Effects of Socio-Environmental Variability and Uncertainty In

EFFECTS OF SOCIO-ENVIRONMENTAL VARIABILITY AND UNCERTAINTY IN

DECISIONS ABOUT FISHING EFFORT OF A SMALL-SCALE TUNA FISHERY IN ENDE,

VICTORIA CONSTANZA RAMENZONI

(Under the Direction of Bram T. Tucker)

ABSTRACT

Fishery research is useful in guiding conservation efforts and implementing quota restrictions based on assessments and simulations of the current state of stocks. Despite the availability of new approaches that account for environmental uncertainty and variability, policy design in small-scale fisheries still relies on a technical understanding of fish and fishermen populations alike (McIlgorm et al. 2010, Acheson and Wilson 1996, McGoodwin 1990).

Ecosystem Based Fishery Models, Adaptive Management, and Socio-Ecological Systems perspectives often fail to incorporate the human dimensions of resource use and bottom- up behavioral approaches. In this dissertation, I study decisions about fishing efforts in a small tuna fishery in Ende, Flores, Eastern Indonesia through ethnographic (participant observation, semi- structure interviews, surveys), ecological (weather monitoring and coastal integrity assessments), and experimental tools (anthropometrics and probability judgment tasks). Relying on a socio- ecological and household-based approach, I use multilevel models, multivariate statistics and regression analysis techniques. My goal is to understand how environmental uncertainty influences fishing intensity and perceptions of catches, creating new behavioral responses that have consequences for the fishery as a whole. My objectives are to: 1) explore local narratives about luck and uncertainty; 2) describe the state of the stocks and present observational evidence to quantify statements of overfishing; 3) study decisions about fishing effort allocation taking into account the role of environmental uncertainty and its effects on the traditional moon cycle fishing calendar; and 4) characterize local perceptions of catchability and uncertainty and their impacts on fishing success. Results indicate that conservation policies need to: 1) understand the role of socioeconomic and environmental uncertainty in fishing effort and how it impacts household patterns of resource use before assuming that an area is overfished; 2) model individual decision making in terms of the economics of resource extraction and incorporate socio-environmental uncertainty to explain current level of fishing effort; 3) consider that fishermen can generate adequate representations that mimic probability distributions of ecological resources even when environmental probability is high; and 4) recognize that flexibility in strategies is a result of enabling structural responses in local communities that go beyond technical solutions.

INDEX WORDS: decision-making in uncertainty, marine anthropology, socio-ecological perspective, time allocation, climate change

EFFECTS OF SOCIO-ENVIRONMENTAL VARIABILITY AND UNCERTAINTY IN

DECISIONS ABOUT FISHING EFFORT OF A SMALL-SCALE TUNA FISHERY IN ENDE,

EASTERN INDONESIA

VICTORIA CONSTANZA RAMENZONI

BA Anthropology, Universidad de Buenos Aires, Argentina, 2006.

A Dissertation Submitted to the Graduate Faculty of The University of Georgia in Partial

Fulfillment of the Requirements for the Degree

DOCTOR OF PHILOSOPHY

ATHENS, GEORGIA

2014

Victoria Constanza Ramenzoni

EFFECTS OF SOCIO-ENVIRONMENTAL VARIABILITY AND UNCERTAINTY IN

DECISIONS ABOUT FISHING EFFORT OF A SMALL-SCALE TUNA FISHERY IN ENDE,

EASTERN INDONESIA

VICTORIA CONSTANZA RAMENZONI

Major Professor: Bram T. Tucker Committee: Ted Gragson Susan S. Tanner Pete Brosius Patricia Yager

Electronic Version Approved:

Maureen Grasso Dean of the Graduate School The University of Georgia May 2014 iv

DEDICATION

To my small Argentine family,

To my midsize American collective,

To my larger Indonesian community,

To all the ocean creatures that contributed their lives to my data.

ACKNOWLEDGEMENTS

Mom, Dad and Vero I acknowledge first, for without them I would not have lived

(literally) these past seven years. Thank you! Thank you! Thank you!

Martin and Virginia, who are always shining in my skylight and keep me on the path to become a good older sister: thank you too guys!

Third and fourth in my list is Bram Tucker who deserves not just one, but two huge and immense acknowledgements. This dissertation is yours as well as mine boss!

To Mike, Patti, Madalena and Rocio, oh man, who are my family, friends, and all of that is in between. To Lexi, Christine, Maria Ruth and Heather i give super and gigantic acknowledgements for they have put up with me in Spanish, English and Indonesian.

Julica, Mama, Intan, Bapak, Anton, Mama Ama, Bapak Nasir, Ahma, Mohammad,

Shakti, Dr. Pujo, Ibu Lala, Pak Chris, Pak Fredy, Agnes, Dr. Titis, and so many others in

Indonesia who have received me and made me dream in Indonesian.

To Ted, Pete, Tish and Susan who have helped me in wonderful and suprising ways, without losing faith, all along this time. Thank you for being patient with me.

And specially to you, Margie, La Bau, Brenda and Deb who made all of this possible, and went further and beyond for me.

To all of those that I did not include but I should have, please forgive me.

Last, I acknowledge my inspiration: Ernesto, Nestor, Rodolfo and Bungkarno.

"Si el presente es de lucha, el futuro es nuestro."

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ...... v

LIST OF TABLES ...... x

LIST OF FIGURES ...... xii

CHAPTER

1 INTRODUCTION ...... 1

2 LITERATURE REVIEW. MODELING DECISION-MAKING IN UNCERTAINTY

AMONG SMALL-SCALE FISHERMEN: A ROADMAP ...... 8

2.1. Introduction ...... 8

2.2. Decision making in fisheries: studying individual fishing effort in small-scale

fisheries ...... 11

2.3. Judgment under uncertainty and bounded rationality ...... 14

2.4. Decision making in uncertainty in anthropology and human behavioral

ecology ...... 17

2.5. Environmental structure based model of decision-making and index of

environmental complexity ...... 21

2.6. General Discussion and Conclusion ...... 31

3 ENDENESE FISHERIES: CASTING THE NET ON LOCAL ECOLOGICAL

KNOWLEDGE. EXPLORATORY FINDINGS ON ENVIRONMENTAL

PERCEPTIONS, FISHING EFFORT, AND OVERFISHING IN EASTERN

vii

INDONESIA...... 33

3.1. Introduction ...... 35

3.2. Ende ...... 38

3.2. Endenese Landscape ...... 41

3.4. Traditional ecological knowledge and climate change: why optimization is

not “rational” ...... 53

3.5. Conclusions ...... 62

4 AN INTEGRATIVE SOCIO-ECOLOGICAL APPROACH TO FISHING

HOUSEHOLDS: COMBINING OBSERVATIONAL, HISTORICAL AND

ENVIRONMENTAL DATA IN POOR INFORMATION FISHERIES TO

ANALYZE ANTHROPOGENIC AND ECOLOGICAL PRESSURES IN ENDE,

FLORES, INDONESIA...... 64

4.1. Introduction ...... 65

4.2. Small-scale fisheries and human dimensions of resource use ...... 67

4.3. Study Area ...... 74

4.4. Data collection and sources ...... 80

4.5. Data analysis ...... 81

4.6. Results ...... 90

4.7. General Discussion and Conclusions ...... 107

5 TEMPORAL PATTERNS IN FISHING EFFORT: USING MULTILEVEL

METHODS TO EXPLORE THE CHANGES IN TIME ALLOCATION

ACCORDING TO THE LUNAR CYCLE OF A SMALL-SCALE FISHERY IN

EASTERN INDONESIA. IS NEW SOCIO-ENVIRONMENTAL VARIABILITY

viii

REDEFINING FISHING PROFILES? ...... 114

5.1. Introduction ...... 115

5.2. Temporal patterns, regularities and decision-making in fisheries ...... 118

5.3. Merging individual data to create profiles: multilevel models and foraging

studies ...... 121

5.4. Etnographic description ...... 126

5.5. Data collection and methods ...... 128

5.6. Data Analysis ...... 131

5.7. Results ...... 136

5.8. General Discussion and Conclusions ...... 155

6 ESTIMATIONS OF CATCHABILITY AND PROBABILITY OF CATCH AMONG

FISHERMEN IN ENDE, FLORES, INDONESIA: DO NARRATIVES AND

SUBJECTIVE PERCEPTIONS OF UNCERTAINTY DETERMINE FISHING

EFFORTS AND RETURNS? ...... 164

6.1. Introduction ...... 165

6.2. Catchability and productive efficiency ...... 167

6.3. Prior expectations of catchability, perceptions of uncertainty and probability

judgments ...... 170

6.4. Ethnographic and environmental description ...... 176

6.5. Methods...... 179

6.6. Results ...... 187

6.7. General Discussion and Conclusions ...... 202

7 CONCLUSIONS...... 209

7.1 Summary of Main Results ...... 209

7.2 Theoretical and methodological implications for behavioral ecology, cognitive

sciences and anthropology ...... 214

7.3 Implications for anthropologists ...... 218

Implications for policy and governance ...... 219

Limitations, challenges and opportunities ...... 221

REFERENCES ...... 223

APPENDICES

ANNEX A: Primary Production ...... 261

ANNEX B: Environmental Structures ...... 264

ANNEX C ...... 268

LIST OF TABLES

Page

Table 2.1: Levels of uncertainty and outputs ...... 25

Table 2.2: Definitions of Environmental Structures ...... 26

Table 2.3: Index of Environmental Complexity ...... 30

Table 4.1: Main characteristics of the fishery ...... 86

Table 4.2: Maximum Sustainable Yield and Maximum Sustainable Effort ...... 91

Table 4.3: Multiple regression yield per month ...... 97

Table 4.4: Seasonal Arima and transfer function ...... 99

Table 4.5: Back cast of consumption values, total yield and CPUE...... 104

Table 5.1: Chart detailing regression analyses and results ...... 132

Table 5.2: One-way ANOVA analyses and Student’s t of mean catches for gear by lunar phase ....

...... 138

Table 5.3: One-way ANOVA of means of catch per lunar day ...... 140

Table 5.4: One-way ANOVA analyses for catches per season ...... 142

Table 5.5: Binomial Logistic Fit for probability of catch by type of boat ...... 144

Table 5.6: Multilevel Mixed Model ...... 147

Table 5.7: One-way Anova of Zscore catch by lunar phase ...... 151

Table 5.8: Fit of mean CV by proximity to full moon by fisherman ...... 153

Table 5.9: One-way Anova of CV by type of gear ...... 154

Table 5.10: Detail of fishing profiles per subject ...... 159

Table 6.1: Medians for canoes for all events ...... 184

Table 6.2: Medians for motorboats for all events ...... 185

Table 6.3: Ranking of factors when deciding to go fishing ...... 187

Table 6.4: Perceptions of future catches and fish distributions ...... 188

Table 6.5: Fish aggregations ...... 190

Table 6.6: Simple regressions of total catch by catchability by gear ...... 192

Table 6.7: Multivariate regression of total catch by catchability for canoes ...... 194

Table 6.8: Multivariate regression of total catch by catchability for motorboats ...... 196

Table 6.9: Mixed effects Canoes ...... 198

Table 6.10: Mixed effects Motorboats ...... 200

Table 7.1: Operationalization of findings ...... 211

Table 9.1: Colwel Index for Ende ...... 264

Table 9.2: Cumulative probabilities for various SPI values and possible interpretation of wet (or

dry) ...... 266

Table 9.3: Comparison between matched means ...... 267

xii

LIST OF FIGURES

Page

Figure 3.1: Maps of Ende City and Pulau Ende...... 39

Figure 3.2: Common Fish Families observed in Ende ...... 45

Figure 3.3: Coral and Fish Families observed in Ende ...... 46

Figure 3.4: Common Coral Species and others observed in Ende ...... 47

Figure 3.5: Common Fish Families and others observed in Ende ...... 48

Figure 3.6: Common Fish Species observed in Ende ...... 49

Figure 3.7: Common Fish Species observed in Ende ...... 50

Figure 3.8: Common Fish Species observed in Ende ...... 51

Figure 3.9: Common Fish Families and others observed in Ende ...... 52

Figure 3.10 System of Climatic Signs and Endenese causality ...... 60

Figure 4.1: Gear and vessels used ...... 75

Figure 4.2: Gears and vessels used 2 ...... 76

Figures 4.3 and 4.4: Maximum Sustainable Yield and CPUE per year ...... 92

Figures 4.5 and 4.6: Trajectories of fish groups (1990s-2010s) ...... 94

Figure 4.7: Common Trajectories of fish groups (1990s-2010s) ...... 95

Figure 4.8: Seasonal Time Series...... 100

Figure 4.9: Changes in monthly averaged CPUE per year (1990s-2010)...... 105

Figures 5.1: Boxplots matching total catch according to the gear and the Lunar Phase ...... 137

Figure 5.2: One-way ANOVA analyses and Student’s t of mean catches for gear by lunar phase

xiii

...... 138

Figure 5.3: Mean catches trajectories by lunar day depending on gear ...... 139

Figure 5.4: Boxplots of total minutes spent fishing per season according to gear ...... 141

Figure 5.5: Canonical Plot ...... 149

Figure 5.6: Profiles of Canoe Users ...... 156

Figure 5.7: Profiles of Motorboat Users ...... 157

Figure 6.1: Distribution of fish species ...... 181

Figure 6.2: Rater’s Consistency ...... 191

Figure 8.1: Reclassification of depths ...... 263

Figure 9.1: Standard Precipitation Index ...... 265

Figure 10.1: Fishing Calendar...... 268

Figure 10.2: Fishing log ...... 269

Figure 10.3: Kulavu or magical objects ...... 270

Figures 10.4 and 5: Endenese landscapes ...... 271

Figures 10.6 and 7: Endenese landscapes ...... 272

CHAPTER 1

INTRODUCTION

“The charm of fishing is that it is the pursuit of what is elusive but attainable, a perpetual series

of occasions for hope.” ~ John Buchan.

