operating characteristic’ plot for the training data (AUCTRAIN) and for the test data (AUCTEST), the difference between AUCTRAIN and AUCTEST (hereafter ‘AUCDIFF’) and the ‘threshold dependent test omission rate’ (hereafter ‘OR’) (Warren & Siefert 2011; Muscarella et al.,
2014; Radosavljevic & Anderson, 2014; Galante et al., 2018). While AUCTRAIN and
AUCTEST can provide model discriminatory power they cannot evaluate model overfitting. On the other hand, AUCDIFF and OR can evaluate model overfitting (Warren and Siefert, 2011; Radosavljevic & Anderson, 2014) but they may also select over permissive models (Galante et al., 2018). Hence, sequential combinations of either OR and AUCTEST (hereafter ORTEST approach) (Galante et al., 2018) or OR, AUCDIFF and
AUCTEST (hereafter AUCDIFF approach) (Radosavljevic & Anderson, 2014) have been used. But there is need to compare directly the performances between these two model selection approaches, and also to try other ORs based on other thresholding rules to advance the field (Galante et al., 2018).
1.2 National perspective
1.2.1 Study area: Bhutan
Bhutan is a small developing country with total area of 38,394 km2, a projected population of 783,909 for 2018 and a gross domestic product (GDP) of 164,627.92 million ngultrum (Nu.) ($2301.23 million with exchange rate of 1USD = Nu.71.54) for the year 2017 (National Statistical Bureau (NSB), 2019). Geographically it is located in the Eastern Himalaya bounded on the East, West and South by the Indian states of
Arunachal Pradesh, Sikkim, and West Bengal and Assam respectively, and whole of the
North by Tibet Autonomous Region of China (Figure 1.1a).
10
Figure 1-1. Bhutan and its neighbours. Indian state of Sikkim to west, Indian state of West Bengal to south-west, Indian state of Assam to south-east, and Tibet Autonomous Region of China the entire north. (Image adopted from Google Earth @ 2020 Google)
1.2.2 Freshwater ecosystem and biodiversity in Bhutan
Freshwater ecosystems and biodiversity in Bhutan are not well documented
(National Biodiversity Centre (NBC), 2014). The freshwater ecosystems of Bhutan comprise rivers, wetlands, lakes and hot springs (NEC, 2016a). However, there is no comprehensive study of freshwater ecosystems other than recent attempts to map high altitude wetlands (Sherub et al., 2010) and glacial lakes (Mool et al., 2001). There is a clear lack of concerted effort and data sharing among different agencies (NEC, 2016b).
For instance, the NEC (2016b) mandated for overall coordination of integrated water resources management at a national level mentions four major river basins while WMD
(unpublished) mandated for specific watershed management mentions five major basins.
There is no adequate freshwater biodiversity data even for groups like fish
(Gurung, Dorji, Tshering, & Wangyal, 2013) or amphibians (Wangyal & Gurung, 2012;
Wangyal, 2013). However, it is encouraging to note the recent growth in the number of studies on groups of freshwater organisms like fish (e.g. Gurung et al., 2013; Dorji & 11
Wangchuk, 2014; Thoni & Gurung, 2014; Thoni, Gurung, & Mayden, 2016) and odonate
(e.g. Mitra, 2008; Brockhaus & Hartmann, 2009; Mitra et al., 2012; Dorji, 2015;
Gyeltshen, 2016; Kalkman & Gyeltshen, 2016; Gyeltshen, Dorji, Nidup, Dorji, &
Kalkman, 2017). Currently 36 species of amphibians (Wangyal, 2013), 109 fish species
(Thoni et al., 2016) and 101 odonate species (Gyeltshen et al., 2017; Gyeltshen &
Kalkman, 2017) are known from Bhutan. A single paper on the Trichoptera of Bhutan by
Malicky (2007) recorded 165 species and described 27 new species though only watercourses from western and central Bhutan were sampled. Others (e.g. Ofenböck et al., 2010; Giri & Singh, 2013; Dorji, 2016) have used macroinvertebrates at a higher taxonomic level (usually Family) to assess river health. However, the recent additions of numerous new records to freshwater species of Bhutan (Gyeltshen et al., 2017), and the discovery of new fish (e.g. Thoni & Gurung, 2014; Thoni et al., 2016) and odonate species
(Gyeltshen, Kalkman, & Orr, 2017) show that freshwater biodiversity in Bhutan is yet to be fully recorded.
1.2.3 Threats to freshwater biodiversity and ecosystem in Bhutan
Along with the lack of freshwater ecosystem and biodiversity studies there has been no concerted effort on identifying threats to freshwater biodiversity and ecosystems.
However, the NEC (2016a) projected water demands for drinking, industries and others, and irrigation will increase from 36.09, 74.39 and 666.9 MCM/yr in 2015 to 77.68, 218.35 and 911.8 MCM/yr by 2030 respectively. There is also increased pressure from waste generation in urban centres as exemplified by lower water quality in the Wangchhu River along vegetable market area over sampling sites away from core Thimphu city area (Giri
& Singh, 2013; NEC, 2016a). As well as these localised impacts hydropower development in Bhutan has been accelerating in recent times (Gurung et al., 2013) and the government considers it the main future development avenue (RGoB, 2016). Also,
12 climate change may compound the effect from the above factors as already glaciers in
Bhutan are melting at high rates (NEC, 2016a) and already some water sources have dried
(NEC, 2016a, 2016b). The threat to freshwater biodiversity can also come from the introduction of alien species, and already alien fishes have been recorded from water bodies in Bhutan (Gurung et al., 2013).
1.2.4 PAs of Bhutan and Gross National Happiness
Bhutan declared the then Manas Wildlife Sanctuary as the first PA in 1966
(Seeland, 1998; 2000; NBC, 2014; DoFPS, 2015). Since then the PA system of Bhutan has changed with major additions in 1974 and 1984, and an overhaul of the PA system in
1993 (Seeland, 1998; 2000). As of 2016 it is officially claimed that Bhutan has five national parks, four wildlife sanctuaries, one strict nature reserve, one botanical park and seven biological corridors covering 51.44% of the country’s total terrestrial area (Wildlife
Conservation Division (WCD), 2016). Bhutan has also declared three Ramsar sites, viz.,
Kotokha, Bumdeling (http://www.ramsar.org/news/bhutans-first-two-ramsar-sites) and
Phobji-Gangtey (http://www.ramsar.org/news/bhutans-third-ramsar-site) (Figure 1.2).
13
Figure 1-2 Existing protected areas of Bhutan along with three Ramsar sites (each PA with unique colour). BWS: Bumdeling Wildlife Sanctuary, JWS: Jomotshangkha Wildlife Sanctuary, PWS: Phibsoo Wildlife Sanctuary, SWS: Sakteng Wildlife Sanctuary, JDNP: Jigme Dorji National Park; JSWNP: Jigme Singye Wangchuck National Park; PNP: Phrumsengla national park, RMNP: Royal Manas National park, WCNP: Wangchuck Centennial National park, JKSNR: Jigme Khesar Strict Nature Reserve; NC: North Corridor; JKSNR-JDNP biological corridor, JDNP-JSWNP biological corridor, PWS-JSWNP-RMNP biological corridor, RMNP-JSWNP- PNP biological corridor, PNP-BWS biological corridor, RMNP-JWS biological corridor, JWS- SWS biological corridor, BR: Bumdeling Ramsar site, KR: Khotokha Ramsar site, GPR: Gangtey-Phobji Ramsar site.
The high percentage of PA coverage can be attributed to a generally pro- environmental stance of Bhutan government (Zurick, 2006; NBC, 2014; Buckley, 2017) as a result of the unique development philosophy known as ‘Gross National Happiness’
(GNH) (Brooks, 2013; Hayden, 2015). However, a recent study (Mascia et al., 2014) found PAs in Bhutan were impacted by PADDD. Given the area of PADDD events is difficult to track, especially in remote areas like Bhutan (Mascia & Pailler, 2010), it is likely the study did not detect all the PADDD events in Bhutan. Besides, studying both gain and lost events would help better understand the PA dynamism (Cook et al., 2017).
There has been no comprehensive review of Bhutan’s PA dynamism despite 50 years of modern conservation history being celebrated in 2016. Knowing both the causes of
PADDD (lost events) and gain events may help design robust PA systems (Symes et al.,
2016; Golden Kroner et al., 2019). Further, there has been no study that assessed any 14 relationship between PA dynamism and tenets of GNH despite the latter being considered the factor in the conservation success story of Bhutan.
1.2.5 Freshwater ecosystem and biodiversity conservation in Bhutan
The protected areas of Bhutan are terrestrial-focused as they were designated to equitably represent different terrestrial ecosystems of Bhutan (DoFPS, 2015). There are no freshwater protected areas in Bhutan besides three Ramsar sites that could be considered equivalent to FPAs (Pittock et al., 2015). These is despite the fact that
Bhutan’s freshwater ecosystems and biodiversity are facing threats that are likely to increase. Despite these known threats, the Bhutan Government plans to designate just a single stretch of river that links to a protected area, as free-flowing by year 2022 (Bhutan
For Life, n.d). Further, Bhutan Government plans to manage a critical watershed each within 10 protected areas by 2023 (Bhutan For Life, n.d). Therefore, there is a need to assess freshwater biodiversity and the ecosystem conservation gap within the existing PA system and also to conduct a systematic freshwater conservation exercise to determine how well freshwater priority areas are represented within existing PAs? The findings may help steer policy change or assist conservationists in identifying appropriate areas for freshwater conservation. But, as mentioned earlier freshwater biodiversity studies are at an early stage aside from a recent increase in fish and odonate studies. Therefore, in order to assist in the gap analysis and conservation planning in Bhutan there is need for species distribution modelling.
1.3 Thesis structure
The overall objective of this thesis is to assess the potential of existing terrestrial- focussed protected areas of Bhutan in conserving freshwater ecosystems and biodiversity within the context of nature needs half target and protected area dynamism.
15
This thesis is comprised of six chapters, with each data chapter formatted as a manuscript. Chapter 1 (this chapter) is an introduction which presents a literature review on global and national level freshwater biodiversity, threats to biodiversity and ecosystems, protected areas as response to the threat, potential of terrestrial-focussed PAs in conserving freshwater biodiversity and ecosystems, need for systematic conservation planning for freshwater, lack of adequate data for conservation planning, and how recent availability of global level data and modelling tools can help meet the data gap. This chapter is followed by four independent research studies, Chapters 2, 3, 4 and 5. Each of these chapters addresses main issues identified in the literature review. The issues are: (1) lack of PA dynamism history and proximate causes of PADDD in Bhutan and relationship between PA dynamism and GNH; (2) freshwater species data gap and key issues while modelling for species distribution using popular SDM tool Maxent; (3) need to assess potential of existing terrestrial-focussed PAs of Bhutan in protecting freshwater biodiversity and ecosystem against NNH target; (4) need for systematic freshwater conservation planning to withstand PADDD and assess how well the existing PAs protect the identified freshwater priority areas.
Chapter 2 presents fifty years of protected area dynamism in Bhutan and how the changes aligned with tenets of GNH;
Chapter 3 deals with difficulty in selecting optimal SDMs and compares different options;
Chapter 4 assess potential of existing PA system in protecting freshwater biodiversity and ecosystems of Bhutan against progressive conservation targets from below Aichi to
NNH;
16
Chapter 5 conducts systematic conservation planning for freshwater species and asses effect on optimal reserve selection when main proximate cause for enacted and proposed
PADDD the hydropower project is accounted during planning;
Chapter 6 presents a general discussion that summarizes the achievements of this thesis and directions for future research.
17
2.1 Abstract
Bhutan is recognized for conservation success under its pro-environmental development philosophy Gross National Happiness (GNH). However, an increase in area coverage alone cannot track the true contribution of protected areas (PAs) to biodiversity conservation. Capturing PA dynamism by tracking PADDD (PA downgrading, downsizing and degazettement) and gain events can be used as a more comprehensive evaluation method. Based on existing data, we tracked gain events, enacted and proposed PADDD events, proximate causes of PADDD, and gain and PADDD events trends in Bhutan from 1966 to 2016. We also compared PADDD events recorded in www.PADDDtracker.org with the primary data sourced for our study. We discussed the findings in light of four tenets of GNH: good governance, sustainable socio-economic development, preservation and promotion of culture, and environmental conservation. We identified 81 gain and 29 PADDD events. All 12 proposed policy downgrading events in 2004 were proximally caused by infrastructure development, while all degazettement (n=6) in 1993 and all downsizing events (2 in 1984 and 1 in 1993) were proximally caused by conservation planning. Overall the gain and PADDD events were episodic but policy downgrading events occurred only from 2002. Based on our country data, we recorded 4.8 times more PADDD events than was recorded in www.PADDDtracker.org where even some degazetted PAs were not recorded. All gain events and even PADDD events, excluding one enacted and 12 proposed policy downgrading events caused by hydropower projects, were aimed to improve conservation thus aligning with tenets of GNH. However, as hydropower is the specific cause for all proposed PADDD, Bhutan should be concerned about PADDD. Our findings provide further evidence for the dynamic nature of PAs, the widespread nature of PADDD and difficulty in detecting it. Further they also suggest the need to conduct archival case studies to better detect PADDD and PA dynamism.
Keywords: Gain events, Hydropower, Pro-environmental development philosophy, Protected area downgrading, downsizing and degazettement, Proximate causes
19
2.2 Introduction
Bhutan recently celebrated half a century of modern conservation since the establishment of its first protected area (PA), the Manas Wildlife Sanctuary (Seeland,
1998). Since 1966, Bhutan has seen the designation of additional PAs in 1974 and 1984, an overhaul of the PA system in 1993 (Seeland, 1998), designation of 12 biological corridors in 1999, designation of Wangchuck Centennial National Park in 2008 (Wildlife
Conservation Division (WCD), 2010), and declaration of three Ramsar sites (wetlands of international importance) since 2012 (Palden, 2016). Since then Bhutan has protected more than 50% of its land (Dudley, 2016; Buckley, 2017). However, the exact figure differs with official sources reporting 51.44% excluding Ramsar sites (National
Biodiversity Centre (NBC), 2014; WCD, 2016), while the United Nations Environment
Programme-World Conservation Monitoring Centre (UNEP-WCMC) (2018) reports total
PA coverage as 48.01%. Either way, this is a considerable area of the country and conservation success is ascribed to a generally pro-environmental stance of the government (Zurick, 2006; NBC, 2014; Buckley, 2017). This results from the unique development philosophy of Bhutan, known as ‘Gross National Happiness’ (GNH)
(Brooks, 2013; Hayden, 2015).
GNH as a concept originated in the 1970’s with the Fourth King of Bhutan stating:
‘Gross National Happiness is more important than Gross National Product (GNP)’
(Hayden, 2015). The Bhutan Government now uses an index based on GNH as an indicator of development success (Ura et al., 2012; Hayden, 2015), it comprises of four pillars: (i) sustainable and equitable socio-economic development, (ii) environmental conservation, (iii) preservation and promotion of culture, and (iv) good governance (Ura et al., 2012). These are further divided into nine domains: psychological wellbeing, health, education, cultural diversity and resilience, time use, good governance, community vitality, living standard, and ecological diversity and resilience (Ura et al., 20
2012). However, despite the inclusion of the environment in both the ‘pillars’ and the
‘domains’, pressure on Bhutan’s environment has been increasing due to population growth and economic development (National Environment Commission (NEC) 2016) both inside and outside of protected areas. As an example, feasibility studies for hydropower development have been undertaken in biological corridors and PAs (WCD,
2010). Further, the increase in PA coverage alone as an indicator of conservation success is questionable; biodiversity worldwide has continued to decline, despite a gross increase in PA coverage (Cook et al., 2017; Barnes et al., 2018). Therefore, to track true progress we need to not only record both increases in PA number and area, but also PA dynamism
– the level of change in both area and protection level in the PA system (Mascia & Pailler,
2011; Cook et al., 2017). Given the increasing pressure on the environment, and therefore the PA’s in Bhutan, it is timely to review the history of PA’s with respect to the history of PA dynamism.
The dynamism of protected areas can be tracked through the study of Protected
Area Downgrading, Downsizing and Degazettement (otherwise known as PADDD)
(Mascia & Pailler, 2011; Mascia et al., 2012; Cook et al., 2017). Downgrading is the
‘legal or policy change that permits an increase in the type, magnitude or extent of human activities within a PA’, downsizing is the ‘loss of legal protection for a portion of a PA as a result of its excision from the PA’, and degazettement is the ‘loss of legal protection for an entire PA’ (Mascia & Pailler, 2011; Mascia et al., 2012). PADDD is a widespread conservation phenomenon spanning almost the whole modern PA history (Mascia &
Pailler, 2011); this history, however, sees a range of very different trends (see Bernard et al., 2014; Mascia et al., 2014; Pack et al., 2016; Cook et al., 2017). Interestingly, PADDD often occurs against a background of increasing PA coverage (Cook et al., 2017) across a range of settings; in countries with a good history of conservation policy like Brazil
(Bernard et al., 2014; Ferreira, 2014), both developed and developing countries (Mascia 21 et al., 2014; Cook et al., 2017) and in iconic PAs like Yosemite National Park (Golden
Kroner et al., 2016). PADDD occurs due to different proximate causes (Mascia et al.,
2014) and risk factors like PA size, local population density (Symes et al., 2016) and ineffective PAs (Tesfaw et al., 2018). PADDD occurrences are often either not reported transparently (Pack et al., 2016) or underreported (Symes et al., 2016; Cook et al., 2017).
However, understanding the causes of PADDD (Mascia & Pailler, 2011) will help design robust PA systems against PADDD, or more specifically those PADDD events that are detrimental to conservation (Mascia & Pailler, 2011). Though literature covering various aspects of PADDD (see Cook et al., 2017; Lewis et al., 2017; World Wildlife Fund
(WWF), 2017) along with media coverage (WWF 2017) is increasing, there is a growing recognition of the need to study PADDD and its proximate causes to understand PA dynamism and conservation implications. PADDD in the PAs of Bhutan has been explored in the context of a multi-regional study (see Mascia et al., 2014). Here, we build on this to demonstrate the difficulty in tracking PADDD events in remote areas, coupled with the lack of transparent reporting, based on a case study in Bhutan, a country with a strong history of PA designation.
Focussing on PADDD alone captures only one element of PA dynamism – the loss of protection. In order to truly capture holistic PA dynamism studies must also include corresponding gains in protection (Cook et al., 2017). Therefore, in this paper we:
(i) study the PA dynamism sensu Cook et al. (2017) of Bhutan’s modern PA system spanning over half a century from 1966 to 2016, (ii) record and discuss temporal trends, and the trends among individual PAs of both ‘gain’ and ‘PADDD’ events, (iii) record proximate and specific causes of PADDD and their trend, and (iv) compare PADDD events recorded by us against those recorded in www.PADDDtracker.org (WWF, 2017).
22
In recognition of Bhutan being a country with a unique development philosophy of GNH, we discuss PA dynamism and its causes in light of the GNH philosophy. We hope our case study will highlight the need to conduct similar PA dynamism case studies, not only in those countries yet to meet the Aichi target 11 (Woodley et al., 2012) of placing 17% terrestrial and inland waters under protected area but also in other countries that have already met the target to keep track of PA dynamism.
2.3 Methods
2.3.1 Study system
Reported data on the current composition of Bhutan’s PA system, in terms of total number, total area and percentage coverage, differs between sources (NBC, 2014; WCD,
2016; DoFPS, 2015; UNEP-WCMC, 2018); although all sources agree on five national parks, four wildlife sanctuaries and one strict nature reserve. Differences occur within the number of biological corridors (e.g. 7 by WCD, 2016; 8 by NBC, 2014; DoFPS, 2015 &
UNEP-WCMC, 2018), inclusion (e.g. by NBC, 2014; WCD, 2016) or exclusion (e.g. by
DoFPS, 2015; UNEP-WCMC, 2018) of the Royal Botanical Park, inclusion (e.g. by
UNEP-WCMC, 2018) or exclusion (e.g. by NBC, 2014; DoFPS, 2015; WCD, 2016) of
Ramsar sites among the PA system, and area calculation. Since the Royal Botanical Park and Bumdeling Ramsar sites are located within the Jigme Dorji National Park-Jigme
Singye Wangchuck National Park Biological Corridor (NBC, 2009) and Bumdeling
Wildlife Sanctuary (Ramsar 2012) respectively, we considered them multiple designations and excluded them from the area calculation and PA number. Collating these sources we considered 20 PAs with a total area of 19,196.51 km2 (Supporting Information
Table S2.3) covering 50% (exact value being 49.999%) of the country’s total land area of 38,394 km2 (NBC, 2014) in 2016 (Figure 2.1).
23
Figure 2-1 Protected area systems of Bhutan in (a) 1984 (legend shows PAs degazetted or downsized over the study period along with other PAs), (b) 1993, (c) 1999, (d) 2008 and (e) 2016. (Figure 2.1 a-c not to scale; Figure 2.1 a digitized after Thinley et al. 2013; Figure 2.1 b after Seeland (1998); Figure 2.1 c after NCD (2004); Figure 2.1 d & 1e redrawn from shapefiles obtained from Watershed Management Division, Department of Forest and Park Services, Ministry of Agriculture and Forest). (BC: Biological Corridor; BWS: Bumdeling Wildlife Sanctuary; DNP: Doga National Park; DWR: Dungsam Wildlife Reserve; GPR: Gangtey-Phobji Ramsar Site; JDNP: Jigme Dorji National Park; JDWS: Jigme Dorji Wildlife Sanctuary; JKSNR: Jigme Khesar Strict Nature Reserve; JSWNP: Jigme Singye Wangchuck National Park; JWS: Jomotshangkha Wildlife Sanctuary; KR: Khotokha Ramsar Site; KRF: Khaling Reserve Forest; KWS: Khaling Wildlife Sanctuary; MWS: Manas Wildlife Sanctuary; NaWR: Namgyel Wangchuck Reserve; NWR: Neoli Wildlife Reserve; PNP: Phrumsengla National Park; PoRF: Pochu Reserved Forest; PRF: Phibsoo Reserve Forest; PWS: Phibsoo Wildlife Sanctuary; RMNP: Royal Manas National Park; SRF: Sinchula Reserve Forest; SWR: Shumar Wildlife Reserve; SWS: Sakteng Wildlife Sanctuary; TNP: Thrumshingla National Park; TSNR: Toorsa Strict Nature Reserve; WCNP: Wangchuck Centennial National Park; ZRF: Zoshing Reserve Forest.)
24
2.3.2 Gain and PADDD events, count and area affected
We identified PADDD events after Mascia & Pailler (2011), Mascia et al. (2012),
Cook et al. (2017) and WWF (2017). We used the decision tree by Mascia et al. (2012) to consistently identify PADDD events and types and followed Cook et al. (2017) for identifying gain events. However, unlike Cook et al. (2017), when defining new PAs we considered the entire area of newly gazetted/designated PAs as new additions to the PA system. This was done despite some PAs incorporating a portion of already protected land, or including a portion of land from degazetted or downsized PAs when newly designated. For instance, Wangchuck Centennial National Park (WCNP) not only contains part of three biological corridors but also includes portions of downsized former
Jigme Dorji Wildlife Sanctuary (WCD, 2010). If we were to follow Cook et al. (2017) the WCNP’s area would need to be categorized into ‘new area’ and ‘reversed PADDDed area’, however, for this study we considered the whole area of WCNP as a ‘new area’ added to the PA system. We did this because spatially explicit data on the PA system of
Bhutan spanning the entire study period was not available.
To describe systemic change or policy change that affected multiple PAs, we considered the total number of PAs that existed in the year of a specific policy change as the total number of PADDD or gain events. When the specific affected region or locality within the PA was not specified, we considered the whole area of a PA as the affected area. When the policy change specified the region or locality within a PA, but did not provide the area of the region or locality, we considered the area affected ‘unknown’. For instance, we considered the area affected by the policy downgrade that allowed harvesting of the Chinese caterpillar fungus (Ophiocordyceps sinensis) - a high value fungal parasite of caterpillars belonging to the ghost moth genus Thitarodes (Cannon et al., 2009) - as unknown. The elevation above which Chinese caterpillar fungus is found was provided,
25 however, the area of the species range falling within PAs was not given in the literature
(see Cannon et al., 2009; Wangchuk et al., 2012). We considered the area affected by enlarged and downsized events as the difference in area before and after the event when sources did not mention the area affected.
2.3.3 Proximate causes and count
We followed the definition of Mascia et al. (2012; 2014) and the details provided by WWF (2017) for the proximate cause of PADDD events. Mascia et al. (2014) identified and defined 16 proximate causes: forestry, mining, oil and gas, industrial agriculture, industrialization, infrastructure, land claims, rural settlement, subsistence, degradation, shifting sovereignty, refugee accommodation, conservation planning, other, unknown and multiple. For each affected PA we counted the number of events caused by a particular proximate cause (Supporting Material, Table S2.1).
2.3.4 Data collection
Following the definitions and operational criteria described above for gain and
PADDD events, and proximate causes we created a comprehensive database through reviewing peer reviewed articles, government documents, PA management plans, media reports and social media between 1966 and 2016. We mainly sourced the documents from websites through Google search engines using various combinations of the search terms
Bhutan, protected area, reserve, national park, biological corridor, Ramsar sites, botanic garden, management, plan, reviews, legalization, policy, acts, regulations, fishing, forestry, history, PADDD, Cordyceps, road, NWFP, collection, designation, removed, gazettement (Bernard et al., 2014), as well as sourcing relevant documents from government organizations where possible. We modelled our PA dynamism record format after the PADDD event record format used in www.PADDDtracker.org (WWF, 2017).
26
We recorded separately details of gain events (Table S2.2) and PADDD events (Table
S2.1).
2.3.5 Data analyses
Using both the median and range of the area affected, we summarized the number of gain and PADDD events, number of separate PAs affected, total area affected and percentage affected. We also calculated the percentage share of individual gain and
PADDD event types over total gain and PADDD events. We summarized types, area impacted and number of gain and PADDD events and status of PADDD events for each year and PA. Additionally, we summarized the number and percentage of proposed and enacted PADDD events, proximate causes of PADDD and types of PADDD events. We highlighted the difference in the number of PADDD events and the area affected in our data compared with that recorded in www.PADDDtracker.org (WWF, 2017) by dividing the corresponding number of PADDD events and area affected recorded in the former by the latter.
2.4 Results
2.4.1 Summary of gain and PADDD events
We recorded a total of 110 gain events which affected 52,847.96 km2 of area excluding the unknown area affected by three policy upgrade events (Table 2.1).The actual area added to the PA system by gain events was only 23,710.30 km2, this was calculated by combining the area added through new (21,964.38 km2) and enlarged PAs
(1746 km2) with the further 24,214.58 km2 in total gain events that were only upgraded to a higher level of protection. Gazettement of new PAs was the most prevalent gain event by number of PAs affected (n=35) and policy upgrade the most prevalent by total area affected (24214.6 km2) (Table 2.1). However, when only the number of gain events was
27 considered then policy upgrade was the most prevalent (n=36, 44.4%) and class upgrade the least prevalent (n=2, 2.5%) (Table 2.1; Supporting Information, Table S2.4).
Table 2-1 Summary of changes to protected area system of Bhutan between 1966 and 2016.
Gain Events PADDDa events
Gazettement Degazettement
No. of events 35 6
No. of PAsb affected 35 6
Total area affected, km2 21964.4 585
Median area affected, km2 (range) 233(1.14-3736) 110(5-180)
Enlargement Downsizing
No. of events 8 3
No. of PAs affected 6 2
Total area affected, km2 1746 3808
Median area affected, km2 (range) 162(20-560) 103(92-3613)
Upgrading Downgrading
Class Policyc Class Policyd
No. of events 2 36 - 20e
No. of PAs affected 2 31 - 7
Total area affected, km2 4923 24214.6 -
Median area affected, km2 (range) 2461.5(1023-3900) 222.5(0-7813) -
Total number of events 81 29
(aProtected Area Downgrading, Downsizing and Degazettement; bProtected areas; cArea affected unknown for 3 policy upgrade events; dArea affected by all policy downgrade events unknown; e14 were proposed and 6 enacted policy downgrade events.)
We recorded 29 PADDD events which removed a total area of 4,393 km2 from the PA system (Table 2.1), excluding areas that were, or may be, subjected to higher disturbance by enacted and proposed policy downgrade events respectively (Table 2.1).
Among PADDD events policy downgrade was the most prevalent by number of events
28
(n=7) and downsizing by area lost (3808 km2) (Table 2.1; Table S2.4). Four policy downgrade events were systemic: two of these resulted from legalization of the collection of Chinese caterpillar fungus and the other two allowed for an increase in the number of people allowed per household to collect the fungus (Appendix 1). Of the total area lost due to PADDD, downsizing removed 86.7% (area = 3,808 km2; n=3) with the remaining
13.3 % (area = 585 km2; n=6) lost to degazettement (Table 2.1; Table S2.4). However, some area lost due to downsizing of Jigme Dorji Wildlife Sanctuary was reversed through the designation of the Bumdeling Wildlife Sanctuary, Wangchuck Centennial National
Park and North Corridor, while some area lost due to the degazettement of Shumar
Wildlife Reserve and Dungsam Wildlife Reserve was reversed through the designation of Royal Manas National Park-Jomotshangkha Wildlife Sanctuary biological corridor
(Appendix 1; Figure 2.1).
29
Table 2-2 Protected areas of Bhutan affected by gain events between 1966 and 2016 with gain event types (enlarged PA and policy upgrade), current, primary and after gain event names of protected areas, year of event and area affected.
Name Name Name (Current)a* (primary)b (after)c Typed Yeare Area(km2) Enlarge 2001 187 BWS BWS Policy 2006 1521 Policy 2009 0 Enlarge 1995 300 Policy 2006 4349 JDNP Enlarge 2008 116 JDNP Policy 2009 0 Policy 1992 7813 JDWS JDNP Class 1993 3900 JKSNR TSNR Policy 2006 650.74 Enlarge 2003 323 JSWNP JSWNP Policy 2006 1400 KRF Policy 1992 233 JWS NWR Policy 1992 40 K/NWS Policy 2006 273 Policy 1992 175 PRF PWS PWS Enlarge 1993 103 PWS Policy 2006 278 Enlarge 2006 137 PNP TNP Policy 2006 905 Enlarge 1984 20 Policy 1992 463 Manas WS RMNP RMNP Class 1993 1023 Enlarge 1993 560 RMNP Policy 2006 1022.84 SWS SWS Policy 2006 650 WCNP WCNP Policy 2009 0 JDNP-JSWNP C JDNP-JSWNP C Policy 2006 275 JKSNR-JDNP C TSNR-JDNP C Policy 2006 147 JSWNP-JDNP C Policy 2006 600 JSWNP-NC Policy 2006 525 JSWNP-WCNP-JDNP- BWS/NC NC Policy 2006 640 BWS-NC Policy 2006 119 TNP-NC Policy 2006 142 JWS-SWS C KWS-SWS C Policy 2006 376 PNP-BWS C TNP-BWS C Policy 2006 79 PWS-JSWNP-RMNP C PWS-RMNP C Policy 2006 160 RMNP-JSWNP-PNP C JSWNP-TNP C Policy 2006 385 RMNP-JWS C RMNP-KWS C Policy 2006 212 DNP Policy 1992 21 DWR Policy 1992 180 30
NWGR Policy 1992 195 PRFf Policy 1992 140 SWR Policy 1992 160 SRF Policy 1992 80 ZRF Policy 1992 5
(aCurrent name of protected area if it exists as in 2016; *BWS: Bumdeling Wildlife Sanctuary, C: corridor, DNP: Doga National Park, DWR: Dungsam Wildlife Reserve, JDNP: Jigme Dorji National Park, JDWS: Jigme Dorji Wildlife Sanctuary, JKSNR: Jigme Khesar Strict Nature reserve, JSWNP: Jigme Singye Wangchuck National Park, JWS: Jomotshangkha Wildlife Sanctuary, KRF: Khaling Reserve Forest, K/NWR: Khaling/Neoli Wildlife Reserve, MRF: Mochu Reserve Forest, NC: North Corridor, NWGR: Namgyel Wangchuck Game Reserve, NWR: Neoli Wildlife Reserve, PRF: Phibsoo Reserve Forest, PWS: Phibsoo Wildlife Sanctuary, PNP: Phrumsengla National Park, SWS: Sakteng Wildlife Sanctuary, SWR: Shumar Wildlife Reserve, SRF: Sinchula Reserve Forest, TSNR: Toorsa Strict Nature Reserve, ZRF: Zoshing Reserve Forest; bPrimary name of protected area if it differed from current name or if it does not exist today; cName of protected area after gain event if it changed at the time of gain event; dGain event type excludes gazettement of new protected area (Enlarge: PA enlarge, Policy: Policy upgrade, Class: Class upgrade); eYear of gain event; fPachu Reserve Forest.)
2.4.2 Enacted versus proposed PADDD events
Out of 29 PADDD events, 17 (58.6%) were enacted against 12 (41.4%) proposed
(Table S2.5). All degazettement (n=6) and downsizing (n=3) events were enacted, while only 8 (40%) of the policy downgrading events were enacted (Table S2.5). All the 12 proposed PADDD events were policy downgrading events which made up 60% of the total policy downgrading events (Table S2.5).
2.4.3 Gain and PADDD events temporal trend
Overall, gain events occurred sporadically with almost decadal gaps between events from 1966 to 1992; however, since 1992 gain events have increased, with a maximum gap of four years between 2012 and 2016 (Figure 2.2 a; Table S2.5). Among the gain events, both types, new PA and PA enlargement, were more spread out. Policy upgrades occurred only in 1992, 2006 and 2009, with both class upgrade events occurring in 1993 (Figure 2a; Table S2.5). The highest number of gain events was recorded in the year 2006 (n=22; 27.2%) (Figure 2.2 a; Table S2.6), with 21 of these being policy upgrade events and only 1 being a PA enlargement event, contributing just 0.58% (137 km2) of the total actual area added to the PA system by gain events between 1966 and 2016 31
(Figure 2.2 b; Table S2.6). The highest number of new PAs was added in 1999 (34.29%) when 12 biological corridors were added, while the highest total area added through new
PA designation was in 1974 (8,772 km2; 39.94%) (Figure 2.2 a; Table S2.7).