In the last twenty years, considering the recent collapse of major stocks like cod and anchovies (Kurlansky 1997, Ellis 2009, McIlgorm et al. 2010) and the current projections for overfishing (FAO 2011), fishery sciences have made significant strides by taking up adaptive management and Ecosystem Based Management approaches. With irreversible changes in ecosystems that are redrafting marine communities and landscapes, better models, forecasting systems and censuses are some of the ways proposed to move forward (NOPIP 2010). But, as this dissertation contends, there is much work left to do in comprehending how people respond to the different kinds of socio-environmental uncertainty to minimize exposure and risks. This endeavor, although critical to creating long-term sustainable strategies, is not without difficulty.

The work presented here contributes to the understanding of changes in behavior and their translation into fishery policy and management. With such goal, this dissertation explores socio- ecological uncertainty and how it affects resource practices at the fishing household.

There is probably no other place in the world where the voracity of big fishing industries and the potential destruction of climate change are more evident than in Southeast Asia. In countries like Indonesia, marine exploitation has not followed a cautious or sustainable path.

While global fish consumption increases at rapid paces, Indonesian shark fins and tuna, once a delicacy in Japanese and Chinese cuisines, have become popularly consumed in restaurants across the world. The future of Indonesian fishermen and fisheries grows more uncertain as the trade of illegal unreported and unregulated fishing products (or IUU) evolves into a multi-billion dollar industry.

Next to IUU fishing, news about the scale of marine degradation and poverty of small- scale fisheries in Indonesia has reached the public media through NGOs’ campaigns and internet blogs. The connections among environmental justice issues, lack of opportunities, and depleted fisheries, are acknowledged by private, political and academic sectors. Scientists have looked at the projections of climate change and how they would impact fish stocks in the tropics (Cheung et al. 2009). Handbooks are being written about socio-ecological resilience and vulnerability of small-island environments (i.e.: SocMonRes 2011). Maps are being drawn including marine protected areas1, sea surface level rise2, and inundation3. Still, there is an important part in all of these accounts about the ocean and its changes that is missing: what does it mean to be a fisherman facing a very difficult and uncertain future?

Unlike affluent fisheries across the world, in Eastern Indonesia, people continue fishing in precarious circumstances. Without GPS or sonars to aid in the detection of stocks, fishermen combine pieces of modern equipment like cell-phones with old plank boats and nets.

Government aid is virtually unknown and no forecasts of weather conditions reach these regions.

Traditional environmental signs that were used by previous generations to avoid danger and to

1: http://wwf.panda.org/what_we_do/how_we_work/conservation/marine/

2 : http://csc.noaa.gov/digitalcoast/tools/slrviewer

3 : http://walrus.wr.usgs.gov/coastal_processes/cosmos/

3 locate fish have lost validity with the increasing changes in the ecosystems. How do fishermen deal with this new uncertainty? How do they decide to go fishing or to stay home? How do they avoid danger?

These are some of the questions that motivated my research in a small tuna fishery in

Flores, Eastern Indonesia. I chose Ende because I was impressed by the largesse of its tuna fishes displayed at the regional market. I was curious about the cognitive skills that a fisherman would use to locate and capture a fish of that size. Later, I discovered that these fish originated in other regencies as local stocks are under intense exploitation.

Over my year and a half of research (June 2011 until January 2013) I met about three hundred fishermen, as well as their wives and children. They taught me about the futile pursuit of income, well-being, and opportunities. Being a fisherman is an unfair challenge, because risk not only makes life interesting, it also shortens it. It is a livelihood that comes at a great individual cost, with long working hours, high exposure to environmental hazards, low returns, and high investments. Fishing households experience high rates of migration, infant mortality, disease and precariousness. Observing these conditions first hand strengthened my determination to document the struggles Endenese lived through and to formulate recommendations to institutional offices and research counterparts.

My interest in exploring the problems associated with governance, precautionary conservation measures and non sustainable practices was the second reason that I came to Ende.

Marine environments are unlike other kinds of ecosystems. Because fish stocks are not observable, the behavior of resources is modeled and intuited to a greater extent than in terrestrial or freshwater ecologies (Acheson and Wilson 1996, Mangel and Clark 1983). From models, recommendations are derived in terms of harvesting and governance policies. But when

4 no information exists and there is reason to suspect future threats, precautionary measures are put in place. These can be in the shape of marine protected areas, fishing closures and quotas.

In Ende, where there are neither complete datasets nor models, the fishing commission operates in the dark. Intensification and subsidies are dominant practices. Corruption and inequity plague efforts to secure distribution of aid. No cooperatives or fishing auction offices exist, and fishermen lack any organizational strength to enforce their own regulations. Fishing practices like dynamite and cyanide are still in use which has turned fishermen and officers against each other. Following main conservation trends from Jakarta, in the course of the year

2012 the government started to draft zonification maps which will result in the exclusion of fishermen from some of their preferred areas. This situation, combined with mining industries and obvious signs of overfishing, furthers threatens the continuity of fishing livelihoods.

In this dissertation, I make a contribution, first, to fishery management research and conservation, and second, to behavioral studies in decision-making.

My primary goal is to understand how socio-environmental uncertainty is influencing fishing practices and creating new behavioral responses that have consequences for the fishery as a whole (ecosystem, fish and fishermen).

I am concerned with exploring the effects of variability (non-patterned and patterned) and randomness at the household level and how it shapes intensity of fishing at different temporal scales: inter-annual, seasonal, monthly, and daily.

One way to characterize randomness is by looking at the different kinds of variation and their predictability in the landscape (Low 1990). In this dissertation, randomness implies a lack of patterned variability in environmental circumstances. That is, changes cannot be predicted in their occurrence or outcome by either by scientists or decision-makers. However, when such

5 changes in variability (even when there is no patterns in the occurrences) can be conceptualized in probabilities, there is a perception of predictability known as risk. Within statistical sciences, randomness has been approached in a slightly different way. It refers to variability that cannot be explained by numerical parameters or models (Ott and Longnecker 2010). Taking into consideration all of these definitions, I explore in this dissertation different ways in which randomness is perceived by fishermen and in relation to statistical methods to analyze measurements.

From a socio-ecological perspective and within a behavioral and cognitive approach, I characterize fishing profiles, perceptions of uncertainty and patterns of resource extraction. In addition, I try to demonstrate how different motivations and rationales explain the use of resources and their exploitation. What makes this dissertation different from other research in livelihoods and fishing communities is that I strive to capture behavioral change by relying on non-traditional methods of analyzing and considering data.

In the course of six chapters, I pose and answer the following questions:

1) What are the limitations in behavioral models of fishing effort? How do we move

forward?

In chapter two, I summarize the literature addressing decision-making under uncertainty in small-scales fisheries and underline the gaps of behavioral approaches. I propose a classification of uncertainty and operationalize a definition of the structure of the environment to facilitate the modeling of behavior in a systematic way.

2) What are the motivations and rationales that explain decisions about fishing effort?

Can we dispense with optimization?

In chapter three, I discuss what drives decision-making about fishing effort by contrasting the narrow bio-economic approaches of managers to the local perspectives of fishermen.

Through ethnographic accounts, I characterize Endenese religious belief on luck as the main determinant of fishing success and problematize optimization/ maximization rationales that are widespread in the recommendations of fishery sciences.

3) What is the role of environmental and anthropogenic induced variations in explaining

the current state of a fishery? Can parametric approaches to stock assessment give a

full picture?

In chapter four, by exploring allocation of effort across different temporal scales and how yield might covary with socio-environmental processes, I demonstrate the importance of selecting a socio-ecologically driven time frame to apprehend causality. I describe the story of fishing stocks for the past 30 years and look at interannual variation in yields and effort. I reconstruct yields over time and potential consumption since 1880s. Most significantly, I establish the importance of household level observations to correct technical parametric models and statements of overfishing in poor information fisheries.

4) How is environmental uncertainty affecting old fishing ways of time allocation

according to the traditional lunar calendar? How do we incorporate individual rich

profiles into research and policy?

In chapter five, I look at how environmental uncertainty has affected the use of the lunar calendar to determine time allocation. I use multilevel models to deal with longitudinal repeated observations, the kinds of datasets that result from time allocation research. By adopting this

7 statistical approach, I isolate the importance of individual characteristics and propose fishing profiles through discriminant and multivariate analyses. These profiles, while general, are not unspecific and can inform policy.

5) How is daily uncertainty in catchability addressed by fishery sciences? Does

fishermen’s perceived uncertainty about yields match reality?

In chapter six, I suggest the relevance of defining coefficients and indicators in terms of a socio-ecological and human dimension (i.e. catchability). Through the study of subjective perceptions of potential catch and matching them to effort data, I discover that fishermen can generate adequate representations that mimic probability distributions of ecological resources even when environmental uncertainty is high.

My dissertation shows that behavioral patterns are undergoing significant changes to respond to climatic and social uncertainties. While vulnerability, adaptability and mitigation studies look at potential modifications in behavior, some of these changes are not projections but realities. Therefore, researchers need to approach the process of adaptation rather than to take it for granted (Dessai and Hulme 2004).

My work is grounded in the conviction that to design better management practices we need both qualitative and quantitative longitudinal datasets that can capture change in processes of resource use and extraction. Part of this effort involves creativity in addressing the future challenges of co-participatory conservation, as well as securing a higher continuity between research and practice. For this reason, the conclusion of my dissertation is devoted to deriving specific recommendations on how my findings can be applied to institutional settings.

CHAPTER 2

LITERATURE REVIEW. MODELING DECISION-MAKING IN UNCERTAINTY AMONG

SMALL-SCALE FISHERMEN: A ROADMAP

“Decision makers found in decision theory should not be confused with real people”

(Fjellman 1976:77).

“Foragers do not eat averages, they eat sums” (Stephens 1992:30).

“Much about risk-sensitive behavior is only nascently and imperfectly understood. For instance, we know little about the actual structure of environmental variation and how it affects

outcome distributions. This severely limits our ability to make conjectures about ecological

rationality” (Winterhalder 2007:443).

2.1. Introduction

A large majority of decisions within fisheries occurs in uncertainty (Holland 2008).

Uncertainty can be understood as a knowledge gap that prevents the assessment of the current or future conditions of events in the world (Gigerenzer et al. 1999). Within fisheries, it originates from the dynamic nature of the marine system which constraints its management and study

(Acheson and Wilson 1996, van Densen 2001). In addition to bio-physical complexity, marine landscapes are dynamic because they include the interaction between coastal communities,

9 fishermen, governments and industries (Mangel & Clark 1983, Acheson & Wilson 1996,

Hillborn & Mangel 1997, van Oostenbrugge 2001, Walker et al. 2003, van Densen 2001:9).

In fishery sciences, scholars have approached uncertainty by exploring management and informational constraints (Sethi et al. 2005, Milner-Gulland 2011, 2012), multi-fleet dynamics

(Hilborn 2006), and choice processes among commercial fishers (Curtis and McConnell 2004,

Bockstael and Opaluch 1983, Eggert and Martinsson 2004). Researchers, however, have seldom analyzed the human dimensions behind decision-making in small subsistence fisheries, often assuming that fishermen behave rationally (Salas et al. 2004, Holland 2008, Coulthardt et al.

2011). When this has not been the case, studies do not provide a modeling component that can be translated to formal simulations of Ecosystem Based Management or Management Strategy

Evaluation perspectives (Holland and Herrera 2009, Milner-Gulland 2011, 2012).

Decision making in uncertainty is a subject of research in economics, management, and behavioral and social sciences (Weber and Johnson 2009). Cognitive psychologists study people's conceptualization of uncertainty, inference mechanisms (Nisbett et al. 1983) and types of rationality (i.e.: bounded ecological rationality, Todd and Gigerenzer 2011, Gigerenzer et al.

1999, Cosmides and Tooby 1996, Nickerson 2012, Kahneman 2003). Other disciplines like anthropology and ecology attend to these topics through ethnographic methods (Quinn 1978) and behavioral studies (Stephens 1990, 2008, Winterhalder 2001). But, approaches have been abandoned for being highly descriptive (i.e.: Gladwin 1989) or for assuming rational action choice (Bettinger 1991, Mangel 1990; but see Mithen 1988, 1990, Mangel 1990, Wilke 2006).

Withal, studies in decision-making under uncertainty have followed a piecemeal viewpoint that lacks continuity. In addition, cognitive psychologists and behavioral economists fail at understanding the ecological situations in which subsistence decisions are made for they

10 have rarely addressed human behavior outside of the experimental setting (Alloy and Tabachnik

1984, Tucker 2007, Henrich et al. 2010). As a consequence, the evidence for judgments of uncertainty and ecological rationality within real-life situations is wanting.

Over the last few years, the importance of understanding the human component in decision-making in fisheries has become more pressing due to rapid environmental change and overfishing. Uncertainty can affect the state of natural resources by constraining levels of effort, influencing equipment and technology use, and increasing the likelihood of non-sustainable practices (Chakravarti-Kaul 1998, Lowe 2006, Halim 2002, Pet-Soede et al. 2000, Mous et al.

2005). Because the application of policy requires in some cases a change in behavior, information on how people make decisions in terms of resource harvesting is central to eliminate errors in the design and implementation of governance measures (Sethi et al. 2005, Bene and

Tewfik 2001, Salas et al. 2004, van Densen 2001).

In the following sections, I suggest that limitations in characterizing and modeling subsistence decision-making in uncertainty across fishery, social and cognitive sciences can be bridged by adopting a behavioral approach (Holland 2008). Behavioral models, by eliciting patterns of fishing effort and rationales governing choices, can provide relevant information to conservation and harvesting policies (Tucker 2007, Colfer et al. 1999, Bene and Tewfik 2001,

Godoy et al. 2009). To that end, I identify a significant gap that has impaired the modeling of decisions in subsistence societies within human behavioral studies. This limitation concerns the definition and study of the characteristics of the environments and is a result of 1) modeling uncertainty as risk (known probability function), 2) not identifying the different levels of variability and uncertainty; and 3) presupposing rational action choice.

Following a review of the literature of decision-making in uncertainty in fishery sciences,

11 cognitive psychology, and behavioral and cultural anthropology, I propose a preliminary taxonomy of uncertainty and a roadmap to identify subsistence decisions based on ecological rationality perspectives. I contend that operationalizing the statistical structure of the environment in which the decisions are made and exploring strategies at individual scales can provide parameters to be included in Agent Based Complex Systems (see Volker Grimm et al.

2005) that can be used in more comprehensive simulations of fisheries.

2.2. Decision making in fisheries: studying individual fishing effort in small-scale

fisheries

Given its high variability in returns, fishing is an activity characterized by uncertainty and randomness (Sosis 2002, Acheson 1989, 1981, Holland and Herrera 2009, Fauzi and Anna 2010, van Oostenbrugge 2001, Allison et al. 2010, 2001). Unlike farmers, who rely on longer temporal spans to make a decision, subsistence fishermen face potentially risky choices on a daily basis

(Fauzi and Anna 2010, Allison et al. 2010). Uncertainty in daily catch can extend over time and constrain household livelihoods by affecting yields (van Oostenbrugge 2001, Allison et al. 2001,

Zwieten et al. 2002). It can obscure the perception of patterns and modify fishing effort.

At a very basic level, subsistence decisions rely on patterns or sequences of states to obtain a preferred goal (Bateson and Kacelnik 1998, Westneat and Fox 2010 or Stephens 2008).

They are “situation-behavior combinations” that bring together alternatives, outcomes, and consequences (Hastie 2001:656-657). When the decision-maker has knowledge specifying the steps to be followed, situations can be clearly ascertained from a complex environmental background, and no unexpected events (accidents) affect the likelihood of achieving the outcome, then decisions are made in certainty. But, when attaining a result is contingent upon other factors and there is only a probability of reaching the goal, decisions are made under the

12 condition of uncertainty (Cashdan 1992, Luce and Raiffa 1957).

At the individual level, uncertainties can be the result of ignoring information about the different elements within an ecosystem and their behavior, inadequate technology or skills, and even cognitive limitations in anticipating future events that might occur (Howell and Clark 1978,

Tversky and Kahneman 1981, Kahneman et al. 1982, Nisbett and Ross 1980, Gigerenzer and

Selten 2002, Nickerson 2012). The decision to go fishing depends on the simultaneous consideration of attributes such as seasonality, characteristics of the fishing area like bathymetry, currents, circulation, and environmental conditions. Choices might also include an estimation of the level of effort needed, of the type of equipment, and an analysis of previous experiences

(Acheson 1981). But, even when a fisherman might meet all of the important conditions to make an informed decision, returns are to some level unpredictable because of the stochastic nature of the marine environment (Mangel and Clarke 1983, van Oostenbrugge 2001, Pet Soede et al.

2001).

In addition to ecological circumstances, uncertainties can emerge from socio-cultural, economic, and political factors (Fauzi and Anna 2010, Holland and Sutinen 2000). Decisions might answer to incentives and reasons not readily apparent such as taboos and ritual practices

(Palmer 1989, Malinowski 1921). They might often engage different actors, like middlemen and market vendors (Gladwin 1970, Quinn 1978). Or, decisions might combine long-term with short- term expectations (Fulton et al. 2011).

Bene and Tewfik (2001) have underlined some of the complexity of characterizing choice in multi-species fisheries, where studying decisions about effort and resource harvesting involves a compromise between systemic and analytical approaches. Other researchers have also emphasized the difficulties in capturing and modeling the socio-economic and cultural factors affecting daily fishing labor patterns (Ota and Just 2008, Staples et al. 2004, Pauly 2006, Salas et al. 2004, Holland and Sutinen 2000, van Oostenbrugge 2001). Models of fisher’s behavior and decision-making have looked at choices of species (Bene and Tewfik 2001, Salas et al. 2004,

Pelletier and Ferraris 2000), seasonal effort (Wiyono et al. 2006), fishing effort allocation (van

Oostenbrugge et al. 2001, Aswani 1998, Begossi 1992, Lopez et al. 2011), and spatial allocation

(Hillborn and Walters 1987, Curtis and McConnell 2004, Pet Soede et al. 2001) among other issues. There are a good number of studies that have modeled uncertainty, risk preferences and utility functions of commercial fishers (Curtis and McConnell 2004, Bockstael and Opaluch

1983, Eggert and Martinsson 2004, Holland and Sutinen 2000, Mistiaen and Strand 2000, van

Densen 2001).

There has not been, however, an explicit consideration or modeling of uncertainty as it affects individual decisions in subsistence fisheries (but see van Oostenbrugge 2001, Pet Soede et al. 2001, van Zwieten et al. 2002). In addition, most studies reviewed above, when not subscribing to a rational action choice perspective, have employed psycho-economic approaches to utility like Prospect Theory, which weights utility functions according to observed behavioral trends like loss aversion (Tversky and Kahneman 1984). Livelihood or household approaches to small-scale fisheries have not provided behavioral characterizations that can be integrated into simulations (Milner-Gulland 2011). As a result, there is a substantial part of comprehending what motivates effort and decisions under uncertainty that remains to be explained.

2.3. Judgment under uncertainty and bounded rationality

The study of judgment under uncertainty has been a productive field of inquiry of human decision making competencies within cognitive sciences (Weber and Johnson 2009, Shafir and

Tversky 1995). A simple Web of Science search on the terms “decision-making under uncertainty” produces over 4880 articles and publications since 1945. Linking recent findings from psychology, neuroscience, economics, biology, and animal behavior, decisions have been approached from different perspectives in this field.

A normative or prescriptive approach concentrates in determining what subjects ought to do given certain constraints and goals (Cancian 1980:161, Garro 1998, Elio 2002). Discussed briefly in the previous sections, it assumes perfect knowledge of the alternatives that a person has when making a choice, and a clear formulation of preferences and utilities for each alternative. It also assumes rational choices, that is, the maximization of the utilities given a set of preferences

(Cancian 1980:163). In this particular view of decision-making, the importance of economic notions has steered the course of research and has shifted the main focus of enquiry to the study of utility, preferences, and risk (Crozier and Ranyard 1997:5).

By contrast, descriptive approaches have concentrated on how the process of decision- making takes place (i.e.: Prospect Theory, Tversky and Kahneman 1984, Shafir and Tversky

1995). Scholars within this subcamp define the normative criteria of prescriptive approaches as oblivious to content and meaning, and concentrate their work in exploring the role of biases, frames, and context in affecting inferences (Kahneman and Tversky 1986, 2000). These studies, however, have placed a substantial weight on demonstrating the lack of use of rationality principles, which has resulted in an overstatement of the irrationality of lay people (Nickerson

2012). In addition, its focus on errors makes the integration of findings in a unified body of

15 theory challenging. Descriptive perspectives have been undermined for their difficulties at making sense of a piecemeal view of the human mind and for not proposing explanatory mechanisms (Gigerenzer et al. 1999).

Bounded rationality and ecological rationality elaborate from the criticisms posed to normative and descriptive approaches to underline the importance of environments in decision- making (Hutchinson and Gigerenzer 2005, Gigerenzer and Brighton 2009). Introduced in the mid

1940s by Simon, bounded rationality proposes that cognitive processes are essentially shaped by the structure of the domain in which they occur and by cognitive limitations (Brunswik 1943,

Simon 1955, 1956, 1987, Anderson 1991a, 1991b). This is a more psychologically plausible notion of how the mind works as it does not imply optimization or exhaustive computations of information.

From the assertion that rationality is bounded it does not follow that humans are less adaptive. Constraints allow for pragmatism and flexibility, and better fit in responses as they narrow the set of factors a decision maker has to pay attention to when making a choice

(Cummins 2002, Gigerenzer and Brighton 2009). The emphasis is placed in frugal decision- making strategies, for environments possess properties that permit the simplification of the choice mechanism (Simon 1956). Ecological rationality assumes that judgments of uncertainty follow simple heuristic rules for information search and choice (Gigerenzer and Selten 2002,

Gigerenzer et al. 1999, Todd and Gigerenzer 2012). These principles are a result of evolutionary processes, with the accuracy of a decision being determined by the degree in which it matches its environment (Cosmides and Tooby 1996, Gigerenzer and Brighton 2009).

For example, in particular classes of events, like highly uncertain situations, the salience of random elements or cues within the structure of the environment makes more likely that no arbitrary connections between data will be established and good decisions be made. Thus, an organism’s choices will be evolutionary adaptive as long as they are tuned to perceive the variability and patterns of regularities in the structural setting. Context will guide the decision and influence the processing of information in a sound ecological way (Kunda 1999, Tversky and Kahneman 1996, Teigen et al. 1996, Winterhalder 2007).

Weather prediction is a context where decisions are highly uncertain. Events are unstable in their occurrence and conditions change rapidly preventing informed choice. As more days are included in the forecast, prediction loses efficiency because error accumulates. For this reason, people tend not to trust weather forecasts and are prone into considering randomness and probabilities if making a decision. Similar domains with high levels of uncertainty are sports, games of chance, and foraging or fishing (Tversky and Kahneman 1984, Kunda 1999).