Figure 2-2 Gain and PADDD events across PA system of Bhutan between 1966 and 2016 (a) number of gain events; (b) area gained from new protected area, enlarged PA and total; (c) number of enacted and proposed PADDD events and (d) area lost to downsizing, degazettement and total.
The highest number of PADDD events was recorded in 2004 (n=14; 48.3%)
(Figure 2.2 c; Table S2.6); however, these were all policy downgrade events (Figure 2.2 c; Table S2.6), two were enacted and the remaining 12 were proposed (Table 2.2). All degazettement events occurred in 1993 (n=6) and all downsizing events in 1984 (n=2) and
1993 (n=1), policy downgrade events occurred almost every two years from 2002 (Figure
32
2.2 b; Table S2.5). Despite this range of PADDD events, an actual area lost from the overall PA system only occurred in two years, 1984 and 1993, due to degazettement and downsizing (Figure 2.2 d; Table S2.7).
2.4.4 Gain and PADDD events trend among individual PAs
Many protected areas were affected multiple times by both gain and PADDD events during the period covered by this study. Out of a total of 34 PAs that gained area, seven gained multiple times from PA enlargement, policy and class upgrades (Table 2.2).
When only currently existing PAs were considered 44.44% (n=18) gained multiple times
(Table 2.2). Among currently existing PAs Jigme Dorji National Park (JDNP) gained the maximum by number of events (n=6), while the highest total area (580 km2) was added to Royal Manas National Park through two PA enlargement events (Table 2.2).
Three PAs were subjected multiples times to PADDD events. JDNP recorded the highest number of PADDD events (n=8, Table 2.3) and also lost the highest total area
(3,716 km2) through two downsizing events (Table 2.3). However, of the six remaining
PADDD events JDNP recorded, three were proposed PADDD events while the other three were policy downgrade events, and hence no actual area was lost.
2.4.5 Proximate causes of PADDD
Only three types of proximate causes were responsible for all 29 PADDD events recorded during the period covered by this study. Infrastructure development was the most frequent proximate cause (n=14, 48.3%) followed by conservation planning (n=9,
31%) and subsistence (n=6, 21.4%) (Table S2.8). However, all degazette and downsize events were caused by conservation planning, while 70% (n=14) of policy downgrade was caused by infrastructure with the remaining 30% (n=6) by subsistence (Table S2.8).
The specific cause for all infrastructure developments was hydropower projects, while
33 specific causes for subsistence were fishery and NWFP (non-wood forest product) collection rights (Table 2.3). Conservation planning was recorded more frequently until
1993, while in the latter part of the study period more of subsistence and infrastructure development events were recorded (Figure 2.3; Table 2.3). Both subsistence and conservation planning were responsible for only enacted PADDD events, but infrastructure development was responsible for only one enacted PADDD while it caused all 12 proposed PADDD events (Table 2.3).
Table 2-3 Protected areas of Bhutan affected by PADDD events between 1966 and 2016 with PADDD events type, area lost, status, year of event, and proximate and specific causes.
PA Namea* Type Areab Status Year Proximate causec Specific caused Enacted Proposed BWS Downgrading - 0 2 2004 Infrastructure Hydropower - 1 0 2004 Subsistence NWFP - 1 0 2008 Subsistence NWFP Conservation 20 1 0 DNP Degazettement 1993 planning Conservation 180 1 0 DWR Degazettement 1993 planning JDNP Downgrading - 1 0 2002 Subsistence NWFP 1 0 2004 Subsistence NWFP - 0 3 2004 Infrastructure Hydropower - 1 0 2008 Subsistence NWFP Conservation 103 1 0 Downsizing 1984 planning Conservation 3613 1 0 1993 planning JSWNP Downgrading - 1 0 2010 Subsistence Fishery - 1 0 2012 Subsistence Fishery - 1 0 2014 Infrastructure Hydropower JSWNP- - 0 1 JDNP C Downgrading 2004 Infrastructure Hydropower Conservation 92 1 0 MRF Downsizing 1984 planning Conservation 20 1 0 PRF Degazettement 1993 planning PWS Downgrading - 0 1 2004 Infrastructure Hydropower PNP Downgrading - 0 3 2004 Infrastructure Hydropower SWS Downgrading - 0 2 2004 Infrastructure Hydropower Conservation 160 1 0 SWR Degazettement 1993 planning Conservation 80 1 0 SRF Degazettement 1993 planning Conservation 5 1 0 ZRF Degazettement 1993 planning
34
Table 2.3 continued: (aCurrent name of PA used if it exists as in 2016; *BWS: Bumdeling Wildlife Sanctuary, DNP: Doga National Park, DWR: Dungsam Wildlife Reserve, JDNP: Jigme Dorji National Park, JSWNP: Jigme Singye Wangchuck National Park, JSWNP-JDNP C: Jigme Singye Wangchuck National Park- Jigme Dorji National Park Corridor, MRF: Mochu Reserve Forest, PRF: Pachu Reserve Forest, PWS: Phibsoo Wildlife Sanctuary, PNP: Phrumsengla National Park, SWS: Sakteng Wildlife Sanctuary, SWR: Shumar Wildlife Reserve, SRF: Sinchula Reserve Forest, ZRF: Zoshing Reserve Forest; bArea affected by downgrade not known; cProximate: Proximate cause, conservation: conservation planning; dSpecific: specific cause, NWFP: Non-wood forest product.)
Figure 2-3 Number of proposed and enacted PADDD events due to different proximate causes across PA system of Bhutan between 1966 and 2016.
2.4.6 PADDD events recorded in www.PADDDtracker.org versus current study
Compared with the PADDD tracker database (www.PADDDtracker.org) our study recorded a higher number of every PADDD event type and area affected, except for the area lost due to downsizing (Table 2.4). We also found differences in specific PAs subjected to PADDD. For example, Mochu Reserve Forest (MRF) and Namgyel
Wangchuck Reserve (NWR) were wrongly recorded as degazetted in the database, while
35 the degazetted Zoshing Reserve Forest and Sinchula Reserve Forest were not recorded at all in the database (Table S2.9). By number of events we recorded 20, 3 and 1.5 times more downgrading, downsizing and degazettement events respectively than recorded in www.PADDDtracker.org (Table 2.4). When we discounted two wrongly listed degazettement events in www.PADDDtracker.org we had recorded three times more degazettement events than recorded in www.PADDDtracker.org (Table 2.4). When all
PADDD events were added we recorded 4.8 times (including wrong records) and 7.3 times (discounting wrong records) more PADDD events than recorded in www.PADDDtracker.org. However, our data was aligned with www.PADDDtracker.org in the area lost to PADDD, both when assessed against specific PADDD events or in total
(Table 2.4). The difference between the datasets was highest for the area lost to degazettement, we recorded 1.7 times more area when we used corrected data (Table 2.4).
When area lost to PADDD was added we recorded 1 and 1.1 times more area lost respectively using uncorrected and corrected www.PADDDtracker.org data (Table 2.4).
Table 2-4 Difference between PADDD events recorded in www.PADDDtracker.org and current study for PA system of Bhutan.
Current study PADDDtrackera Difference (Number of times)b
Event type No.c Area (km2) No. Area (km2) No. Area (km2)
Downgradingd 20 0 1 0 20 NAe
Downsizing 3 3808 1 3534 3 1.1
Degazettement 6 585 4(2) 670(335) 1.5(3) 0.9(1.7)
Total 29 4393 6(4) 4204(3869) 4.8(7.3) 1.0(1.1)
(aValues within parenthesis ( ) are after discounting 2 wrongly included PAs as degazetted in www.PADDDtracker.org; bValues within parenthesis ( ) are difference (Number of times) derived using corrected values of www.PADDDtracker.org data; cNo.: Number of events; dAll were policy downgrade events and area affected unknown; eNA: Not applicable since area affected unknown.)
36
2.5 Discussion
2.5.1 Overall gain and PADDD events
Our study showed that the PA system of Bhutan was as dynamic as in any other country where PADDD has been studied (Bernard et al., 2014; Mascia et al., 2014; Cook et al., 2017). This is despite Bhutan being recognized for conservation success ascribed to its unique (Brooks 2013; Hayden, 2015) and pro-environmental development philosophy of GNH (Zurick 2006; Buckley, 2017). Inclusion of gain events in this PA dynamism study revealed more dynamism than in either a PADDD study alone or in a study focussed just on an increase in the area covered (Cook et al., 2017). The total area captured by PAs grew in Bhutan during the period of study, this reflected the total area added through new PA and PA enlargement events (23,710.4 km2), these were far greater than the area lost to enacted degazettement and downsizing (4,393 km2) events. This may suggest that the occurrence of PADDD amidst major PA growth (Cook et al., 2017) is a prevalent phenomenon among countries of any size or any particular economy, however this requires further study.
We found that although the total number and area affected by PADDD events was naturally smaller in Bhutan than in bigger countries (e.g. Pack et al., 2016; Cook et al.,
2017) it impacted a higher proportion of PAs. Degazettement of just 6 PAs with a total area of 585 km2 accounted for 15.7% of the total number of PAs designated (n=35) and
2.6% of the total area added to the PA system through gazettement (21,964.38 km2) in
Bhutan, while 108 degazetted PAs with 4,122.3 km2 accounted for only 1.8% and 0.3% by number and area respectively in Australia (Cook et al., 2017). Similarly, removal of
95,768 km2 accounted for only 6% of potential national park in Brazil (Pack et al., 2016).
However, Bhutan has not fallen back on its Aichi target despite only having designated
35 PAs over 50 years. This may reflect the role played by its strong pro-environmental
37 development philosophy of GNH which considers preservation of environment as one of its four pillars (Zurick 2006; Ura et al., 2012). This may also explain why all degazettement and downsize events we recorded in Bhutan were caused by conservation planning (Seeland, 1998). Also, all recorded policy upgrade events (n=36), which contributed to the highest number of gain events in Bhutan were aligned to its GNH philosophy. Thirty-three policy upgrade events were related to an increase in the number of days on which fishing was restricted, a blanket rule applied throughout the country including within PAs (Forest Department 1974; Ministry of Agriculture 1992). This aligned with the ‘preservation and promotion of culture’ pillar of GNH (Ura et al., 2012) by restricting fishing on Bhutanese holy days or months. While, reduction in the number of persons allowed per household to harvest Chinese caterpillar fungus in Jigme Dorji
National Park, Wangchuck Centennial National Park and Bumdeling Wildlife Sanctuary
(Cannon et al. 2009) reflects the ‘environmental preservation’ pillar of GNH (Ura et al.,
2012).
2.5.2 Gain and PADDD events temporal trend
Our study showed that similar to Australia (Cook et al., 2017) both gain and
PADDD events in Bhutan were episodic. While PAs were designated sporadically over almost every decade since the first PA was designated in 1966, PA enlargements were recorded only from 1993. The PA designations in 1974 and 1984 may have been influenced by informal relationships between some officials of WWF(US) and the Bhutan
Government, along with the recognition that PA creation can be a mechanism to protect wildlife and forest produce from poaching and illegal harvesting (Seeland, 1998). The PA system revision in Bhutan in 1993 was the consequence of recommendations from PA system reviews done by the World Conservation Union (IUCN) in 1984, the Food and
Agriculture Organization of the United Nations (FAO) in 1988 and 1989, Bhutan’s
38
Master Plan for Forestry Development Project in 1991 and the World Wildlife Fund
(WWF) in 1993 (NCD, 2004).
The more recent gain events recorded in our study can be argued to be the result of scientific findings; for example, the declaration of biological corridors after surveys of tiger occurrence (NCD, 2004) and the designation of Khotokha and Gangtey-Phobji
Ramsar sites as important winter habitat for the globally vulnerable black- necked crane
(Grus nigricollis) (Ramsar 2012; 2016). The PA enlargement events in 1993 were part of a total PA system overhaul to include better ecosystem representation (Seeland, 1998;
NBC 2009), while latter enlargement events were probably due to the adaptive management of PAs, with the actual management of PAs beginning only from 1995 onwards (NBC, 2009). For example, new areas were added to the Bumdeling Wildlife
Sanctuary, and the then Thrumshingla National Park, to bring additional tiger and snow leopard habitats into the PA system in 2001 and 2006 respectively (NBC, 2009). Also, an increase in the area of Jomotshangkha Wildlife Sanctuary from 337 to 1160 km2 occurred with the inauguration of its independent park management on 5th February, 2017 (Lhamo,
2017) and reflects the government’s strong commitment to environmental conservation and willingness to enlarge PAs. Historical events also led to the designation of PAs, for example the Wangchuck Centennial National Park was designated in 2008 to celebrate
100 Years of Wangchuck Dynasty’s reign (DoFPS, 2015).
Policy downgrades influencing PAs in Bhutan were recorded only after 2002 and occurred almost every two years thereafter until 2014. These affected more PAs in terms of number, and occurred at multiple times, compared with any other PADDD or gain event. Policy downgrade events, 70% of which were due to the proximate cause infrastructure development, with both enacted and proposed hydropower being the specific cause for all of these, may signify a shift in the Bhutanese government’s
39 development policy. A similar shift in policy was recorded in Brazil where hydropower development became a major proximate cause of PADDD (Bernard et al., 2014; Pack et al., 2016). The Bhutanese government also considers hydropower as one of the five jewels of economic development (RGoB 2016). Although the Bhutan government is guided by the overarching philosophy of GNH in the pursuit of economic development, the need for greater transparency with respect to hydropower development plans has been raised
(Premkumar, 2016). All proposed policy downgrades recorded in 2004 were due to hydropower development compared to just one enacted policy downgrade due to hydropower in 2014. Although all the proposed hydropower projects may not actually be implemented (WBG, 2014) there are chances many could be enacted eventually, given
Bhutan is under pressure for continued economic growth coupled with increasing materialistic wealth in Bhutan (Hayden, 2015). Although hydropower projects in Bhutan are run-of-the-river schemes which do not involve large water storage dams and therefore absolute area lost within PAs will be minimal (WBG, 2014), hydropower projects, if enacted, will have adverse impacts on biodiversity conservation.
A further 30% of the policy downgrade events were the result of subsistence, for example, the legalization of Chinese caterpillar fungus harvesting and fishing through the formation of community capture fishery groups. Legalization of both the Chinese caterpillar fungus harvesting and fishing aimed to reduce illegal harvesting and increase the practice of sustainable management of resources, to bring equitable socio-economic development and recognize communities’ traditions (RNR-RDC Bajo, 2011; Thapa
2017). Reflecting GNH philosophy these subsistence based policy downgrades would actually help conserve biodiversity by garnering support from local communities (Mascia
& Pailler, 2011). However, care must be taken as Chinese caterpillar fungus harvesting has not only increased degradation of the alpine environment (Wangchuk et al., 2012), but has also harmed the traditional yak husbandry by rendering it to second place in 40 income generation after that of Chinese caterpillar fungus sale and also by placing the burden of yak herding on female members of the household while traditionally male members herded the yaks (Wangchuk & Wangdi, 2015). In another example, the increased human activities associated with the legalization of fishing in Berti and
Harachhu villages could disturb the habitat of critically endangered White Bellied Heron
(Ardea insignis) in the Berti and Harachhu rivers (Wangdi et al., 2017).
All six degazettements occurred in 1993, while two of the three downsizing events occurred in 1984 and the other one in 1993. Thus, PADDD events which resulted in actual area lost from PAs occurred episodically. However, these events were the result of conservation planning. Designation of additional PAs and changes made to existing PAs for better management led to both recorded downsizing events in 1984 (Seeland, 1998), while a review to ensure adequate ecological representation within the PA system of
Bhutan caused all six degazettement and one downsizing events in 1993 (Seeland, 1998).
The PADDD events of 1993 resulted in a PA system considered to be the most comprehensive by area coverage and representativeness of all ecoregions across the country (NBC, 2009). Thus, PADDD can be used as tool to make the PA system more effective in biodiversity conservation (Fuller et al., 2010). However, when we visually compared degazetted PAs and the current PA system it revealed some PAs were degazetted from the under-represented South-Western region of Bhutan (Figure 2.1).
Further, though some portions of degazetted and downsized PAs were re-designated, or added to existing PAs, they occurred in already well represented regions like northern
Bhutan or South Eastern Bhutan (Figure 2.1) and not where there is a need for adequate representation of biological corridors such as in the South-Western region of Bhutan
(WCD, 2010). Hence, there is need for further studies that employ spatially explicit methods to study PA dynamism in Bhutan to truly assess conservation gain and loss since our findings suggest the PA system in Bhutan will remain dynamic and can be subject to 41 both gain and PADDD events in future as were the PAs in other countries (Mascia et al.,
2014; Pack et al., 2016; Cook et al., 2017).
2.5.3 Gain and PADDD events trend among individual PAs
Many protected areas were affected multiple times by both gain and PADDD events (Cook et al., 2017) with bigger PAs subjected a greater number of times to both gain and PADDD events. For example, Jigme Dorji National Park, which was the largest
PA until the designation of Wangchuck Centennial National Park in 2008 and is till currently the second largest PA (Thinley et al., 2015), was impacted the most frequently by both gain and PADDD events. Jigme Dorji National Park also lost the highest total area to PADDD, due to conservation planning, and was also subjected to three policy downgrade events to allow collection of Chinese caterpillar fungus (Cannon et al., 2009).
Likewise, Jigme Singye Wangchuck National Park, the third largest PA, was the only PA where two local communities were granted legal rights for a community capture fishery
(Jigme Singye Wangchuck National Park, 2013). This supports the earlier finding where larger PAs were more prone to PADDD due to higher opportunity cost (Symes et al.,
2016).
2.5.4 Proximate cause of PADDD
Our study detected only three types of proximate causes for PADDD in Bhutan - infrastructure development, subsistence and conservation planning- against 16 recognized categories (Mascia et al., 2014). A fewer number, and a smaller total area of
PA system in Bhutan, compared to other countries where PA dynamism has been studied might have exposed PAs in Bhutan to fewer types of proximate causes. For instance,
Bhutan has designated only 35 PAs covering only 21,964.4 km2 during 50 years (1966 to
2016) compared with larger countries like Australia which had 7,281 PAs covering 1.22
42 million km2 in 2014 (Cook et al., 2017) and Brazil which had 1,762 conservation units that covered 1.5 million km2 in 2013 (Bernard et al., 2014). Although the country is small with a small number of PAs there is strong support from the Bhutan Government for conservation (Wangchuk et al., 2017) which is reflected and embedded in its legislation; for example, the provision of ‘Article 5 on Environment’ under which the Bhutan
Government is mandated to protect a minimum of 60% of the country under forest cover for all times (The Constitution of the Kingdom of Bhutan, 2008) or maintain a ban on mining within PAs (Department of Forest, 2006). While legal rights accorded to local communities inside or around PAs for traditional landuse in Bhutan (NBC, 2009; 2014;
Wangchuk et al., 2017) may explain absence of rural human settlements as proximate cause which accounted for 20% of PADDD events in Brazil (Pack et al., 2016).
2.5.5 PADDD recorded in www.PADDDtracker.org versus current study
Our study recorded 29 PADDD events against 6 recorded in www.PADDDtracker.org for Bhutan. Further, when corresponding individual PADDD types were compared we recorded very large differences in the number of events although the area impacted by PADDD did not differ that much. For instance, we recorded 20 times more policy downgrade events than recorded in www.PADDDtracker.org, and when we discounted 2 PAs wrongly reported as degazetted in www.PADDDtracker.org we recorded 3 times more degazettement events than that recorded in www.PADDDtracker.org. These differences may reflect additional years included in our study (see Mascia et al., 2014) or they may reflect the difference in identifying an event as a PADDD event. For instance, we did not consider the road constructed through Jigme
Dorji National Park in 2011 as a PADDD event (Thinley et al., 2015) while it was recorded as a proposed downgrade in www.PADDDtracker.org. Road construction is legal within buffer and multiple use zones, and not in core zones, but in the absence of
43 official zonation within PAs there is no legal basis to object to road construction (World
Wildlife Fund Bhutan & Sakteng Wildlife Sanctuary, 2011). However, even degazetted
PAs were not recorded or some PAs were wrongly recorded as degazetted in www.PADDDtracker.org. Probably, the relatively small size and remote location of PAs like Zoshing Reserve Forest (5km2) and Sinchula Reserve Forest (80km2) (Symes et al.,
2016) might have made it difficult to detect and record in the broader database. Also, a lack of transparent records of PADDD events could have contributed to the incorrect record of Mochu Reserve Forest (MRF) and Namgyel Wangchuck Reserve (NWR) as degazetted. Actually, MRF was downsized and a portion of it was renamed Phibsoo
Reserve Forest (Seeland, 1998), while NWR was amalgamated with Manas Wildlife
Sanctuary as Royal Manas National Park (RMNP) (RMNP, 2015). These examples add evidence to a global difficulty in detecting PADDD events and the lack of transparent reporting. Further these differences may suggest the need for specific case studies from individual countries, or the need to include professionals from native countries in multiregional studies (e.g. Mascia et al., 2014). While our study was not designed to be comprehensive archival research, we feel differences in PADDD recorded between studies could also be the result of involving a co-author from the relevant country, this might have resulted in better access to available data sources or key persons from whom to source the data. This may suggest the importance of archival research, where possible, on PA dynamism studies over opportunistic or crowd sourcing methods if PA dynamism in general, and PADDD in particular, are to be better detected. Of course, we feel the current ability of individuals to report or correct PADDD events in www.PADDDtracker.org (WWF, 2017) is a positive development in understanding
PADDD as a conservation phenomenon.
Our study shows inclusion of both gain and PADDD events, even when a spatially explicit method is not used, can reveal higher PA dynamism than a PADDD study alone. 44
Our findings provide further evidence about the episodic, widespread nature of PADDD and the difficulty in detecting it. While gain events were all designed to improve biodiversity conservation, excluding proposed and enacted policy downgrade events caused by infrastructure development with hydropower as its specific cause, all other
PADDD events caused by conservation planning and subsistence were aimed at improving biodiversity conservation in Bhutan. This may reflect Bhutan’s commitment to environmental conservation guided by its GNH philosophy.
45
3.1 Abstract
Maxent is commonly used species distribution modelling (SDM) program due to its better performance over other SDM programs. But model complexity and selecting optimal models are two important concerns while using it. In order to help advance the field we modelled distribution for 10 fish and 28 odonate species of Bhutan with small occurrence data. We built 44 sets of models for each species using combination of 11 regularization multipliers (RMs) and four feature classes (FCs). We then selected optimal models using four sequential optimal model selection approaches: two ORTEST approaches which combined test omission rate (test OR) followed by area under receiver operating curve for test data (test AUC) and two AUCDIFF approaches that combined test OR followed by diference between training AUC and test AUC(AUC difference) and then test AUC. We then selected ecologically plausible optimal model from optimal models selected by the sequential approaches or from remaining models using expert knowledge (EXP approach). We then compared different features and the predicted binary habitat area of the optimal models selected by the five approaches. ORTEST approaches matched better with EXP approach despite them selecting more complex models compared to AUCDIFF approaches. However, models selected through AUCDIFF approaches overpredicted the habitat more often than the models selected through ORTEST approaches compared to models chosen by EXP approach. We recommend use of ORTEST approaches as first line of model screening or by its own when less restrictive thresholds are used to produce binary habitat maps as we did here. First, this would reduce time required to select ecologically plausible models by expert screening when models are built for many species. Second, when used alone, ORTEST approach can prevent from selecting models that restrict predicted habitats around occurrence points or spread across whole study area.
Keywords: Maxent, SDM, sequential model selection approach, test AUC, AUC difference, data poor
47
3.2 Introduction
Species distribution models (SDMs) are used increasingly in different stages of conservation decision making (Guisan, et al. 2013). Among numerous SDM methods, techniques that use presence-only occurrence data have become more popular with readily available georeferenced presence data (for example from GBIF) and environmental variables (Gomez et al., 2018). Among the presence only modelling methods Maxent has become very popular (Merow, Smith, & Silander Jr, 2013; Phillips et al., 2017; Morales, Fernández, & Baca-González, 2017). Maxent identifies the suitable geographic areas for species given the set of environmental variables and known occurrence records by applying a maximum entropy model (Phillips & Dudík, 2008). The increased use of Maxent has been ascribed to its better performance over other methods when used with low occurrence data (Elith et al., 2006; Elith et al., 2011; Coxen, Frey,
Carleton, & Collins, 2017) and also for its ease of use through its graphical user interface
(GUI) (Phillips et al., 2006; Morales et al., 2017, Kass et al., 2018).
The Maxent output can be overfit or under fit to the occurrence localities which results in models that under-predict or overpredict the suitable area respectively
(Shcheglovitova & Anderson, 2013). Maxent output quality depends on model complexity (Shcheglovitova & Anderson, 2013; Syfert, Smith, & Coomes, 2013; Morales et al., 2017; Phillips, 2017), run type (Phillips, 2017), bias in occurrence data (Phillips et al., 2009; Syfert et al., 2013) and its correction methods (Kramer-Schadt et al., 2013), background selection methods (Merow et al., 2013; Vollering, Halvorsen, Auestad, &
Rydgren, 2019), and output types (Phillips, 2017; Phillips et al., 2017). Once other factors are taken into consideration, model complexity becomes the most important issue (Syfert et al., 2013). Model complexity is controlled by the types of feature class (hereafter: ‘FC’) and the value of the regularization multiplier (hereafter: ‘RM’) used (Radosavljevic &
Anderson, 2014; Morales et al., 2017; Phillips, 2017). 48
Maxent feature classes are the predictors - derived from continuous and categorical environmental variables (Phillips et al., 2006). The current version of Maxent v 3.4.1 has linear (L), quadratic (Q), product (P), and hinge (H) features as default FCs and threshold (T) as optional FC (Phillips et al., 2017). All continuous environmental variables are linear features, square of continuous variables the quadratic features, product of two continuous variables the product feature (Phillips et al., 2006), and the hinge and threshold features are piecewise constant splines and piecewise linear splines respectively
(Phillips & Dudík 2008). When Maxent is run with its default setting the FCs used in model building are determined by the number of occurrence records used to build the model (Shcheglovitova & Anderson, 2013; Morales et al., 2017; Phillips et al., 2017).
The models are expected to be more complex and hence overfit if built using more complex FCs, for instance like H over L, when the number of occurrence records are small (Phillips & Dudík, 2008). Therefore, for the species with occurrence records 80 or more all FCs are used, while only L, Q and H are used for the species with 15 to 79, L and Q for 10 to 14, and only L for the species with below 10 records (Phillips & Dudík,
2008; Shcheglovitova & Anderson, 2013).
Regularization is a penalty for including additional feature classes, and it increases with an increase in the weight of the feature class (Phillips et al., 2006). Specific feature classes have default regularization values corresponding to the number of occurrence records (Phillips & Dudík, 2008). However, for practical purposes, Maxent uses a blanket regularization multiplier (RM) which controls the intensity of the regularization across all the feature classes used in the model building (Shcheglovitova & Anderson, 2013).
The default RM used is 1, and though the models built, using either larger or smaller RM compared with the default, are expected to be over-complex or over-simplistic respectively (Phillips et al., 2017), it was actually the models built using default RM and the FCs that were found to be either over-complex or over-simplistic (Shcheglovitova & 49
Anderson, 2013; Morales et al., 2017). Therefore, it is prudent to build multiple models using different combinations of RM values and FCs and then choose the optimal model for use in conservation decisions (Muscarella et al., 2014; Morales et al., 2017; Phillips et al., 2017; Galante et al., 2018).
Choosing the optimal model is a key consideration for all Maxent users (Warren
& Seifert, 2011; Shcheglovitova & Anderson, 2013; Muscarella et al., 2014; Galante et al., 2018). Maxent models have been selected using two model selection approaches: (i) information criteria, specifically Akaike information criteria corrected for small sample size (AICc) and (ii) performance in predicting withheld data (Galante et al., 2018). Out of the two approaches AICc was more robust when species occurrence had sampling bias, but both performed well when the bias was corrected (Galante et al., 2018). However, there is a question on the AICc’s fit for model selection in Maxent despite its better performance (Muscarella et al., 2014; Galante et al., 2018). Whereas, the withheld data selection approach is open to the use of multiple selection criteria (Muscarella et al., 2014) which are either used independently (Warren & Siefert, 2011) or in various combinations
(Shcheglovitova & Anderson, 2013; Radosavljevic & Anderson, 2014; Galante et al.,
2018) to select the optimal model.
The commonly used model selection criteria under the ‘withheld data’ approach include the ‘area under the curve of the receiver operating characteristic’ plot for the training data (AUCTRAIN) and for the test data (AUCTEST), the difference between
AUCTRAIN and AUCTEST (hereafter ‘AUCDIFF’) and the ‘threshold dependent test omission rate’ (hereafter ‘OR’) (Warren & Siefert, 2011; Muscarella et al., 2014;
Radosavljevic & Anderson, 2014; Galante et al., 2018). When used independently
AUCTRAIN and AUCTEST can provide model discriminatory power. However, further, use of AUCTRAIN for model evaluation is criticized, with the use of AUCTEST preferred
50
(Radosavljevic & Anderson, 2014). While, AUCDIFF and OR can evaluate overfitting
(Warren & Siefert, 2011; Radosavljevic & Anderson, 2014), they may also select over permissive models (Galante et al., 2018). Therefore, recent literature has either used a sequential combination of OR and AUCTEST (hereafter ‘ORTEST approach’) (Galante et al., 2018) or OR, AUCDIFF and AUCTEST (hereafter ‘AUCDIFF approach’) (Radosavljevic
& Anderson, 2014) but performances have not been compared. There is also a need to assess the performance of different thresholding rules and their corresponding omission rates, given Maxent produces multiple ORs corresponding to multiple thresholding rules, for optimal model selection along with the use of multiple taxonomic groups to help derive general patterns (Galante et al., 2018), if present.
In this study we developed multiple SDMs with different FC and RM combinations for two taxonomic groups of freshwater organisms with different life history traits. We used 10 fish species which complete their whole lifecycle in water and
28 odonate species whose nymphal stage is aquatic but the adults are terrestrial (Bybee et al., 2016). We then selected optimal models for each species using four sequential approaches comprised of two ORTEST and two AUCDIFF approaches. We also selected optimal models through expert screening for ecologically plausible models using binary suitable habitat maps (hereafter ‘EXP approach’). Though it is sensible to tune models and then screen for ecologically plausible models in every study (Muscarella et al., 2014;
Morales et al., 2017; Phillips et al., 2017; Galante et al., 2018), this option may be either very time consuming or outright impractical if multiple species are involved within a time bound project. Therefore, we aimed to assess which of the sequential approaches best matched the EXP approach in selecting the optimal models, and hence helped reduce the time required for the optimal model selection. We did this by comparing (i) model complexity and (ii) the predicted suitable habitats of the optimal models selected through the five selection approaches. 51
3.3 Methods
3.3.1 Species occurrence data
We collated and cleaned the occurrence data of the fish and odonate species of
Bhutan (see Appendix 1 Table S3.1). We then rarefied the occurrence of all the fish and odonate species with five or more occurrences to Euclidean distance of 5km using the
SDMtoolbox v2.4 (available from http://sdmtoolbox.org/, Brown, 2014; Brown, Bennett,
& French, 2017). We aimed to reduce the negative influence of the spatial autocorrelation among occurrences on the species distribution models (Brown, 2014; Brown et al. 2017;
Galante et al., 2018), but also to retain the maximum possible number of species with at least five occurrences. This resulted in 10 fish and 28 odonate species with small occurrences (n= 5 to 21) for which we built SDMs and used these for the current study.
3.3.2 Environmental variables
We used 19 bioclimatic variables with 30s resolution (~ 1km2 near the equator) that gives current climate data (Fick & Hijmans, 2017 available in http://worldclim.org/version2) as the environmental variables. Bioclimatic variables are derived from the monthly temperature and rainfall values to make them biologically more meaningful; for example, mean annual temperature, maximum temperature of the warmest month, annual precipitation and precipitation of the wettest quarter etc. (Fick &
Hijmans, 2017). In addition, we also used elevation (AsterDEM_Version3_drukref03.im obtained from Watershed Management Division, Ministry of Agriculture and Forest) as an environmental variable. We used bioclimatic variables as environmental variables even for the fish as previous studies have found SDMs built using bioclimatic and hydrological variables did not differ for fish (McGarvey et al., 2018). We also did not reduce bioclimatic variables since our study is exploratory in nature, and collinearity among environmental variables was not an issue in a machine learning environment like
52
Maxent (Elith et al., 2011; Marco Júnior & Nóbrega, 2018) though some literature (e.g.
Merow et al., 2013) suggests being cautious when interpreting SDMs resulting from the use of correlated environmental variables.
3.3.3 Bias files
The ‘presence only’ SDMs are affected by bias sampling, to overcome this one of the common methods employed is the use of a target group bias file to restrict background sampling area (Phillips et al., 2009; Syfert et al., 2013; Vollering et al., 2019). We pooled all occurrence coordinates of the fish species and adult odonate species with 2 or more occurrence coordinates to produce target group bias files for both fish and odonate species respectively. However, we excluded Epiophlebia laidlawi from the odonate target group since its distribution in Bhutan is known only from its larval distribution while all other odonate species are mainly known from their adult occurrence (Appendix 1). Therefore, we developed a separate bias file for the E. laidlawi using all the sampling sites from
Dorji (2015) along with occurrence coordinates from Brockhaus and Hartmann (2008) and Dupchu, S (Personal communication). We used Hydrosheds (hybas_as_lev12_v1c available in www.hydrosheds.org, Lehner & Grill, 2013) clipped to the Bhutan boundary to bound the occurrence instead of regular sized grid cells of the environmental variables
(Phillips et al., 2009; Syfert et al., 2013) to develop sampling bias grids.