Despite the relevance of judgment and decision-making to understanding behavior, it is only recently that emphasis has been placed in exploring how individuals deal with real-world decision tasks in ecological environments or ill-defined field settings. Relatively few studies have been conducted in non-Western societies until the last ten years (Yates et al. 1998, Nisbett

& Miyamoto 2005, Uskul et al. 2008, Wilke & Barret 2009) and investigators have yet to address the cognitive significance of cultural beliefs and social factors in decisions (Weber &

Johnson 2009, Alloy & Tabachnik 1984). In all, this translates into ecologically sound studies of behavior in universities, laboratories and firms (Tucker 2007, Henrich et a. 2010), and models of decision behavior in real ecological settings that rely on non-realistic tenets (i.e.: rational action choice, Holland 2008).

2.4. Decision making in uncertainty in anthropology and human behavioral ecology

Recognizing the importance of contextual factors, anthropologists have investigated decision-making in subsistence societies. Studies such as those carried out by Gladwin (1976,

1979, 1980, 1989), Murtaugh and Gladwin (1980), Chibnik (1980), Boster (1984) and Quinn

(1978) have achieved relative success in accounting for choice patterns across varied productive domains. Formulations, however, suffer from shortcomings related to the difficulty of eliciting cognitive processes with oral protocols (see Gladwin 1971) and have been highly descriptive

(Boster 1984).

Within Human Behavioral Ecology, Optimal foraging theory (Smith 1983, Smith and

Winterhalder 1992, Mithen 1988, 1990) has a strong empirical component, but it disregards a psychological or information based approach (but see Winterhalder 1981, 2007). Excluding a handful of studies (Mithen 1988, 1990; Wilke 2006, Wilke and Barrett 2009), the decision process itself is treated as a black box and left unexplored. The lack of attention to the mechanism behind decisions implies that there is little knowledge of the cognitive processes governing foraging patterns and subsistence behaviors (Kaplan and Hill 1992).

For example, classic models have explained the individual decision to allocate time to a productive activity as a function of the net acquisition rate (Foley 1985, Stephen and Krebs 1986,

Smith and Winterhalder 1992, Winterhalder and Smith 2000). This is a calculation of energetic gains minus energetic costs over time. No consideration has been paid to relevant structures in the environment that can cue organisms into foraging in richer patches or the temporal dimension of resource abundance (i.e: seasonality; but see Wilke and Barrett 2009, Wike 2006). Or to cultural normatives that can aid decision processes (Boyd and Richerson 2002).

In addition to a narrow energetic perspective in what motivates decisions, behavioral studies have not been effective at distinguishing the different scales of uncertainty and have defined decision environments in a restrictive way in terms of variation. These gaps are byproducts of presupposing a rational action choice paradigm (Bettinger 1991) and have led modelers to conceptualize uncertainty as risk (Caraco 1981, Alden Smith et al. 1983,

Winterhalder et al.1999, Stepehens 1992).

The focus on risk derives from equating ecological uncertainty (the probability of experiencing an energetic shortfall) to economic definitions of uncertainty (Cashdan 1990,

Caraco 1981, Smith et al. 1983). As presented in previous sections, in more classical explanations, uncertainty is considered as lack of information about the states of the world to achieve a desired outcome (Bernstein 1996, Hacking 1990). If uncertainty is complete, the organism does not have enough knowledge of the structure of the environment to assign the probabilities of occurrence of different alternatives in a decision problem. Lack of knowledge, however, can be more broadly illustrated as a result of ignorance and stochasticity4. When the individual can formulate an estimation of the likelihood of outcomes, it is said that it can formulate an assessment of risks (Luce and Raiffa 1953, Cashdan 1992, Low 1992, Bernstein

1996).

Optimal Foraging Theory considers that decisions that do not involve maximization of returns can be better modeled as decisions made in risk. In these contexts, the problem lies in

4 Important to observe is that risk is different from incomplete knowledge (Winterhalder 2007). Incomplete knowledge approaches to uncertainty imply that increasing information eliminates unpredictability. However, in ecological scenarios where stochasticity and emergent events are the norm, outcomes are only probable to happen. That is, even when we have enough information about the solar system, we cannot say with complete certainty that the sun will rise tomorrow. Uncertainty might be better defined, then, as unpredictability in achieving a predetermined outcome.

19 choosing “among different probability distributions” (Ludovico et al. 1991, Stephens and Krebs

1986) and selecting the one that brings the maximum utility. For instance, if a fisher has to decide among different patches, he will consider the probabilities within each patch of acquiring the maximum return. If he is risk prone, he will choose the patch that reports the highest cost/benefit ratio even when probabilities of success are low. If he is risk averse, he will select the patch that has the higher probability of a successful outcome disregarding whether this choice will provide the best cost/benefit relation.

Because probabilities are based on an objective interpretation5 of the principle behind probability theory (Gigerenzer and Murray 1987, Bernstein 1996, Hacking 1990) and OFT assumes that risks have a known probability distribution (Winterhalder 1999, 2007, Stephens and

Krebs 1986), it predominantly relies on general linear models or generalized linear models

(GLM) to capture variability6. GLMs are ways in which relations between variables can be modeled (i.e. regression techniques) and presupposes ways of representing variance (i.e.: errors follow a multivariate normal distribution). This does not imply that OFT cannot allow for some level of stochasticity as it is considering probabilistic functions, and this is attained through the generalized family of models like Poisson regression (Stephens 1992). However, its treatment of variability is through known parametric distributions (for example, it uses sample means as population parameters and Z scores; Cashdan 1992, Stephens 1992, Winterhalder et al. 1999,

Winterhalder 2007, Kuznar et al. 2002).

5 An objective interpretation of probability defines the probability of an event as the relative frequency of occurrence of an incident within its class. 6 “We presume that the outcome is unpredictable and, for simplicity, characterize it by a probability distribution, such as the normal distribution. This formalizes the outcome expectation as the mean value ….” Winterhalder 2007:434.

Unfortunately, the kind of environmental uncertainty affecting subsistence decisions cannot always be subjected to parametrization. Error and non-patterned variability are significant components to the understanding of change which in turn can affect the modeling of risk (see

Winterhalder 2007). Furthermore, in ecological environments not many resources are randomly distributed like OFT assumes (Sutton et al. 2004). They might approach clumped aggregations for prey can vary in their temporal and spatial availability (Wilke and Todd 2009, Begossi 1992,

Allison et al. 2001, Bird et al. 2009) or they might look like highly autocorrelated events in a time series (Winterhalder 2007). In brief, the statistical structure of the environment should not be assumed as possessing a particular distribution of variability or errors7. It should be a matter of study on its own before creating a model and deciding on a behavioral rule (Simon 1947,

1955).

In addition, there are important differences between economic and ecological notions of risk. First, causes and consequences in both scenarios are extremely diverse. In economics, theorists discuss uncertainty about preferences and utilities; whereas, in behavioral models, risks imply potential variability in fecundity or survival rates (Cashdan 1990, Stephens 1990,

Winterhalder 2007). The two sets of decisions involve processes of choice that play out at different temporal scales, with different rewards and pay-offs. It is reasonable to assume that framing effects will modify how risk is interpreted by the decision-maker (Tversky and

Kahneman 2000, Winterhalder 1996) and the information value (Stephens 1990, Dove 1993).

7 “We assume the environment is well behaved, at least to the degree that the parameters of the outcome distribution are stable over time periods encompassing action and result.… Finally, we assume that the resource in question is divisible in increments small enough to make comparison of sigmoid positions meaningful (Henrich and McElreath 2002).” Winterhalder 2007:435.

Taking into account bounded rationality perspectives, in the next paragraphs I suggest that some of the limitations encountered in modeling decision-making within small-scale fisheries can be resolved with a specification of the different levels of uncertainty (Kujala et al.

2012, Walker et al. 2003, Howell and Clark 1978), of the informational structure of the environment (Todd and Gigerenzer 2012, Wilke and Todd 2010), and its dimension and components (Duncan 1972). By narrowing the definitions used, and by starting with fine resolution questions, modelers can operationalize structural constructs in order to formulate an understanding of how a pattern of behavior can emerge (“pattern oriented modeling”, Volker

Grimm et al. 2005).

2.5. Environmental structure based model of decision-making and index of environmental

complexity

Like many other theoretical constructs, environment is a concept whose use has been non-consistent across human behavioral studies and ecology (Jax 2006). Characterized as “... the

[biotic and abiotic] surroundings in which an organism interacts” (Sutton and Anderson

2006:31), it is often confused with the terms ecosystem8, niche, and habitat. The latter refer to functional or interactive ways of looking at environments and do not perfectly overlap with more informational perspectives of behavior.

Management and administrative sciences have considered environments as the combination between physical and social factors that are taken into account to make a decision

(Duncan 1972: 314). This definition establishes a distinction between epistemic and ontological ways of looking at the environment and ecological concepts (Jax 2006). Epistemic perspectives

8 Ecosystems have been defined as: “an assemblage of organisms of different types (species, life forms) together with their abiotic environment in space and time” (Jax 2006:240).

22 focus on the informational attributes and abstract characteristics of the units within the environment instead of their more material, physical or natural properties. They can also focus on the relation between the organism and its perception of the objects in the environ (Brunswick

1943). The latter perspective is the one that will be adopted in this article.

As mentioned above, making a decision involves selecting among different alternatives the one that is most likely to produce the desired outcome given a particular situation or context.

In terms of decision-making processes, the environment of the decision-maker can be characterized as a set of informational structures. Structures can be treated as statistical constructs depending on their dimension and scale. Informational structures represent the characteristics, behaviors and relations between the units of a decision problem and setting. In this case, the units of the decision problem are elements, factors, dimensions or domains that determine the different alternatives in the choice scenario. The types of information that conform environmental structures and units within a decision problem originate from physical, biological, social, and cultural dimensions (Wilke and Todd 2010:5).

Informational structures are patterns that the decision-maker deduces from observing the behavior and the characteristics of the component units in the decision problem or that emerge from the ecological environment. Informational patterns can be summarized in cues or environmental signs, relative frequencies and classification systems or taxonomies. For instance, to minimize uncertainty when making a choice, individuals learn about the prior probability distributions of alternatives through repeated experiences, formulate beliefs about conditions that might make outcomes more likely, and implement routines of actions. If a farmer has to decide, for instance, whether to plant maize this month or the next instead of barley, he will consider: 1) how often other alternative crops have reported successful outcomes, 2) how many times his

23 crops have been successful if the rains occurred in the first month of the season; and, based on this knowledge, he would define when he should plant. Along this process, experiential knowledge can be condensed in cultural beliefs and narratives, and form a body of “shared understanding” or “schemas” to guide behavior (Garro 1998, D’Andrade 1989).

According to Wilkes and Todd (2010:5-6) informational patterns can have a certain validity that indicates how often it produces successful decisions, redundancy if patterns have a high correlation among themselves, and discrimination rates (how often particular cues distinguish between alternatives, regardless of their accuracy). In addition, Todd and Gigerenzer

(2012:16) mention three types of structures that can be found in epistemic environments: uncertainty, number of alternatives, and sample size9. Other candidate structures are distribution types, predictability, ambiguity, types of variability10 (Duncan 1972, Cashdan 1992, Low 1992,

Wilke and Barrett 2009, Mithen 1988, 1990). It should be noted that many of the informational structures listed overlap as they can be descriptions of the same process, characteristic or behavior.

9 Uncertainty is a gap in information. The number of alternatives indicates the number of potential options from which a decisor has to choose. Sample size refers to the number of previous similar situations that the individual has experienced. These experiences aid the individual in the decision of a course of action and signal the familiarity of the subject with the domain in which such decisions are to be made.

10 Distribution types are a description of arrays of elements in an environment (evenly, aggregated, clumped, etc.). Statisticians and ecologists have studied the forms in which elements are distributed in a particular setting. This has allowed them to formulate a taxonomy, that when quantified, can help the researcher estimate the likelihood that a future element will be found in a certain position. Predictability indicates the extent to which a certain event can be estimated to occur. Ambiguity refers to cases when informational cues do not permit to discriminate with certainty whether one outcome is more likely to occur than other. Types of variability are ways in which the behavior of elements can vary within a system. If variability is non-patterned, researchers cannot distinguish patterns or behavioral rules to explain the occurrences of events. By contrast, when patterns can be discerned it is said that variability is patterned.

Within this classification, uncertainty can be broadly summarized as the absence of knowledge arising from ignorance or unpredictable non-patterned variability (Kujala et al. 2012,

Cashdan 1990:2). Similar to probabilities and environments, uncertainty has a dual nature, possessing epistemic and ontological dimensions. Subjective aspects of uncertainty are estimations of chance, quantifiable or non-quantifiable, that might reflect a judgment on the likelihood of future events (Hacking 1990). They might affect behavior as they can guide the course of actions (Kunda 1999). But, they are different from objective dimensions of uncertainty in that they might confound confidence of prediction (how certain I am that this prediction will happen) with an actual prediction (Duncan 1970).