3.3.4 Model setting
We used Maxent v 3.4.1 (Phillips, Dudík, & Schapire) to model the species distributions. We used Maxent’s current default output format Cloglog since it gives a better result over logistic when bias correction is used (Phillips, 2017; Phillips et al.,
2017). We used four sets of FCs resulting from the use of individual FCs independently or in combination with other FCs, namely (i) linear (L), (ii) linear-quadratic (LQ), (iii) hinge (H) and (iv) linear-quadratic-hinge (LQH) after Galante et al. (2018) to build 53 models given our species had only a small number of occurrence records (Table S3.1).
When the default model setting is used the number of occurrences determines the FCs used; higher FCs were found to produce less complex (i.e., lesser number of parameters) and less overfitting (i.e., lower omission rates) models when occurrences were small
(Radosavljevic & Anderson, 2014). Specifically, the hinge feature was better for species with small occurrence values (Galante et al., 2018); further, the use of product feature is discouraged and it has been suggested it should be replaced by hinge feature (Phillips et al., 2017).
RMs with values less than default produce models which are overfit to occurrence data and are not well generalized, while larger RMs would produce spread out and less localized models (Phillips, 2017). Though Radosavljevic and Anderson (2014) also observed a slight peak in the model discriminatory ability around the default RM, they found substantial reduction in overfitting when RMs of two to four times that of the default were used. However, they also found both the model quality and the overall discriminatory power declined when RMs were above 4 (Radosavljevic & Anderson,
2014). Hence, we used conservatively 11 different RM values, namely .25, .5, 1, 1.5, 2,
2.5, 3, 3.5, 4, 4.5 and 5.
We used Maxent’s default replication method of cross-validation since it is a better replicate option with small occurrence data. It randomly splits occurrence data into folds and uses all the folds in turn to build and evaluate models (Phillips, 2017). We set the number of iterations for each FC-RM combination equal to the number of occurrence
(n) for each species thus making it equivalent to n-1 jackknife folds, which is a good approach for species with small occurrence data (Warren & Seifert, 2011; Shcheglovitova
& Anderson, 2013; Radosavljevic & Anderson, 2014; Galante et al., 2018). Crossing 11
RM values and four FC sets we built 44 sets of models for each species. Maxent generates
54
(n+1) models including one composite (average) model for each set of RM-FC combination.
3.3.5 Selecting optimal models using sequential approaches
Out of the 11 omission rates corresponding to the 11 thresholds that Maxent produces in its output we used “10th percentile training presence test omission” (hereafter
‘percentile OR’) and “balance training omission, predicted area and threshold values test omission” (hereafter ‘balance OR’) for the sequential model selection approaches. We chose percentile OR (Radosavljevic & Anderson, 2014; Galante et al., 2018) over the
“minimum training presence test omission” (Shcheglovitova & Anderson, 2013;
Radosavljevic & Anderson, 2014) since the latter is more sensitive to extreme localities and over predicts when calibration localities are many (Radosavljevic & Anderson,
2014). We used balance OR to assess utility of a new thresholding rule and its OR in selecting optimal model. Further, we also used it to assess if the use of OR (here balance
OR), corresponding to the threshold used to develop binary habitat maps (here balance threshold), selected better optimal models compared with the use of other OR (here percentile OR) to select the optimal models. Through different sequential combinations of the two ORs, AUCTEST and AUCDIFF we formulated four sequential approaches. They were: (i) sequential combination of percentile OR followed by AUCTEST (hereafter
ORTEST_PER), (ii) sequential combination of balance OR followed by AUCTEST (hereafter
ORTEST_BAL), (iii) sequential combination of percentile OR followed by AUCDIFF and then by AUCTEST (hereafter AUCDIFF_PER), and (iv) sequential combination of balance OR followed by AUCDIFF and then by AUCTEST (hereafter AUCDIFF_BAL) approaches.
We used composite models instead of jackknife iterations (Galante et al., 2018) for each RM-FC combination to select the optimal model. However, Maxent averages all the jackknife iterations to produce the composite model irrespective of whether some
55 individual jackknife models have good model discrimination (AUCTEST>.5), marginal discrimination (AUCTEST<.5) or no discrimination at all (AUCTEST=.5) (Figure S3.1).
When the composite models are comprised of jackknife models with no discrimination they would have lower average ORs since Maxent assigns zero OR to the models with no discriminatory power, and thereby favours these as optimal models. Therefore, we first sorted composite models into four hierarchical groups beginning with (i) the composite models with all jackknife iterations with AUCTEST>.5, (ii) followed by ones with some jackknife iteration models with AUCTEST<.5, (iii) then with some jackknife iteration models with AUCTEST=.5 and (iv) ended with composite models with all their jackknife iteration models having AUCTEST=.5.
Following the hierarchical groups, we then ranked the ORs, AUCDIFF and
AUCTEST of the composite models. We accorded the highest rank to the models with the lowest OR since ORs higher than the theoretically expected value indicate overfitting
(Radosavljevic & Anderson, 2014). Similarly, we accorded the highest rank to the models with the lowest AUCDIFF since less overfitting models are expected to have lower
AUCDIFF (Warren & Seifert, 2011; Radosavljevic & Anderson, 2014). Here, we also considered negative AUCDIFF as equal to zero, the lowest AUCDIFF for model selection
(Muscarella et al., 2014), though we used raw values for general analysis. For the
AUCTEST we accorded the highest rank to the models with the highest AUCTEST since higher AUCTEST means better model performance or discriminatory ability
(Radosavljevic & Anderson, 2014).
Once thus ranked, we followed the steps outlined in Figure 3.1. We chose the composite model or subset of composite models with the highest OR rank (corresponding to Step 1 of Figure 3.1). Since we used OR as the first criteria to select the optimal models if only a single composite model had the highest OR rank (i.e., the lowest ORs among the
56 models) we considered it the optimal model for both ORTEST and AUCDIFF approaches
(Figure 3.1). If Step 1 resulted in a subset of composite models we chose either a composite model or subset of models with the highest AUCTEST rank for the two ORTEST approaches (corresponding to Step 2b, Figure 3.1) (Shcheglovitova & Anderson, 2013,
Galante et al., 2018). Whereas, for AUCDIFF approaches we chose the model or models with the best ranked AUCDIFF (corresponding to Step 2a, Figure S3.2) followed by Step
2b (Radosavljevic & Anderson, 2014) depending on the outcome of Step 2a (Figure 3.1).
After Step 2b, depending on the outcome, we followed Steps 3 to 5 for both the ORTEST and AUCDIFF approaches (Figure 3.1). In Step 3 we chose the models with the lowest average number of parameters since models with lower numbers of parameters are considered less complex and better models (Galante et al., 2018). We derived the average number of parameters for each candidate composite optimal model by dividing the sum of the number of parameters with non-zero lambda coefficients for each individual model extracted from the LAMBDA text file (Galante et al., 2018) by the number of iterations used for building SDM since Maxent does not provide directly the average number of parameters in the result for the composite models unlike it does for the threshold values and ORs. Further, when multiple optimal models had equal average number of parameters, we then chose models with the lower average lambda coefficients obtained by dividing the sum of the absolute value of lambda coefficients by the total number of parameters. However, for some species multiple optimal models with same RM values but different FCs had equal average numbers of parameters as well as the average absolute lambda coefficients. In such cases we used the composite models with simpler FC as the final optimal model since lower FCs are considered better for species with smaller occurrence in Maxent (Shcheglovitova & Anderson, 2013).
57
Figure 3-1 Schematic diagram of selecting optimal models. Steps with red arrows are unique to AUCDIFF approaches while all other steps are common to all the four sequential approaches. Green boxes with “Single model selected” are the optimal models for either ORTEST approaches or AUCDIIF approaches. If Step 2a is followed then one will derive optimal models for AUCDIFF after Step 2, otherwise optimal models derived will be for ORTEST approaches. While purple box with “Single model selected” is the optimal model for only AUCDIFF approaches. Step 5 is the ultimate step wherein models with lower feature class is chosen as the optimal model when multiple models have same numbers of parameters and average absolute value.
3.3.6 Selecting optimal models using expert approach
We visually screened and compared which of the suitable/unsuitable binary maps produced from the optimal models selected using the four sequential approaches were ecologically plausible. If none of the optimal models selected through the four sequential approaches resulted in ecologically plausible binary maps we then moved to the next best choice and continued to screen till we arrived at the best possible ecologically plausible map using expert knowledge (Galante et al., 2018). We qualitatively assessed ecological plausibility by considering the known elevation range and the predicted suitable habitat.
We also compared the predicted map with any available literature, such as IUCN 58 distribution maps or authors’ field experience. However, we acknowledge here that given the relatively high number of fish and odonate species modelled, compounded by very poor literature on the species from Bhutan, some of the chosen optimal models could be subjective. We developed the binary maps using the “balance training omission, predicted area and threshold value Cloglog threshold” (hereafter ‘balance threshold’) out of 11
Cloglog thresholds generated by Maxent for each model iteration. Though lower thresholds can overpredict suitable habitat, they are better for species with few occurrence data (Pearson et al., 2007; Radosavljevic & Anderson, 2014) and can also uncover potentially informative distribution areas (Pearson et al., 2007). We also found using the
10th percentile threshold restricted the predicted suitable habitats around occurrence points, or predicted the whole study area as unsuitable, in some cases (See ‘Predicted
Habitat’ in Result) (Pearson et al., 2007; Radosavljevic & Anderson, 2014; Galante et al.,
2018). We produced binary maps using ArcGIS10.2.2 (ESRI 2014).
3.3.7 Data analysis
We calculated the percentages of optimal models sharing the same RM-FC combination among all the five model selection approaches, and between different pairs, to check for any general trend in RM-FC combination among optimal models. We also statistically tested the correlation among the optimal models chosen by the selection approaches for the model features, viz., ORs, AUCDIFF, AUCTEST, and average number of parameters, and also the area of suitable habitat predicted. We further checked for the correlation between the number of occurrence records used for model building, the model features and the area predicted. We interpreted the strength of correlation based on
Mukaka (2012). We used nonparametric Spearman’s correlation since our variables were not normally distributed, except for few (Table S3.2) when we performed a preliminary analysis. We performed all the statistical analysis in IBM SPSS Statistics 23 (IBM Corp.).
59
3.4 Results
3.4.1 Model complexity
3.4.1.1 Model setting
The most common model setting or regularization multiplier and feature class
(RM-FC) combinations for fish was 5xH (Figure 3.2a, Table S3.3), while for odonates it was 5xL (Figure 3.2b, Table S3.3). However, different model selection approaches selected optimal models with different frequency of RM-FC combinations for fish and odonates, though a noticeably greater number of optimal models had RM-FC combinations of larger RM values and more complex feature classes (Figure 3.2a-b, Table
S3.3). This suggests all selection approaches selected some complex optimal models given the small number of occurrence records used for model building (Table S3.3). All five selection approaches selected the same optimal models for three fish and four odonate species (Table 3.1). ORTEST_PER chose the same optimal models as EXP approach for seven fish species, while AUCDIFF_PER and ORTEST_BAL chose same optimal models as
EXP approach for 10 odonate species each (Table 3.1; Table S3.3).
60
Figure 3-2 Regularization multiplier and feature class (RM-FC) combinations for fish and odonate of the optimal models chosen by five model selection approaches.
Table 3-1 Number of fish and odonate species of Bhutan with same model setting (Regularization Multiplier-Feature Class (RM-FC) combination) of their optimal models as selected by all five model selection approaches (ALL) and by the different pairs of the selection approaches.
Selection approaches Fish Odonata All 3 4 EXP-ORTEST_PER 7 6 EXP-AUCDIFF_PER 3 10 EXP-ORTEST_BAL 5 10 EXP-AUCDIFF_BAL 3 8 ORTEST_PER-AUCDIFF_PER 6 11 ORTEST_PER-ORTEST_BAL 6 14 ORTEST_PER-AUCDIFF_BAL 5 10 AUCDIFF_PER-ORTEST_BAL 7 12 AUCDIFF_PER-AUCDIFF_BAL 8 25 ORTEST_BAL-AUCDIFF_BAL 8 13
61
EXP: Expert approach based on ecological plausibility of binary suitable/unsuitable model; ORTEST_PER: sequential approaches using 10 percentile training presence test omission and test AUC; AUCDIFF_PER: 10 percentile training presence test omission, AUC difference and test AUC; ORTEST_BAL: balance training omission, predicted area and threshold value test omission and test AUC; AUCDIFF_BAL: balance training omission, predicted area and threshold value test omission, AUC difference and test AUC.
3.4.1.2 Omission rates
We found all optimal models chosen by all five selection approaches had ‘10th percentile training presence test omission’ (percentile ORs) equal to or more than theoretically expected 10% (Radosavljevic & Anderson, 2014) for both fish and odonates
(Figure 3a-b). On the other hand, except for four fish and 10 odonate species, all species had optimal models with 0 ‘balance training omission, predicted area and threshold values test omission’ (balance ORs) (Figure 3e-d). Overall, as expected, ORTEST_PER and
AUCDIFF_PER selected a greater number of optimal models with smaller percentile ORs over EXP, ORTEST_BAL and AUCDIFF_BAL for both the fish (Figure 3a) and odonates
(Figure 3b). On the other hand, again as expected, ORTEST_BAL and AUCDIFF_BAL selected a greater number of optimal models with smaller balance ORs over EXP, ORTEST_PER and
AUCDIFF_PER for both fish (Figure 3c) and odonates (Figure 3d). But, there was a very high and statistically significant correlation for the percentile ORs of the optimal models chosen by the EXP approach to the optimal models chosen by all four sequential approaches for fish, while the odonate EXP approach had statistically significant but very low correlations only with ORTEST_PER, AUCDIFF_PER and ORDIFF_BAL (Table S3.4).
However, for balance the OR EXP approach had statistically significant correlations only with ORTEST_PER and AUCDIFF_PER for fish and with AUCDIFF_PER for odonates (Table
S3.4). We also found negative correlations between the percentile ORs of the optimal models chosen by all five approaches with the occurrence records used for model building for both fish and odonates, though not statistically significant for ORTEST_PER and
62
AUCDIFF_PER approaches for odonates. This suggest model overfitting may decrease with an increase in the number of occurrence records used for model building. Whereas, there was no statistically significant correlations for balance OR along with no clear direction in correlation coefficients (Table S3.4).
63
Figure 3-3 Summary of different features of the optimal models selected through five optimal model selection approaches for the fish and odonate species of Bhutan. 10 percentile training presence test omission for the (a) fish and (b) odonate; balance training omission, predicted area and threshold value test omission for the (c) fish and (d) odonate; AUC difference for the (e) fish and (f) odonate; average number of parameters for the (g) fish and (h) odonate; test AUC for the (i) fish and (j) odonate species of Bhutan. EXP: Expert approach based on ecological plausibility of binary suitable/unsuitable model; ORTEST_PER: sequential approaches using 10 percentile training presence test omission and test AUC; AUCDIFF_PER: 10 percentile training presence test omission, AUC difference and test AUC; ORTEST_BAL: balance training omission, predicted area and threshold value test omission and test AUC; AUCDIFF_BAL: balance training omission, predicted area and threshold value test omission, AUC difference and test AUC.
3.4.1.3 AUC Difference
AUC differences (AUCDIFF), as expected, were comparatively lower for the optimal models chosen by AUCDIFF approaches over ORTEST approaches for both fish and odonates (Figure 3e-f). While, the EXP approach chose a greater number of optimal models with comparatively larger AUCDIFF over any of the four sequential approaches for both fish and odonates (Figure 3e-f) suggesting the EXP approach might have selected more overfit models. However, there were statistically significant correlations for
64
AUCDIFF of the optimal models selected by the EXP approach to all the sequential selection approaches (Table S3.4). The highest positive correlation was found with
ORTEST_PER for fish and with ORTEST_BAL for odonates though correlation strength was only moderate for the latter (Table S3.4). Though only the EXP approach for the odonates had a statistically significant correlation between the AUCDIFF of the optimal models and the number of occurrences used for model building, there were negative correlations for both fish and odonates by all five approaches (Table S3.4) suggesting model overfitting may decrease with an increase in occurrence records used for model building.
3.4.1.4 Number of parameters
The greater number of optimal models selected by the EXP approach had a greater number of parameters among all the selection approaches for both fish and odonates
(Figure 3g, h) suggesting the EXP approach might have selected a greater number of complex optimal models. Further, optimal models chosen by the EXP approach had a statistically significant correlation for a number of parameters only with optimal models chosen by ORTEST_PER for fish (Table S3.4). All approaches selected optimal models with the number of parameters greater than the number of occurrences used for model building for at least some species (Figure 3g-h). Among the four sequential approaches, the ORTEST approaches selected a greater number of optimal models with a greater number of parameters over the AUCDIFF approaches (Figure 3g-h). There was no clear relationship between the number of parameters of optimal models and the number of occurrence records used for model building, except for a very low and statistically significant negative correlation for the optimal models chosen by the EXP approach for odonates
(Table S3.4).
65
3.4.1.5 AUC test
A greater number of optimal models selected by the EXP approach for fish had a greater AUCTEST over other approaches, while for the odonates the ORTEST approaches had greater number of optimal models with greater AUCTEST followed by the EXP approach (Figure 3i-j). Overall the AUCTEST values of optimal models chosen by all five approaches were high (Figure 3i-j) except for one fish species Schizothorax progastus which had an AUCTEST value of .382 for the optimal models selected by ORTEST_PER and
AUCDIFF_PER approaches (Figure 3i-j). There was also a high to very high statistically significant positive correlation for the AUCTEST among the optimal models chosen by all five approaches (Table S3.4) suggesting all five selection approaches may select equally those optimal models with good discriminatory power. We did not find any statistically significant correlation between the AUCTEST of the optimal models selected by all five approaches and the occurrence records used for model building (Table S3.4).
3.4.2 Predicted habitat
The EXP approach selected the maximum number of optimal models with comparatively smaller predicted habitat area among all the approaches; this was expected as the expert approach tried to avoid over prediction while selecting the optimal models
(Figure 3.4a-b). In comparison, AUCDIFF approaches predicted a comparatively larger suitable habitat area for a greater number of both fish and odonate species over the ORTEST approaches (Figure 3.4a-b). Most often suitable habitats were over predicted by AUCDIFF approaches and also for some species by ORTEST approaches when compared to the area predicted by the EXP approach (Figure 3.4a-b). For instance, AUCDIFF_PER and
2 AUCDIFF_BAL predicted habitat area above 38,000 km for four of the odonate species and for one fish species by AUCDIFF_BAL (Figure 3.4a-b).
66
Figure 3-4. Area of the predicted habitats of the optimal models selected through the five model selection approaches for the (a) fish and (b) odonate species of Bhutan.
The choice of threshold used to derive binary suitable habitat maps from the optimal models chosen may explain some of the variation in the area of habitat predicted.
Though we did not assess for the effect of thresholds used on the habitat areas predicted we present the binary maps derived using percentile and balance thresholds for four species with varying occurrence records used for the model building as an example
(Figure 3.5). In general, we observed the restrictive percentile threshold seemed to restrict
67 the predicted suitable habitat around occurrence data used for model building, while the less restrictive balance threshold seemed to overpredict the habitat for most of the optimal models selected by sequential approaches (Figure 3.5).
Figure 3-5. Examples of binary suitable habitat maps derived using 10 percentile training presence Cloglog threshold (Left panel) and balance training omission, predicted area and threshold value Cloglog threshold (Right panel) among the optimal models chosen by the Expert and the four sequential optimal model selection approaches. (a) Cyprinion semiplotum (5 occurrence; EXP, ORTEST_PER, ORTEST_BAL, AUCDIFF_PER and AUCDIFF_BAL), (b) Neolissochilus 68 hexagonolepis (12 occurrence; EXP, ORTEST_PER); (c) Neolissochilus hexagonolepis (12 occurrence; ORTEST_BAL, AUCDIFF_PER and AUCDIFF_BAL); (d) Aristocypha quadrimaculata (8 occurrence; EXP, ORTEST_PER, ORTEST_BAL, AUCDIFF_PER and AUCDIFF_BAL); (e) Diplacodes trivialis (21 occurrence; EXP); (f) Diplacodes trivialis (21 occurrence; ORTEST_PER); (g) Diplacodes trivialis (21 occurrence; ORTEST_BAL, AUCDIFF_PER and AUCDIFF_BAL). 0: Unsuitable habitat; 1: Suitable habitat.
However, there were statistically significant correlations between the areas of the optimal models chosen by the EXP approach, with optimal models chosen by all four sequential approaches (Table S3.4). Comparatively, the strongest correlation was with
ORTEST_PER and ORTEST_BAL for fish and with ORTEST_PER and AUCDIFF_PER for odonates, though correlation strength differed with very high correlations for the former and moderate for the latter (Table S3.4). There was no statistically significant correlation between the predicted habitat area and the number of occurrence records used for model building (Table S3.4) suggesting predicted habitat may not depend on the number of occurrence records used for model building.
3.5 Discussion
Our study compared, for the first time, the model complexity and predicted suitable habitat of the optimal models selected by ORTEST and AUCDIFF approaches.
Futher, we used two omission rates derived from one restrictive and other less restrictive thresholds. We also used the less restrictive threshold to derive predicted suitable habitats and to assess for their ecological plausibility. The focus of this study, Bhutan, is also interesting as freshwater species distributions are poorly documented (National
Biodiversity Centre, 2014).
3.5.1 Model complexity
Overall, our results showed 5xH was the most common RM-FC combinations for fish which agreed well with the earlier findings of optimal models with higher RM values combined to more complex FCs like H as better model setting for species with small occurrence data (Shcheglovitova & Anderson, 2013; Galante et al., 2018). While, for the 69 odonates the most common RM-FC combination was 5xL which also agrees with the need to use RM values greater than the default setting (Galante et al., 2018). However, our result also showed other optimal models had different RM-FC combinations for both fish and odonates (Figure 3.2a-b). Further, our results also showed poor agreement among the five model selection approaches with regard to RM-FC combinations of the optimal models selected by them for both the fish and odonate. The EXP approach had the highest agreement with ORTEST_PER for fish, while for odonates the EXP approach agreed the most with AUCDIFF_PER and ORTEST_BAL. When these findings are considered together, they may suggest the need for taxon-specific model tuning, thus agreeing with the recognized need for model tuning for the specific species in a given study (Galante et al., 2018).
Our results showed that the use of corresponding omission rates (ORs) to select the optimal models resulted in selecting optimal models with the smaller corresponding
ORs. However, our results may suggest the overly relaxed balance OR used in our case is not a good criterion to assess model overfit since the majority of the optimal models had 0 balance ORs. Whereas, the restrictive percentile OR varied more among the optimal models and thus may be able to better evaluate the model overfit. Though percentile ORs were equal to or greater than the theoretically expected 10%, suggesting generally overfit models (Galante et al., 2018), other studies have also found percentile ORs greater than
10% (see Muscarella et al., 2014; Radosavljevic & Anderson, 2014; Galante et al., 2018) when a small number of occurrences were used. Further, there was no statistically significant correlation between balance ORs and occurrence records of the optimal models selected by all five approaches for both fish and odonates. On the other hand, there were statistically significant negative correlations between percentile ORs and occurrence records for the optimal models chosen by all five selection approaches, except for ORTEST_PER and AUCDIFF_PER for odonates. These findings may further support the case since ORs are expected to reduce with an increase in the number of occurrence 70 records used for model building (Pearson et al., 2007), and we expect a better model evaluation criterion should reflect this relationship.
Our result showed the EXP approach chose a larger number of optimal models with larger percentile ORs, larger AUCDIFF and a larger number of parameters over four sequential approaches for both fish and odonates. These findings suggest the EXP approach might have selected overfit and over-parameterized optimal models (Muscarella et al., 2014; Radosavljevic & Anderson, 2014; Galante et al., 2018). However, an earlier study found over-parameterization a lesser issue than under-parameterization (Warren &
Seifert, 2011). Further, our use of a relaxed balance threshold to first generate binary suitable/unsuitable habitat area and then choosing the EXP optimal model might have overcome model overfitting and under predicting (Pearson et al., 2007; Radosavljevic &
Anderson, 2014). Whereas, models chosen solely based on smaller percentile ORs,
AUCDIFF and number of parameters may choose overly relaxed (over predicting) models
(Galante et al., 2019) which can be aggravated by our use of a relaxed balance threshold to generate the binary habitat map. Therefore, though AUCDIFF approaches chose the optimal models with small AUCDIFF and smaller number of parameters over ORTEST and
EXP approaches, in our context these optimal models are not necessarily the best. Further, optimal models chosen by AUCDIFF approaches had comparatively lower AUCTEST values over the EXP approach followed by ORTEST approaches. Therefore, AUCDIFF approaches might have chosen a greater number of over predicting optimal models with lower model discriminatory power (Warren & Seifert, 2011). However, to trully conclude our findings comparative studies that build model by removing highly correlated bioclimatic variables may be required since we used all the bioclimatic varibales.
71
3.5.2 Predicted habitat
The EXP approach selected optimal models with generally smaller predicted habitat among all the approaches given we avoided over prediction while selecting ecologically plausible models. Among the four sequential approaches AUCDIFF approaches generally predicted a larger area as suitable habitat for both the taxonomic groups compared to the two ORTEST approaches when the balance threshold was used.
For some species the habitat predicted was several times bigger than that predicted by the
EXP approach, thus over predicting. When the less restricted thresholds, considered robust for the species with small occurrence records (Pearson et al., 2007, Warren &
Seifert, 2011), are used to develop binary suitable habitat maps, as we did here, use of
ORTEST approaches may be a better choice over AUCDIFF approaches for optimal model selection. While the use of AUCDIFF approaches may not be useful when a restrictive threshold, like the percentile threshold (Galante et al., 2018), is used, since optimal models selected through AUCDIFF approaches under predicted for some species (Figure
3.5). Further studies may be required to confirm if there is a true relationship between the thresholds used to derive binary suitable maps and the corresponding model selection approaches used to select the optimal models.
When evalauated using RM values, ORs, AUCDIFF and average number of parameters the expert approach selected more complex models, but the use of a less restrictive threshold value for producing binary suitable habitat maps would have overcome this and help select ecologically plausible models (Pearson et al., 2007). Our findings may suggest not to use balance OR as a criterion for assessing model complexity, though it can be used for optimal model selection. We suggest use of ORTEST approaches for model slecting since optimal models chosen by ORTEST approaches had higher correlation with the models chosen by EXP approach over that by AUCDIFF approaches
72 for almost all the parameters we used. Further, optimal models chosen by AUCDIFF approaches over predicted the habitat area when the balance threshold was used.
However, further studies that use other thresholding rules available in the Maxent result and also including a broader range of taxonomic groups are required to assess the generality of our findings.
73
4.1 Abstract
The Nature Needs Half (NNH) movement aims to protect 50% of the earth. That said protected area designation usually neglects freshwater ecosystems and biodiversity. We assess protection of the surface area of lakes, length of river reaches, habitat area of fish and odonata species within the terrestrial-focused protected areas of Bhutan that meets NNH target. We categorise percentage protection into four progressive levels: (i) Below Aichi (≤16.9%), (ii) Aichi and above (17-34.9%), (iii) Near NNH (35-49.9%) and (iv) NNH and above (≥50%). Overall, we found both freshwater ecosystems and biodiversity are well represented within PAs of Bhutan. 1080 out of 1181 lakes had ‘NNH and above’ percentage surface area protection against only 99 with ‘below Aichi’ protection. Also, 1388 out of 3418 river reaches had ‘NNH and above’ percentage river length protection, though comparatively smaller against 1926 reaches with ‘below Aichi’ protection. No fish or odonate species had ‘below Aichi’ percentage habitat area protection, but only one fish and no odonate species had ‘NNH and above’ protection. However, lakes and river reaches when considered by agro-ecological zone and river reach types respectively had no equitable and adequate protection. 14 of the 19 lakes within the five lower elevation agro-ecological zones had ‘below Aichi’ protection, while 1075 of the 1162 lakes inside alpine zone had ‘NNH and above’ protection. Similarly, the river reach types with fewer number of reaches had a greater number of the reaches with ‘below Aichi’ protection. Also, majority of the fish (n = 7 of 10) and odonate (n = 24 of 28) species had only ‘Aichi and above’ percentage habitat area protection. Further, shorter river reaches and odonate species with smaller habitat area had lower percentage protection. Our findings imply a need for a priori consideration of freshwater ecosystems and biodiversity in PA designation even within NNH paradigm.
Keywords: Aichi target; fish; lake; odonate; protected area; river reach.
75
4.2 Introduction
Globally, biodiversity continues to decline at an alarming rate (World Wide Fund for Nature (WWF), 2016; Darwall et al., 2018). This is despite an increase in the number of protected areas (PAs) after the Convention on Biological Diversity (CBD) set out the
‘Aichi target 11’, stipulating designation of at least 17 per cent of terrestrial and inland water areas and 10 per cent of coastal and marine areas as PAs by 2020 (United Nations
Environment Programme World Conservation Monitoring Centre & International Union for Conservation of Nature, 2016). However, the Nature Needs Half (NNH) movement considers the 17% PA designation by the CBD as a politically agreed target that has no scientific basis and argues for the protection of 50% of the planet (Locke, 2013;
Dinerstein et al., 2017). However, in designating PAs, land-based PAs focus on terrestrial organisms and ecosystem representation (Pittock et al., 2015). The freshwater ecosystems and their associated biodiversity –amongst the most threatened ecosystems globally
(WWF, 2016; Darwall et al., 2018) - are included in PAs only by coincidence (Hermoso,
Abell, Linke, & Boon, 2016) and are not specifically prioritised for management (Pittock et al., 2015; Darwall et al., 2018).
Globally, 15% of open surface inland waters (by area, Bastin et al. 2019), 16% of river length (Abell, Lehner, Thieme, & Linke, 2017) and 11.3% of seasonal inland wetlands (Reis et al., 2019) are protected or –in the case of rivers- locally protected (Abell et al., 2017). The nature of the protection differs among basins (Abell et al., 2017), countries (Bastian et al., 2019) and continents (Bastin et al., 2019; Reis et al., 2016).
Although the current percentage protection of broad freshwater ecosystems is close to
‘Aichi target 11’ it is still way below meeting the NNH target; in comparison, 98 (12%) of the 846 terrestrial ecoregions have 50% protection with another 26 (3%) having at least
40% protection (Dinerstein et al., 2017). Further, a global analysis of the distribution of
132 species of freshwater megafauna found 84% of their total range remains unprotected, 76 with only two species having more than 50% of their range protected (Carrizo, Jāhnig et al., 2017). Likewise, in Europe 23% of those catchments critical for freshwater species conservation were unprotected, another 70% had less than 20% protection, while only 6% had more than 70% protection (Carrizo, Lengyel et al., 2017).
To improve the status quo, studies have suggested either expansion of protected area size (e.g., Rodríguez–Olarte, Taphorn & Lobón–Cerviá, 2011; Holland et al., 2012), modification of PA boundaries (e.g., Rodríguez–Olarte et al., 2011; Holland et al, 2012) or to consider freshwater components as important factors when designating PAs
(Hermoso et al., 2016); or, that freshwater specific PAs be designated (Januchowski-
Hartley et al., 2011; Hermoso et al., 2016) to adequately protect freshwater ecosystems and biodiversity. Most of the study areas had PA coverage less than 50%, the exception being the wet tropics of Australia which had more than 60% PA coverage (Januchowski-
Hartley et al., 2011). However, even in this high percentage PA coverage region, freshwater ecosystems and specific freshwater habitats were found to be not well represented (Januchowski-Hartley et al., 2011).
Moving towards a post-2020 biodiversity framework (Visconti et al., 2019) - along with the movement for NNH (Dinerstein et al., 2017) - there is need to assess freshwater ecosystem and biodiversity protection inside terrestrial-focused PAs at country level in the context of NNH paradigm. Bhutan, one country which met NNH target through designation of terrestrial-focused PAs (Wildlife Conservation Division
(WCD), 2016; Wangchuk, Lham, Dudley, & Stolton, 2017, Dorji, Linke, & Sheldon,
2019), provides the opportunity. Therefore, using Bhutan as a case study we assess freshwater ecosystem and biodiversity protection in the context of NNH paradigm by categorizing percentage protection into four progressive protection levels based on both the Aichi and NNH targets. Unfortunately, there are no comprehensive national level data
77 on any of the freshwater ecosystems in Bhutan which are broadly classified as rivers, lakes, wetlands and groundwater (NEC, 2016). Therefore, we used only lakes and rivers to assess for the freshwater ecosystems representation. Similarly, biodiversity in Bhutan is poorly documented though there has been a recent increase in the number of studies of freshwater fish and odonate species (Gyeltshen et al., 2018). We therefore used the modelled habitat area (Guisan et al., 2013) of the fish and odonate species of Bhutan to assess the freshwater biodiversity protection.
4.3 Materials and method
4.3.1 Protected areas of Bhutan
Bhutan’s PA system consists of five national parks, four wildlife sanctuaries, one strict nature reserve and eight biological corridors (Department of Forests and Park
Services (DoFPS), 2015) and covers 51.44% of the country (WCD, 2016). In addition, there are three wetlands of international importance (Ramsar sites) that are not officially regarded as protected areas (Dorji et al. 2019). Here, we excluded the Ramsar sites from the PA systems since we consider them as freshwater protected areas (Pittock et al., 2015) and not terrestrial-focused PAs. The rationale was to assess freshwater ecosystems and biodiversity protection within terrestrial-focused PAs in the context of NNH. We used the shapefiles of the revised protected area and biological corridors of Bhutan (obtained from Watershed Management Division, Department of Forest and Park Services, Ministry of Agriculture and Forest, Royal Government of Bhutan, Dorji et al., 2019) (Figure 4.1) and the Bhutan country boundary (obtained from Geographical Information System
Laboratory, College of Natural Resources, Royal University of Bhutan) for the analysis and found the terrestrial-focused PAs cover only 49.2%. This is below the NNH target and the officially reported PA coverage (WCD, 2016), although only by 0.8%.
78
Figure 4-1. Terrestrial protected areas of Bhutan. Drawn from shapefiles obtained from Watershed Management Division, Department of Forest and Park Services, Ministry of Agriculture and Forestry, Royal Government of Bhutan. JDNP: Jigme Dorji National Park, JSWNP: Jigme Singye Wangchuck National Park, PNP: Phrumsengla National Park, RMNP: Royal Manas National Park, WCNP: Wangchuck Centennial National Park, JKSNR: Jigme Khesar Strict Nature Reserve, BWS: Bumdeling Wildlife Sanctuary, JWS: Jomotshangkha Wildlife Sanctuary, SWS: Sakteng Wildlife Sanctuary, PWS: Phibsoo Wildlife Sanctuary. The biological corridors, except for NC (North corridor), are named after the PAs they connect. For example, JKSNR-JDNP is biological corridor connecting Jigme Khesar Strict Nature Reserve and Jigme Dorji National Park.
4.3.2 Freshwater ecosystems representation
We did not use the global lake dataset HydroLAKES to obtain data on lakes in
Bhutan since it covers only the lakes with surface areas of 10 hectares and above though it is readily accessible open access datasets (Messager et al., 2016) unlike some other datasets which are not readily accessible (e.g., Global Water Bodies database by
Verpoorter, Kutsr, Seekell, & Tranvik, 2014; global land cover facility inland surface water datasets by Feng, Sexton, Channan, & Townshend, 2015). Instead, we used the landuse and land cover map of Bhutan 2016 (Forest Resources and Management Division
79
(FRMD), 2017) to create lake shapefile using the selection function of ArcGIS 10.2.2
(ESRI 2014). We obtained the surface area of each lake falling within PAs after intersecting between the lake shapefile and the PA shapefile using ArcGIS 10.2.2 (ESRI
2014). Further, we matched the elevations of the lakes with the elevation ranges of the six agro-ecological zones of Bhutan: wet sub-tropical (0-600 m a.s.l.), humid sub-tropical
(600-1200 m a.s.l.), dry sub-tropical (1200-1800m a.s.l.), warm temperate (1800-2600m a.s.l.), cool temperate (2600-3600m a.s.l.) and alpine (>3600m a.s.l.) (NEC, 2016) to assess the distribution of the lakes within the agro-ecological zones.
We created the shapefile of the river reaches falling within Bhutan by intersecting the Bhutan boundary shapefile with the global river reach classification (GloRiC) shapefile (dataset available from http://www.hydrosheds.org/page/gloric, Ouellet
Dallaire, Lehner, Sayre, & Thieme, 2019) using ArcGIS 10.2.2 (ESRI 2014). We then intersected the resulting river reaches shapefile of Bhutan with the PA shapefile of Bhutan to obtain the river reaches falling within PAs using ArcGIS 10.2.2 (ESRI 2014). We used the river reach types derived through supervised classification out of the five river reach classifications provided by Ouellet Dallaire et al. (2019). The 127 reach types derived from the supervised method would be ecologically more meaningful since it combines all the three sub-classifications based on hydrologic, physio-climatic and geomorphic properties and also used expert input on deciding thresholds to designate the river reach types (Quellet Dallaire et al., 2019).
4.3.3 Freshwater biodiversity representation
In order to overcome the patchiness of the occurrence data of the fish and odonate species we used predicted habitat maps of 10 fish and 28 odonate species of Bhutan to assess freshwater biodiversity protection. The predicted habitat maps were the optimal
80 models selected from the multiple species distribution models developed for each fish and odonate species of Bhutan with ≥5 georeferenced occurrence data (Dorji, Sheldon &
Linke, unpublished data) using the Maxent ver. 3.4.0 (Phillips et al., 2017). The predicted habitat maps were produced as part of the work which assessed the performance of multiple optimal model selection approaches in selecting the optimal models for each species (Dorji et al., unpublished data). A set of 44 models were developed for each species using the combination of four feature class combinations and 11 regularization multipliers (Radosavljevic & Anderson, 2014; Galante et al., 2018) out of numerous combinations possible in order to reduce the model complexity (Morales, Fernández &
Baca-González, 2017). We overlaid and intersected the PA map with the species habitat maps to assess their protection using ArcGIS 10.2.2 (ESRI, 2014).
4.3.4 Percentage protection levels
In order to assess freshwater ecosystem and biodiversity protection in the context of the NNH paradigm we categorised percentage protection into four progressive protection levels: (1) Below Aichi (≤16.9%), (2) Aichi and above (17%-34.9%), (3) Near
NNH (35%-49.9%) and (4) NNH and above (≥50%). We considered 35% as the cut-off between the ‘Aichi and above’ and the ‘Near NNH’ targets since it is almost at the mid- point between Aichi and NNH targets. We see increased room for the dialogue when it comes to percentage representation of PA coverage when four levels are used against either Aichi or NNH targets though the use of percentage protected may be misleading for species (Hoek, Zuckerberg, & Manne, 2015) or ecosystem conservation (Dinerstein et al., 2017). Our four protection categories differ, both by name and cut-off percentages, from those used in assessing terrestrial ecoregion protection (Dinerstein et al., 2017). We assess the current protection against the Aichi and NNH targets, whereas Dinerstein et al.
81
(2017) assessed current and potential protection of the terrestrial ecoregions against only the NNH target.
4.3.5 Data analysis
We calculated the number of lakes with different percentage surface area protection after segregating into different agro-ecological zones. Similarly, we calculated the number of river reaches with different percentage river reach length protection after segregating into different reach types based on GloRiC. While we calculated the number of fish and odonate species with different percentage habitat area protection. We then performed correlation analysis between the surface area of the lakes, length of the river reaches, and the predicted habitat area of the fish and odonate species and their percentage protected. Since our exploratory analysis revealed that only fish and odonate habitat areas were normally distributed (Table A1) we used Kendall’s tau-b to assess the correlations
(Field, 2013). We used IBM SPSS Statistics 23 (IBM Corp.) for all the statistical analysis.
4.4 Results
4.4.1 Freshwater ecosystem protection
The database contained a total of 1181 lakes with total surface area of 63.14 km2
(M = 0.05, SD = 0.11). The lakes were very unequally distributed within six agro- ecological zones of Bhutan. The alpine zone alone had 1162 lakes against the combined total of 19 lakes within the other five lower elevation agro-ecological zones including the humid sub-tropical zone with no lakes at all (Table 4.1). Overall, we found 1080 lakes with ‘NNH and above’ percentage surface area protection against only 99 lakes with
‘below Aichi’ percentage surface area protection (Table 4.1). Though 1072 (92%) lakes with ‘NNH and above’ protection had 100% of their surface area protected, the 96 (97%) lakes with ‘below Aichi’ protection had zero percentage of their surface area protected
82
84
Figure 4-2. Relationship between the percentages protected and surface area of the lakes (a), length of river reaches (b), habitat area of fish species (c), and habitat area of odonate species (d) of Bhutan.
85
We found 22 river reach types with a total of 3418 river reaches measuring total length of 9901.4 km (M = 2.9, SD = 2.2) were represented in Bhutan (Table 4.2) out of the 127 river reach types measuring 35.9 million kms at the global scale contained in the data set of Quellet Dallaire et al. (2019). However, the number of river reaches differed among the river reach types with the number of reaches ranging from 2 to 878 (Table
4.2). Overall, the number of river reaches with ‘below Aichi’ river reach length percentage protection (n = 1926) was greater than the number of reaches with ‘NNH and above’ percentage protection (n = 1388) (Table 4.2). Further, 1863 (97%) of the river reaches with ‘below Aichi’ protection had zero percentage of their length protected against only 1220 (88%) river reaches with ‘NNH and above’ protection with 100% of their river length protected (Table 4.2). Also, the river reach types with fewer river reaches tended to have more number of the river reaches with lower percentage protection though some reach types with fewer river reaches also had all of their river reaches with
‘NNH and above’ protection (Table 4.2). Overall, the river reaches with shorter reach length tended to have lower percentages of their length protected, τ (3418) = .058, p<.001,
95% CI [.03, .08] (Figure 2b).
86
4.4.2 Biodiversity protection
Eight fish species out of the 10 had ‘Aichi and above’ percentage habitat area protection against one each with ‘near NNH’ and ‘NNH and above’ protections. For the odonates 24 out of the 28 species had ‘Aichi and above’ percentage habitat area protection with the remaining four species with ‘Near NNH’ protection. But the predicted habitat area of most fish and odonate species was small and hence the absolute area protected were small (Table A2). Fish species with smaller predicted habitat area tended to have higher percentage of their habitat protected (Figure 2c) though there was no statistically significant correlation between the habitat areas of the fish species and their percentages protected, τ (10) =-0.022, p=.93, 95% CI [-.85, .60]. However, for the odonates, species with smaller habitat area tended to have lower percentages of their habitats protected, τ
(28) = .347, p=.01, 95% CI [.11, .58] (Figure 2d).
4.5 Discussion
Our study is the first case study in the world which assessed the freshwater ecosystems and biodiversity protection within the terrestrial-focused PAs in the context of the NNH paradigm. Our findings provide a glimpse into how freshwater ecosystems and biodiversity may be represented within terrestrial-focused PA systems when the NNH target is achieved through the designation of terrestrial-focused PAs. Our study revealed that overall the freshwater ecosystems in Bhutan are well represented within the PA system of Bhutan. 1080 (91%) lakes and 1220 (41%) river reaches were completely protected in Bhutan by the terrestrial-focussed PAs. In comparison only 17% of the lakes by the number were protected in British Columbia, Canada which has only 14% PA coverage across the province (Poppe et al., 2016). Also, only 16% of the total river reach length is protected at global level with only 14% global PA coverage (Abell et al., 2017).
Thus, our findings suggest that even meeting the NNH target through designation of
88 terrestrial-focussed PAs may provide greater overall protection to the freshwater ecosystems at country level.
On the other hand, our results also show the lakes and river reaches within different agro-ecological zones and river reach types respectively are not equitably and adequately protected in Bhutan. For example, maximum number of lakes in the lower altitude agro-ecological zones, which face higher anthropogenic threats (Wangdi, 2014;
Wangda & Gyeltshen, 2018), are not protected although these agro-ecological zones contain already fewer lakes. While the alpine zone with the highest number of lakes had
1075 (96%) lakes with ‘NNH and above’ percentage surface area protection. Also, the river reaches with shorter length tended to have lower percentage of their length protected. Further, the river reach types with fewer number of river reaches had fewer river reaches protected at higher percentage protection levels. Our country level findings agree with the global scale findings of inadequate wetlands (Reis et al., 2017) and river reach types (Abell et al., 2017) protection. Further, we only analysed the local protection of the river reaches which does not reflect true river reach protection as the disturbances in the upstream river reaches will impact the river reaches within PA boundary (Abell et al., 2017).
Overall, we also found terrestrial-focussed PAs in Bhutan afforded Aichi and above or higher levels of percentage habitat area protection for all the fish and donate species included in the study. However, only one fish species had NNH and above percentage habitat area protection, while there was none among the odonate species
(Januchowski-Hartley et al., 2011). But use of blanket thresholds like percentage habitat protected can be misleading for the species conservation (Hoek, Zuckerberg, & Manne,
2015) since different species will require different habitat area to be protected for their persistence (Camaclang, Maron, Martin, & Possingham, 2014). Further, our analysis
89 showed odonate species with smaller predicted habitat areas had smaller percentages of protection, this is relevant as the species may not occupy the entirety of the predicted habitat (Merow, Smith & Silander Jr, 2013). Further for the fishes, a higher percentage of habitat occurring within PA may not guarantee effective conservation (Watson et al.,
2010) especially for the migratory fishes which may be threatened by the hydropower dams in Bhutan (Jha & Rayamajhi, 2010; Dorji et al., 2019).
Though further studies from the countries or regions with PA coverage equal to, or more than the NNH target is required, our findings coupled with the finding of inadequate freshwater ecosystems and species habitat protection in the wet tropics of
Australia, despite more than 60% PA coverage (Januchowski-Harley et al., 2011), imply the need for a priori consideration of freshwater ecosystems and biodiversity while designating PAs even when the NNH target is met. Though use of percentage protection may be misleading for species (Hoek et al., 2015) or ecosystems conservation (Dinerstein et al., 2017), the NNH target is considered the ultimate conservation goal if implemented properly (Dinerstein et al., 2017). We hope our study increased the interest of freshwater conservationists to use the NNH target as the new yardstick to measure freshwater ecosystems and biodiversity protection and move beyond CBD’s 2020 ‘Aichi target 11’ as has already been undertaken by our terrestrial counterparts (Dinerstein et al., 2017). In this way we will bring freshwater ecosystems and biodiversity conservation to the same level as terrestrial ecosystems and biodiversity (Darwall et al., 2018).
90
5.1 Abstract
Nature Needs Half (NNH) aims to protect 50% of the earth by 2050, but increase in PA coverage alone does not guarantee adequate protection of species. Identifying conservation priority areas that meet specific species conservation targets rather than setting aside random parcels of land to meet PA target is suggested as a way forward. Further, protected areas (PAs) are dynamic entities and the loss event termed protected area downgrading, downsizing and degazettement (PADDD) can hamper efficacy of PAs in biodiversity conservation. Considering known threats to both biodiversity and PAs during systematic conservation planning can help identify robust conservation priority areas. We identified freshwater priority areas for Bhutan using the widely used systematic conservation planning tool Marxan. We assessed two different Marxan scenarios: scenario 1 did not exclude planning units (PUs) with hydropower sites (hydropower is the main cause of PADDD in Bhutan) and scenario 2 excluded. We found using freshwater species as well as forest types during conservation planning afforded better protection to the species and forest types by the reserves over that by the existing PA system, though the reserves and existing PA system had similar total area that meets NNH target. Further, the reserve from the Marxan scenario 2 that excluded PUs with hydropower sites accorded similar protection to the freshwater species and forest types as to that provided by the reserve from Marxan scenario 1. We suggest considering target species and known threats to better design PA system or identify priority areas even within NNH paradigm.
Keywords: Hydropower, Marxan, NNH target, Odonata, PADDD, Fish
92
5.2 Introduction
Proponents of Nature Needs Half propose to protect 30% of the earth’s continental area by 2030, followed by 50% by 2050 in order to halt biodiversity loss (Dinerstein et al., 2017; Pimm, Jenkins, & Li, 2018; Dinerstein et al., 2019). Recent studies have not only shown it is possible to set aside half of the earth’s surface for nature (Dinerstein et al., 2017; Pimm et al., 2018; Tallis et al., 2019), they have suggested that at the same time it is still possible to meet the food and energy demands of a growing human population
(Tallis et al., 2019). However, recent global increases in protected area coverage in order to meet the Aichi Target 11 (17% terrestrial and inland water protection by 2020, see
Convention on Biological Diversity, 2010) has not seen corresponding benefits to biodiversity (Venter et al., 2014; Pimm et al., 2018; Visconti et al., 2019). Further, even protecting half the earth may not guarantee biodiversity protection (Pimm et al., 2018), especially if PAs are designated in wild places with low biodiversity and less opportunity cost, as has been done so far (Venter et al., 2014, Pimm et al., 2018; Visconti et al., 2019).
On the other hand, if PAs are designated to cover biodiversity priority areas they can better protect biodiversity even within smaller total areas of PA coverage (Venter et al.,
2014; Pimm et al., 2018).
Designation of terrestrial protected areas has previously focused on terrestrial ecosystems and biodiversity (Roux et al., 2008; Pittock et al., 2015; Hermoso, Abell,
Linke, & Boon, 2016). This is despite the fact Aichi target 11 mandated 17% PA coverage for inland waters (CBD, 2010). Multiple studies have assessed the potential of terrestrial- focused PAs to protect freshwater biodiversity and ecosystems (Acreman et al., 2020).
Acreman et al., (2020) found that out of 75 case studies with quantitative evidence 38 reported positive, 12 negative and 25 neutral outcomes for freshwater biodiversity conservation within terrestrial protected areas. Given freshwater biodiversity and ecosystems are threatened more than their terrestrial and marine counterparts (WWF, 93
2016; 2018), recent studies have employed systematic conservation planning to identify freshwater priority areas for freshwater conservation (e.g. Hermoso, Filipe, Segurado &
Beja, 2017; Tognelli et al., 2018; Cañedo-Argüelles, Hermoso, Herrera‐Grao, Barquín, &
Bonada, 2019). The findings suggest prior inclusion of freshwater biodiversity and ecosystems while designing terrestrial protected areas (Roux et al., 2008; Pittock et al.,
2015; Hermoso et al., 2016) is a more efficient option when trying to protect freshwaters compared with designating terrestrial-focused PAs and then considering freshwater representation. A freshwater ecosystem focused strategy has a better potential to conserve freshwater ecosystems and biodiversity within similar sized PA systems or with only a minimal increase over the total area of the terrestrial-focused PAs (Roux et al., 2008;
Venter et al., 2014; Bush et al., 2014).
The NNH target may provide another opportunity to expand PAs going beyond
2020 and hence provide a chance to plan for freshwater representation for those countries with terrestrial PA coverage below 50% (Pimm et al., 2018). However, countries like
Bhutan, which have already designated 50% of the total land area as PA system
(Wangchuk et al., 2017; Dorji et al., 2019), may have less opportunity to increase the PA coverage to better protect freshwater ecosystems. Systematic conservation planning to identify freshwater priority areas could help Bhutan increase equitable representation in the context of the dynamic nature of protected areas (Mascia & Pailler, 2011; Cook et al.,
2017; Dorji et al., 2019). That said – there may be a need to adapt PA systems as additional information becomes available and also to reflect the changing conservation needs overtime (Martín-García, Sangil, Brito, & Barquín-Diez, 2015; Pittock et al., 2008;
Bush et al., 2014).
PA dynamism can be a mixed blessing for a PA’s long term efficacy for biodiversity conservation especially from the loss events termed protected area
94 downsizing, downgrading and degazettement (PADDD) (Mascia & Pailler, 2011; Symes et al., 2016; Golden Kroner et al., 2019). Therefore, considering the drivers and proximate causes of PADDD while designing PA systems can help designate robust PA systems
(Symes et al., 2016; Golden Kroner et al, 2019). In Bhutan the main cause for PA degazettement and downsizing was found to be conservation planning, however, hydropower development was found the main cause for proposed PADDD (Dorji et al.,
2019). Another specific cause for PADDD designation in Bhutan was the community fishery rights given to two communities within Jigme Singye Wangchuck National Park
(Dorji et al., 2019). Freshwater conservation planning should consider human impact and hydropower development, since these will impact freshwater ecosystems and biodiversity conservation in Bhutan (Dorji et al., 2019). The existing PA system, despite covering
50% of total area, does not provide adequate protection to freshwater species and ecosystems in Bhutan (Dorji, Linke, & Sheldon, Chapter 4). We propose a need for systematic conservation planning to identify freshwater priority areas that can help better redesign existing PA systems (Wildlife Conservation Division, 2010) or help prioritize areas within existing protected areas for better freshwater conservation in Bhutan (Bhutan for Life & WWF, n.d).
The aim of this study is to identify conservation priority areas with Marxan, a widely used conservation planning software (Loos, 2011), using the modeled distributions of 10 fish, 28 odonata, 3 birds and 1 amphibian species along with 8 forest types within Bhutan as conservation feature. We run two Marxan scenarios and identify one best solution that has a total area of all its selected PUs equivalent to 50% of Bhutan’s total terrestrial area as the optimal reserve system. We compare the percentage habitat area of different species included within optimal reserve systems from the two Marxan scenarios and within the existing PA. Finally, we also show that many PUs selected by the two scenarios have different levels of protection accounted for by the existing PA 95 system and which part of Bhutan has freshwater priority areas not well represented within existing PAs.
5.3 Methods
5.3.1 Conservation features
We used modelled suitable habitat maps of 10 fish and 28 odonate species of
Bhutan (Dorji et al., Chapter 3) as conservation features. Additionally, we used the IUCN range maps of black-necked crane (Grus nigricollis) a vulnerable bird species (BirdLife
International, 2017), white-bellied heron (Ardea insignis) a critically endangered bird species (BirdLife International, 2018), Blyth’s kingfisher (Alcedo hercules) a near threatened bird species (BirdLife International, 2016) and Annandale’s Paa frog
(Nanorana annandalii) a near threatened amphibian (Ohler, Shrestha, & Bordoloi, 2008) as conservation features. We clipped the IUCN maps of these species with the Bhutan boundary map in ArcGIS 10.2.2 (ESRI, 2014) to use as conservation features. Further, we also used the distribution area of eight forest types of Bhutan, namely alpine scrub, blue pine forest, broadleaf forest, Chirpine forest, fir forest, meadows, mixed conifer forest and shrubs as coarse filter targets for terrestrial ecosystems (Roux, 2008). We extracted the distribution of forest types from the landuse and landcover map of Bhutan
2016 (Watershed Management Division, Ministry of Agriculture and Forestry) in ArcGIS
10.2.2 (ESRI, 2014).
5.3.2 Planning units
Planning units are the basis for systematic conservation planning using Marxan and PUs of different shape and sizes are used for different purposes (Loos, 2011). In this study we used subcatchments as PUs (see Linke, Norris, & Pressey, 2008; Klein et al.,
2009; Hermoso, Linke, Prenda, & Possingham, 2011). We first selected subcatchments
96 from HydroSHEDS Level 12 (available in www.hydrosheds.org, Lehner & Grill, 2013) using rivers that drain through Bhutan as the source layer using selection by location function in ArcGIS 10.2.2 (ESRI, 2014). From this selection, we retained all the subcatchments both outside and inside Bhutan that had rivers draining into Bhutan, while we removed the subcatchments with less than or equal to 2/3rd of their area outside Bhutan if they were not part of the major basins and had rivers draining out of Bhutan. This step was used to avoid selecting these particular PUs due to their lower connectivity cost
(Fischer et al., 2010). This resulted in 405 subcatchments as our final planning units
(Figure 5.1).
Figure 5-1. Planning units (PUs) with different status for the two Marxan scenarios. Available: PUs available for both scenarios; Locked in: PUs with three Ramsar sites locked in for both scenarios; Locked out: PUs with hydropower sites (of existing, under construction and detailed project report stages) and one preceding and one following the PUs with hydropower sites along the river flow direction were locked out for Scenario 2. We used PUs outside Bhutan’s international boundary to emphasis importance of longitudinal connectivity in identifying freshwater reserve design.
97
5.3.3 Reserve design
Marxan (Ball & Possingham, 2000) aims to find the optimal reserve network by minimizing an objective function comprised of cost, feature penalties weighted by species penalty factor (SPF) and spatial design (connectivity) weighted by connectivity penalty
(CP) (Eqn. 1) (Hermoso et al., 2011). We ran Marxan version 2.43 and used its simulated annealing with iterative improvement algorithm.
Objective function = ∑푝푙푎푛푛푖푔 푢푛푖푡 Cost + SPF ∑푓푒푎푡푢푟푒푠 Feature Penalty +
CP ∑퐶표푛푛푒푐푡푖푣푖푡푦 퐶표푠푡 Eqn. 1
5.3.3.1 Connectivity penalty
Connectivity is a crucial component of freshwater conservation (Linke et al.,
2008; Hermoso et al., 2011; Bush et al., 2014; Linke, Turak, Asmyhr, & Hose, 2019). We used longitudinal connectivity (Linke et al., 2007; Hermoso eta al., 2011) to emphasize our aim to avoid hydropower projects that would impact longitudinal connectivity. We followed the method of Hermoso et al., (2011) which weighs distances among stream segments of the planning units and applies higher penalties to the solutions that miss PUs nearer to the upstream and downstream of the PUs included in the solution than when distant PUs’ are missed. We calibrated using 13 connectivity penalties (0, 0.01, 0.05,
0.1,1, 5, 10, 50, 100, 500, 1000, 5000, 10000) to find the optimal CP for the final Marxan runs (Figure 5.2). We aimed to find the solution with the highest connectivity with comparable costs among the best solutions (Hermoso et al., 2011).
5.3.3.2 Planning unit cost
Varied methods are used to assign cost to planning units (Fischer et al., 2010).
Freshwater planners have used either ‘same cost’ for all PUs (e.g. Hermoso et al., 2011,
98
Hermoso, Kennard, & Linke, 2012) or ‘used costs’ based on different factors (see Loos,
2011; Bush et al., 2014). However, assigning cost based on real world constraints is suggested as a better option (Cameron et al., 2010; Nhancale & Smith, 2011; Hermoso et al, 2012) since this can even offset the negative impact of planning unit characteristics on
Marxan solutions (Nhancale & Smith, 2011). We used the method by Linke et al. (2012), who weighted the size of a planning unit by its human impact, which we derived using the 2009 human footprint index, release 2018 (Venter et al., 2016; Venter et al., 2018).
5.3.3.3 Marxan scenarios
We considered two Marxan scenarios for the current study. For Scenario 1 we left all the PUs available, while for Scenario 2 we locked out PUs which contained hydropower sites and one each of the PUs preceding and succeeding it in a continuum along the direction of river flow to account for minimum river distance that would be impacted by hydropower development. We considered the PUs as hydropower affected if the PUs either contained existing hydropower plants, hydropower plants under construction or planned hydropower projects which are at an advanced planning stage
(DPR stage in Bhutan). To identify the PUs with hydropower plants we obtained the shapefile of hydropower plants from the GIS laboratory, College of Natural Resources,
Royal University of Bhutan and updated the status of the hydropower plants using current literature (Department of Hydropower & Power Systems, 2018; Rai, 2016).
In addition, we locked in three PUs which contained the three Ramsar sites (Dorji et al., 2019) of Bhutan for the four scenarios, this was done since Ramsar sites are considered equivalent to freshwater protected areas (Pittock et al., 2015). We did not lock in terrestrial-focused existing PAs of Bhutan since we aimed to identify alternative freshwater priority areas. The existing PA system already covers half the total area of
99
Bhutan (Dorji et al., 2019) and locking in the PUs falling within the PA system would constrain Marxan from finding optimal solutions (Fischer et al., 2010).
5.3.3.4 Target setting
We calibrated the targets towards our goal of identifying reserves with total area equal or close to 50% (i.e., 19,197 km2) of Bhutan’s total land area of 38,394 km2
(National Statistical Bureau, 2019). When we obtained a target that had a best solution with total reserve area close to our 50% goal, we reran the Marxan and selected the solutions that were closer to 50% and used these for further analysis.
5.3.4 Analysis
We intersected the best solutions of the Marxan scenario 1 (hereafter Reserve 1) and scenario 2 (hereafter Reserve 2), and the existing PA system with the species distribution map in ArcGIS 10.2.2 (ESRI, 2014). We then compared the percentage of species habitat protected within the two reserves and the existing PA system. We also visually assessed if the PUs in Reserve 1 had hydropower sites within them. Further, we intersected the Reserve 1 and Reserve 2 with the existing PA system and then calculated what percentage of the individual PU’s area were protected (hereafter percentage protection). We then categorized the percentage protection into 5 ranges (0, 1-25, 25-50,
50-75 and 75-100) and then compared between the two reserves the number and percentage of PUs falling within each percentage protection range. We also visually assessed how the PUs in the two reserves with different percentage protections were distributed within Bhutan.
5.4 Results
The optimal CP with the highest drop in connectivity, but with minimal increase in cost, was CP 50 (Figure 5.2). From the range of uniform targets calibrated, best
100 solutions with targets 740 km2 and 675 km2 for Scenario 1 and 2 respectively, had the total area of the reserves close to 50% of Bhutan (Table S5.1). Since we could set a higher target for scenario 1, we found Reserve 1 protected higher percentages of species distribution areas for the maximum number of species compared to the Reserve 2 (Figure
5.3), but it had only 19,136 km2 total reserve area against 19,182 km2 for the Reserve 2
(Table S5.1). More importantly, species with smaller modelled habitat area seemed to be better protected by Reserve 1 over that protected by Reserve 2. That said, the lowest percentage protected among all the species by Reserve 1 was lower (37% for odonate species Epiophlebia laidlawi) than that by Reserve 2 (46% for fish species Garra annandalei) (Figure 5.3). This could suggest the PUs that were locked out from scenario
2 have high percentages of habitat available for the species with smaller habitat area. This may also explain why Reserve 1 included PUs with hydropower sites (Figure 5.4) despite the fact we used values derived from human footprint index as the PU costs for both the scenarios.
Figure 5-2. Connectivity penalty (CP) calibration. We set a medium uniform species target of 650 km2 and ran Marxan with 10 repeat runs for 13 different CPs (0, 0.01, 0.05, 0.1, 1, 5, 10, 50, 100, 500, 1000, 5000 and 10000) using fixed SPF of 100. CP 50 has the highest increase in connectivity with the lowest increase in cost.
101
Figure 5-3. Percentage of modelled species distribution and forest type distribution areas protected within the two reserves and existing protected area system of Bhutan.
Figure 5-4. Reserve 1 and Reserve 2 with the location of hydropower sites and the common and unique PUs selected. Reserve 1(top panel): all PUs available; Reserve 2 (bottom panel): PUs with hydropower sites locked out. Note we locked in three Ramsar sites for both the reserves. 102
Overall, both the modelled reserves provided a very high percentage of protection compared to the existing PA system (Figure 5.3). For example, the lowest percentage of species habitat protected for freshwater species was 36% for odonate species Epiophlebia laidlawi in Reserve 1. This is significantly higher than just 24% coverage for the three fish and one odonate species in the existing PA system (Figure 5.3; Table S5.2). The existing PA system performed poorly in even protecting forest types (Figure 5.3). For example, existing PA covered only 7% and 13% of Blue pine and Chirpine forest distributions, compared to 75% and 72% covered by Reserve 1 and 66% and 61% by
Reserve 2, respectively. Biasing reserves to upland areas, the existing PA system covered
90% of the Alpine scrub (Figure 5.3; Table S5.3). Overall, both Marxan solutions tended to protect a higher percentage of species distributions for the species with smaller distribution areas compared with the existing PA system (Figure 5.3).
We found 32% (n=50) and 35% (n=56) of the PUs comprising the optimal modelled reserves 1 and 2 respectively had 75 to 100% coverage of the existing PA systems (Figure 5.5; Table S5.3). Again, an additional 12% and 13% of the PUs had 50 to 75% protection respectively for scenarios 1 and 2 (Figure 5.5; Table S5.3). However, the best protected PUs from all the scenarios 1 and 2 tended to be distributed towards northern Bhutan – reinforcing the existing bias - while the least protected PUs were distributed towards south-west and south-east Bhutan (Figure 5.5).
103
Figure 5-5. Percentage protection level of PUs of Reserve 1 (top panel) and Reserve 2 (bottom panel).
5.5 Discussion
Our study demonstrates the clear need to include freshwater biodiversity in conservation planning for increased protection – even in a country with a high percentage of protection. Reserves 1 and 2 from our Marxan solutions afforded better protection to freshwater species compared to the existing PA system of Bhutan (Dorji et al., Chapter
4). Even some forest types were better represented within the reserves compared to their representation within existing PAs. This is despite existing PAs being designated to better represent terrestrial ecosystems (WCD, 2016; Wangchuk et al., 2017; Dorji et al., 2019).
Overall, both modelled reserves tended to protect a greater percentage of the habitat areas for the species with smaller total habitat area compared to the relative poor performance of the existing PA system; this is despite similar total areas for the modelled two reserves
104 and the existing PA system. Our findings for Bhutan are similar to other countries where inclusion of freshwater components in reserve design could improve PA performance with little or no increase of PA size (Roux et al., 2008). Further, our findings suggest there is a need to include conservation features during the systematic conservation planning stage even within the NNH paradigm to better protect them (Pimm et al., 2018; Tallis et al., 2019).
Our findings not only suggest additional benefits to freshwater species conservation but even to the forest types of Bhutan – if the PA system is redesigned using systematic methods. The need to modify existing biological corridors which are part of the PA system, and to include connected river systems within biological corridors is officially recognized in Bhutan (WCD, 2010). A high number of PUs within the two modelled reserves are included within existing PA systems (Figure 5.5; Table S5.3). This suggests there is no need for a drastic redesign of the existing PA system, particularly as the PA system in Bhutan was overhauled in the past to better represent terrestrial ecosystems (Seeland, 2000; Dorji et al., 2019). However, PUs with a higher percentage protection were concentrated towards northern Bhutan reflecting the unbalanced PA system of Bhutan which covers nearly half of northern Bhutan (Department of Forest and
Park Services (DoFPS), 2015). PUs with lower protection were located towards the south- west and south-east of Bhutan (Figure 5.5), and our finding agrees with the need to designate biological corridors towards south-west Bhutan for better PA system representation (WCD, 2010). The inclusion of these PUs within Bhutan’s PA system would not only increase the percentage habitat protection of freshwater species but would also protect whole river systems against poor performance of the existing PA system
(Dorji et al., Chapter 3). Further, PUs which contain entire small river basins could also be used as integrated watershed management units since Bhutan already has a national strategy for integrated watershed management (Tshering, 2011). 105
Reserve 1 selected through Marxan scenario 1 included PUs with both existing and planned hydropower sites (Figure 5.4), thus exposing the system to the threat from hydropower development, though we used human footprint index to derive the PU cost.
Hydropower development is the main cause for proposed PADDD in Bhutan (Dorji et al.,
2019) and also an emerging global threat to freshwater ecosystems (Reid et al., 2019).
Interestingly, Reserve 2 also provided a high level of protection that was comparable to that provided by the modelled Reserve 1 despite locking out PUs with hydropower sites during Marxan runs. These findings suggest the need to consider specific threats, if known, at the planning stage to develop robust PA systems that can avoid PADDD
(Symes et al., 2016; Reid et al., 2019) and at the same time also provide adequate protection to conservation features. On the other hand, Reserve 1, despite slightly lower total reserve area over that of Reserve 2, tended to provide better protection to the species with smaller habitat area (Figure 5.3). This may suggest if all proposed hydropower developments are carried through, they would impact disproportionately the species with already smaller modelled habitats. Further, migratory fish species would be disproportianetly impacted by lack of river connectivity due to hydropower dams (Jha &
Rayamajhi, 2010; Dorji et al., 2019).