Objective estimates of uncertainty are, on the other hand, numerical assessments based on parametric, non-parametric and complex ways of representing variance. They can be operationalized, for example, through different indexes measuring predictability, contingency, constancy, ranges and extreme variation like means, coefficients of variation, spectral analysis, time-series, and component analysis (Low 1990, van Densen 2001, van Oostenbrugge 2001).

Given that taxonomies and types of uncertainty vary depending on the domain in which the decision is made, subjective and objective assessments will be domain-specific (Johnson and

Baksh 1990, Gigerenzer 2008, Todd and Gigerenzer 2012, Wilke and Barrett 2009). Tables 2.1 and 2.2 exemplify the definitions with time allocation decisions to fishing.

Table 2.1 Levels of uncertainty and outputs

Definition Example Data sets

Subjective Uncertainty  “Everything depends on Choice protocols Estimations of uncertainty or luck” and interviews. confidence in prediction:  The probability that I will Judgments of probability catch fish today is 50%. Narratives of chance

Objective Uncertainty  Colwell Index, SPI. Datasets on atmospheric, Quantifiable estimations of variability  Transfer functions and physical and through parametric, non-parametric ARIMA models. and complex methods biological  Hazard and Cox models. phenomena. Ranges, CV, CI, SD, Predictability, contingency, constancy, times-series,  Bayesian models. covaration series, time to event, etc.  Simulations.

Table 2.2 Definitions of Environmental Structures

Definition Example Data sets

Decision: Time allocation to fishing Longitudinal repeated Selecting between  Go fishing different alternatives, measurements or also a combination  Rest simulation between outcomes and  Work in the garden behaviors  Work on equipment

Decision environment: A set of informational structures that represents the behavior and the relations between the component units of the decision setting.

Units:  Physical (type of fishing area,  Oceanographic Factors, dimensions or currents, oceanographic conditions, datasets etc) domains that shape the  Weather different alternatives in a  Atmospheric (weather for that day, datasets decision problem and season, climate) their probabilities of  FAO species achieving a successful  Biological (energetic needs) distribution and biomass outcome.  Ecological (prey) assessments,  Social (family obligations, fishing white papers rules if any) and reports  Cultural domains (prescriptions,  Cultural and taboos), etc. social ethnographies

Table 2.2 Definitions of Environmental Structures (cont.)

Definition Example Data sets

Information  Distribution of prey species (clustered, even,  Oceanographic structures: semi aggregated) datasets Characteristics,  Tidal patterns  Weather associations and datasets behaviors of the units  Temperature of the water and changes  FAO species in the decision  Clouds as a cue for wind problem and its distribution and setting; also known  Precipitation today, yesterday, (up until 9 biomass as cues, relative days as a cue for storms) assessments, white papers frequencies,  Barometric pressure as a cue for storms taxonomies, etc. and reports  Type of Season (dry, wet, intermediate) Inferred by the  Cultural and decision-maker or  Seasonality strength (comparison to past social emerging from the seasons) ethnographies ecological environment.  Wind strength, direction  Archival research,  Type of gear Various types: anecdotal statistical  Catchability of gear yesterday, day before sources, census distributions, types of yesterday, etc. documents and variability and statistical  Lunar phase (proximity to full moon) change (time series), reports ambiguity, degree of  Experience (number of days when uncertainty, types of experiencing similar conditions, also sample uncertainty, size of size) learning sample, patchiness,  Subjective uncertainty (beliefs about luck, predictability, randomness, certainty of value of density, diversity predictions, risk perception) index, constraints,  Objective uncertainty (potential ranges, etc. stochasticity, emergent properties of ecological processes)  Fishing calendar (cue or taxonomy of best fishing months according to fish species)  Density of prey (number of successful encounters with prey over t time), information about previous yields, stock biomass, etc.  Behavior of prey (cue about association with other species or where to locate it)

Behavioral pattern:  Fish if wind is blowing from the south.  Heuristics, Repeated set of observations,  Do not go fishing for rays in March actions based in or simulations. particular cues. Rule according to fishing calendar. that connects information structure with action.

To model a decision, it is necessary to define the environment of the decision problem by considering its informational structures and component units. Once the structures are identified, the complexity of the decision problem can be assessed. For example, to study and model the decision to allocate time to fishing, these initial steps should be considered:

1) Identify the decision problem. Characterize the decision environment. Identify

alternatives and potential preferences.

2) Identify factors, domains and elements that affect alternatives.

3) Identify informational structures for the different alternatives by focusing on cues,

statistical properties, ecological processes, times-series, regularities and other

patterns.

4) Quantify uncertainty: subjective and objective. Create composite index of complexity

(Duncan 1970).

5) Assign cue values and weights based on frequency of co-occurrence, redundancy,

ambiguity and uncertainties (Duncan 1970).

6) Choose a decision heuristic (Todd and Gigerenzer 2011) or define individual

behavioral patterns and agents rules by observation of choice or simulation of

outcomes. Define cognitive costs and restrictions placed on agent.

7) Simulate decision scenario by including cues, informational structures and weights in

model.

It should be noted that unlike OFT models, this process of approaching decision modeling does not assume a priori the statistical structure of the environment or the rule that the individual will employ to choose. The particular informational mechanism that is used to make that inference can only be determined by looking at the level of uncertainty in the environment

30 and the information that is available to the individual. Therefore, the rule will only emerge after and not before studying the complexity of the environment.

Composite index of environmental complexity

Finally, in order to formulate an assessment of the complexity of the decision in terms of uncertainty and environmental structures that is comparable to other processes of choice or across time, it is possible to combine the objective and subjective estimates of uncertainty into one environmental factor or index of environmental conditions (see Duncan 1970 for a similar index). The process of aggregation of measures into the index requires standardization of the different variables analyzed and, when possible, the use of weights leveraging the quality of the original datasets (Colburn and Jepson 2012). Also of note is that the selection of the factors or structures to include will depend on the kind of ecosystem in which the decision is made.

Table 2.3 Index of Environmental Complexity

Formula (after Duncan 1970:316) Index of environmental complexity = (F).(C)2 Where F is the number of environmental structures for which uncertainty is estimated, and C the number of component variables used for each factor.

Biotic (catch and productivity) Atmospheric (rain, drought, winds)

 Changes in primary productivity  Precipitation at different levels of variability (monthly, seasonal,  Cumulative changes in CPUE at different inter-annual, decadal) levels of variability (monthly, seasonal, inter-annual)  Wind averages and extremes  Total amount of days fished last season  Barometric pressure oscillations  Average number of fishermen (mean fishing  Contingency, constancy and effort) predictability (Colwell Index).  Catchability or aggregation index

2.6. General Discussion and Conclusion

The main objective of this article is to propose a solution to limitations in the behavioral modeling of decision-making in uncertainty of small-scale fisheries that could further advance the incorporation of such decisions into more complex Ecosystem Based Models. The previous section introduced a roadmap for said process, as well as an operationalization of how to estimate the uncertainty as level of complexity in the environment. In addition, I also proposed several definitions derived from the literature on bounded ecological rationality and studies in decision making in managerial sciences.

Considering the structure of the environment before determining the actual behavioral rule can improve characterizations of decision-making in uncertainty in at least two ways. First, it can help achieve explanations and models of behavior that are more realistic. By concentrating on the structures that are critical in terms of stochasticity and unpredictability, the modeler does not need to consider relatively large amounts of information to capture ecological settings. Also, a focus on change and thresholds of change can limit the onerous analytical requirements manifested by other approaches (compensatory decision-making modeling, Shafir and Tversky

2003, Gigerenzer 2008).

Second, by pointing to the significant environmental structures in terms of its informational and statistical characteristics, it can guide the detection of behavioral rules that might seem counterintuitive from a more generalistic perspective. Consequently, models have the potentiality of explaining behaviors from other kinds of rationales not considered by optimization approaches. This is more in consonance with classical ecological studies in anthropology like Rappaport’s Tsembaga ritual (1968) or Dove’s Kantu augury system (1993).

The following chapters of this dissertation address some of the issues raised in this article. Chapter three offers a problematization of optimization and rational action choice in the context of a small-scale fishery undergoing rapid socio-economic and environmental change.

Chapter four corrects parametric or more linearized approaches to variability in terms of interannual yields with house-level data. It shows the importance of environmental structures in explaining causality and underlines the relevance of studying decisions at the small scale.

Chapter five looks into monthly variability in catches and how it affects fishing effort. It also demonstrates the restrictions in behavioral approaches that do not consider the autocorrelation of observations when analyzing datasets. Chapter six explores the relations between subjective and objective uncertainty by studying fishermen’s judgments of probability and chance and their effect on daily fishing success.

In all, this dissertation addresses decision-making in uncertainty across different environmental structures and scales (interannual, seasonal, monthly and daily) and ponders its effects on fishery management. Future articles will offer an operationalization of the change in decision complexity regarding fishing effort as a function of environmental change.

CHAPTER 3

ENDENESE FISHERIES: CASTING THE NET ON LOCAL ECOLOGICAL KNOWLEDGE.

EXPLORATORY FINDINGS ON ENVIRONMENTAL PERCEPTIONS, FISHING EFFORT,

AND OVERFISHING IN EASTERN INDONESIA11.

11: Ramenzoni, V.C. 2013, Journal of Ethnobiology Letters 4:39–51. Reprinted here with permission of the publisher.

Abstract

Fishing fleets in South East Asia have recently experienced unprecedented expansion.

Consequently, catches and regional diversity have dramatically decreased throughout the Indian

Ocean. Regional governments and conservation organizations blame the local fishermen and their use of damaging fishing practices for the present state of resources. However, many of these institutions endorse a narrow perspective on bio-economic governance and human action

(rational action choice) that compromises the understanding of resource use and exploitation among small-scale fisheries. Over the last few decades, there is an increasingly recognized tradition that points to the importance of ecological systems of knowledge, uncertainty representation, and traditional skills, in conceptualizing processes of environmental decision- making and the likelihood of introducing successful sustainability practices. In line with this perspective, this article presents preliminary findings regarding resource use decision-making processes among Endenese fishing villages in central Flores Island, Indonesia. Grounded on 22 months of ethnographic, experimental and ecological research (semi-structured interviews, participant observation, visual surveys, probability and uncertainty assessments), and exploring local cognitive representations of marine processes, climate, ichthyology and the role of luck, this article discusses the current economic representations of small-scale fishers as avid maximizers. It concludes by emphasizing the need to further explore the role of mental models and beliefs regarding uncertainty in motivating fishing effort to design adequate conservation and governance programs.

Key Words: small-scale fisheries, luck, uncertainty representation, decision-making, traditional ecological knowledge

3.1. Introduction

Fisheries in South East Asia have experienced in the last half-century an unprecedented expansion (Semedi 2001). As a consequence, the catch per unit of effort has dropped significantly in many regions of the Indian and Pacific Oceans (Butcher 2004, 2005, Boomgard

2005, Henley and Osseweijer 2005). Reports from conservation and intergovernmental organizations attribute stock depletion to overfishing and damaging fishing practices (UNEP

2008, Ingles et al. 2008). In an attempt to regulate endangered resources, countries like Indonesia have engaged in decentralization, community-based and ecosystem management approaches

(Williams and Staples 2010, Satria and Matsuda 2004).

Many of these efforts have encountered difficulties in dealing with the large-scale illegal trade of aquatic resources (Heazle and Butcher 2007, Fox 2005). They have also failed at recognizing the inequities in fishing capacity that are so common in Eastern Indonesia. In the province of Nusa Tenggara Timur, where the current research takes place, poverty extends to one third of the population (Monk et al. 1997, Resosudarno and Jotzo 2009). But, most significantly, limitations in management and development approaches have impaired the understanding of local fishermen's role in environmental degradation. A strict bio-economic perspective has prevented the eradication of damaging fishing practices such as bombs and cyanide-potassium

(Lowe 2006). The continuous use of non sustainable practices has, in return, resulted in very limited foreign investment, a condition that further exacerbates poverty and environmental pressure (Halim 2002).

Over the last decades, scholars have noted that fishery managers and government officers often work under the assumption that maximization and self-interest are the main motivations behind the allocation of fishing effort (Cordell 1974, Allison and Ellis 2001, Perry et al. 2003).