Our study showed the importance of including conservation features in a systematic fashion to better protect them (Roux et al., 2008) even within the NNH paradigm (Pimm et al., 2018; Tallis et al., 2019). Our result also showed the importance of including known threats to help avoid PUs that will be impacted by PADDD (Symes et al., 2016; Reid et al., 2019). We conclude that while there is need for more comprehensive planning, no drastic redesign of Bhutan’s PA system is needed to better protect both freshwater as well as forest types.
106
6 General Discussion
6.1 Overall Summary
The overall objective of this thesis was to assess the potential of existing protected areas (PAs) of Bhutan - which are currently terrestrial-focused - to conserve freshwater biodiversity and ecosystems within the context of Nature Needs Half (NNH) and protected area dynamism. To achieve this objective, this thesis used systematic archival research method for country specific case study, species distribution modelling and systematic conservation planning techniques to address four main issues when assessing the potential of freshwater conservation within existing PA system:
(1) The lack of a country specific case study on PA dynamism and its proximate causes, and the possible link between PA dynamism and Bhutan’s pro-environmental developmental philosophy of Gross National Happiness (GNH). To fill this gap, a case study of PA dynamism in Bhutan spanning 50 years (1966-2016) of its modern conservation history against the backdrop of four tenets of GNH philosophy was presented in Chapter 2.
(2) The lack of adequate freshwater species distribution data in Bhutan. In Chapter
3, Maxent, the widely used species distribution modelling tool, was used to derive species distributions. At the same time the thesis helped advance the field of species distribution modelling by investigating the performance of four sequential model selection approaches when compared to an expert approach in selecting an ecologically plausible model.
(3) The need to assess the potential of the existing terrestrial-focused PA system of Bhutan in protecting freshwater biodiversity and ecosystems, and the general need to
107 contextualize freshwater conservation against Nature Needs Half target going beyond
2020. Chapter 4 evaluated progress against these targets.
(4) The need for systematic freshwater conservation planning to identify freshwater priority areas within the context of NNH and PA dynamism. In Chapter 5 freshwater priority areas that cover 50% of Bhutan’s total land area were identified under two different scenarios using the conservation planning decision tool Marxan, and the percentage protection of species within the reserves and the existing PA system, which also meets NNH target, were compared.
6.2 Protected area dynamism
Protected areas are considered the cornerstone of biodiversity conservation, and in the face of the rapid decline in biodiversity, the number and total area of PAs globally has increased in recent times (Watson et al., 2016; Venter et al., 2017; Pimm et al., 2018).
On the other hand, though PAs are intended to protect biodiversity in perpetuity (Golden
Kroner et al., 2019), a growing numbers of studies (e.g. Mascia & Pailler, 2011; Mascia et al., 2014; Pack et al., 2014; Symes et al., 2016; Cook et al., 2017; Dorji et al., 2019;
Golden Kroner et al., 2019) are not only showing that PAs are dynamic entities but are subjected to legal changes termed PADDD that can hinder PA sustainability and efficacy
(Golden Kroner et al., 2019). Knowing the drivers and proximate causes of PADDD can help design robust PA systems (Mascia & Pailler, 2011; Symes et al., 2016; Golden
Kroner et al., 2019).
PADDD is a mixed blessing for conservation (Mascia & Pailler, 2011) since it results from numerous proximate causes which include both detrimental land uses and also conservation planning (Mascia & Pailler, 2011; Mascia et al., 2012). The main proximate cause that lead to actual loss in land from the PA system of Bhutan through
PA degazettement and downsizing was conservation planning, which was used to
108 redesign PA systems to better represent terrestrial ecosystems (Figure 2.3, Table 2.3). On the other hand, infrastructure development was the main proximate cause for proposed
PADDD with hydropower development being the specific cause (Figure 2.3, Table 2.3).
Another proximate cause was subsistence, with the specific cause being fishery rights given to two communities living inside Jigme Singye Wangchuck National Park to prevent illegal fishing and encourage sustainable harvesting (Table 2.3). Overall, the proximate causes (Figure 2.3; Table 2.3) of PADDD events over the 50 years of modern conservation history reflect Bhutan’s commitment to environmental conservation guided by its GNH philosophy. These findings may suggest scope for redesigning the PA system in Bhutan to better represent conservation features of interest (e.g., freshwater biodiversity and ecosystems). Further, hydropower development, identified here as the main specific cause for PADDD in Bhutan (Figure 2.3; Table 2.3) could be incorporated into systematic freshwater conservation planning to help design robust PA systems or identify robust priority areas as done in here (Chapter 5).
PADDD events, though a global conservation phenomenon, are difficult to track particularly in remote areas (Mascia & Pailler, 2011; Symes et al., 2016). The systematic archival literature review used at the country level case study (Chapter 2) detected a far greater number of PADDD events in Bhutan over what was previously recorded from a regional level study (Mascia et al., 2014) (Table 2.4). Further, there were differences between the two approaches in in detecting the types of PADDD events and which PAs these events affected (Table 2.4; Table S2.9). These findings suggest the need for more archival literature reviews (Golden Kroner et al., 2019) at country level case studies to better detect PADDD events for a country.
109
6.3 Species distribution modelling
Species distribution models are used at different stages of conservation planning in data poor areas (Guisan et al., 2013). Maxent is a widely used SDM technique when presence only data is available (Merow, Smith, & Silander Jr, 2013; Phillips et al., 2017) due to its better performance over other similar SDM techniques (Elith et al., 2011;
Coxen, Frey, Carleton, & Collins, 2017). However, resulting species distribution maps depend on the quality of the species occurrence data (Aubry, Raley & McKelvey, 2017).
In data poor countries like Bhutan (National Biodiversity Centre, 2014) a lack of cleaned and collated georeferenced occurrence data has been a problem so far. This thesis undertook the curation of fish and odonata species of Bhutan recorded until 2017 (Chapter
3, Appendix 3.1, Table S3.1). This can help increase data collation of other species to facilitate the use of other SDMs to help make better conservation decisions in Bhutan.
The multiple SDMs built here (Chapter 3) using 44 different model settings for each of the 10 fish and 28 odonate species (Galante et al., 2018; Hallgren, Santana, Low-Choy,
Zhao, & Mackey, 2019) following all the possible recommended best practices for
Maxent (Merow et al., 2013) could be a good starting point for freshwater conservation planning in Bhutan until additional occurrence data and environmental covariates become available to build better models.
Maxent users face important issues when selecting optimal models (Warren &
Seifert, 2011; Shcheglovitova & Anderson, 2013; Muscarella et al., 2014; Radosavljevic
& Anderson, 2014; Galante et al., 2018). Out of the two main approaches used for model selection, namely information criteria and withheld data (Galante et al., 2018) the users of latter approach have multiple model evaluation criteria available like the ‘area under the curve of the receiver operating characteristic’ plot for the training data (AUCTRAIN) and for the test data (AUCTEST), the difference between AUCTRAIN and AUCTEST
(hereafter ‘AUCDIFF’) and the ‘threshold dependent test omission rate’ (hereafter ‘OR’) 110
(Warren & Siefert, 2011; Muscarella et al., 2014; Radosavljevic & Anderson, 2014;
Galante et al., 2018). When used independently AUCTRAIN and AUCTEST can provide model discriminatory power but cannot evaluate model overfitting. Of the two AUCTEST is preferred as model evaluation criteria (Radosavljevic & Anderson, 2014). While,
AUCDIFF and OR can evaluate model overfitting (Warren & Siefert, 2011; Radosavljevic
& Anderson, 2014) but they may also select over permissive models (Galante et al.,
2018). Hence, sequential combinations of either OR and AUCTEST (hereafter ORTEST approach) (Galante et al., 2018) or OR, AUCDIFF and AUCTEST (hereafter AUCDIFF approach) (Radosavljevic & Anderson, 2014) have been used, but their performances have not been compared directly. For the first time the performance of ORTEST and
th AUCDIFF model selection approaches are compared directly, using ‘10 percentile training presence test omission’ a commonly used model evaluation criterion (e.g.
Radosavljevic & Anderson, 2014; Galante et al., 2018) and ‘balance training omission, predicted area and threshold values test omission’ which has not been used so far as the model evaluation criterion (Chapter 3).
ORTEST approaches performed better over AUCDIFF approaches when compared to the expert (EXP) approach that selected optimal models by screening the models for ecological plausibility (Galante et al., 2018). A range of studies (e.g. Galante et al., 2018;
Hallgren et al., 2019) suggest performing species specific model setting and then screening for ecologically plausible models (Galante et al., 2018; Hallgren et al., 2019).
Findings in Chapter 3 suggest the use of ORTEST over AUCDIFF approaches as a good first line of model screening that can reduce the time taken by the EXP approach. Further, even when used alone ORTEST approaches would select models which are neither too overfit nor too permissive when compared to the AUCDIFF approaches which were chosen for many species the models which were over permissive that predicted whole of Bhutan as suitable (Chapter 3, Figure 3.4). This was especially true when less restrictive 111 thresholds, which are considered better for species with small records (Pearson,
Raxworthy, Nakamura, & Peterson, 2007), like the ‘balance training omission, predicted area and threshold value Cloglog threshold’ were used (Chapter 3, Figure 3.5).
6.4 Freshwater biodiversity and ecosystem protection against NNH target
Freshwater biodiversity and ecosystems continue to be the most threatened, compared to their terrestrial and marine counterparts (WWF, 2018; Reid et al., 2019).
They are usually included inside terrestrial-focused PAs by coincidence and not because of their significance to the protection of biodiversity (Darwall et al., 2011; Pittock et al.,
2015; Flitcroft, Cooperman, Harrison, Juffe-Bignoli, & Boon, 2019). A recent review of studies that assessed the potential of terrestrial-focused PAs in conserving freshwater biodiversity and ecosystems found mixed results with some findings suggesting the positive role of PAs on freshwater conservation while others suggested a negative role
(Acreman et al., 2020). However, most studies (e.g. Juffe-Bignoli et al., 2016; Poppe,
Biffard, & Stevens, 2016; Abell et al., 2017; Reis et al., 2017) have assessed the potential of freshwater conservation against Aichi target 11 of protecting 17% terrestrial and inland water areas. Moving beyond 2020 this might not be enough - instead a need to assess against NNH (50%) target is emerging (e.g. Dinerstein et al., 2017; Pimm et al., 2018;
Dinerstein et al., 2019).
The four progressive levels of percentage targets based on both Aichi target 11 and NNH target categorized and used here (Chapter 4) could help drive freshwater conservation assessments within the NNH context. Despite covering 50% of Bhutan’s total land area, freshwater ecosystems and biodiversity were inadequately represented within the existing PA system of Bhutan, when assessed against individual targets
(Wangchuk et al., 2017; Dorji et al., 2019). These findings (Chapter 4) suggest the need to consider freshwater biodiversity and ecosystems while designing PAs even within the 112
NNH paradigm since the existing PAs of Bhutan were designated for terrestrial ecosystems representation (DoFPS, 2015).
6.5 Systematic conservation planning in light of NNH and PA dynamism
An increase in the total area of land placed under protection has not halted biodiversity decline (Venter et al., 2014; Pimm et al., 2018; WWF, 2018; Visconti et al.,
2019) and the suggested way forward is the need for systematic conservation planning to identify priority areas followed by designation of new PAs (Venter et al., 2014, Pimm et al., 2018; Visconti et al., 2019). However, PAs have traditionally been designated to represent terrestrial ecosystems and biodiversity, and a recent literature review that assessed the potential of terrestrial-focused PAs in freshwater conservation found mixed results (Acreman et al., 2020). In recognition of these shortcomings, systematic conservation planning to identify freshwater priority areas have grown over the years despite it lacking behind terrestrial and marine counterparts (Flitcroft et al., 2019;
Tognelli et al., 2019).
In Bhutan, even with 50% of the country under PA system, freshwater ecosystems and biodiversity were not adequately represented within PAs (Chapter 3). Therefore, systematic conservation planning to identify freshwater conservation priority areas was conducted using Marxan (Chapter 5). Reflecting findings in previous studies that assessed freshwater conservation in other areas (e.g. Hermoso et al., 2017; Tognelli et al., 2019), reserves identified as freshwater priority areas not only protected higher percentages of modeled habitats for almost all freshwater species but also for forest types over that protected by the existing PA system (Figure 5.3). More importantly reserves protected higher percentages of habitat for the species with smaller total habitat (Figure 5.3) against poor performance of the existing PA system (Chapter 3). This is despite the reserves and existing PAs having equivalent total areas (Chapter 5, Table S5.1). This agrees with a 113 global analysis which found placing PAs where there are species could provide better protection to the terrestrial mammals, terrestrial birds and amphibians even within areas less than NNH target (Pimm et al., 2018).
Considering known threats during systematic conservation planning can help design robust PA systems (Symes et al., 2016; Golden Kroner et al., 2019; Hermoso et al., 2017). When hydropower sites were considered during systematic conservation planning the resulting reserve system protected comparable habitat areas of freshwater species against the reserve identified without considering it within a similar total area
(Chapter 5). This finding agrees with an earlier similar finding (Hermoso et al., 2017).
On the other hand, the modelled reserve from the Marxan scenario that did not lock out planning units with hydropower sites protected a higher percentage of habitats for the species with smaller habitat area compared to the modelled reserve from the scenario that locked out PUs with hydropower sites (Figure 5.3). This finding may suggest that if
Bhutan implements all the planned hydropower projects there are chances that freshwater species with already smaller habitats would be threatened disproportionately. But, high percentages of PUs within each reserve were accorded 75 to 100% protection by the existing PA system (Figure 5.5). This may suggest Bhutan may not need to change drastically the existing PA system if PAs are redesigned to better represent both freshwater species and forest types, while redesign could benefit if south-west and south- east regions are given priority (Figure 5.5).
6.6 Significance of this thesis and future research
Moving beyond 2020 there is a need for new conservation targets (Pimm et al.,
2018; Dinerstein et al., 2019; Visconti et al., 2019) and NNH is considered the ultimate target (Dinerstein et al., 2017; Pimm et al., 2019; Dinerstein et al., 2019). Though protection of terrestrial ecoregions (Dinerstein et al., 2017) and the associated terrestrial
114 mammals, terrestrial birds and amphibians (Pimm et al., 2018) have been assessed against
NNH target, freshwater systems have been assessed against only Aichi target 11 (e.g.
Abell et al., 2017; Reis et al., 2018). The progressive conservation target categories that contextualized the freshwater conservation against NNH target here (Chapter 4) can hopefully inspire similar studies in freshwater conservation, which lags behind its terrestrial and marine counterparts (Darwall et al., 2019). However, only increasing the parcel of land protected will not guarantee conservation benefits if PAs are misplaced
(Pimm et al., 2018; Visconti et al., 2019). The gap analysis (Chapter 4) provides a real world example of the shortcomings of the PA system in conserving freshwater even when
NNH target is met through designating terrestrial-focused PAs, as is the case in Bhutan.
Freshwater priority areas identified through systematic conservation planning that included freshwater species and forest types (Chapter 5) showed reserves of similar total areas were possible. However, the proposed redesign provided far better protection to both freshwater species and forest types over that provided by the existing PAs. This highlights the need for systematic conservation planning even within NNH paradigm
(Pimm et al., 2018).
PAs have remained dynamic and are subjected to both gain and loss events (Cook et al., 2017; Dorji et al., 2019). Especially the loss events, termed PADDD, can reduce the land under protection and also affect efficacy of PAs in biodiversity and ecosystem conservation (Golden Kroner et al., 2019). Therefore, identifying and considering proximate causes and drivers of PADDD can help designate robust PA systems, but detecting PADDD events in remote areas is difficult (Mascia & Pailler, 2011; Symes et al., 2106). Use of archival research methods (Golden Kroner et al., 2019) at country level case study for PA dynamism (Chapter 2) could help better detect PADDD events and their proximate causes (Dorji et al., 2019). However, there is urgent need for more country specific PA dynamism studies to better understand the drivers and proximate causes of 115
PADDD at global level. Even for Bhutan there is need to study what are the drivers of
PADDD.
The reserve system identified through systematic conservation planning which considered hydropower sites, the specific cause for proposed PADDD in Bhutan (Dorji et al., 2019), provided comparable protection to freshwater biodiversity as that by the reserve system that did not consider hydropower sites. While, the later included PUs with hydropower sites as priority areas which can be detrimental to freshwater conservation
(Hermoso et al, 2017; Dorji et al., 2019). This was despite the use of PU cost derived from the human footprint index, thereby suggesting the need to consider any known specific threats while designing PA system (Chapter 5). On the other hand, higher protection accorded to species with smaller habitat area by reserve system that did not lock out PUs with hydropower suggest hydropower could harm species with smaller habitat disproportionately. There is an urgent need to better document freshwater species distribution in Bhutan to better plan their conservation. Furture studies should consider including more number of species belonging to different taxonomic groups and also include other known proximate causes of PADDD in conservation planning.
SDMs area used to fill species distribution data gap for conservation planning.
For the widely used SDM tool Maxent model setting and optimal model selection have remained a constant challenge (Galante et al., 2018). Recent advances in optimal model selection approaches used different sequential combinations of different model evaluation criteria within the withheld data technique to select the optimal model (Galante et al.,
2018). Findings in Chapter 3 outline the better performance of ORTEST approaches over
AUCDIFF approaches when assessed against expert approach advances the model selection field. Also, the use of balance omission as evaluation criteria in the sequential model selection approach is new to the field (Galante et al., 2018). However, there is a need to
116 assess the utility of other thresholding rules and their corresponding omission rates in sequential optimal model selection approaches, and to compare these approaches with the information criteria approach, specifically Akaike information criteria corrected for small
th sample size (AICc) which was found better over ORTEST approach using 10 percentile omission rate (Galante et al., 2018).
6.7 Conclusion
The main contributions of the thesis are contextualizing the potential of terrestrial- focused PAs to conserve freshwater within the NNH paradigm, and recognizing the need for systematic conservation planning even within the NNH paradigm to better protect freshwater. This thesis also contributed to the study of PADDD, a widely prevalent but unrecognized conservation phenomenon, by showing that a country level archival literature review method is the way forward to detect PADDD. For Bhutan, the thesis provides a PA dynamism record for almost the entire modern conservation history. The thesis not only filled the freshwater species distribution data gap of Bhutan by developing
SDMs for the fish and odonates species but also advanced the SDM field, specifically for
Maxent through the use of a new omission rate and comparing two broad sequential approaches of model selection within the withheld data technique. The thesis linked a PA dynamism study with freshwater conservation planning by using the findings of the earlier study in helping design robust reserves. Further, thesis findings clearly show the need for freshwater conservation planning, and if there was to be a redesign of the PA system where it could be done to better protect both freshwater and forest types in Bhutan.
117
7 References
Abell, R., Lehner, B., Thieme, M. & Linke, S. (2017). Looking beyond the fenceline:
assessing protection gaps for the World’s Rivers. Conservation Letters, 10, 384–394.
https://doi.org/10.1111/conl.12312
Abell, R., Thieme, M. L., Revenga, C., Bryer, M., Kottelat, M., Bogutskaya, N., Coad,
B., Mandrak, N., Balderas, S. C., Bussing, W., Stiassny, M. L. J., Skelton, P., Allen,
G. R., Unmack, P., Naseka, A., Ng, R., Sindorf, N., Robertson, J., Armijo, E.,
…Petry, P. (2008). Freshwater ecoregions of the world: a new map of
biogeographic units for freshwater biodiversity conservation. Bioscience, 58, 403-
414.
Acreman, M., Hughes, K. A., Arthington, A. H., Tickner, D., & Dueñas, M-A. (2020).
Protected areas and freshwater biodiversity: a novel systematic review distils eight
lessons for effective conservation. Conservation Letters, 13, e12684.
https://doi.org/10.1111/conl.12684
Amis, A.M., Rouget, M., Lotter, M., & Day, J. (2009). Integrating freshwater and
terrestrial priorities in conservation planning. Biological Conservation, 142, 2217-
2226.
Araújo, M. B., Anderson, R. P., Barbosa, A. M., Beale, C. M., Dormann, C. F., Early, R.,
Garcia, R. A., Guisan, A., Maiorano, L., Naimi, B., O’Hara, R. B., Zimmermann,
N. E., & Rahbek, C. (2019). Standards for distribution models in biodiversity
assessments. Sciences Advances, 5, eaat4858.
https://doi.org/10.1126/sciadv.aat4858
118
Aubry, K. B., Raley, C. M., & McKelvey, K. S. (2017). The importance of data quality
for generating reliable distribution models for rare, elusive, and cryptic species.
PLoS ONE, 12, e0179152. https://doi.org/10.1371/journal.pone.0179152
Balaguera-Reina, S. A., Espinosa-Blanco, A. S., Morales-Betancourt, M. A., Seijas, A.
E., Lasso, C. A., Antelo, R., & Densmore, L. D. (2017). Conservation status and
regional habitat priorities for the Orinoco crocodile: past, present, and future. PLoS
ONE, 12, e0172439. https://doi.org/10.1371/journal.pone.0172439.
Ball, I. R., & Possingham, H. P. (2000). MARXAN (V1.8.2): Marine Reserve Design
Using Spatially Explicit Annealing, a Manual.
Barnes, M. D., Glew, L., Wyborn, C., & Craigie, I. D. (2018). Prevent perverse outcomes
from global protected area policy. Nature Ecology and Evolution, 2, 759-762.
Bastin, L., Gorelick, N., Saura, S., Bertzky, B., Dubois, G., Fortin, M-J., & Pekel, J-F.
(2019). Inland surface waters in protected areas globally: current coverage and 30-
year trends. PLoS ONE, 14, e0210496.
https://doi.org/10.1371/journal.pone.0210496
Bernard, E., Penna, A. O., & Araújo, E. (2014). Downgrading, downsizing,
degazettement, and reclassification of protected areas in Brazil. Conservation
Biology, 28, 939-950.
Betts, M. G., Wolf, C., Ripple, W. J., Phalan, B., Millers, K. A., Duarte, A. Butchart, S.
H. M., & Levi, T. (2017). Global forest loss disproportionately erodes biodiversity
in intact landscapes. Nature, 547, 441–444. doi:10.1038/nature23285
Bhattarai, S. & Thinley, P. (2005). Project report: survey of fish fauna along with basic
water quality in some rivers of Bumdeling Wildlife Sanctuary in eastern Bhutan.
119
Bhutan for Life & WWF. (n.d). Bhutan for Life prospectus. Bhutan for Life, WWF.
http://www.bfl.org.bt/bflsite/public/upload/resources/19071lo2056e5.pdf.
Downloaded on 01 March 20120.
BirdLife International. (2016). Alcedo hercules. The IUCN Red List of Threatened
Species 2016: e.T22683024A92973795.
https://dx.doi.org/10.2305/IUCN.UK.2016-3.RLTS.T22683024A92973795.en.
Downloaded on 17 February 2020.
BirdLife International. (2017). Grus nigricollis (amended version of 2016 assessment).
The IUCN Red List of Threatened Species 2017: e.T22692162A110659467.
https://dx.doi.org/10.2305/IUCN.UK.2017-1.RLTS.T22692162A110659467.en.
Downloaded on 17 February 2020.
BirdLife International. (2018). Ardea insignis. The IUCN Red List of Threatened Species
2018: e.T22697021A134201407. https://dx.doi.org/10.2305/IUCN.UK.2018-
2.RLTS.T22697021A134201407.en. Downloaded on 17 February 2020.
Bower, S. D., Lennox, R. J., and Cooke, S. J. (2014). Is there a role for freshwater
protected areas in the conservation of migratory fish? Inland Waters, 5, 1-6. DOI:
10.5268/IW-5.1.779
Brockhaus, T., & Hartmann, A. (2009). New records of Epiophlebia laidlawi Tillyard,
1921 in Bhutan with notes on its biology, ecology, distribution, zoogeography and
threat status (Anisozygoptera: Epiophlebiidae). Odonatologica, 38, 203–215.
Brooks, E. M. G. E., Holland, R. A., Darwall, W. R. T., & Eigenbrod, F. (2016). Global
evidence of positive impacts of freshwater biodiversity on fishery yields. Global
Ecology and Biogeography, 25, 553–562. https://doi.org/10.1111/geb.12435
120
Brooks, J. S. (2013). Avoiding the Limits to Growth: Gross National Happiness in Bhutan
as a model for sustainable development. Sustainability, 5, 3640-3664.
https://doi.org/10.3390/su5093640
Brown, J. L. (2014). SDMtoolbox: a python-based GIS toolkit for landscape genetic,
biogeographic, and species distribution model analyses. Methods in Ecology and
Evolution, 5, 694-700. DOI: 10.1111/2041-210X.12200.
Brown, J. L., Bennett, J. R., & French, C. M. (2017). SDMtoolbox 2.0: the next generation
Python-based GIS toolkit for landscape genetic, biogeographic and species
distribution model analyses. PeerJ, 5, e4095. DOI 10.7717/peerj.4095.
Buckley, M. (2017). Saving half the earth in Bhutan. Available from
http://www.planetexperts.com/saving-half-the-earth-in-bhutan/ Accessed January
2018.
Bush, A., Hermoso, V., Linke, S., Nipperess, D., Turak, E., & Hughes, L. (2014).
Freshwater conservation planning under climate change: demonstrating proactive
approaches for Australian Odonata. Journal of Applied Ecology, 51, 1273-1281.
doi: 10.1111/1365-2664.12295
Bybee, S., Cόrdoba-Aguilar, A., Duryea, M. C., Futahashi, R., Hansson, B., Lorenzo-
Carballa, M. O., Schilder, R., Stoks, R., Suvorov, A., Svensson, E. I., Swaegers, J.,
Watts, P. C., & Wellenreuther, M. (2016). Odonata (dragonflies and damselflies)
as a bridge between ecology and evolutionary genomics. Frontiers in Zoology, 13,
46. DOI 10.1186/s12983-016-0176-7.
Camaclang, A. E., Maron, M., Martin, T. G., & Possingham, H. P. (2014). Current
practices in the identification of critical habitat for threatened species. Conservation
Biology, 29, 482-492. DOI: 10.1111/cobi.12428. 121
Cameron, S. E., Klein, C. J., Canessa, R., & Geselbracht, L. (2010). Addressing
socioeconomic objectives. In Ardron, J. A., Possingham, H. P., & Klein, C.J. (Eds.),
Marxan Good Practices Handbook, Version 2. Victoria, BC, Canada: Pacific Marine
Analysis and Research Association.
Cañedo‐Argüelles, M., Hermoso, V., Herrera‐Grao, T., Barquín, J., & Bonada, N. (2019).
Freshwater conservation planning informed and validated by public participation: The
Ebro catchment, Spain, as a case study. Aquatic Conservation: Marine and Freshwater
Ecosystems, 29, 1253-1267. https://doi.org/10.1002/aqc.3108
Cannon, P. F., Hywel-Jones, N. L., Maczey, N., Norbu, L., Tshitila, Samdrup, T., &
Lhundup, P. (2009). Steps towards sustainable harvest of Ophiocordyceps sinensis
in Bhutan. Biodiversity Conservation, 18, 2263- 2281.
Carrizo, S. F., Jähnig, S. C., Bremerich, V., Freyhof, J., Harrison, I., He, F., …Darwall,
W. (2017). Freshwater megafauna: flagships for freshwater biodiversity under threat.
BioScience, 67, 919-927. https://doi.org/10.1093/biosci/bix099
Carrizo, S. F., Lengyel, S., Kapusi, F., Szabolcs, M., Kasperidus, H. D., Scholz, M.
…Darwall, W. (2017). Critical catchments for freshwater biodiversity conservation in
Europe: identification, prioritisation and gap analysis. Journal of Applied Ecology, 54,
1209-1218. https://doi.org/10.1111/1365-2664.12842
Ceballos, G., & Ehrlich, P. R. (2006). Global mammal distributions, biodiversity
hotspots, and conservation. Proceedings of the National Academy of Sciences, 103,
19374–19379.
Clausnitzer, V., Kalkman, V. J., Ram, M., Collen, B., Baillie, J. E. M., Bedjanič, M.,
Darwall, W. R. T., Dijkstra, K. B., Dow, R., Hawking, J., Karube, H., Malikova,
E., Paulson, D., Schütte, K., Suhling, F., Villanueva, R. J., von Ellenrieder, N., & 122
Wilson, K. (2009). Odonata enter the biodiversity crisis debate: the first global
assessment of an insect group. Biological Conservation, 142, 1864-1869.
College of Natural Resources, Royal University of Bhutan (CNR-FISH). Available in
http://bhutanbiodiversity.net/collections/misc/collprofiles.php?collid=3&stat=taxo
nomy Downloaded on 21.01.2019.
Collen, B., van Strien, A., Scharlemann, J. P. W., Almond, R. E. A., Baillie, J. E. M.,
Bomhard, B., Brown, C., Bruno, J., Carpenter, K. E., M. Carr, G. M., Chanson, J.,
Chenery, A. M., Csirke, J., Davidson, N. C., Dentener, F., Foster, M., Galli,
A.,…Watson, R. (2010). Global biodiversity: indicators of recent declines. Science,
328, 1164-1168. DOI:10.1126/science.1187512.
Convention on Biological Diversity. (2010). Decision adopted by the Conference of the
Parties to the Convention on Biological Diversity at its Tenth Meeting. Decision
X/2. The Strategic Plan for Biodiversity 2011–2020 and the Aichi Targets. Nagoya,
Japan: Secretariat for the Convention on Biological Diversity.
Cook, C. N., Valkan, R. S., Mascia, M. B., & Mc Geoch, M. A. (2017). Quantifying the
extent of protected-area downgrading, downsizing and degazettement in Australia.
Conservation Biology, 31, 1039-1052.
Coxen, C. L., Frey, J. K., Carleton, S. A., & Collins, D. P. (2017). Species distribution
models for a migratory bird based on citizen science and satellite tracking data.
Global Ecology and Conservation, 11, 298-311.
https://doi.org/10.1016/j.gecco.2017.08.001
Cumberlidge, N., Ng, P. K. L., Yeo, D. C. J., Magalhaes, C., Campos, M. R., Alvarez, F.,
Naruse, T., Daniels, S. R., Esser, L. J., Attipoe, F. Y. K., Clotilde-Ba, F. L., Darwall,
W., McIvor, A., Ram, M.,& Collen, B. (2009). Freshwater crabs and the 123
biodiversity crisis: importance, threats, status, and conservation challenges.
Biological Conservation, 142, 1665–1673.
Darwall, W., Bremerich, V., De Wever, A., Dell A. I., Freyhof, J., Gessner, M. O.,
Grossart, H-P., Harrison, I., Irvine, K., Jähnig, S. C., Jeschke, J.M., Lee, J. J., Lu,
C., Lewandowska, A. M., Monaghan, M. T., Nejstgaard, J. C., Patricio, H.,
Schmidt-Kloiber, A., Stuart, s. N., …Weyl, O. (2018). The Alliance for Freshwater
Life: A global call to unite efforts for freshwater biodiversity science and
conservation. Aquatic Conservation: Marine and Freshwater Ecosystems, 28,
1015-1022. https://doi.org/10.1002/aqc.2958
Darwall, W., Holland, R. A., Smith, K. G., Allen, D., Brooks, E. G. E., Katarya,
V., Pollock, C. M., Shi, Y., Clausnitzer, V., Cumberlidge, N., Cuttelod,
A., Dijkstra, K.‐D. B., Diop, M. D., García, N., Seddon, M. B., Skelton, P. H.,
Snoeks, J., Tweddle, D., & Vié, J-C. (2011). Implications of bias in conservation
research and investment for freshwater species. Conservation Letters, 4, 474– 482.
https://doi.org/10.1111/j.1755-263X.2011.00202.x
Deguignet, M., Juffe-Bignoli, D., Harrison, J., MacSharry, B., Burgess, N. & Kingston,
N. (2014). 2014 United Nations List of Protected Areas. Cambridge, UK: UNEP-
WCMC.
Department of Forest and Park Services. (2015). Protected areas of Bhutan: National
parks, wildlife sanctuaries, strict nature reserves and biological corridors.
Thimphu, Bhutan: Department of Forest and Park Services.
Department of Forest. (2006). Forest and nature conservation rules of Bhutan, 2006.
Thimphu, Bhutan: Ministry of Agriculture, Royal Government of Bhutan.
124
Department of Hydropower & Power Systems. (2018). National transmission grid master
plan (NTGMP) of Bhutan-2018. Thimphu, Bhutan: Ministry of Economic Affairs,
Royal Government of Bhutan.
Department of Water Resources. (2015). Macroinvertebrates and fish assessment in
Kurichhu hydropower plant: part of EIA for augmentation project. Lamigoempa,
Bumthang, Bhutan: Ugyen Wangchuck Institute for Conservation and
Environment.
Dinerstein, E., Olson, D., Joshi, A., Vynne, C., Burgess, N. D., Wikramanayake, E., Hahn,
N., Palminteri, S., Hedao, P., Noss, R., Hansen, M., Locke, H., Ellis, E. C., Jones,
B., Barber, C. V., Hayes, R., Kormos, C., Martin, V., Crist, E.,… Saleem, M.
(2017). An ecoregion-based approach to protecting half the terrestrial realm.
BioScience, 67, 534–545.
Dinerstein, E., Vynne, C., Sala, E., Joshi, A. R., Fernando, S., Lovejoy, T.E., Mayorga,
J., Olson, D., Asner, G. P., Baillie, J. E. M., Burgess, N. D., Burkart, K., Noss, R.
F., Zhang, Y. P., Baccini, A., Birch, T., Hahn, N., Joppa, L. N., & Wikramanayake,
E. (2919). A Global Deal for Nature: Guiding principles, milestones, and targets.
Science Advances, 5, eaaw2869. 10.1126/sciadv.aaw2869
Domisch, S., Friedrichs, M., Hein, T., Borgwardt, F., Wetzig, A., Jähnig, S. C., &
Langhans, S. D. (2019). Spatially explicit species distribution models: A missed
opportunity in conservation planning? Diversity and Distributions, 25, 758–769.
doi:10.1111/ddi.12891
Domisch, S., Jähnig, S. C., Simaika, J. P., Kuemmerlen, M., & Stoll, S. (2015).
Application of species distribution models in stream ecosystems: the challenges of
spatial and temporal scale, environmental predictors and species occurrence data.
125
Fundamental and Applied Limnology / Archiv Für Hydrobiologie, 186, 45–61.
doi:10.1127/fal/2015/0627
Dorji, K. (2016). Utility of an existing biotic score method in assessing the stream health
in Bhutan. (Unpublished doctoral thesis). Brisbane, Australia: Queensland
University of Technology.
Dorji S. & Wangchuk T. (2014). Freshwater fishes of Royal Manas National Park.
Gelephu, Bhutan: Royal Manas National Park, Department of Forests and Park
Services.
Dorji, T. (2015). New distribution records of Epiophlebia laidlawi Tillyard, 1921
(Insecta: Odonata) in Bhutan. Journal of Threatened Taxa, 7, 7668–7675.
Dorji, T., Linke, S., & Sheldon, F. (2019). Half century of protected area dynamism in
the country of Gross National Happiness, Bhutan. Conservation Science and
Practice, 1, e46. https://doi.org/10.1111/csp2.46
Drouineau, H., Durif, C., Castonguay, M., Mateo, M., Rochard, E., Verreault, G.,
Yokouchi, K, & Lambert, P. (2018). Freshwater eels: A symbol of the effects of
global change. Fish and Fisheries, 19, 903-930. doi:10.1111/faf.12300
Dudgeon, D. (2010). Prospects for sustaining freshwater biodiversity in the 21st century:
linking ecosystem structure and function. Current Opinion in Environmental
Sustainability, 2, 422–430. DOI 10.1016/j.cosust.2010.09.001.
Dudgeon, D. (2019). Multiple threats imperil freshwater biodiversity in the
Anthropocene. Current Biology, 29, R942-R995.
Dudgeon, D., Arthington, A. H., Gessner, M. O., Kawabata, Z-I., Knowler, D. J.,
Lévêque, C., Naiman, R. J., Prieur Richard, A-H., Soto, D., Stiassny, M. L. J., & 126
Sullivan, C. A. (2006). Freshwater biodiversity: importance, threats, status and
conservation challenges. Biological Reviews, 81, 163–182.
Dudley, N. (2016). Bhutan: more than half. Available from
https://www.iucn.org/content/bhutan-more-half. Accessed on January 2018.
Elith, J., Graham, C. H., Anderson, R. P., Dudík, M., Ferrier, S., Guisan, A., Hijmans, R.
J., Huettmann. F., Leathwick, J. R., , Lehmann, A., Li, J., Lohmann, L. G., Loiselle,
B. A., Manion, G., Moritz, C., Nakamura, M., Nakazawa, Y., Overton, J. McC.,
Peterson, A. T., Phillips, S. J., … Zimmermann, N. E. (2006). Novel methods
improve prediction of species' distributions from occurrence data. Ecography, 29,
129-151.
Elith, J., Phillips, S. J., Hastie, T., Dudík, M., Chee, Y. E., & Yates, C. J. (2011).
A statistical explanation of Maxent for ecologists. Diversity and Distribution, 17,
43-57.
Esselman, P. C., & Allan, J. D. (2011). Application of species distribution models and
conservation planning software to the design of a reserve network for the riverine
fishes of northeastern Mesoamerica. Freshwater Biology, 56, 71–88.
Feng, M., Sexton, J. O., Channan, S., & Townshend, J. R. (2015). A global, high-
resolution (30-M) inland water body dataset for 2000: first results of a topographic–
spectral classification algorithm. International Journal of Digital Earth, 9, 113–
133. doi: 10.1080/17538947.2015.1026420
Ferreira, J., Aragão, L. E. O. C., Barlow, J., Barreto, P., Berenguer, E., Bustamante, M.,
Gardner, T. A., Lees, A. C., Lima, A., Louzada, J., Pardini, R., Parry, L., Peres, C.
A., Pompeu, P. S., Tabarelli, M., & Zuanon, J. (2014). Brazil's environmental
leadership at risk. Science, 346, 706–707. DOI: 10.1126/science.1260194 127
Fick, S. E., & Hijmans, R. J. (2017). worldClim 2: new 1-km spatial resolution climate
surfaces for global land areas. International Journal of Climatology, 37, 4302-4315.
DOI: 10.1002/joc.5086
Field, A. (2013). Discovering statistics using IBM SPSS statistics (4th ed.). London:
SAGE Publications Inc.
Fischer, D. T., Alidina, H. M., Steinback, C., Lombana, A. V., Ramirez de Arelano, P. I.,
Ferdana, Z, & Klein, C. J. (2010). Ensuring robust analysis. In Ardron, J.A.,
Possingham, H.P., & Klein, C.J. (Eds.), Marxan Good Practices Handbook,
Version 2. Victoria, BC, Canada: Pacific Marine Analysis and Research
Association.
Flitcroft, R., Cooperman, M. S., Harrison, I. J., Juffe-Bignoli, & Boon, P. J. (2019).
Theory and practice to conserve freshwater biodiversity in the Anthropocene.
Aquatic Conservation: Marine and Freshwater Ecosystems, 29, 1013-1021.
https://doi.org/10.1002/aqc.3187.
Forest Department. (1974). Notification dated 1st January 1974: Bhutan Fishing Rules.
Director General of Forest, Bhutan. Available from
https://www.ecolex.org/details/legislation/bhutan-fishing-rules-1974-lex-
faoc040223/. Accessed December 2017.
Forest Resource and Management Division. (2017). Land use and land cover of Bhutan
2016: maps and statistics. Thimphu, Bhutan: Department of Forest and Park
Services, Ministry of Agriculture and Forests.
Fuller, R. A., McDonald-Madden, E., Wilson, K. A., Carwardine, J., Grantham, H. S.,
Watson, J. E. M., Klein, C. J., Green, D. C., & Possingham, H. P. (2010). Replacing
128
underperforming protected areas achieves better conservation outcomes. Nature,
466, 365-367.
Galante, P. J., Alade, B., Muscarella, R., Jansa, S. A., Goodman, S. M., & Anderson, R.
P. (2018). The challenge of modelling niches and distributions for data-poor
species: a comprehensive approach to model complexity. Ecography, 41, 726-736.
DOI:10.1111/ecog.02909
Gatti, R. C. (2016). Freshwater biodiversity: a review of local and global threats.
International Journal of Environmental Studies, 73, 887–904.
https://doi.org/10.1080/00207233.2016.1204133
Giri, N., & Singh, O. P. (2013). Urban growth and water quality in Thimphu, Bhutan.
Journal of Urban and Environmental Engineering, 7, 82-95.
Golden Kroner, R. E., Krithivasan, R, & Mascia, M. B. (2016). Effects of protected area
downsizing on habitat fragmentation in Yosemite National Park (USA), 1864-2014.
Ecology and Society, 21, 22. http://dx.doi.org/10.5751/ES-08679-210322.
Golden Kroner, R. E., Qin, S., Cook, C. N., Krithivasan, R., Pack, S. M., Bonilla, O. D.,
Cort-kansinally, K. A., Coutinho, B., Feng, M., Garcia, M. I. M., He. Y., Kennnedy,
C. J., Lebreton, C., Ledezma, J. C., Lovejoy, T. E., Luther, D. A., Parmanand, Y.,
Ruíz-Agudelo, C. A., Yerena, E., ...Mascia, M. B. (2019). The uncertain future of
protected lands and waters. Science, 364, 881– 886. DOI: 10.1126/science.aau5525
Gomez, V. H. F., IJff, S. D., Raes, N., Amaral, I. L., Salomão, R. P., Coelho, L. d. S…,
Matos, F. D. d. A., Castilho, C. V., Filho, D. d. A. L., López, D. C., Guevara, J. E.,
Magnusson, w. E., Phillips, O L., Wittmann, F., Carim, M. d. J. V., martins, M. P.,
Irume, M. V., Sabatier, D., Molino, J-F., …Steege, H. T. (2018). Species
129
distribution modelling: contrasting presence-only models with plot abundance data.
Scientific Reports, 8, 1003. DOI:10.1038/s41598-017-18927-1
Grill, G., Lehner, B., Thieme, M., Geenen, B., Tickner, D., Antonelli, F., Babu, S.,
Borrelli, P., Cheng, L., Crochetiere, H., Ehalt Maedo, H., Filgueiras, R., Goichot,
M., Higgins, J., Hogan, Z., Lip, B., McClain, M. E., Meng, J., Mulligan, M., …
Zarfl, C. (2019). Mapping the world’s free-flowing rivers. Nature, 569, 215–221.
doi:10.1038/s41586-019-1111-9
Guisan, A., Tingley, R., Baumgartner, J. B., Naujokaitis-Lewis, I., Sutcliffe, P. R.,
Tulloch, A. I. T., Regan, T. J., Brotons, L., McDonald-Madden, E., Mantyka-
Pringle, C., Martin, T. G., Rhodes, J. R., Maggini, R., Setterfield, S. A., Elith, J.,
Schwartz, M. W., Wintle, B. A., Broennimann, O., Austin, …Buckley, Y. M.
(2013). Predicting species distributions for conservation decisions. Ecology Letters,
16, 1424–1435.
Gurung, D. B. & Thoni, R. J. (2015). Fishes of Bhutan: a preliminary checklist. Punakha,
Bhutan: Centre for Rural Development Studies (CRDS), College of Natural
Resources, Royal University of Bhutan.
Gurung, D. B., Dorji, S., Tshering, U. & Wangyal, J. T. (2013). An annotated checklist
of fishes from Bhutan. Journal of Threatened Taxa, 5, 4880–4886.
http://dx.doi.org/10.11609/JoTT.o3160.4880-6
Gyeltshen, C., Tobgay, K., Gyeltshen, N., Dorji, T., & Dema, S. (2018). New species
discoveries and records in Bhutan Himalaya. In M. Hartmann, M. V. L. Barclay
and J. Weipert. (Eds.), Biodiversität und Naturausstattungim Himalaya /
Biodiversity and Natural Heritage of the Himalaya, vol. VI. Erfurt:
Naturkundemuseum Erfurt.
130
Gyeltshen, T. (2016). A survey of Odonata from eastern Bhutan, with nine new national
records. Notulae odonatologicae, 8, 2017: 319-374.
Gyeltshen, T., Dorji, P., Nidup, T., Dorji, T., & Kalkman, V. J. (2017). Bibliography and
checklist of the dragonflies and damselflies of Bhutan. Journal of Threatened Taxa,
9, 9743-9747.
Gyeltshen, T., Kalkman, V. J., & Orr, A. G. (2017). Honouring His Royal Highness the
Crown Prince of Bhutan: Megalestes gyalsey (Odonata: Synlestidae). Zootaxa,
4244, 588-594. DOI: http://dx.doi.org/10.11646/zootaxa.4244.4.9
Hallgren, W., Santana, B., Low-Choy, S., Zhao, Y., & Mackey, B. (2019). Species
distribution models can be highly sensitive to algorithm configuration. Ecological
Modelling, 408, 108719. https://doi.org/10.1016/j.ecolmodel.2019.108719
Harrison, I., Abell, R., Darwall, W., Thieme, M. L., Tickner, D., & Timboe, I. (2018).
The freshwater biodiversity crisis. Science, 362, 1369.1–1369.
doi:10.1126/science.aav9242
Hayden, A. (2015). Bhutan: blazing a trail of a postgrowth future? Or stepping on the
treadmill of production? Journal of Environment and Development, 24, 161-186.
He, F., Zarfl, C., Bremerich, V., David, J. N. W., Hogan, Z., Kalinkat, G., Tockner, K.,
& Jähnig, S. C. (2019). The global decline of freshwater megafauna. Global Change
Biology, 25, 3883-3892. doi:10.1111/gcb.14753
He, F., Zarfl, C., Bremerich, V., Henshaw, A., Darwall, W., Tockner, K., & Jähnig, S. C.
(2017). Disappearing giants: a review of threats to freshwater megafauna. Wiley
Interdisciplinary Reviews: Water, 4, e1208. doi:10.1002/wat2.1208
131
Hermoso, H., Kennard, M. J., & Linke, S. (2012). Integrating multidirectional
connectivity requirements in systematic conservation planning for freshwater
systems. Diversity and Distribution, 18, 448-458. 10.1111/j.1472-
4642.2011.00879.x
Hermoso, V., Abell, R., Linke, S., & Boon, P. (2016). The role of protected areas for
freshwater biodiversity conservation: challenges and opportunities in a rapidly
changing world. Aquatic Conservation: Marine and Freshwater Ecosystems, 26
(Suppl. 1), 3-11. https://doi.org/10.1002/aqc.2681
Hermoso, V., Cattarino, L., Kennard, M. J. & Linke, S. (2015). Catchment zoning for
freshwater conservation: refining plans to enhance action on the ground. Journal of
Applied Ecology, 52, 940– 949. https://doi.org/10.1111/1365-2664.12454
Hermoso, V., Filipe, A. F., Segurado, P., & Beja, P. (2017). Freshwater conservation in
a fragmented world: Dealing with barriers in a systematic planning framework.
Aquatic Conservation: Marine and Freshwater Ecosystems, 28, 17-25,
https://doi.org/10.1002/aqc.2826
Hermoso, V., Linke, S., Prenda, J. & Possingham, H. P. (2011). Addressing longitudinal
connectivity in the systematic conservation planning of fresh waters. Freshwater
Biology, 56, 57–70.
Hoek, Y., Zuckerberg, B., & Manne, L. L. (2015). Application of habitat thresholds in
conservation: considerations, limitations, and future directions. Global Ecology and
Conservation, 3, 736-743. https://doi.org/10.1016/j.gecco.2015.03.010
Holland, R. A., Darwall, W. R. T., & Smith, K. G. (2012). Conservation priorities for
freshwater biodiversity: The Key Biodiversity Area approach refined and tested for
continental Africa. Biological Conservation, 148, 167-179. 132
IPBES. (2019). Summary for policymakers of the global assessment report on
biodiversity and ecosystem services of the Intergovernmental Science-Policy
Platform on Biodiversity and Ecosystem Services. Díaz, S., Settele, J., Brondízio,
E. S., Ngo, H. T., Guèze, M., Agard, J., Arneth, A., Balvanera, P., Brauman, K. A.,
Butchart, S. H. M., Chan, K. M. A., Garibaldi, L. A., Ichii, K., Liu, J., Subramanian,
S. M., Midgley, G. F., Miloslavich, P., Molnár, Z., Obura, D., …Zayas, C. N. (Eds.).
Bonn, Germany: IPBES secretariat.
Janßen, H., Göke, G., & Luttmann, A. (2019). Knowledge integration in Marine Spatial
Planning: A practitioners' view on decision support tools with special focus on
Marxan. Ocean and Coastal Management, 168, 230-138.
https://doi.org/10.1016/j.ocecoaman.2018.11.006
Januchowski-Hartley, S. R., Pearson, R. G., Puschendorf, R., & Rayner, T. (2011). Fresh
Waters and Fish Diversity: Distribution, Protection and Disturbance in Tropical
Australia. PLoS ONE, 6, e25846. https://doi.org/10.1371/journal.pone.0025846.
Jha, B. R., & Rayamajhi, A. (2010). Tor putitora. The IUCN Red List of Threatened
Species, e.T166645A6254146. http://dx.doi.org/10.2305/IUCN.UK.2010-
4.RLTS.T166645A6254146.en. Retrieved from
https://www.iucnredlist.org/species/166645/6254146
Jigme Singye Wangchuck National Park. (2013). Conservation management plan: Jigme
Singye Wangchuck National Park January 2014-December 2018. Trongsa, Bhutan:
Department of Forests and Park Services.
Juffe-Bignoli, D., Harrison, I., Butchart, S. H., Flitcroft, R., Hermoso, V., Jonas, H.,
Lukasiewicz, A., Thieme, M., Turak, E., Bingham, H., Dalton, J., Darwall, W.,
Deguignet, M., Dudley, N., Gardner, R., Higgins, J., Kumar, R., Linke, S., Milton,
133
G. R., …van Soesbergen, A. (2016). Achieving Aichi Biodiversity Target 11 to
improve the performance of protected areas and conserve freshwater biodiversity.
Aquatic Conservation: Marine and Freshwater Ecosystems, 26, 133–151.
doi:10.1002/aqc.2638
Kalkman, V. J. & Gyeltshen, T. (2016). Records of dragonflies from western Bhutan
collected in October 2015. International Dragonfly Fund Report, 94, 1-15.
Kass, J. M., Vilela, B., Aiello-Lammens, M. E., Muscarella, R., Merow, C., & Anderson,
R. P. (2017). WALLACE: a flexible platform for reproducible modelling of species
niches and distributions built for community expansion. Methods in Ecology and
Evolution, 9, 1151-1156. DOI.10.1111/2041-210x.12945.
Kingsford, R. T., Biggs, H. C., & Pollard, S. R. (2011). Strategic Adaptive Management
in freshwater protected areas and their rivers. Biological Conservation, 144, 1194–
1203.
Klein, C., Wilson, K., Watts, M., Stein, J., Berry, S., Carwardine, J., Smith, M. S.,
Mackey, B., & Possingham, H. (2009). Incorporating ecological and evolutionary
processes into continental scale conservation planning. Ecological Applications,
19, 206-217. https://doi.org/10.1890/07-1684.1
Kopnina, H. (2016). Half the earth for people (or more)? Addressing ethical questions in
conservation. Biological Conservation, 203, 176-185.
http://dx.doi.org/10.1016/j.biocon.2016.09.019
Kramer-Schadt, S, Niedballa, J., Pilgrim, J. D., Schroder, B., Lindenborn, J., Reinfelder,
V., Stillfried, M., Heckmann, I., Scharf, A.K., Augeri, D. M., Cheyne, S. M., Hearn,
A. J., Ross, J., Macdonald, D. W., Mathai, J., Eaton, J., Marshall, A. J., Semiadi,
G., Rustam, R., …Wilting, A. (2013). The importance of correcting for sampling 134
bias in MaxEnt species distribution models. Diversity and Distribution, 19, 1366-
1379. DOI: 10.1111/ddi.12096
Lehner, B., & Grill, G. (2013). Global river hydrography and network routing: baseline
data and new approaches to study the world’s large river systems. Hydrological
Processes, 27, 2171–2186. Data is available at www.hydrosheds.org.
Levin, N., Mazor, T., Brokovich, E., Jablon, P.E., & Kark, S. (2015). Sensitivity analysis
of conservation targets in systematic conservation planning. Ecological
Application, 25, 1997–2010. https://doi.org/10.1890/14-1464.1
Lewis, E., MacSharry, B., Juffe-Bignoli, D., Harris, N., Burrows, G., Kingston, N., &
Burgess, N. D. (2017). Dynamics in the global protected-area estate since 2004.
Conservation Biology, 33, 570-579. https://doi.org/10.1111/cobi.13056.
Lhamo, R. (2017). MoAF inaugurates Jomotshangkha Wildlife Sanctuary. Bhutan Times.
Available from https://www.pressreader.com/bhutan/bhutan-
times/20170212/281560880544735. Accessed January 2018.
Lhamo, D. (2018). The Rufford Foundation final report: Educating communities on the
conservation benefit of Golden Mahseer (Tor putitora) in Sherichu River, Mongar.
Available in https://www.rufford.org/files/20118-1Final%20Report.pdf. Accessed
January 2019.
Lindenmayer, D., Gibbons, P., Bourke, M., Burgman, M., Dickman, C., Ferrier, S.,
Fitzsimons, J., Freudenberger, D., Garnett, S., Groves, C., Hobbs, R., Kingsford,
R., Krebs, C., Legge, S., Lowe, A., McLean, R., Montambault, J., Possingham, H.,
Radford, J., ... Zerger, A. (2012). Improving biodiversity monitoring. Austral
Ecology, 37(3), 285-294. https://doi.org/10.1111/j.1442-9993.2011.02314.x
135
Linke, S., Hermoso, V., & Januchowski‐Hartley, S. (2019). Toward process‐based
conservation prioritizations for freshwater ecosystems. Aquatic Conservation:
Marine and Freshwater Ecosystems, 29, 1149-1160.
https://doi.org/10.1002/aqc.3162
Linke, S., Kennard, M. J., Hermoso, V., Olden, J. D., Stein, J. & Pusey, B. J. (2012)
Merging connectivity rules and largescale condition assessment improves
conservation adequacy in a tropical Australian river. Journal of Applied Ecology,
49, 1036– 1045.
Linke, S., Norris, R. H., & Pressey, R. L. (2008). Irreplaceability of river networks:
towards catchment-based conservation planning. Journal of Applied Ecology, 45,
1486-1495. doi: 10.1111/j.1365-2664.2008.01520.x
Linke, S., Turak, E., Asmyhr, M. G., & Hose, G. (2019). 3D conservation planning:
Including aquifer protection in freshwater plans refines priorities without much
additional effort. Aquatic Conservation: Marine and Freshwater Ecosystems, 29,
1063– 1072. https://doi.org/10.1002/aqc.3129
Linke, S., Turak, E., & Nel, J. (2011). Freshwater conservation planning: the case for
systematic approaches. Freshwater Biology, 56, 6-20. doi:10.1111/j.1365-
2427.2010.02456.x
Liu, C., Newell, G., & White, M. (2016). On the selection of thresholds for predicting
species occurrence with presence-only data. Ecology and Evolution, 6, 337–348.
Locke, H. (2013). Nature needs half: a necessary and hopeful new agenda for protected
areas. Parks, 19, 13-22. DOI: 10.2305/IUCN.CH.2013.PARKS-19-2.HL.en
136
Loos, S. (2011). Marxan analyses and prioritization of conservation areas for the Central
Interior Ecoregional Assessment. BC Journal of ecosystems and Management, 12,
88-97.
Malicky, H. (2007). Kocherfliegen aus Bhutan (Insecta, Trichoptera). Linzer Biologische
Beiträge, 39, 475–517.
Marco Junior, P. D., & Nóbrega, C. C. (2018). Evaluating collinearity effects on species
distribution models: an approach based on virtual species simulation. PLoS ONE,
13, e0202403. https://doi.org/10.1371/journal.pone.0202403
Martín-García, L., Sangil, C., Brito, A., & Barquín-Diez, J. (2015). Identification of
conservation gaps and redesign of island marine protected areas. Biodiversity and
Conservation, 24, 511-529. doi: 10.1007/s10531-014-0833-0
Mascia M. B., Pailler, S., Krithivasan, R., Roshchanka, V., Burns, D., Mlotha, M. J.,
Murray D. R., & Peng, N. (2014). Protected area downgrading, downsizing, and
degazettement (PADDD) in Africa, Asia, and Latin America and the Caribbean,
1900–2010. Biological Conservation, 169, 355–361.
Mascia, M. B., & Pailler, S. (2011). Protected area downgrading, downsizing, and
degazettement (PADDD) and its conservation implications. Conservation Letters,
4, 9-20. DOI: 10.1002/aqc.2638.
Mascia, M. B., Pailler, S., & Krithivasan, R. (2012). PADDDtracker.org technical guide
(version 1). Washington, DC: World Wildlife Fund.
Mascia, M. B., Pailler, S., Krithivasan, R., Roshchanka, V., Burns, D., Mlotha, M. J.,
Murray, D. R., & Peng, N. (2014). Protected area downgrading, downsizing, and
137
degazettement (PADDD) in Africa, Asia, and Latin America and the Caribbean,
1900-2010. Biological Conservation, 169, 355–361.
McGarvey, D. J., Menon, M., Woods, T., Tassone, S., Reese, J., Vergamini, M., &
Kellogg, E. (2018). On the use of climate covariates in aquatic species distribution
models: are we at risk of throwing out the baby with the bath water? Ecography,
41, 695-712.
Merow, C., Smith, M. J., & Silander, Jr J. A. (2013). A practical guide to Maxent for
modeling species’ distributions: what it does, and why inputs and settings matter.
Ecography, 36, 1058-1069.
Messager, M.L., Lehner, B., Grill, G., Nedeva, I., & Schmitt, O. (2016): Estimating the
volume and age of water stored in global lakes using a geo-statistical approach.
Nature Communications, 7, 13603. doi:10.1038/ncomms13603.
Ministry of Agriculture. (1992). Bhutan fishing (Amendment) rules 1992. Ministry of
Agriculture, Royal Government of Bhutan. Available from
https://www.ecolex.org/details/legislation/bhutan-fishing-amendment-rules-1992-
lex-faoc023793/. Accessed December 2017.
Mitra, A. (2008). Dragonfly fauna of Bhutan – an annotated and updated check-list with
ten new records. Fraseria (N.S.), 7, 105-109.
Mitra, A., Choden, K., Dorji, Y., Penjor, T., Dorji, R., Subedi, K., & Dorji, P. (2012).
Odonata of Samdrup Choling Dungkhag in Samdrup Jongkhar, Bhutan. Bhutan
Journal of Research & Development, 1, 125-141.
Mitra, A., Dow, R., Subramanian, K. A., & Sharma, G. (2010). Chapter 5. The status and
distribution of dragonflies and damselflies (Odonata) of the Eastern Himalaya. In
138
D. J. Allen, S. Molur & B. A. Daniel. (Compilers). The status and distribution of
freshwater biodiversity in the Eastern Himalaya (pp. 54-66). Cambridge, UK and
Gland, Switzerland: IUCN, and Coimbatore, India: Zoo Outreach Organisation.
Mool, P. K., Wangda, D., Bajracharya, S. R., Kunzang, K., Gurung, D. R. & Joshi, S.P.
(2001). Inventory of glaciers, glacial lakes and glacial lake outburst floods:
Monitoring and early warning systems in the Hindu Kush-Himalayan Region
Bhutan. Kathmandu, Nepal: International Centre for Integrated Mountain
Development.
Morales, N. S., Fernández, I. C., & Baca-González, V. (2017). MaxEnt's parameter
configuration and small samples: are we paying attention to recommendations? A
systematic review. PeerJ, 5, e3093. doi:10.7717/peerj.3093
Mukaka, M. M. (2012). Statistics corner: a guide to appropriate use of correlation
coefficient in medical research. Malawi Medical Journal, 24, 67-71.
Muscarella, R., Galante, P. J., Soley-Guardia, M., Boria, R. A., Kass, J. M., Uriarte, M.
and Anderson, R. P. (2014). ENMeval: an R package for conducting spatially
independent evaluations and estimating optimal model complexity for Maxent
ecological niche models. Methods in Ecology and Evolution, 5, 1198-1205.
Doi:10.1111/2041-210X.12261.
National Biodiversity Centre. (2014). National Biodiversity Strategies and Action Plan
of Bhutan. Thimphu, Bhutan: Ministry of Agriculture and Forests, Royal
Government of Bhutan.
National Environment Commission. (2016). Bhutan state of the environment report 2016.
Thimphu, Bhutan: NEC, Royal Government of Bhutan.
139
National Environment Commission. (2016). National integrated water resources
management plan 2016. Thimphu, Bhutan: NEC, Royal Government of Bhutan.
National Research Centre for Riverine & Lake Fisheries (DOL-NRCRLF). Available in
http://bhutanbiodiversity.net/collections/misc/collprofiles.php?collid=15.
Accessed January 2019.
National Research Centre for Riverine and Lake Fisheries. (2017). Fishes of Western
Bhutan: assessment of species composition and distribution of fish in Bhutan’s
three major river basins (Amochhu, Punatsangchhu & Wangchhu). Haa, Bhutan:
Department of Livestock, Ministry of Agriculture & Forests.
National Statistical Bureau. (2019). Bhutan at a glance 2019. Thimphu, Bhutan: National
Statistical Bureau.
National Biodiversity Centre. (2009). Bhutan-Biodiversity Action Plan 2009. Thimphu,
Bhutan: Ministry of Agriculture, Royal Government of Bhutan.
National Biodiversity Centre. (2014). National biodiversity strategies and action plan of
Bhutan, 2014. Thimphu, Bhutan: Ministry of Agriculture and Forests, Royal
Government of Bhutan.
Nature Conservation Division. (2004). Bhutan Biological Conservation Complex: A
Landscape Conservation Plan - Way Forward. Thimphu, Bhutan: Department of
Forestry Services, Ministry of Agriculture, Thimphu.
Nel, J. L., Reyers, B., Roux, D. J., & Cowling, R. M. (2009). Expanding protected areas
beyond their terrestrial comfort zone: identifying spatial options for river
conservation. Biological Conservation, 142, 1605-1616.
140
Nhancale, B. A., & Smith, R. J. (2011). The influence of planning unit characteristics on
the efficiency and spatial pattern of systematic conservation planning assessments.
Biodiversity Conservation, 20, 821–1835. https://doi.org/10.1007/s10531-011-
0063-7
Norris, D. (2014). Model thresholds are more important than presence location type:
understanding the distribution of lowland tapir (Tapirus terrestris) in a continuous
Atlantic forest of southeast Brazil. Tropical Conservation Science, 7, 529-547.
Ofenböck, T., Moog, O., Sharma, S., & Korte, T. (2010). Development of the HKHbios:
a new biotic score to assess the river quality in the Hindu Kush-Himalaya.
Hydrobiologia, 651, 39-58.
Ohler, A., Kumar Shrestha, T., & Bordoloi, S. (2008). Nanorana annandalii. The IUCN
Red List of Threatened Species 2008: e.T58417A11777219.
https://dx.doi.org/10.2305/IUCN.UK.2008.RLTS.T58417A11777219.en.
Accessed February 2020.
Ouellet-Dallaire, C. O., Lehner, B., Sayre, R., & Thieme, M. (2019). A multidisciplinary
framework to derive global river reach classifications at high spatial resolution.
Environmental Research Letters, 14, 024003. https://doi.org/10.1088/1748-
9326/aad8e9
Pack, S. M., Ferreira, M. N., Krithivasan, R., Murrow, J., Bernard, E., & Mascia, M. B.
(2016). Protected area downgrading, downsizing, and degazettement (PADDD) in
the Amazon. Biological Conservation, 197, 32–39.
Palden, T. (2016). Gangtey-Phobji becomes third Ramsar site. Available from
http://www.kuenselonline.com/gangtey-phobji-becomes-third-ramsar-site/
Accessed January 2018. 141
Pearson, R. G., Raxworthy, C. J., Nakamura, M., & Peterson, A. T. (2007). Predicting
species distributions from small numbers of occurrence records: a test case using
cryptic geckos in Madagascar. Journal of Biogeography, 34, 102-117.
Pearson, K. D. (2018). Rapid enhancement of biodiversity occurrence records using
unconventional specimen data. Biodiversity and Conservation, 27, 3007-3018.
https://doi.org/10.1007/s10531-018-1584-0
Pereira, D. G., Afonso, A., & Medeiros, F. M. (2015). Overview of Friedman’s test and
post-hoc analysis. Communications in Statistics - Simulation and Computation, 44,
10, 2636-2653, DOI: 10.1080/03610918.2014.931971
Phillips, A. (2004). The history of the international system of protected area management
categories. Parks, 14, 4-14.
Phillips, S. J. (2017). A brief tutorial on Maxent. Available from
http://biodiversityinformatics.amnh.org/open_source/maxent/. Accessed February
2018.
Phillips, S. J., Anderson, R. P., Dudík, M., Schapire, R. E., & Blair, M. E. (2017). Opening
the black box: an open-source release of Maxent. Ecography, 40, 887-893.
https://doi.org/10.1111/ecog.03049
Phillips, S. J., Dudík, M., Elith, J., Graham, C. H., Lehmann, A., Leathwick, J., & Ferrier,
S. (2009). Sample selection bias and presence-only distribution models:
implications for background and pseudo-absence data. Ecological Applications, 19,
181-197. DOI: 10.1890/07-2153.1
Phillips, S. J. & Dudík, M. (2008). Modeling of species distributions with Maxent: new
extensions and a comprehensive evaluation. Ecography, 31, 161-175.
142
Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximum entropy modeling of
species geographic distributions. Ecological Modelling, 190, 231–259.
Phillips, S.J., Dudík, M., & Schapire, R. E. [Internet] Maxent software for modeling
species niches and distributions (Version 3.4.1). Available from url:
http://biodiversityinformatics.amnh.org/open source/maxent/. Accessed February
2018.
Phuntsho. (n.d.). Assessment of Nika Chu freshwater ecology under Jigme Singye
Wangchuck National Park through using fresh water fish diversity and habitats as
bio-indicator. Available in https://www.rufford.org/files/16189-
1%20Detailed%20Final%20Report.pdf. Accessed January 2018.
Pimm, S. L., Jenkins, C. N., & Li, B. V. (2018). How to protect half of Earth to ensure it
protects sufficient biodiversity. Science Advances, 4, eaat2616 (2018).
Pimm, S. L., Jenkins, C. N., Abell, R., Brooks, T.M., Gittleman, J. L., Joppa, L. N.,
Raven, P. H., Roberts, C. M., & Sexton, J. O. (2014). The biodiversity of species
and their rates of extinction, distribution, and protection. Science, 344, 1246752.
DOI: 10.1126/ science.1246752
Pittock, J., Finlayson, M., Arthington, A. H., Roux, D., Matthews, J. H., Biggs, H.,
Harrison, I., Hermoso, V., Junk, W., Kumar, R., Linke, S., Nel, J., Cunha, C. N. d.,
Pattnaik, A., Pollard, S., Rast, W., Thieme, M., Turak, E., Turpie, J., …Viers, J.
(2015). Managing freshwater, river, wetland and estuarine protected areas. In G. L.
Worboys, M. Lockwood, A. Kothari, S. Feary & I. Pulsford (eds.), Protected area
governance and management. Canberra: ANU press.