This assumption is deeply rooted in the idea that fisheries, when not regulated, are open access systems where everybody's property becomes nobody's (Gordon 1954, McCay 1981, Feeny et al.

1990). It also stems from the way human behavior is characterized by economic formalizations.

Bio-economic models of Maximum Sustainable Yield and Optimal Foraging Theories or

Marginal Value Theorem (A. Smith 1983, Winterhalder and Smith 2000) explain individual decisions and conservation practices through rational action choice (Gowdy 2008). These models have been relatively successful in generating simple, parsimonious, and generalized explanations consistent, in some cases, with field observations and ethnography (Winterhalder 1981, 1996).

At the same time, Optimal Foraging Theories (OFT) have been widely criticized for remaining inattentive to the social embeddedness of decision-making processes concerning subsistence practices. Critics have targeted OFT's assumptions about optimality and rational action, stressing its restrictions in dealing with dynamic choices (Gigerenzer 2008, Gigerenzer &

Brighton 2009, Gigerenzer & Selten 2001, Foley 1985, Mc Cay 1981, Mangel & Clark 1986,

Houston et al. 1988). Remaining for the most part inattentive to advances in the studies of decision making under uncertainty (but see Mithen 1989, 1990, Mangel 1990, Wilke 2006),the role of information as constraining efficiency has been left unexplored and OFT has marginally addressed psychological and social preferences (Aswani 1998).

Although it is indisputable that commercial fisheries in Southeast Asia are creating unnatural pressures on fish stocks (Ellis 2009, Helfman 2007, Butcher 2004), the responsibility of small-scale fisheries in the current decline of marine biodiversity cannot be established with certainty. Because decision making processes explaining fishing effort are multifaceted and extend beyond simple economics (Mc Goodwin 1990, Bene and Tewfik 2001), it is necessary to address local interests, systems of values, and adaptation strategies in order to fully comprehend

37 the impact of fishermen in their environment (Ludwig et al. 1993, Allison and Ellis 2001, Mc

Ilgorm et al. 2010).

To that end, building from a cognitive and ecological anthropology perspective, this article presents preliminary findings regarding information, local ecological knowledge and decision making processes explaining fishing effort of Endenese fishing communities in the

Island of Flores, Indonesia. Positioned on the northern margins of the Savu sea, Ende has been known for its prodigious catch and marine biodiversity (Roos 1877, Weber 1899, Van Suchtelen

1921, Monk et al. 1997, Fox 1977). It has remained marginalized from investment and economic development (Butcher 2004). But, with drops in production landings throughout the Indo Pacific ocean, coral bleaching, and climate change, new plans have been drafted that include the creation of one of the largest marine protected areas in the Coral Triangle (TNC 2009). Unfortunately, information on the state of marine resources in the Savu is very fragmented. There is a dearth of knowledge on the way local communities use and represent the marine ecosystem (Munasik et al. 2011) and a wide propensity to blame local fishermen for the current state of environmental degradation.

In order to explore perceptions and decisions about the environment, resource use, and climate change, I conducted ethnographic research, using semi-structured interviews and participant observation, in June-July 2009, November 2010-January 2011, and June 2011-March

2012 in Pulau Ende, Ipy and in the village of Arubara. Preliminary findings indicate that the quantity of fish has decreased in Ende Bay over the last 50 years and that significant changes have been observed by the local fishermen in sea surface temperature and wave activity (Badan

Pusat Statistik Kabupaten Ende 1985-2011). In addition, findings suggest that decisions regarding fishing effort combine assessments of sailing conditions, knowledge of prey

38 availability, and weather patterns. Interviews regarding traditional knowledge and ecological assessments have showed that decision making is not conducted under conditions of perfect knowledge. The major explanation given for variability in resource exploitation and motivations to go fishing is luck (rezeki)12. There is not a clear notion of risk or of probability quantification.

This latter finding challenges the univocal characterization of fishermen as optimizers and rational actors. It also suggests that studying local perceptions of environmental uncertainty is crucial when assessing the patterns of ecological variability of an area to design sustainable management strategies.

3.2. Ende

Ende city is a middle size port surmounting to approximately 17,000 people (Badan Pusat

Statistik Kabupaten Ende 2010). Across the bay from the city is Pulau Ende, a small island that includes seven villages with a total of 8,000 people (see figure 3.1).

Coastal Endenese have a complex origin. They reflect a mix between local hinterland groups ("Ata Lio" and "Ata Keo"), Javanese and Chinese traders, Bimanese warriors, Sumbanese slaves, and migrant Bugis, Butonese and Makassarese fishermen from Sulawesi (Tule 2004,

Nakagawa 1984, 1996, Sareng Orin Bao 1969, Dietrich 1983, Knaap and Sutherland 2004,

Needham 1968, 1980). Islam spread in the 16th century through trade and resulted in the consolidation of Buginese cultural traits to the expense of local characteristics (Edjid 1979).

Buginese traits include a unique syllabic alphabet system named Bahasa Lota (Van Suchtelen

1921, Roos 1877, Banda 2005), complex descent myths (Pelras 1996), food prescriptions, birth and wedding ceremonies, and an intricate symbolism and set of ritual practices that link the

12 : I use italics underscored to distinguish words in Bahasa Indonesia. I use italics and accent marks ‘’ to distinguish words in Endenese. In all cases translation is provided within the text.

39 social representations of the house and the boat (perahu or ‘sampa’; Chou 2003, Southon 1995,

Sopher 1965). Also among these traits is the practice of mencari rezeki or the search for fortune

('nggae ka') as a way to explain one's decisions in all aspects of life (Acciaioli 2004, Pelras

1996).

Figure 3.1: Maps of Ende City and Pulau Ende. Attribution: copyright by Ewesewes at id.wikipedia under Creative Commons Attribution-Share Alike 3.0 Unported License and ESRI, Inc. under creative commons licenses CC By-NC-SA 3.0. Modified by Victoria Ramenzoni.

Anthropologists have explored Coastal Endenese groups incidentally while studying kinship rules, magic, and agricultural practices of hinterland communities (Nakagawa 1984,

1996, Needham 1968, Tule 2004, Forth 1998). Historians have devoted some attention to the illegal trade of slaves and pirating activities carried out by the Endenesein the eighteenth and nineteenth centuries (Needham 1968, 1980, Dietrich 1983, Knaap and Sutherland 2004). During this time, the Endenese were a powerful force that engaged in commerce activities throughout the entire eastern Indo-Pacific region. After Dutch military intervention in the early twentieth century, Ende became famous as Sukarno’s exile destination. At that time, Endenese had already endured the transition to a local agricultural economy under colonial pressure and became both politically and commercially isolated. Nowadays little seems to have changed.

In comparison to other parts of Indonesia like Kalimantan or Java, development programs have progressed at a slower rate in Flores (Resosudarmo and Jotzo 2009). In Ende, fishing is still carried out by traditional boats (sampans) or smaller motor boats with 4 ½ to 1 inch fishing nets.

Activities are mostly for subsistence or small-scale trade as there is no industry operating in the district or external investment to support the improvement of the fishing gear.

Bigger fish are sold at the town markets of Mbongawani, Potulando and Wolowona along with octopus (Octopodidae, Octopus spp.), squids and scalops (Pectinidae, Amusium spp.), manta rays (Dasyatidae, Dasyatis spp., Mobulidae, Mobula spp., Myliobatidae spp.) and sharks

(Alopiidae, Alopias spp.,Charcharinidae,Charcharinus spp., Lamnnidae, Isurus spp.), anchovies and sardines (Clupeidea, Sardinella gibbosa, Sardinella lemuru, Dussumeria acuta). A common list of families of species includes flying fishes (Exocoetidae, Cypselurus spp.), sail fishes and marlins (Istiophorus, Makaira Indica, Makaira mazarra, Xiphias gladius, Istiophorus platypterus), tuna (Thunnus maccoyii, Thunnus obesus, Thunnus tonggol), skipjack (Euthynnus

41 affinis, Katsuwonu spelamis), needle fishes (Belonidae, Tylosorus spp.), scad (Caesionidae,

Caesio caerularea, Caesio cuning), snapper (Lutjanidae, Lutjanus spp.), and groupper

(Serranidae, Cromileptes altivelis, Eponephelus tauvina); (see Figure 3.2 and following).

3.3 The Endenese Landscape

The research area is in the Ende Regency located in the southern coasts of Central Flores.

The region can be characterized as a set of “specialized [environments], with a higher proportion of endemic species in an overall depauperate community” (Auffenberg 1980:45). In terms of its morphology, Flores is an inner volcanic arc island, originating no more than 15myr ago during the late Mio-Pliocene (Monk et al. 1997:26-38). Its coastal landforms are still ongoing uplifting processes; which restrict formations of mangrove forests, and there are very limited alluvial plains that support agriculture. It is also one of the driest regions in Indonesia. The landscape combines low precipitation, high intense solar radiation and strong winds and it drastically varies depending on orography. The city of Ende, in the fringes of the Sawu sea, had in 2011 an annual total rainfall of 961 mm in the lowlands, while in Wolowaru, in the uplands, precipitation was

2169 mm (BPS Ende 2012: 38-39). Heaviest precipitation occurs during the monsoon season spanning from December to March. Average annual temperatures are at 27, 8 C at sea level (BPS

Ende 2012:42).

3.3.1 Morphology/Climate

The Sawu basin is delimited by a set of inner-arc volcanic islands comprising Flores,

Lembata, and Alor, and by the outer-arc volcanic islands of Timor, Rote, Sumba and Savu. It covers 105,000 square kilometers (Costello et al. 2010) and is a fore-arc deep sea basin formed by the Sunda-Banda subduction system (Van Weering et al. 1989). The Sawu Basin can reach maximum depths of 3,893 m (mean depth of 1796 m) and surface temperatures on average

42 around 28 C. Unlike other tropical basins, the Sawu high temperatures (2.9 C), relatively low salinity (34.54-34.59%), low oxygen concentration (2.26-1.86) and low density (Wyrtki 1961,

Humphries and Webb 2008, Potemra et al. Weber 1901) in the deeper strata are of significance for coral conservation (TNC 2009).

As in other regions in Eastern Indonesia, the high stratification of the water column prevents upwelling in areas without deep currents and constraints internal water circulation. The discontinuity layer, the transition between surface masses and colder and denser deeper layers of water, lies between 70 to 160 ms and might be relatively uniform (Wyrtki 1961). The basin

“does not show a clear mixed layer in the annual mean (…), maximum change in density with depth occurs at 75 m” (Potemra et al. 2003:page). Small upwelling processes probably occur via

Ekman pumping along the southern coasts of Flores, and intensify during the peak of the dry season. At this time, the maximum productivity is around 0.35 m d-1 (Potemra et al. 2003).

Salinities and temperatures, thus, vary in relation to orography and seasonally with the action of the monsoons and water circulation. Nutrient content, primary production and transparency are affected extensively by these gradients (Tomascik et al. 1997:1009).

3.3.2 Circulation

Tidal patterns differ from other regions west of Lombok and Kalimantan, having two low and two high tides a day (mixed tide, prevailing semidiurnal, Tomascik et al. 1997). It is influenced by the strong currents originating in the western Pacific Ocean that redistribute the heat absorbed into the Indian Ocean. The circulation pattern is known as the Indonesian throughflow and has global significance for climate regulation (Fieux et al. 1994, Tomascik et al.

1997, Tapper 2002). Maximum transport of water between oceans occurs in the Southeast monsoon (Potemra et al. date).

3.3.3 Atmospheric components

The Sawu Sea is positioned in what is termed the “Maritime Continent,” the most important tropical area in terms of atmospheric and convective activity during the austral summer, where “massive exchanges of energy” occur (Tapper 2002:5). Two important climatic variabilities in interannual and monthly scales affect the spatial and temporal distribution of rainfall regimes at the local level.

The Asian-Australian Monsoon is driven by the thermal differences between surface temperatures of continental masses. During the austral summer (December to March), cold masses flow from continental Asia to the equatorial Northwest Australian heated landmass following differences in pressure. A low pressure cell that draws air from the equatorial zone displacing the Inter Tropical Convergence Zone south of the equator and bringing high humidity in subequatorial regions. Precipitation is most intense during this period with means around 200 to 400 mm (Tapper 2002, Aldrian 2003). In March and April the Northwest winds subside as the trade southeasterly winds become stronger. During July and August, southeast continental winds bring dry cold masses of air from the Australian winter (Adrian & Djamil 2007:437), reducing the intensity of rains throughout the archipelago. As a result, extreme droughts occur that can expand for 4 to 5 months in Sumba and Flores. September and October are another transitional period where the trade winds loose strength. In November and December southwest Ocean winds enter Eastern Indonesia, and progressively shift to Northwest winds as the Wet Monsoon cycle begins again.