143
Poppe, K. L., Biffard, D. A., & Stevens, V. (2016). Freshwater aquatic ecosystem
representation in British Columbia's provincial protected area system. Natural
Areas Journal, 36, 268-276.
Premkumar, L. (2016). A study of the India-Bhutan energy cooperation agreements and
the implementation of hydropower projects in Bhutan. New Delhi, India: Vasudha
Foundation.
Radosavljevic, A., & Anderson, R. P. (2014). Making better Maxent models of species
distributions: complexity, overfitting and evaluation. Journal of Biogeography, 41,
629-643. https://doi.org/10.1111/jbi.12227
Raghavan, R., Das, S., Nameere, P.O., Bijukumar, A., & Dahanukar, N. (2016). Protected
areas and imperilled endemic freshwater biodiversity in the Western Ghats Hotspot.
Aquatic Conservation: Marine and Freshwater Ecosystems, 26 (Suppl. 1), 78–90.
DOI:10.1002/aqc.2653
Rai, M. R. (2017). The Rufford Foundation final report: Study on Natural Habitat of
Globally Endangered Golden Mahseer (Tor putitora) and its conservation for
ecological, economical, and cultural importance. Available in
https://www.rufford.org/files/19791-1%20Final%20Report.pdf.
Rai, R. (2016, June 21). Work on Druk Bindu hydropower begins. Kuensel.
https://kuenselonline.com/work-on-druk-bindu-hydropower-begins/
Ramsar. (2012). Bhutan’s first two Ramsar Sites. https://www.ramsar.org/news/bhutans-
first-two-ramsar-sites.
Reid, A. J., Carlson, A. K., Creed, I. F., Eliason, E. J., Gell, P. A., Johnson, P. T. J., Kidd,
K. A., MacCormack, T. J., Olden, J. D., Ormerod, S. J., Smol, J. P., Taylor, W.
144
W., Tockner, K., Vermaire, J. C., Dudgeon, D., & Cooke, S. J. (2019). Emerging
threats and persistent conservation challenges for freshwater biodiversity.
Biological Review, 94, 849-873. DOI:10.1111/brv.12480
Reis, V., Hermoso, V., Hamilton, S. K., Bunn, S. E., & Linke, S. (2019). Conservation
planning for river-wetland mosaics: A flexible spatial approach to integrate
floodplain and upstream catchment connectivity. Biological Conservation, 236,
356-365. https://doi.org/10.1016/j.biocon.2019.05.042
Reis, V., Hermoso, V., Hamilton, S. K., Ward, D., Fluet-Chouinard, E., Lehner, B., &
Linke, S. (2017). A global assessment of inland wetland conservation status.
BioScience, 67, 523-533. https://doi.org/10.1093/biosci/bix045
Renewable Natural Resource Research and Development Centre - Bajothang. (2012).
Annual Report 2011-2012. Wangdue Phodrang, Bhutan: Department of
Agriculture, Ministry of Agriculture and Forests.
Rodríguez–Olarte, D. Taphorn, D. C., & Lobón–Cerviá, J. (2011). Do protected areas
conserve neotropical freshwater fishes? A case study of a biogeographic province in
Venezuela. Animal Biodiversity and Conservation, 34, 273–285.
Roux, D.J., Nel, J.L., Ashton, P.J., Deacon, A.R., de Moor, F.C., Hardwick, D., Hill, L.,
& Scholes, R.J. (2008). Designing protected areas to conserve riverine biodiversity:
Lessons from a hypothetical redesign of the Kruger National Park. Biological
Conservation, 141, 100 -117. doi:10.1016/j.biocon.2007.09.002
Royal Government of Bhutan and World Bank. (2016). Managing environmental and
social impacts of hydropower in Bhutan: final report. Washington DC, USA: World
Bank Group.
145
Royal Manas National Park. (2015). Conservation Management Plan, Royal Manas
National Park (July 2015- June 2020). Thimphu, Bhutan: Department of Forest and
Park Services, Ministry of Agriculture and Forest.
Seeland, K. (1998). The national park management regime in Bhutan: historical
background and current problems. Worldviews: Environment, Culture, Religion, 2,
139-153.
Seeland, K. (2000). National park policy and wildlife problems in Nepal and Bhutan.
Population and Environment: A Journal of Interdisciplinary Studies, 22, 43-62.
Shcheglovitova, M., & Anderson, R. P. (2013). Estimating optimal complexity for
ecological niche models: a jackknife approach for species with small sample sizes.
Ecological Modelling, 269, 9-17.
http://dx.doi.org/10.1016/j.ecolmodel.2013.08.011
Sherub, Wangdi, N., & Norbu, N. (2010). Saving wetland sky high: inventory of high
altitude wetlands in Bhutan. Bumthang, Bhutan: Ugyen Wangchuck Institute of
Conservation and Environment.
Stuart, S. N., Hoffmann, M., Chanson, J. S., Cox, N. A., Berridge, R. J., Ramani, P., &
Young, B. E. (Eds.) (2008). Threatened Amphibians of the World. Barcelona,
Spain: Lynx Edicions; Gland, Switzerland: IUCN; and Arlington, Virginia, USA:
Conservation International.
Syfert, M. M., Smith, M. J., & Coomes, D. A. (2013). The effects of sampling bias and
model complexity on the predictive performance of MaxEnt species distribution
models. PLoS ONE, 8, e55158. Doi:10.1371/journal.pone.0055158.
146
Symes, W.S., Rao, M., Mascia, M.B., & Carrasco, L.R. (2016). Why do we lose protected
areas? Factors influencing protected area downgrading, downsizing and
degazettement in the tropics and subtropics. Global Change Biology, 22: 656-665.
Tallis, H. M., Hawthorne, P. L., Polasky, S., Reid, J., Beck, M. V., Brauman, K., Bielicki,
J. M., Binder, S., Burgess, M. G., Cassidy, E., Clark, A., Fargione, J., Game, E. T.,
Gerber, J., Isbell, F., Kiesecker, J., McDonald, R., Metian, M., Molnar, J.
L.,…McPeek, B. (2018). An attainable global vision for conservation and human
well-being. Frontiers in Ecology and the Environment, 16, 563–570. DOI:
10.1002/fee.1965
Tesfaw, A. T., Pfaff, A., Golden Kroner, R. E., Qin, S., Medeiros, R., Mascia, M.
B. (2018). Land‐use and land‐cover change shape the sustainability and impacts of
protected areas. Proceedings of the National Academy of Sciences, 115,
2084– 2089.
Thapa, S. (2017). Bhutan’s Cordyceps collectors eye return to yak herding. Available
from https://www.thethirdpole.net/2017/09/21/bhutans-cordyceps-collectors-eye-
return-to-yak-herding/. Accessed on February 2018
The Constitution of the Kingdom of Bhutan 2008, adopted on 18th July, 2008.
Thinley, P., Tharchen, L., & Dorji, R. (2015). Conservation management plan of Jigme
Dorji National Park for the period January 2015 to December 2019: biodiversity
conservation in pursuit of Gross National Happiness. Thimphu, Bhutan:
Department of Forest and Park Services.
Thoni, R. J. & Gurung, D. B. (2014). Parachiloglanis bhutanensis, a new species of
torrent catfish (Siluriformes: Sisoridae) from Bhutan. Zootaxa, 3869, 306-312.
147
Thoni, R. J. & Gurung, D. B. (2018). Morphological and molecular study of the torrent
catfishes (Sisoridae: Glyptosterninae) of Bhutan including the description of five
new species. Zootaxa, 4476, 040-068.
Thoni, R. J., Gurung, D. B., & Mayden, R. L. (2016). A review of the genus Garra
Hamilton, 1822 of Bhutan, including the descriptions of two new species and three
additional records (Cyprineformes: Cyprinidae). Zootaxa, 4169, 115-132.
Tickner, D., Opperman, J. J., Abell, R., Acreman, M., Arthington, A. H., Bunn, S. E.,
Cooke, S. J., Dalton, J., Darwall, W., Edwards, G., Harrison, I., Hughes, K., Jones,
T., Leclere, D., Lynch, A. LJ., Leonard, P., McClain, M. E., Muruven, D., Olden,
J. D., …Young, L. (2020). Bending the Curve of Global Freshwater Biodiversity
Loss: An Emergency Recovery Plan. BioScience, doi:10.1093/biosci/biaa002
Tognelli, M. F., Anderson, E. P., Jiménez-Segura, L. F., Chuctaya, J., Chocano, L.,
Maldonado-Ocampo, J. A., Mesa-Salazar, L., Mojica, J. I., Carvajal-Vallejos, F.
M., Correa, V., Ortega, H., Romero, J. F. R., Sánchez-Duarte, P., Cox, N. A.,
Hidalgo, M., Prado, P. J., Lasso, C. A., Sarmiento, J., Velásquez, M.A., & Villa-
Navarro, F. A. (2019). Assessing conservation priorities of endemic freshwater
fishes in the Tropical Andes region. Aquatic Conservation: Marine and Freshwater
Ecosystems, 29, 1123-1132. https://doi.org/10.1002/aqc.2971
Tshering, K. (2011). A Roadmap for watershed management in Bhutan: From policy to
practice. Thimphu, Bhutan: Watershed Management Division, Department of
Forest and Park Services, Ministry of Agriculture and Forest.
Turak, E., Harrison, I., Dudgeon, D., Abell, R., Bush, A., Darwall, W., Finlayson, C. M.,
Ferrier, S., Feyhof, J., Hermoso, V., Juffe-Bignoli, D., Linke, S., Nel, J., Patricio,
H. C., Pittock, J., Raghavan, R., Revenga, C., Simaika, J. P., & De Wever, A.
148
(2017). Essential Biodiversity Variables for measuring change in global freshwater
biodiversity. Biological Conservation, 213, 272–279.
https://doi.org/10.1016/j.biocon.2016.09.005
United Nation Environment Programme-World Conservation Monitoring Centre. (2018).
Protected area country profile for Bhutan from the World Database of protected
areas, February 2018. Available from www.protectedplanet.net. Accessed on
January 2018.
United Nations Environment Programme-World Conservation Monitoring Centre, &
International Union for Conservation of Nature. (2016). Protected Planet Report
2016. Cambridge UK: UNEP-WCMC, and Gland, Switzerland: IUCN.
Ura, K., Alkire, S., Zangmo, T., & Wangdi, K. (2012). An extensive analysis of GNH
index. Thimphu, Bhutan: Centre for Bhutan Studies.
UWICER. (2017). Aquatic biodiversity assessment: a baseline study in Mangdechhu,
Central Bhutan. Lamai Goempa, Bumthang, Bhutan: UWICER Press.
http://www.uwice.gov.bt/admin uwice/publications/publication files/Reports/201
7/MHPA_Report_2017.pdf
Venter, O., Fuller, R .A., Segan, D. B., Carwardine, J., Brooks, T., Butchart, S. H. M.,
Marco, M. D., Iwanmur, t., Joseph, L., O՛Grady, D., Possingham, H. P., Rondinini,
C., Smith, R. J., Venter, M., & Watson, J. E. M. (2014). Targeting global protected
area expansion for imperiled biodiversity. PLoS Biology, 12, e1001891.
https://doi.org/10.1371/journal.pbio.1001891
Venter, O., Magrach, A., Outram, N., Klein, C. J., Possingham, H. P., Di Marco, M., &
Watson, J. E. M. (2017). Bias in protected-area location and its effects on long-term
149
aspirations of biodiversity conventions. Conservation Biology, 32, 127–134.
https://doi.org/10.1111/cobi.12970
Venter, O., Sanderson, E. W., Magrach, A., Allan, J. R., Beher, J., Jones, K. R.,
Possingham, H. P., Laurance, W. F., Wood, P., Fekete, B. M., Levy, M. A., &
Watson, J. E. (2018). Last of the Wild Project, Version 3 (LWP-3): 2009 Human
Footprint, 2018 Release. Palisades, NY: NASA Socioeconomic Data and
Applications Center (SEDAC). https://doi.org/10.7927/H46T0JQ4.
Venter, O., Sanderson, E.W., Magrach, A., Allan, J. R., Beher, J., Jones, K. R.,
Possingham, H. P., Laurance, W. F., Wood, P., Fekete, B. M., Levy, M. A., &
Watson, J. E. (2016). Global terrestrial Human Footprint maps for 1993 and 2009.
Scientific Data, 3, 160067 (2016). https://doi.org/10.1038/sdata.2016.67
Verpoorter, C., Kutser, T., Seekell, D. A., & Tranvik, L. J. (2014). A global inventory of
lakes based on high-resolution satellite imagery, Geophysical Research Letters, 41,
6396–6402. https://doi.org/10.1002/2014GL060641
Visconti, P., Butchart, S. H. M., Brooks, T. M., Langhammer, P. F., Marnewick, D.,
Vergara, S., Yanosky, A., & Watson, J. E. M. (2019). Protected area targets post-
2020. Science, 364, 239-241. DOI: 10.1126/science.aav6886
Vollering, J., Halvorsen, R., Auestad, I., & Rydgren, K. (2019). Bunching up the
background betters bias in species distribution models. Ecography, 42, 1717-1727.
Doi:10.1111/ecog.04503
Wangchuck Centennial National Park. (n.d.) Report on fish diversity in Wangchuck
Centennial National Park. Nasiphel, Bumthang, Bhutan: Ministry of Agriculture
and Forest, Department of Forest and Park Services.
150
Wangchuk, S., Lham, D., Dudley, N., & Stolton, S. (2017). Half-Bhutan: the evolution
and effectiveness of protected areas in a country recognizing nature needs half.
International Journal of Wilderness, 23, 43-46.
Wangchuk, S., Norbu, N., & Sherub. (2012). Impacts of Cordyceps collection on
livelihoods and alpine ecosystems in Bhutan as ascertained from questionnaire
survey of Cordyceps collectors. Bumthang, Bhutan: UWICE Press.
Wangchuk, S., & Wangdi, G. (2015). Mountain pastoralism in transition: consequences
of legalizing Cordyceps collection on yak farming practices in Bhutan.
Pastoralism: Research, Policy and Practice, 5, 1-10. DOI 10.1186/s13570-015-
0025-x.
Wangda, P. & Gyeltshen, D. (2018). Peling-Tso, a low-altitude wetland and its
management practices, Southern Bhutan. In Bhatta, L. D., Wu, N., Udas, E.,
Agrawal, N. K., Ranabhat, S., & Basnet, D. (Eds.), Wetlands in the Himalaya:
securing services for livelihoods. Kathmandu: ICIMOD.
Wangdi, S. (2014, December 27). Shrinking Kharul lake saddens locals. Bhutan
Broadcasting Service. Retrieved from http://www.bbs.bt/news/?p=35313
Wangyal, J. T. & Gurung, D. B. (2012). Amphibians of Punakha-Wangdue Phodrang
valley, Bhutan. Frog leg: Newsletter of the Amphibian Network of South Asia and
Amphibian Specialist Group - South Asia, 18, 31-44.
Wangyal, J. T. (2013). New records of reptiles and amphibians from Bhutan. Journal of
Threatened Taxa, 5, 4774–4783.
151
Warren, D. L., & Seifert, S. N. (2011). Ecological niche modelling in Maxent: the
importance of model complexity and the performance of model selection criteria.
Ecological Applications, 21, 335-342. https://doi.org/10.1890/10-1171.1
Watson, J. E. M., Darling, E. S., Venter, O., Maron, M., Walston, J., Possingham, H. P.,
Dudley, N., Hockings, M., Barnes, M., & Brooks, T. M. (2016). Bolder science
needed now for protected areas. Conservation Biology, 30, 243–248.
Watson, J. E. M., Evans, M. C., Carwardine, J., Fuller, R. A., Joseph, L. N., Segan, D. B.,
Taylor, M. F., Fensham, R J., & Possingham, H. P. (2010). The capacity of
Australia’s protected-area system to represent threatened species. Conservation
Biology, 25, 324-332. https://doi.org/10.1111/j.1523-1739.2010.01587.x
Watson, J. E. M., Dudley, N., Segan, D. B. & Hockings, M. (2014). The performance and
potential of protected areas. Nature, 515, 67-73, doi:10.1038/nature13947.
Wildlife Conservation Division. (2010). Regulatory framework for biological corridors
in Bhutan. WCD, Department of Forests and Park Services, Thimphu, Bhutan.
Wildlife Conservation Division. (2016). Summary: Bhutan’s state of parks report 2016.
Thimphu, Bhutan: WWF and WCD, Department of Forest and Parks Services,
Ministry of Agriculture and Forest.
Woodley, S., Bertzky, B., Crawhall, N., Dudley, N., Londoño, J. M., MacKinnon, K.,
Redford, K., & Sandwith, T. (2012). Meeting Aichi target 11: what does success
look like for protected area systems? Parks, 18, 23-36.
World Bank Group. (2016). Managing environmental and social impacts of hydropower
in Bhutan: Final report. Washington DC, USA: WBG.
152
World Wide Fund for Nature. (2016). Living Planet Report 2016. Risk and resilience in
a new era. WWF International, Gland, Switzerland.
World Wide Fund for Nature. (2018). Living Planet Report 2018: Aiming Higher. Gland,
Switzerland: WWF International.
World Wide Fund for Nature. (2017). PADDDtracker: tracking protected
area downgrading, downsizing, and degazettement [Beta version]. Available from
www.PADDDtracker.org. Accessed on January 2017.
World Wildlife Fund Bhutan, & Sakteng Wildlife Sanctuary. (2011). Participatory
zoning for Sakteng Wildlife Sanctuary: Balancing Conservation and Development
goals. Thimphu, Bhutan: WWF Bhutan and Department of Forests and Park
Services, Ministry of Agriculture and Forests.
Worm, B. (2017). How to heal an ocean. Nature, 543, 630-631. doi:10.1038/nature21895.
Young, N., Carter, L., & Evangelista, P. (2011). A maxent model v3.3.3e tutorial (ArcGIS
v10). Last modified on September 1, 2011. Natural Resources Ecology Laboratory
at Colorado State University and the National Institute of Invasive Species Science.
Available in
http://ibis.colostate.edu/webcontent/ws/coloradoview/tutorialsdownloads/a maxe
nt_model_v7.pdf. Accessed on January 2018.
Zurick, D. (2006). Gross National Happiness and environmental status in Bhutan. The
Geographic Review, 4, 657-681.
153
8 Supplementary materials
8.1 Supplementary materials for Chapter 2
Table S2.1 Summary of changes to protected area system of Bhutan between 1966 and 2016. (aProtected Area Downgrading, Downsizing and Degazettement; bProtected areas; cArea affected unknown for 3 policy upgrade events; dArea affected by all policy downgrade events unknown; e14 were proposed and 6 enacted policy downgrade events.) Avaialble from https://conbio.onlinelibrary.wiley.com/doi/abs/10.1111/csp2.46 as Supporting Information csp246-sup-0001-AppendixS1.xlsx
Table S2.2 Protected areas of Bhutan affected by gain events between 1966 and 2016 with gain event types (enlarged PA and policy upgrade), current, primary and after gain event names of protected areas, year of event and area affected. Available from https://conbio.onlinelibrary.wiley.com/doi/abs/10.1111/csp2.46 as Supporting Information csp246-sup-0002-AppendixS2.xlsx
Table S2.3. Protected areas and their areas from different data sources and consolidated area.
Sl. No. PA Namea WMDb NBCc WDPAd WDPAe Consolidatedf 1 BWS 1523.9 1520.6 1528.7 1521.0 1523.9 2 JDNP 4377.1 4316.0 4338.4 4316.0 4377.1 3 JSWNP 1733.7 1730.0 1740.4 1730.0 1733.7 4 JWS 351.3 334.7 336.8 334.0 351.3 5 PWS 269.5 268.9 270.6 269.0 269.5 6 RMNP 1059.6 1057.0 1035.7 1057.0 1059.6 7 SWS 741.6 740.6 744.9 741.0 741.6 8 PNP 907.2 905.1 910.5 905.0 907.2 9 JKSNR 610.8 609.5 613.0 609.0 610.8 10 WCNP 4921.5 4914.0 4939.6 4921.0 4921.5 11 JDNP-JSWNPg 276.0 NAh 277.0 275.0 276.0 12 JSWNP-RMNP-PWS 376.6 NA 378.1 376.0 376.6 13 KWS-SWS 160.4 NA 160.9 160.0 160.4 14 NC 935.6 NA 939.1 720.0 935.6 15 RMNP-KWS 212.1 NA 212.9 212.0 212.1 16 TNP-BWS 79.2 NA 79.5 79.0 79.2 17 TNP-JSWNP-RMNP 499.9 NA 503.5 501.0 499.9 18 JKSNR-JDNP 149.8 3307.14i 150.3 149.0 149.8 19 RBP NA 47.0 NA NA OMj 20 GPR 9.8 0.0 9.8 9.7 9.8 21 KR 1.2 0.0 NA 1.1 1.2 22 BR 1.4 0.0 NA 1.4 OM
Total (km2) 19197.9 19750.6 19169.7 18887.3 19196.5 154
Percentage (%) 50.0 51.4 49.9 49.2 50.0 aBWS: Bumdeling Wildlife Sanctuary; JDNP: Jigme Dorji National Park; JSWNP: Jigme Singye Wangchuck National Park; JWS: Jomotshangkha Wildlife Sanctuary; PWS: Phibsoo Wildlife Sanctuary; RMNP: Royal Manas National Park; SWS: Sakteng Wildlife Sanctuary; PNP: Phrumsengla National Park; JKSNR: Jigme Khesar Strict Nature Reserve; WCNP: Wangchuck Centennial National Park; NC: North Corridor; RBP: Royal Botanical Park; GPR: Gangtey-Phobji Ramsar Site; KR: Khotokha Ramsar Site; BR: Bumdeling Ramsar Site. bWMD: Watershed Management Division, Department of Forest and Park Services, Ministry of Agriculture and Forests, Royal Government of Bhutan provided GIS shapefiles for protected areas and biological corridors, and Ramsar sites. cNBC: National Biodiversity Centre. 2014. National biodiversity strategies and action plan of Bhutan, 2014. NBC, Ministry of Agriculture and Forests, Royal Government of Bhutan, Thimphu, Bhutan. dWDPA: GIS area data from World Database for protected area. Available from https://protectedplanet net/search?country=Bhutan&q=Bhutan accessed on 31.12.2018. eWDPA: Reported data from World database for protected area. Available from https://protectedplanet net/search?country=Bhutan&q=Bhutan accessed on 31.12.2018. fAfter comparing all data sources we consolidated data mainly from WMD and reported in methods section. gJDNP-JSWNP: Jigme Dorji National Park-Jigme Singye Wangchuck National Park Biological corridor; All biological corridors are named after which PAs they link. hNA: Not available; Areas are not reported. iArea is the total area of all biological corridors as reported in source (NBC, 2014). jOM: Omitted; We excluded the area and also did not include these protected areas while counting the number of current PAs as they are designated inside existing protected areas.
Table S2.4 Percentage distribution of number of gain and PADDD events and area affected across PA system of Bhutan between 1966 and 2016.
Gain events No. Percent (%) Area (km2)a Percent (%) PADDD events No. Percent (%) Area (km2)a Percent (%) type type Gazettement 35 43.2 21964.4 41.6 Degazettement 6 20.7 585 13.3 Enlargement 8 9.9 1746.0 3.3 Downsizing 3 10.3 3808 86.7 Policy upgrade 36 44.4 24214.6 45.8 Policy downgrade 20 69.0 0 0 Class upgrade 2 2.5 4923.0 9.3 Class downgrade 0 0.0 0 0 Total 81 100.0 52848.0 100.0 Total 29 100.0 4393 100
*Area for all policy downgrade and 3 policy upgrade events were unknown.
Table S2.5. Summary of number and percentage of proposed and enacted PADDD events across PA system of Bhutan between 1966 and 2016.
PADDD Type Degazettement Downsizing Downgrading* Total Event status No. Percent (%) No. Percent (%) No. Percent (%) No. Percent (%) Enacted 6 100 3 100 8 40 17 58.6 Proposed 0 0 0 0 12 60 12 41.4 Total 6 100 3 100 20 100 29 100
* All were policy downgrade.
155
Table S2.6. Distribution of gain events (A) and PADDD events (B) and percentage composition across PA system of Bhutan between 1966 and 2016.
(A)
Gain New PA Enlarged PA Policy Class upgrade Total Type upgrade Year No Percent No Percent No Percent No Percent No Percent . (%) . (%) . (%) . (%) . (%) 1966 1 2.9 1 1.2
1974 8 22.9 8 9.9
1984 6 17.1 1 12.5 7 8.6
1992 12 33.3 12 14.8
1993 5 14.3 2 25.0 2 100 9 11.1
1995 1 12.5 1 1.2
1999 12 34.3 12 14.8
2001 1 12.5 1 1.2
2003 1 12.5 1 1.2
2006 1 12.5 21 58.3 22 27.2
2008 1 2.9 1 12.5 2 2.5
2009 3 8.3 3 3.7
2012 1 2.9 1 1.2
2016 1 2.9 1 1.2 Total 35 100 8 100 36 100 2 100 81 100
(B)
PADDD type Degazettement Downsizing Policy downgrade Class downgrade Total Year No. Percent No. Percent No. Percent No. Percent No. Percent (%) (%) (%) (%) (%) 1984 0 2 66.7 2 6.9
1993 6 100 1 33.3 7 24.1
2002 1 5 1 3.4
2004 14 70 14 48.3
2008 2 10 2 6.9
2010 1 5 1 3.4
2012 1 5 1 3.4
2014 1 5 1 3.4 Total 6 100 3 100 20 100 0 0 29 100
156
Table S2.7. Change in area and percentage across PA system of Bhutan between 1966 and 2016 due to (A) new and enlarged PAs and (B) degazettement and downsizing events.
(A)
Gain type New PA Enlarged PA Total Year Area (km2) Percent (%) Area (km2) Percent (%) Area (km2) Percent (%)
1966 419 1.9 419 1.8
1974 8772 39.9 8772 37.0 1984 605 2.8 20 1.2 625 2.6 1993 4762 21.7 663 38.0 5425 22.9
1995 300 17.2 300 1.3
1999 3660 16.7 3660 15.4
2001 187 10.7 187 0.8
2003 323 18.5 323 1.4
2006 137 7.9 137 0.6 2008 3736 17.0 116 6.6 3852 16.3
2012 1 0.0 1 0.0
2016 10 0.0 10 0.0 Total 21964 100.0 1746 100 23710 100.0
(B)
PADDD type Degazettement Downsizing Total Year Area (km2) Percent (%) Area (km2) Percent (%) Area (km2) Percent (%) 1984 195 5.1 195 4.4 1993 585 100 3613 94.9 4198 95.6 Total 585 100 3808 100.0 4393 100.0
Table S2.8. Number and percentage of different proximate causes for different PADDD types across PA system of Bhutan between 1966 and 2016.
PADDD Type Degazettement Downsizing Downgrading Total Proximate cause No. Percent (%) No. Percent (%) No. Percent (%) No. Percent (%) Conservation Planning 6 100 3 100 0 0 9 31.0 Infrastructure 0 0 0 0 14 70 14 48.3 Subsistence 0 0 0 0 6 30 6 20.7
157
Table S2.9. Summary of PADDD recorded in www.PADDDtracker.org across PA system of Bhutan.
PA Namea PADDD Type Status Year Cause
Enacted Proposed
No. Area (km2) No. Area (km2) JDNP Downgrading 0 0 1 unknown 2011 Infrastructure JDNP Downsizing 1 3534 0 0 1993 Conservation Planning Mochu WRb Degazettement 1 175 0 0 1993 Conservation Planking Namgyel Wangchuck Degazettement 1 195 0 0 1993 Conservation Reserveb Planning Pachu WR Degazettement 1 140 0 0 1993 Conservation Planning Shumar WR Degazettement 1 160 0 0 1993 Conservation Planning Total 5 4204 1 0 aJDNP: Jigme Dorji National Park; WR: Wildlife Reserve bWe found these two protected areas were not degazetted.
8.2 Supplementary materials for Chapter 3
8.2.1 Appendix 3.1 Detail method of fish and odonata species occurrence data
cleaning and collation for Bhutan
There are no collated georeferenced occurrence data on any freshwater species of
Bhutan. Therefore, we collated and cleaned data on fish and odonata of Bhutan which are among the better studied freshwater groups (Gyeltshen, Tobgay, Gyeltshen, Dorji, &
Dema, 2018). We searched for literatures on fish and odonata species of Bhutan using various combinations of search terms like Bhutan, fish, diversity, checklist, golden mahseer, odonata, dragonfly, damselfly, Anisoptera, Zygoptera, adult, larva and
Epiophlebia laidlawi the Google search and the Google Scholar. We also contacted individuals or organizations working on odonata and fish of Bhutan.
We obtained 17 recent literatures with fish occurrence records including journal articles, project reports, books and specimen checklist data sets. But only these eight
158 literatures (Wangchuck Centennial Park (WCP), University of Calcutta (UoC) & World
Wide Fund Bhutan (WWF Bhutan), 2012; Dorji & Wangchuk, 2014; Department of
Water Resources (DoWR), 2015; Thoni, Gurung, & Mayden, 2016; Thoni & Gurung,
2018; Phuntsho, n.d; Wangchuck Centennial National Park (WCNP), n.d) and one fish specimen record dataset (National Research Centre for Riverine and Lake Fishery,
Department of Livestock (NRCRLF-DOL) available in http://www.bhutanbiodiversity.net/)) had georeferenced occurrence records. In addition, a single occurrence coordinate for Tor putitora was obtained from Rai, M.R. (Forest
Range Office, Samtse Territorial Division, personal communication, May 12, 2018).
We obtained 15 journal articles with odonata occurrence or checklist records. But only nine papers (Mitra & Thinley, 2007; Mitra et al., 2012; Mitra, 2013; Mitra et al.,
2014; Dorji, 2014; Kalkman & Gyeltshen, 2016; Gyeltshen, 2017; Gyeltshen & Kalkman,
2017; Gyeltshen, Kalkman, & Orr, 2018) had georeferenced occurrence coordinates on adult and some larva distribution of odonata in general and two papers (Brockhaus et al.,
2014; Dorji, 2015) on larval distribution of E. laidlawi in particular. In addition, two occurrence coordinates for E. laidlawi larva were obtained from Dupchu, S. (Forest
Range Office, Monggar Territorial Division, personal communication, March 22, 2018).
We entered details for each species from all the literature sources and compiled the final master list for the fish and odonates. We then corrected typos for species and genus names, removed subspecies names which are not applicable, and updated the genus and species names based on the latest revisions for every entry and recorded the changes under ‘Name remark’ columns in Table S3.1a-b. We also added remarks for species with wrong names, species not identified to species level, species names that mentioned requires confirmation (cf.), species with unavailable scientific name and the species which we suspected are wrongly identified or the literatures mentioned so. We also added
159 remarks under ‘Occurrence remark’ columns whether the entry has no georeferenced coordinates or if the entry’s coordinates fall outside geographic boundary of Bhutan, and if they are multiple entries from different sampling dates.
We then segregated the entire entries into two portions, first portion consisting of the entries with adequate species level identification and georeferenced coordinates, and second portion either without species identification or without georeferenced or without both. We also included the introduced species within second portion for the fish and for the odonata the occurrence records for larva. However, we included the larval distrbuiton records of the odonate species E. laidlawi within first portion since its occurrence in
Bhutan is known only from its larva (Dorji, 2015). Within unidientiifed species we included species mentioned as confirmation required (cf.), species (only Anisopleura bella) or subspecies (only Aciagrion olympicum aruni) with unavailable scientific name, the species which we suspected are wrongly identified and the species which literatures mentioned as wrongly identified. For example, we placed fish Oreoglanis insignis in
WCP, UoC & WWF Bhutan (2012) to the unidentified group as we doubt its identity given O. insignis is currently recorded only in upper Irrawaddy and Sulween basins in
China (Allen, 2017). We also added another fish Exostoma labiatum in Dorji &
Wangchuk (2014) to the unidentified group since its identity is considered doubtful by
Thoni & Gurung (2018). We got a total of 386 entries for the fish including 232 within first portion and the remaining 153 entries in second portion (Table S3.1a), and 724 entries for odonate including 616 wihtin first portion and remaining 108 wihtin second portion (Table S3.1b).