3.3.4 Intraseasonal/Interseasonal variability

Onset of the monsoon is influenced by ENSO events (delay onset of rains, early start of dry season), convective and synoptic-scale events. El Niño phenomenon refers to the differences

44 in the gradient of pressure measured between Tahiti and Darwin, the Southern Oscillation Index

(SOI), altering meteorological conditions at a global scale. Changes in the strength of the convective activity of the Walker cell (east to west Pacific Ocean) driven by differences in sea surface temperature (mostly the Pacific Warm Pool) affect precipitation events and the frequency of droughts. In Eastern Indonesia this translates to below normal rainfall indexes and extremely dry conditions (Tapper 2002, Monk et al. 1997).

Figure 3.2: Common Fish Families observed in Ende: Common Tropical Coral Fish families (from top left to right): Chaetodontidae, Holocentridae, Nemipteridae, Ostraciidae, Caesionidae, Kyphosidae, Balistidae, Pomacentridae, Serranidae, Achanturidae, Lethrinidae, Scorpionidae, Pomacentridae, Serranidae, Diodontidae.All pictures by Victoria Ramenzoni.

Figure 3.3: Coral and Fish Families observed in Ende: Coral families: Pocillophora, Acroporidae, Fungiidae, Agariciidae, Siderastreidae, Oculinidae, Merulinidae, Mussidae, Caryophylliidae, Dendrophyllidae. Fish families: Sparidae and Syngnathidae. Pictures by Victoria Ramenzoni,and The Nature Conservancy, Taman Laut Sawu website.

Figure 3.4: Common Coral Species and others observed in Ende: From top left to right: sea star, octopus (Octopoda), sea star, sea cucumber, porcupine fish, crab, soft coral, giant sea clam (Tridacnanidae), lobster (Nephropidae), porcupine fish, moray eel (Muranidae), sea horse (Sygnathidae) and sea turtle (Chaelonidae). All pictures by Victoria Ramenzoni.

Figure 3.5: Common Fish Families and others observed in Ende: Fish families from top left to bottom: squid (Teuthida), needlefish (Belonidae), Surgeonfish (Acanthuridae), and flying fish (Exocoetidae). All pictures by Victoria Ramenzoni.

Figure 3.6: Common Fish Species observed in Ende: Common fish families from top left to bottom: Jack (Carangidae), ray (Dasyatidae), sardines (Clupeidae). All pictures by Victoria Ramenzoni.

Figure 3.7: Common Fish Families observed in Ende: Common fish families from top to bottom: halfbeaks (Hemiramphidae), groupper (Serranidae). All pictures by Victoria Ramenzoni.

Figure 3.8: Common Fish Families observed in Ende: Common fish species from top left to bottom: coral fishes, skipjack (Katsuwonus Pelamis). All pictures by Victoria Ramenzoni.

Figure 3.9: Common Fish Species observed in Ende: Common fish families from top left to bottom: yellow stripe scad (Selaroides leptolepis), grunts (Lethrinidae). All pictures by Victoria Ramenzoni.

3.4. Traditional ecological knowledge and climate change: why optimization is not

“rational”

One of the key criteria among Optimal Foraging Models and Rational Action Choice is the idea that decisions are always made considering the whole set of alternatives at hand.

Optimization is the result of a sound evaluation of outcomes in terms of all possible options and their assigned probability (Gigerenzer et al. 1999). From a cognitive approach, however, rational action choice entails a set of psychological skills and preferences that is far from being realistic

(Quinn 1978, Gladwin 1980, 1971, Gigerenzer 2008). For example, it implies the ability to have perfect knowledge about the environment or to clearly conceptualize the probability values of different choices and alternatives in terms of risk perception (Mithen 1989, 1990). This misconstruction of skills and preferences is the result of a lack of studies on the cognition of fishing decision-making processes (Bene and Tewfik 2001, Colfer et al. 1999).

In marine environments, choice is always riddled with uncertainty (Mangel & Clark

1983, Acheson & Wilson 1996, Hillborn & Mangel 1997). The amount of fish present in a particular fishing spot cannot be readily or accurately ascertained, weather conditions are hard to predict, and probabilities are not always easily perceived (Quinn 1978, Gladwin 1970). Dynamic ecosystems, rapid choices, and changing conditions in the socioeconomic environment all constrain the structure in which decisions need to be made and render the idea of an exhaustive consideration of alternatives implausible.

Within this context, far from perfect knowledge, research has shown that people rely on local mechanisms of prediction and ecological knowledge to secure livelihoods and adaptation

(Godoy et al. 2009, Tucker 2007b, Orlove et al. 2002). Much of this knowledge has been formalized in systems of predictive cues that encompass fishermen’s experiences and

54 observations over centuries (Cordell 1974, Bjarnason & Thorlindsson 1993, Paolisso 2002). In other cases, knowledge has remained implicit or embedded in cultural practices (Dove 1993,

Rappaport 1968).

Over the last half century, with climate change and advanced environmental degradation due to intensification of extractive practices, ecological patterns have been altered. While uncertainty has affected the efficacy of local belief systems, in some regions this has not undermined their use. Predictive cues are consistently incorporated into scientific forecasts among African and Indian farmers to anticipate droughts and plan crops (see Roncoli et al. 2001,

2002, Acharia 2010, Pareek & Trivedi 2010).

This has not been the case in Ende. Despite the fact that there are no available forecasts even at the regional level, former predictive mechanisms have become unreliable and their use by younger generations less frequent. But, as it will be argued later, this does not indicate that fishermen do not rely on environmental cues or that they remain unaware of environmental patterns like it has been suggested for the Ambonese small purse seine fishery (van

Oostenbrugge et al. 2001).

Through interviews and surveys among Endenese fishermen, I was able to determine that an informal system of weather forecasting and maritime conditions was in place well before the introduction of engines and fishing intensification in the 1980s. In conversations and fishing trips, I was able to record a thorough body of environmental and climatic information in terms of cues or signs of the marine ecosystem. The association of environmental indicators to fish stocks would permit a fisherman to estimate presence or absence of fish, weather events, and currents.

In spite of being frequently used, this knowledge remains fragmented and to some level implicit making elicitation an arduous process.

Difficulties might be rooted in the fact that even older fishermen have now begun to challenge the certainty of predictions. Thirty to forty years ago weather conditions could be determined with moderate exactitude before going to sea, and predictions on stocks and climate could extend to longer periods of time like seasons. Nowadays, such knowledge is rare and might only be applicable if the frame in which decisions are made is modified or new patterns of variability can be detected that encompass previous cues.

One good example of the changes in the efficacy of predictive knowledge can be found in the use of fishing calendars. According to most fishermen, it is widespread knowledge that fishing patches are selected on the basis of an annual calendar regulated by the monsoon seasons and moon phases that permits to calculate the presence and abundance of certain species. In this system, winds and sea water temperature might be the most important factors determining catch, unit of effort, and sailing conditions. But as a consequence of increased climatic alterations, the onset of the dry and wet monsoon seasons has changed (see Badan Pusat Statistik Kabupaten

Ende for climatic data, Hamada et al. 2002, Aldrian and Susanto 2004). This has brought many interviewees to mention the impossibility of relying on calendars anymore to establish with certainty the availability of fish species (“ikan tidak kenal musim lagi” or fish do not know seasons anymore).

In fact, in the 1980s, precipitation events would commonly start in October and continue until late march (Badan Pusat Statistik Kabupaten Ende 1984-2010). These were preceded by a reduction of the strength in the Trade Eastern Winds (angin timur), an intensification of Western and Northern winds (angin barat, angin utara). With the wet monsoon, changes in currents and sea water temperatures would increase the availability of species like small tuna, squids, and anchovies. However, in the last two years, the Western winds which inaugurate the wet season,

56 lack strength reducing precipitation events. The onset of the rainy season has been delayed until

December and shortened its duration. This indicates a significant change in climatic patterns that affect marine species in terms of life histories and biomass. Most significantly, it is the opposite of what would be normally expected as a result of the current transitional period (2010-2011) between El Niño and La Niña conditions, signaling the beginning of new precipitation and temperature patterns.

These environmental and climatic alterations not only affect coastal communities by increasing the frequency of extreme events such as cyclones, destructive storms and beach abrasion. But they have also resulted in increased crop failures and reduced catches that have long term impacts on the population's morbidity and mortality rates. With changes in biomass affecting total catches and ultimately reducing incomes, families have lower possibilities of diversifying their diets and paradoxically consume less and less fish. Environmental uncertainty combined to economic instability has created new challenges that many fishermen do not feel prepared to deal with. Under these conditions, it would be reasonable to assume that the change in patterns of variability has affected the competency of traditional forecasting cues and contribute to their progressive disappearance as fishermen perceive their fallibility.

Yet, far from a simple interpretation, these interviews also suggest that previous weather- related knowledge and fishing experience have been reformulated and are still being consolidated in new associations and re-associations of cues. Some fishermen indicated that they pay attention to stars and clouds (shapes, positions, movements and colors) and atmospheric phenomena like lightning to determine wind conditions that might affect fishing.

In some cases, fishermen pay attention to the presence of marine life (zooplankton) to predict currents and winds, and to fishing feeding behavior to anticipate possible fishing spots.

These cues might not be new, though the temporal decision making frame in which they are applied has changed.

With fishing seasons presenting a higher uncertainty on the occurrence of winds and certain fish species, fishermen have begun to target multiple species by diversifying fishing tools. They have also incorporated some small innovations like the use of colorful baits, a practice that is common in other areas in Sulawesi. And most significantly, they have altered their pattern of activities in the wet season. Before, fishermen would remain at home for a period of forty days (December-January-February) while strong western and northern winds would prevent navigation. Nowadays, fishermen go fishing throughout the year, staying occasionally for periods of one or two weeks when storms hit the region. The frequency of their trips has, thus, changed. In addition, with the changes in marine activities from trade to a more fishing based subsistence, their trips and duration have shortened considerably.

However, the reason why optimization might not account for behavior in Ende is not only in terms of cognitive skills and the demands that perfect knowledge imposes in dynamic contexts

(high cognitive costs when decisions need to be quick in a fast changing environment). Indeed, one might argue that changes in predictive systems might reflect an ongoing process of adaptation to develop more accurate representational beliefs and towards achieving optimization.

One could also even argue that optimization towards catch maximization might occur under constraints, or that fishing effort could be best explained by satisficing or ameliorating principles

(Simon 1957, Mithen 1989, 1990). But, as it has been the case for OFT, such line of reasoning cannot be readily tested or empirically assessed (Foley 1985, Gigerenzer & Gaissmaier 2011).

Optimization might not be a rational choice according to Endenese standards, as the main factor explaining the motivation to go fishing might lie not in a profit-driven mentality or in a risk-reduction perspective, but in a more comprehensive approach to uncertainty and life that defies a clear cut probability conceptualization.

As a matter of fact, the most important decision an Endenese fisherman has to face is to determine whether to stay fishing or to return given climatic conditions. This process, which combines the analysis of a number of cues like clouds, current strengths, and the behavior of other fishermen, is not single handedly explained by expertise or by the expectation of the fish to be caught that day (harapan). Similarly, tools or fishing gear do not seem to be the main cause behind catch numbers. Many interviewees when inquired about the role of previous experience and type of fishing equipment indicated that even those that have many years at sea or that employ motor boats with many nets can from time to time return empty handed.

Previous research has established that risk reduction and the avoidance of losses can be an important motivation behind the time spent fishing (van Osteenbrugge et al. 2001, Ammarell

2002). But in Ende, some fishermen are willing to stay at sea under adverse conditions if the catch might be certain, whereas others might favor an early return even when conditions are safe and the fish are eating. Therefore, evidence collected so far suggests that risk preference, experience, expertise, and gear do not completely account for the motivations inspiring fishing effort and decision-making.

The major explanation that is willingly given for variability of fishing effort and success is luck (rezeki). This concept is rooted in Islamic, Endenese, and Buginese-traditional beliefs and rituals (Ammarell 2002, Acciaioli 2004, Pelras 1996). Its causality is complex. According to most interviewees, only God can determine the conditions in which luck occurs (“peraturan

59 dikirim oleh Allah”) and only he knows (“hanya Allah yang tahu”). Because marine environments, as well as any other ecosystem, are the result of God’s creation, they remain unpredictable or random in terms of human perception (“laut sembarang”). The ocean is but a big puzzle (“taka teki”).