160
234 Amblyceps mangois 1 No coordinates 235 Anguilla bengalensis 1 No coordinates 236 Badis 89.3064 26.9193 3 No species identification 237 Badis 89.3064 26.9193 3 No species identification 238 Badis badis 3 No coordinates 239 Badis badis 3 No coordinates 240 Badis cf. senegensis 90.9811 26.7916 1 Species name cf. 241 Bangana dero 1 No coordinates 242 Barilius 89.9271 26.7545 3 No species identification 243 Barilius 89.913 26.7988 3 No species identification 244 Barilius vagra 3 No coordinates 245 Batasio merianiensis 1 No coordinates 246 Botia lohachata 1 No coordinates 247 Chaca chaca 1 No coordinates 248 Chagunius chagunio 1 No coordinates 249 Chagunius chagunio 3 No coordinates 250 Channa gachua 3 No coordinates 251 Channa gachua 3 No coordinates 252 Channa gachua 3 No coordinates 253 Channa gachua 3 No coordinates 254 Channa gachua 3 No coordinates 255 Channa gachua 3 No coordinates 256 Channa orientalis 89.3726 26.8876 3 Endemic to Sri Lanka 257 Channa orientalis 89.8886 26.7471 3 Endemic to Sri Lanka 258 Channa punctata 3 No coordinates 259 Channa punctata 3 No coordinates 260 Clarias gariepinus 89.3726 26.8876 3 Introduced species 261 Cyprinion semiplotum 1 No coordinates 262 Cyprinus carpio 3 Subspecies communis removed No coordinates 263 Danio cf. assamila 89.9271 26.7545 3 Species name cf. 264 Danio dangila 1 No coordinates 265 Danio rerio 1 No coordinates 266 Danio rerio 1 No coordinates 267 Devario aequipinnatus 3 No coordinates 268 Devario aequipinnatus 3 No coordinates 269 Esomus danricus 1 No coordinates 270 Exostoma labiatum 90.9011 26.8761 1 Doubtful identification 271 Exostoma labiatum 90.6739 26.8148 1 Doubtful identification 272 Exostoma labiatum 8 No coordinates 273 Garra annandalei 1 No coordinates 274 Garra annandalei 3 No coordinates 275 Garra annandalei 3 No coordinates Duplicate 276 Garra gotyla 89.4102 27.0427 3 Subspecies gotyla removed coordinates 277 Garra gotyla 3 Subspecies gotyla removed No coordinates 278 Garra gotyla 1 Subspecies gotyla removed No coordinates Duplicate 279 Garra gravelyi 89.3518 27.4942 3 coordinates 280 Garra gravelyi 3 No coordinates 281 Garra lamta 90.6739 26.8148 1 Doubtful identification 166
282 Garra lamta 90.9011 26.8761 1 Doubtful identification 283 Garra lamta 90.9811 26.7916 1 Doubtful identification 284 Glyptothorax glyptothorax 89.2985 26.9376 3 Wrong species name 285 Glyptothorax glyptothorax 89.4353 27.4558 3 Wrong species name 286 Glyptothorax glyptothorax 3 No coordinates 287 Glyptothorax glyptothorax 3 No coordinates 288 Glyptothorax glyptothorax 3 No coordinates 289 Glyptothorax sp. 91.2038 27.2164 2 No species identification 290 Lepidocephalichthys guntea 1 No coordinates 291 Mastacembelus armatus 3 No coordinates 292 Mastacembelus armatus 3 No coordinates 293 Mastacembelus armatus 1 No coordinates 294 Mastacembelus armatus 1 No coordinates 295 Mystus vittatus 1 No coordinates 296 Neolissochilus 89.21 27.0194 3 No species identification 297 Neolissochilus hexagonolepis 3 No coordinates 298 Neolissochilus hexagonolepis 3 No coordinates 299 Neolissochilus hexagonolepis 3 No coordinates 300 Neolissochilus hexagonolepis 89.3064 36.9193 3 Outside Bhutan 301 Neolissochilus hexasticus 3 No coordinates 302 Olyra longicaudata 1 No coordinates 303 Onchorhynchus mykiss 3 No coordinates 304 Opsarius barna 3 Genus changed from Barilius No coordinates 305 Opsarius barna 3 Genus changed from Barilius No coordinates 306 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 307 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 308 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 309 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 310 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 311 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 312 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 313 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 314 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 315 Opsarius bendelisis 3 Genus changed from Barilius No coordinates 316 Oreochromis mossambica 89.3744 26.8678 3 Introduced species 317 Oreochromis mossambica 89.3744 26.8678 3 Introduced species 318 Oreoglanis insignis 90.4909 27.7327 9 Doubtful identification 319 Oreoglanis insignis 90.4424 27.5405 9 Doubtful identification 320 Pangio pangia 1 No coordinates Genus changed from 321 Paracanthocobitis botia 1 No coordinates Acanthocobitis 322 Parachiloglanis hodgarti 1 No coordinates 323 Parachiloglanis sp. 8 No species identification No coordinates 324 Pethia conchonius 1 No coordinates 325 Pethia ticto 1 No coordinates 326 Pseudecheneis 89.2985 26.9376 3 No species identification 327 Pseudecheneis 89.2985 26.9376 3 No species identification 328 Pseudecheneis sulcata 8 No coordinates 329 Pseudolaguvia ribeiroi 1 No coordinates
167
330 Pterocryptis cf. barakensis 90.1162 26.8225 1 Species name cf. 331 Pterocryptis cf. barakensis 90.1286 26.8216 1 Species name cf. 332 Puntius 89.8886 26.7471 3 No species identification 333 Raiamas bola 1 Typo Rajamas corrected No coordinates 334 Raiamas bola 1 Typo Rajamas corrected No coordinates 335 Salmo trutta 3 No coordinates 336 Salmo trutta 90.283 27.5189 4 Introduced species 337 Salmo trutta 90.2859 27.5149 4 Introduced species 338 Salmo trutta 90.2904 27.5136 4 Introduced species 339 Salmo trutta 90.2979 27.509 4 Introduced species 340 Salmo trutta 90.303 27.5065 4 Introduced species 341 Salmo trutta 90.3135 27.4982 4 Introduced species 342 Salmo trutta 90.3216 27.4914 4 Introduced species 343 Salmo trutta 90.3303 27.4887 4 Introduced species 344 Salmo trutta 90.3319 27.4874 4 Introduced species 345 Salmo trutta 90.3404 27.4859 4 Introduced species 346 Salmo trutta 90.3424 27.4826 4 Introduced species 347 Salmo trutta 90.3457 27.4798 4 Introduced species 348 Salmo trutta 90.3484 27.4753 4 Introduced species 349 Salmo trutta 90.35 27.4729 4 Introduced species 350 Salmo trutta 90.3513 27.4721 4 Introduced species 351 Salmo trutta 90.3522 27.4692 4 Introduced species 352 Salmo trutta 90.3537 27.4669 4 Introduced species 353 Salmo trutta 90.3544 27.4642 4 Introduced species 354 Salmo trutta 90.356 27.4614 4 Introduced species 355 Salmo trutta 90.3586 27.4599 4 Introduced species 356 Salmo trutta 90.3633 27.4566 4 Introduced species 357 Salmo trutta 90.3659 27.4544 4 Introduced species 358 Salmo trutta 90.3665 27.4536 4 Introduced species 359 Salmo trutta 90.3685 27.4511 4 Introduced species Introduced & subspecies fario 360 Salmo trutta 90.2998 27.5271 8 removed Introduced & subspecies fario 361 Salmo trutta 90.6754 27.6118 8 removed Introduced & subspecies fario 362 Salmo trutta 90.6846 27.6089 8 removed Introduced & subspecies fario 363 Salmo trutta 90.726 27.6809 8 removed Introduced & subspecies fario 364 Salmo trutta 90.8854 27.6236 8 removed Introduced & subspecies fario 365 Salmo trutta 90.8854 27.6236 8 removed Introduced & subspecies fario 366 Salmo trutta 90.8932 27.6245 8 removed Introduced & subspecies trutta 367 Salmo trutta 90.6728 27.6108 9 removed Introduced & subspecies trutta 368 Salmo trutta 90.6751 27.6118 9 removed Introduced & subspecies trutta 369 Salmo trutta 90.6984 27.6049 9 removed Introduced & subspecies trutta 370 Salmo trutta 90.6984 27.6063 9 removed Introduced & subspecies trutta 371 Salmo trutta 90.7142 27.6016 9 removed Introduced & subspecies trutta 372 Salmo trutta 90.8837 27.6722 9 removed 373 Schizothorax cf. plagiostomus 90.6896 27.1549 1 Species name cf. 374 Schizothorax cf. plagiostomus 90.7395 27.1169 1 Species name cf. 168
635 Anisopleura bella sp. nov. 91.4876 27.6246 10 Species name not available 636 Anisopleura bella? 90.1358 27.0122 13 Species name not available 637 Cephalaeschna spec. A. 89.3901 73.9697 5 No species identification 638 Cephalaeschna spec. A. 89.5756 27.3381 9 No species identification 639 Cephalaeschna spec. A. 90.5550 26.9704 9 No species identification 640 Cephalaeschna spec. A. 89.7288 27.7784 9 No species identification 641 Cephalaeschna spec. A. 90.1844 27.4555 9 No species identification 642 Cephalaeschna spec. A. 89.5467 27.2928 9 No species identification 643 Cephalaeschna spec. B. 89.5756 27.3381 9 No species identification 644 Ceriagrion sp. 13 No species identification No coordinates 645 Coeliccia sp. 91.4952 26.9798 6 No species identification 646 Coeliccia sp. 91.6810 26.9391 12 No species identification 647 Cordulegastridae sp. 90.1844 27.4555 9 No species identification 648 Cordulegastridae sp. 89.5467 27.2928 9 No species identification 649 Cordulegastridae sp. 9 No species identification No coordinates 650 Cord./Chloro. gen. sp.* 89.5737 27.3543 5 No species identification (Larva) 651 Cord./Chloro. gen. sp.* 89.2862 27.3581 5 No species identification (Larva) 652 Cord./Chloro. gen. sp.* 89.3194 27.3807 5 No species identification (Larva) 653 Cord./Chloro. gen. sp.* 89.3535 27.3717 5 No species identification (Larva) 654 Cord./Chloro. gen. sp.* 89.3580 27.4794 5 No species identification (Larva) 655 Crocothemis erythraea/servilia 89.6338 27.4890 5 Doubtful identification 656 Crocothemis erythraea/servilia 89.3754 27.4549 5 Doubtful identification 657 Crocothemis erythraea/servilia 89.8957 27.5943 5 Doubtful identification 658 Crocothemis erythraea/servilia 90.1253 27.0198 9 Doubtful identification 659 Crocothemis erythraea/servilia 90.6712 27.1486 9 Doubtful identification 660 Crocothemis erythraea/servilia 90.5803 27.3488 9 Doubtful identification 661 Crocothemis erythraea/servilia 89.3970 26.8685 9 Doubtful identification 662 Gomphidae gen. sp. 89.2883 27.3571 5 No species identification (Larva) 663 Gomphidae gen. sp. 89.3096 27.3707 5 No species identification (Larva) 664 Gomphidae gen. sp. 89.6430 27.5167 5 No species identification (Larva) 665 Gomphidae gen. sp. 89.6253 27.5414 5 No species identification (Larva) 666 Gomphidae gen. sp. 90.0099 27.5034 5 No species identification (Larva) 667 Gomphidae sp. 90.9452 26.8433 6 No species identification 668 Gomphidae spec 9 No species identification No coordinates 669 Macromia spec 9 No species identification No coordinates 670 Megalestes major 89.3754 27.4549 5 Identification doubtful 671 Megalestes major 90.0099 27.5034 5 Identification doubtful 672 Megalestes spec. 89.7249 27.7868 9 No species identification 673 Aciagrion olympicum 13 No coordinates 674 Aciagrion olympicum 10 Subspecies aruni removed No coordinates 675 Aciagrion olympicum 10 Subspecies aruni removed No coordinates 676 Agriocnemis clauseni 13 No coordinates 677 Agriocnemis clauseni 13 No coordinates 678 Agriocnemis femina 13 No coordinates 679 Anisopleura bella sp. nov. 10 Species name not available No coordinates 680 Aristocypha cuneata 13 Genus changed from Rhynocypha No coordinates 681 Aristocypha cuneata 10 Genus changed from Rhynocypha No coordinates 682 Bayadera indica 10 Duplicate 683 Bayadera indica 10 Duplicate 684 Calicnemia eximia 10 Duplicate 182
685 Calicnemia eximia 10 Duplicate 686 Calicnemia mortoni 10 No coordinates 687 Ceriagrion fallax 10 Subspecies cerinomelas removed No coordinates 688 Coeliccia svihleri 13 No coordinates 689 Coenagrion exclamationis 13 Changed from Himalagrion exclamatione No coordinates 690 Diplacodes trivialis 13 No coordinates 691 Diplacodes trivialis 10 No coordinates 692 Indolestes cyaneus 13 No coordinates 693 Ischnura rubilio 13 No coordinates 694 Neurobasis chinensis 13 No coordinates 695 Neurothemis fulvia 13 No coordinates 696 Orthetrum internum 10 Changed from O. japonicum internum No coordinates 697 Orthetrum luzonicum 10 No coordinates 698 Orthetrum luzonicum 13 No coordinates 699 Orthetrum pruinosum 10 Subspecies neglectum removed No coordinates 700 Orthetrum pruinosum 13 Subspecies neglectum removed No coordinates 701 Orthetrum pruinosum 13 Subspecies neglectum removed No coordinates 702 Orthetrum sabina 13 Subspecies sabina removed No coordinates 703 Orthetrum sabina 10 No coordinates 704 Orthetrum taeniolatum 10 No coordinates 705 Orthetrum triangulare 10 Subspecies triangulare removed No coordinates 706 Orthetrum triangulare 13 Subspecies triangulare removed No coordinates 707 Palpopleura sexmaculata 10 Subspecies sexmaculata removed No coordinates 708 Pantala flavescens 10 No coordinates 709 Pantala flavescens 13 No coordinates 710 Pantala flavescens 13 No coordinates 711 Polycanthagyna erythomelas 11 No coordinates 712 Pseudagrion rubriceps 13 Subspecies rubriceps removed No coordinates 713 Rhinocypha unimaculata 10 No coordinates 714 Tramea basilaris 13 No coordinates 715 Trithemis aurora 10 No coordinates 716 Trithemis aurora 10 Duplicate 717 Trithemis aurora 13 Duplicate 718 Trithemis festiva 13 No coordinates 719 Vestalis gracilis 13 Subspecies gracilis removed No coordinates 720 Indolestes cyaneus 90.1826 27.4581 5 Larvae 721 Indolestes cyaneus 89.5748 27.3546 9 Larvae 722 Indolestes cyaneus 89.5467 27.2928 9 Larvae 723 Indolestes cyaneus 89.5521 27.1391 9 Larvae 724 Indolestes cyaneus 89.5204 26.9226 9 Larvae
Lat.: latitude (degree); Long.: longitude (degree); Ref.: (Data source)
Reference list for fish (1: Dorji & Wangchuk, 2014; 2: Department of Water Resources, 2015; 3: National Research Centre for Riverine and Lake Fishery, Department of Livestock (fish specimen dataset); 4: Phutsho, n.d; 5: Rai, 2018 (Personal communication); 6: Thoni & Gurung, 2018; 7: Thoni, Gurung, & Mayden, 2016; 8: Wangchuck Centennial National park, n.d.; 9: Wangchuck Centennial Park, University of Calcutta, & World Wildlife Fund Bhutan, 2002)
183
Reference list for odonata (1: Brockhaus & Hartmann, 2009; 2: Dorji, 2014; 3: Dorji, 2015; 4: Dupchu, 2018 (Personal communication); 5: Gyeltshen & Kalkman, 2017; 6: Gyelsthaen, 2017; 7: Gyeltshen et al., 2017; 8: Gyeltshen, Kalkman, & Orr, 2017; 9: Kalkman & Gyeltshen, 2016; 10: Mitra & Thinley, 2007; 11: Mitra, 2013;12: Mitra et al., 2012; 13: Mitra et al., 2014)
*Cor./Chloro. gen. spe: Cordulegastridae/Chlorogomphidae gen. sp.
Table S3.2 Shapiro Wilk test for normality of different model features of the optimal models chosen through five optimal model selection approaches and the number of occurrence record for the fish and odonate species of Bhutan. Percentile OR: 10 percentile training presence test omission; Balance OR: Balance training omission, predicted area and threshold value test omission; No. of parameters: Number of parameters with non- zero lambda coefficients; Habitat area: predicted habitat area; Occurrence: Number of occurrence records of the species used for model building.
Features Taxon group Fish Odonata Selection approaches Statistic df p Statistic df p Percentile OR Expert 0.611 10 0.000 0.822 28 0.000
ORTEST_PER 0.807 10 0.018 0.779 28 0.000
AUCDIFF_PER 0.807 10 0.018 0.779 28 0.000
ORTEST_BAL 0.740 10 0.003 0.823 28 0.000
AUCDIFF_BAL 0.785 10 0.010 0.734 28 0.000
Balance OR Expert 0.750 10 0.004 0.590 28 0.000
ORTEST_PER 0.750 10 0.004 0.520 28 0.000
AUCDIFF_PER 0.538 10 0.000 0.188 28 0.000
ORTEST_BAL 0.366 10 0.000 0.188 28 0.000
AUCDIFF_BAL 0.366 10 0.000 NA NA NA
AUC difference Expert 0.653 10 0.000 0.771 28 0.000
ORTEST_PER 0.582 10 0.000 0.883 28 0.005
AUCDIFF_PER 0.497 10 0.000 0.778 28 0.000
ORTEST_BAL 0.655 10 0.000 0.848 28 0.001
AUCDIFF_BAL 0.671 10 0.000 0.636 28 0.000
No. of parameters Expert 0.765 10 0.005 0.639 28 0.000
ORTEST_PER 0.670 10 0.000 0.240 28 0.000
AUCDIFF_PER 0.633 10 0.000 0.233 28 0.000
ORTEST_BAL 0.690 10 0.001 0.421 28 0.000
184
AUCDIFF_BAL 0.623 10 0.000 0.252 28 0.000
Test AUC Expert 0.870 10 0.100 0.974 28 0.689
ORTEST_PER 0.721 10 0.002 0.970 28 0.570
AUCDIFF_PER 0.735 10 0.002 0.961 28 0.375
ORTEST_BAL 0.886 10 0.154 0.984 28 0.932
AUCDIFF_BAL 0.877 10 0.119 0.961 28 0.359
Habitat area Expert 0.871 10 0.103 0.979 28 0.821
ORTEST_PER 0.862 10 0.081 0.874 28 0.003
AUCDIFF_PER 0.901 10 0.225 0.874 28 0.003
ORTEST_BAL 0.827 10 0.031 0.913 28 0.023
AUCDIFF_BAL 0.827 10 0.031 0.895 28 0.009
Occurrence 0.822 10 0.027 0.864 28 0.002
NA: not applicable since Balance test omission was constant for AUCDIFF_BAL and omitted from the normality test.
Table S3.3 Regularization multiplier and feature class (RM-FC) combinations of the optimal models chosen by five model selection approaches for the fish and odonate species of Bhutan. Number of occurrence points used for model building is also given.
Group Species Occurrence points EXP ORTEST_PER AUCDIFF_PER ORTEST_BAL AUCDIFF_BAL Fish Schizothorax progastus 5 .25xLQ 1.5xH 1.5xH .25xLQ 2.5xLQ Garra annandalei 5 1.5xLQH 1.5xLQH 2xLQH 3.5xH 3 5xH Barilius barna 5 3xH 5xH 5xH 5xH 5xH Cyprinion semiplotum 5 5xH 5xH 5xH 5xH 5xH Badis badis 6 4.5xH 4.5xH 4.5xH 4.5xH 4 5xH Devario aequipinnatus 7 1.5xH 1.5xH 5xLQ 1.5xH 5xLQ Barilius bendelisis 10 2xH 2xH 2xH 2xH 2xH Barilius vagra 10 4xLQ 5xL 5xL 5xL 5xL Schizothorax richardsonii 11 .5xLQ .5xLQ 2.5xL 2.5xL 2.5xL Neolissochilus hexagonolepis 12 1.5xL 1.5xL 5xH 5xH 5xH Odonata Orthetrum internum 5 .25xLQ 2xLQ 3xLQ 1xH 1xH Pseudagrion rubriceps 5 4.5xL 4.5xL 4.5xL 4.5xL 4.5xL Ischnura aurora 5 2.5xLQH 4xL 4xL 4xL 4xL Calicnemia miniata 5 3xL 3xH 3xH .5xH 5xL Aciagrion olympicum 5 .5xLQH 1xLQ 5xLQ .5xLQH 5xLQ Neurothemis fulvia 5 5xLQ 5xLQ 5xLQ 5xLQ 5xLQ Anisopleura comes 6 .5xLQH 1xLQ 1.5xL .5xLQH 1.5xL Anotogaster nipalensis 6 .5xLQ .5xLQ 2.5xL 2.5xL 2.5xL Megalestes major 6 .25xLQH 2.5xLQH 4.5xLQ 2.5xLQH 4.5xLQ Orthetrum pruinosum 7 1xL 1xL 1xL 1xL 1xL Indolestes cyaneus 7 .5xLQ 2xL 4.5xL 2xL 4.5xL Orthetrum luzonicum 7 1xLQ 5xLQ 5xLQ 1xLQ 5xLQ Ceriagrion fallax 8 2.5xLQ 2xLQ 2.5xLQ 2xLQ 2.5xLQ Aristocypha quadrimaculata 8 5xL 5xL 5xL 5xL 5xL
185
Aristocypha cuneata 8 5xH 1.5xL 5xH 3xL 5xLQ Trithemis aurora 8 2xL 3xL 5xLQH 5xL 5xLQH Sympetrum hypomelas 9 1xLQH 4xLQ 3xH 4xLQ 3xH Orthetrum triangulare 9 5xH .5xL 4.5xL .5xL 4.5xL Trithemis festiva 9 3.5xLQH 3.5xLQH 5xLQ 5xL 5xLQ Epiophlebia laidlawi 10 1xLQ 5xLQ 5xLQ 1xLQ 5xLQ Orthetrum sabina 11 2.5xL 4.5xL 4.5xL 4.5xL 4.5xL Neurobasis chinensis 11 5xL 4.5xL 5xL 4.5xL 5xL Orthetrum glaucum 11 2.5xLQ 5xL 5xL 5xL 5xL Crocothemis servilia 12 4.5xH .25xL 4.5xH 4.5xH 4 5xH Palpopleura sexmaculata 13 5xH .5xL 5xH 5xH 5xH Pantala flavescens 13 3xL 5xL 5xL 5xL 5xL Calicnemia eximia 15 5xH 1xL 5xH 2xH 5xH Diplacodes trivialis 21 5xH .5xLQH 1xLQH 1xLQH 1xLQH
Table S3.4 Summary of Spearman’s correlation coefficient (rs) and two-tailed significance (p) value among the optimal models chosen through different model selection approaches for the different model features for the fish and odonate species of
Bhutan. Features: 10 percentile OR: 10 percentile training presence test omission;
Balance OR: Balance training omission, predicted area and threshold value test omission;
AUCDIFF: Difference between training AUC and test AUC; Parameters: number of parameters with non-zero lambda coefficient values; Test AUC; Area: predicted habitat area.
Taxon Selection approach Expert ORTEST PER AUCDIFF PER ORTEST BAL AUCDIFF BAL Occurrence Features Expert rs 1.00 0.98 0.98 0.99 0.98 -0.71 p . 0.00 0.00 0.00 0.00 0.02 OR TEST_PER rs 0.98 1.00 1.00 0.97 0.98 -0.72 p 0.00 . . 0.00 0.00 0.02
AUC DIFF_PER rs 0.98 1.00 1.00 0.97 0.98 -0.72 p 0.00 . . 0.00 0.00 0.02 OR TEST_BAL rs 0.99 0.97 0.97 1.00 1.00 -0.70
p 0.00 0.00 0.00 . 0.00 0.02 OR 10 percentile AUC Fish DIFF_BAL rs 0.98 0.98 0.98 1.00 1.00 -0.71 p 0.00 0.00 0.00 0.00 . 0.02 Occurrence rs -0.71 -0.72 -0.72 -0.70 -0.71 1.00 p 0.02 0.02 0.02 0.02 0.02 . Expert rs 1.00 1.00 0.79 0.46 0.46 0.21
p . . 0.01 0.18 0.18 0.56 OR TEST_PER rs 1.00 1.00 0.79 0.46 0.46 0.21
p . . 0.01 0.18 0.18 0.56 OR Balance AUC DIFF_PER rs 0.79 0.79 1.00 0.58 0.58 -0.26 186
p 0.01 0.01 . 0.08 0.08 0.48 OR TEST_BAL rs 0.46 0.46 0.58 1.00 1.00 0.06 p 0.18 0.18 0.08 . . 0.87 AUC DIFF_BAL rs 0.46 0.46 0.58 1.00 1.00 0.06 p 0.18 0.18 0.08 . . 0.87 Occurrence rs 0.21 0.21 -0.26 0.06 0.06 1.00 p 0.56 0.56 0.48 0.87 0.87 . Expert rs 1.00 1.00 0.87 0.87 0.86 -0.05 p . . 0.00 0.00 0.00 0.89 OR TEST_PER rs 1.00 1.00 0.87 0.87 0.86 -0.05 p . . 0.00 0.00 0.00 0.89 AUCDIFF_PER
rs 0.87 0.87 1.00 1.00 0.99 -0.19 p 0.00 0.00 . . 0.00 0.61 OR TEST_BAL rs 0.87 0.87 1.00 1.00 0.99 -0.19 AUC diff AUC p 0.00 0.00 . . 0.00 0.61 AUC DIFF_BAL rs 0.86 0.86 0.99 0.99 1.00 -0.19 p 0.00 0.00 0.00 0.00 . 0.61 Occurrence rs -0.05 -0.05 -0.19 -0.19 -0.19 1.00 p 0.89 0.89 0.61 0.61 0.61 . Expert rs 1.00 0.88 0.36 0.54 0.43 -0.39 p . 0.00 0.30 0.11 0.21 0.26 OR TEST_PER rs 0.88 1.00 0.41 0.62 0.43 -0.32 p 0.00 . 0.24 0.06 0.21 0.37 AUCDIFF_PER rs 0.36 0.41 1.00 0.70 0.95 0.00 p 0.30 0.24 . 0.02 0.00 0.99 OR TEST_BAL rs 0.54 0.62 0.70 1.00 0.75 0.14 Parameters p 0.11 0.06 0.02 . 0.01 0.70 AUC DIFF_BAL rs 0.43 0.43 0.95 0.75 1.00 0.12 p 0.21 0.21 0.00 0.01 . 0.74 Occurrence rs -0.39 -0.32 0.00 0.14 0.12 1.00 p 0.26 0.37 0.99 0.70 0.74 . Expert rs 1.00 1.00 1.00 0.93 0.93 -0.32 p . . . 0.00 0.00 0.37 OR TEST_PER rs 1.00 1.00 1.00 0.93 0.93 -0.32 p . . . 0.00 0.00 0.37 AUCDIFF_PER
rs 1.00 1.00 1.00 0.93 0.93 -0.32 p . . . 0.00 0.00 0.37 OR TEST_BAL rs 0.93 0.93 0.93 1.00 1.00 -0.10 Test AUC Test p 0.00 0.00 0.00 . . 0.78 AUC DIFF_BAL rs 0.93 0.93 0.93 1.00 1.00 -0.10 p 0.00 0.00 0.00 . . 0.78 Occurrence rs -0.32 -0.32 -0.32 -0.10 -0.10 1.00 p 0.37 0.37 0.37 0.78 0.78 . Expert rs 1.00 0.90 0.93 0.84 0.93 0.23 p . 0.00 0.00 0.00 0.00 0.53 ORTEST_PER
rs 0.90 1.00 0.90 0.90 0.92 0.37 p 0.00 . 0.00 0.00 0.00 0.29 Area AUC DIFF_PER rs 0.93 0.90 1.00 0.86 0.92 0.25 p 0.00 0.00 . 0.00 0.00 0.50 OR TEST_BAL rs 0.84 0.90 0.86 1.00 0.92 0.25 187
p 0.00 0.00 0.00 . 0.00 0.48 AUC DIFF_BAL rs 0.93 0.92 0.92 0.92 1.00 0.06 p 0.00 0.00 0.00 0.00 . 0.88 Occurrence rs 0.23 0.37 0.25 0.25 0.06 1.00 p 0.53 0.29 0.50 0.48 0.88 . Expert rs 1.00 0.44 0.44 0.54 0.34 -0.56 p . 0.02 0.02 0.00 0.08 0.00 OR TEST_PER rs 0.44 1.00 1.00 0.48 0.92 -0.37 p 0.02 . . 0.01 0.00 0.06 AUC DIFF_PER rs 0.44 1.00 1.00 0.48 0.92 -0.37 p 0.02 . . 0.01 0.00 0.06 OR TEST_BAL rs 0.54 0.48 0.48 1.00 0.55 -0.38
p 0.00 0.01 0.01 . 0.00 0.05 OR 10 percentile AUC DIFF_BAL rs 0.34 0.92 0.92 0.55 1.00 -0.39 p 0.08 0.00 0.00 0.00 . 0.04 Occurrence rs -0.56 -0.37 -0.37 -0.38 -0.39 1.00 p 0.00 0.06 0.06 0.05 0.04 . Expert rs 1.00 0.36 0.39 -0.11 . -0.21 p . 0.06 0.04 0.58 . 0.29 OR TEST_PER rs 0.36 1.00 0.45 0.33 . 0.11 p 0.06 . 0.02 0.08 . 0.59
AUC DIFF_PER rs 0.39 0.45 1.00 -0.04 . -0.26 p 0.04 0.02 . 0.85 . 0.17 OR TEST_BAL rs -0.11 0.33 -0.04 1.00 . 0.26 Balance OR Balance p 0.58 0.08 0.85 . . 0.17 AUC DIFF_BAL rs ...... p ...... Occurrence Odonata rs -0.21 0.11 -0.26 0.26 . 1.00 p 0.29 0.59 0.17 0.17 . . Expert rs 1.00 0.39 0.49 0.69 0.49 -0.45 p . 0.04 0.01 0.00 0.01 0.02 OR TEST_PER rs 0.39 1.00 0.75 0.56 0.76 -0.01 p 0.04 . 0.00 0.00 0.00 0.98 AUCDIFF_PER
rs 0.49 0.75 1.00 0.60 1.00 -0.18 p 0.01 0.00 . 0.00 0.00 0.35 OR TEST_BAL rs 0.69 0.56 0.60 1.00 0.61 -0.21 AUC diff AUC p 0.00 0.00 0.00 . 0.00 0.28 AUC DIFF_BAL rs 0.49 0.76 1.00 0.61 1.00 -0.18 p 0.01 0.00 0.00 0.00 . 0.37 Occurrence rs -0.45 -0.01 -0.18 -0.21 -0.18 1.00 p 0.02 0.98 0.35 0.28 0.37 . Expert rs 1.00 0.28 0.21 0.24 0.15 -0.39 p 0.15 0.28 0.21 0.46 0.04 OR TEST_PER rs 0.28 1.00 0.74 0.49 0.71 0.12
p 0.15 0.00 0.01 0.00 0.55 AUC DIFF_PER rs 0.21 0.74 1.00 0.40 0.86 0.07
p 0.28 0.00 0.04 0.00 0.73 Parameters OR TEST_BAL rs 0.24 0.49 0.40 1.00 0.54 -0.10 p 0.21 0.01 0.04 0.00 0.60 AUC DIFF_BAL rs 0.15 0.71 0.86 0.54 1.00 0.01 188
8.4 Supplementary materials for Chapter 5
Table S5.1. Varying conservation feature targets calibrated to obtain Marxan best solutions with reserve area of 50% Bhutan’s total land area.
Scenario 1 Scenario 2 Target area set Reserve area Percentage Target area set Reserve area Percentage (%) (km2) (km2) (%) (km2) (km2) 700 16286 42 700 19474 %) 750 20972 55 700 19526 51 745 20392 53 650 18918 49 740* 19136 50 675 19134 50 740 19506 51 675* 19182 50 735 18485 48
*The best solutions chosen for the final analysis.
191
Table S5.2. Conservation feature category, total area of conservation features and their percentage protected within Reserve 1 and Reserve 2, and the existing protected area system (PA) of Bhutan.
Category Conservation feature Total area Percentage protected (%) 2 (km ) Reserve 1 Reserve 2 Existing PA Odonata Aciagrion olympicum 12579.3 40 49 33 Anisopleura comes 14656.7 48 52 25 Anotogaster nipalensis 11611.0 44 52 32 Aristocypha cuneata 15203.2 54 53 26 Aristocypha quadrimaculata 2798.8 77 62 34 Calicnemia eximia 17044.6 53 53 27 Calicnemia miniata 21157.7 49 50 28 Ceriagrion fallax 25488.3 48 50 33 Crocothemis servilia 16613.4 53 53 27 Diplacodes trivialis 24137.6 49 51 31 Epiophlebia laidlawi 10051.3 36 47 38 Indolestes cyaneus 17884.4 41 49 35 Ischnura aurora 24331.9 48 51 31 Megalestes major 8874.8 50 56 31 Neurobasis chinensis 10138.0 61 57 24 Neurothemis fulvia 4331.4 74 61 31 Orthetrum glaucum 17725.7 52 53 27 Orthetrum internum 18327.1 50 53 47 Orthetrum luzonicum 16082.5 52 52 27 Orthetrum pruinosum 15246.5 54 55 27 Orthetrum sabina 21708.8 50 52 29 Orthetrum triangulare 16613.4 53 53 27 Palpopleura sexmaculata 13878.1 56 55 26 Pantala flavescens 22400.7 50 51 29 Pseudagrion rubriceps 805.7 92 62 33 Sympetrum hypomelas 17654.5 40 48 36 Trithemis aurora 9617.9 63 56 26 Trithemis festiva 14525.9 55 53 25 Fish Badis badis 1947.4 80 63 39 Barilius barna 4517.9 73 61 30 Barilius bendelisis 4524.9 74 61 29 Barilius vagra 9086.9 64 56 24 Cyprinion semiplotum 4736.2 73 60 29 Devario aequipinnatus 9172.1 58 48 24 Garra annandalei 12384.2 54 46 24 Neolissochilus hexagonolepis 22992.1 49 50 30 Schizothorax progastus 24040.1 49 53 56 Schizothorax richardsonii 20030.7 52 54 31 Bird Alcedo hercules 5687.6 71 59 28 Ardea insignis 16454.0 52 55 42 Grus nigrocollis 3772.3 49 68 46
192
Amphibian Nanorana annandalii 18954.1 49 48 31 Forest Alpine Scrubs 1317.4 58 52 90 types Blue pine 1021.2 75 66 7 Broadleaf 17807.9 44 48 34 Chir pine 1032.2 72 61 13 Fir 2330.2 54 51 78 Meadows 974.4 76 69 61 Mixed conifer 5248.1 48 51 58 Shrubs 3778.4 56 51 65
Table S5.3. Number of planning units and their percentage falling within different percentage protected range by the existing protected area system for the Reserve 1 and Reserve 2.
Percent protected range (%) Reserve 1 Reserve 2 Number Percentage (%) Number Percentage (%) 0 49 31 43 27 1-25 21 13 20 13 25-50 18 12 20 13 50-75 18 12 20 13 75-100 50 32 56 35 Total 156 100 159 100
193
194