In spite of the highly variable conditions surrounding fishing, fishermen can still try to grasp a limited understanding of the ocean that permits them to catch what has been granted for their subsistence (see Figure 3.10). To that end, luck, catch and climate are all related in a system of signs that is given by god to interpret. These climatic signs, described previously as a system of traditional knowledge, are not straightforward and their predictive validity is not fixed. They are effective only with a certain probability. Thus, natural events are not completely predictable as such in this narrative of luck.

The decision to go fishing is indeed inspired in the idea that luck cannot be procured by other means but being a hard worker (harus berusaha) and diligent (rajin). But, overall, one cannot do anything to increase luck with certainty, but go to sea and search for fortune (“Rezeki tidak bias tambah, hanya mencari cari ikan”).

Formal practices that might result in better luck refer to respecting the daily five prayers

(sholat) as established in the Qur'an, and having a pure heart (hati murni). Luck can also be favored from prayers on Monday, Thursday, and Friday (Jum'at) nights that involve the burning of wood in front of the house ("kemenyan"). Furthermore, fishermen follow the adat (rules) set by the ancestors when building boats or venturing on new enterprises to sea, these are all connected to luck. Dreams also hold an important place among some fishermen as they are considered an indication of future success sent by god.

Figure 3.10: System of Climatic Signs and Endenese Causality

Endenese system of luck and causality: As described by informant, God is responsible for all life on earth (natural systems) and weather. Behavior of events can be anticipated and predicted through symbols and signs like dreams, but predictions are only probable. God sends blessings to humans in direct or indirect ways. Blessings (berkah) are interpreted as luck (rezeki). Therefore, luck has only a positive connotation and is predetermined. There is no method to improve luck through sanctioned ways. People can only work hard, be diligent in prayers and endeavor to be good with others. However, through the means of witchcraft individuals can manipulate their luck. This is dangerous as spells and encantations go against religious practices and have costs.

Other ways in which luck is sent by god include the finding of precious objects ("kulavu," barang gaib), though in some cases these might be connected to demons (djins). This practice is associated by more religious fishermen to pagan beliefs (kafir) from the time when the ancestors where around (nenek moyan) and is considered very close to sin (termasuk sirik, dosa). In fact, some informants indicated that they would rather have nothing to do with precious objects as they might provide short term luck at the expense of a huge loss (sometimes human life).

According to them, the devil (iblis) walked the earth way before humanity, and has clever ways of deceiving people. If one transgresses God’s rules by engaging with magic objects risks eternal damnation for there is no forgiveness for such sin. The belief in magic objects as such is common among Endenese that have connections with Lionese groups or that reflect a mixed

Endenese-Lionese descent.

Finally, luck is also associated with following old adat rules when fishing for some species of coral fish ("ikan asa," Serranidae and Lutjanida espp.). According to such prescriptions, fishermen cannot talk, smoke, cook, or eat when fishing on one of these patches or they would risk making the fish angry.

Overall, it is interesting to observe, that next to the use of weather cues, this traditional body of knowledge and rules related to luck has become sparse among the newer fishermen who do not believe (“Orang tidak percayaa lagi”) or follow the established rules (“Tidak ikut peraturan dari dulu”). As one of the elder fishermen states, the lack of fish or failure in the catch can be the result of not respecting the former ways: “Harus percayaa atau tidak dapat ikan.

Dulu biasa per bulan perahu penuh, sekarang tidak yakin. Dua atau tiga hari lagi, habis” (One must believe in order to catch fish. Before, boats used to be full throughout the month when returning from fishing. Now, after 2 or 3 days there is no more fish).

In conclusion, one could say that the evidence presented here is not entirely incompatible to explanations of fishing effort by maximization practices. At the individual level, risk preferences and non-verbal processes of probability perception (unconscious) might still result in optimization over the long term. However, it is crucial to emphasize that luck as the main motivation behind fishing effort places rewards in a future after life and not in the achievement of material success. In addition, this narrative of luck implies a certain attitude towards nature that shapes the perception of ecological patterns. But luck also defines suitable rules on how to interact with an environment and which expectations are valid. This, in turn, constraints decision-making processes and resource use practices.

Therefore, local perceptions of environmental uncertainty and nature are key to understanding what lies behind resource exploitation, along with religious beliefs and cultural values. They should be addressed by government agencies and conservation institutions to design culturally sound management practices. I will further discuss the implications of these findings for rational theory and environmental policies in future articles.

3.5. Conclusions

In summary, preliminary findings suggest the importance of ecological knowledge in fishing effort decision-making and the existence of different attitudes towards the use of marine resources in Ende. Exploratory interviews indicate so far that neither conservation organizations nor the local government actively incorporate local ecological knowledge when drafting management plans for Ende, and they assume that fishermen are mostly driven by their own maximization of interests.

Nonetheless, in a world where climate change threatens to reshape the global ecology and economy of marine-human ecosystems (Badjeck et al. 2009, Cheung et al. 2009), conservation

63 and management initiatives need to look at the local to understand why certain choices are made before assuming, as they usually do, that cost-benefit rationales apply uniformly. Because complex problems require insightful solutions, conservation and governmental institutions should forge a multidisciplinary methodological and theoretical perspective to engage local needs and vulnerabilities.

Cognitive and behavioral studies of decision-making in small societies can inform such endeavors by telling about the local impacts of overarching policies and the strategies devised to represent environmental uncertainty (Tucker 2007a, Colfer et al. 1999). Future research will explore these issues and ponder the importance of how different conceptions of the marine environment across generations and stakeholders (baselines) ultimately constrain local responses and livelihoods.

CHAPTER 4

AN INTEGRATIVE SOCIO-ECOLOGICAL APPROACH TO FISHING HOUSEHOLDS:

COMBINING OBSERVATIONAL, HISTORICAL AND ENVIRONMENTAL DATA IN

POOR INFORMATION FISHERIES TO ANALYZE ANTHROPOGENIC AND

ECOLOGICAL PRESSURES IN ENDE, FLORES, INDONESIA.13

13: Ramenzoni, V. C. 2014. To be submitted to Conservation Biology.

Abstract

Ecosystem Based Fishery Management, Adaptive Management and Socio-Ecological

Systems approaches are becoming more common in fishery studies. Yet, they do not include an account of resource use and practices from a human perspective. To that end, this article reconstructs current resource exploitation in a small-scale tuna fishery in the Sawu Sea, Eastern

Indonesia using a socio-ecological and historical ecology approach. It underlines the importance of household studies to characterize the drivers of change in an ecosystem. Analyzing nutritional questionnaires (n: 110), market surveys (n: 33), fishing effort surveys (n: 132) and observations

(n: 2633), it qualifies regional statistical data to provide a more accurate picture of resource exploitation. Through parametric models and projections from ethnographic, observational and nutritional data, the studied area is characterized as intensely exploited with significant changes in trophic levels. A Transfer Function analysis is also used to model the covariation in fluctuations between changes in yields and precipitation. Findings emphasize the impossibility of disaggregating anthropogenic components from environmental conditions in their effect on stocks and challenge simplistic perceptions of fishermen as responsible for the current state of the fishery. In order to manage resources adequately, more research should be conducted on the patterns of resource use and livelihoods in small-scale fishing communities. This project should be pursued, as conservation policies affect the survival of local coastal populations.

Key Words: small-scale fisheries, livelihoods, decision-making, fishing effort, parametric models, MSY

4.1. Introduction

In the last 30 years, fishery scientists and managers have dealt with contradictory results

(Charles 2011, 2012). Landings increase as programs target the development of communities at

66 local scales (Panaĭotov and Panayotou 1986, McGoodwin 1990, Satia et al. 2004, Wiber et al.

2004, Jacquet and Pauly 2008). But, overall, overfishing, excessive mechanization, unregulated catches and illegal practices affect productivity and deepen poverty structures among subsistence fisheries (Butcher 2004, Heazle and Butcher 2007, Wiyono et al. 2006, Varkey et al. 2010).

Given the restricted success in fishery management policies (McIlgorm et al. 2010), researchers have identified limitations in tools, perspectives, and information. Studies point to the combination of economic and biological models with top-down managerial approaches as one of the most significant factor in governance failures (Miller et al. 2010, Perry et al. 2005, Perry et al. 2010, Charles 2012).

Through a case study in Ende, Flores, Indonesia, I contend that it is the lack of an integrative socio-ecological approach at the household level that constrains both the design of adaptive responses to fishery harvesting regulation, and the understanding of anthropogenic influences in overfishing (Charles 2012). I show these limitations by combining multiple sources of information (ethnographic, observational, nutritional, and historical; Zeller et al. 2007, Zeller et al. 2006, Palomares and Heymans 2006, Teh et al. 2007, 2011, Shackeroff and Campbell

2011) to reconstruct the story of stocks, fishing effort, and livelihoods at small domestic scales

(Ellis et al. 2001, Coulthard et al. 2011, Britton and Coulthard 2013).

My objectives were to: 1) formulate a parametric model for the fishery based on fishing effort and yields to determine the current state of stocks and baselines; 2) model the relation between changes in yields across time and environmental factors to determine if the latter could explain the fluctuations experienced by fish species; 3) qualify and correct regional-level data with household observations and market and fishing effort surveys; 4) to generate an estimate of yields from household observations; and 5) to back cast potential Catch per unit of effort (CPUE)

67 and extraction demands to conceptualize what these changes might imply in terms of subsistence strategies and options.

The article is organized in 8 sections. The first two discuss the theoretical and practical underpinnings behind the research. Section three, four and five present the ethnographic and methodological context for the study, while section six, seven and eight introduce results and a general discussion on their implications in fishery governance.

4.2. Small-scale fisheries and humans dimensions of resource use

Ecosystem Based Fishery Management, Management Strategy Evaluation and Socio-

Ecological Systems perspectives

Since the late 1990s, ecosystem based fisheries management (EBFM) and adaptive management approaches like Management Strategy Evaluation (MSE) aim to correct previous informational and methodological limitations in coastal and marine sciences by capturing the complexity of multi-scalar fishing environments (Link 2005, Perry et al. 2010, Holland and

Herrera 2009). Comprehensive statistical models that include ecosystem and socio-economic variables are more common (Ecopath with Ecosym, SEAPODYM, etc.). The assessment of harvesting strategies incorporates uncertainties from rule implementations and trade-offs

(Milner-Gulland 2012). Even so, sparse data, insufficient technological development, and lack of scientific input constrain the application of these framesets to small-scale fisheries (Zeller et al.

2007, Silvestre and Pauly 1997, Alfaro-Shigueto et al. 2010, Milner-Gulland 2012).

In addition to EBFM and MSE, the socio-ecological systems framework (SES) is gaining prevalence in fishery sciences (Charles 2009, 2012, Kittinger et al. 2012). SES recognizes the importance of processes of interdependence between coastal ecosystems and human communities. It devotes substantial effort to the design of tools to anticipate the impacts of

68 climate change and to reduce vulnerabilities (Perry et al. 2010, 2012). Because most of these changes affect small-scale fisheries, the need to understand household patterns of resource use and fishing practices is pressing (Satia and Gardiner 2004, Allison and Ellis 2001, van

Oostenbrugge et al. 2002, 2004, Fox 2005, Allison and Kelling 2009, Charles 2011, Stanford et al. 2013).

In terms of analysis, SES, EBFM and MSE attend to large-scale systems, measuring and projecting the impacts of climate change to fishing stocks at a transnational/national level (Zeller et al. 2007, Cheung et al. 2009, 2010). Great progress is made in representing resource extraction and harvesting practices among industrialized fishing fleets and commercial fisheries (Milner-

Gulland 2011). Unfortunately, none of these approaches explores the role of fishing households as socio-ecological units where changes in resource use take place (Charles 2012:351, Kittinger et al. 2012:321, Miner-Gulland 2012). Models do not effectively incorporate the behavioral and social components of fishing-effort and subsistence decisions (Charles 2012, Fulton et al. 2011,

Coulthard et al. 2011, Symes and Phillipson 2009, Bene and Tewfil 2001, Milner-Gulland 2011,

2012).

Parametric models

One of the explanations offered for this absence of subsistence decision-making perspectives in fishery management tools and policies lies in the difficulties of procuring long- term measures of domestic level production (Silvestri and Pauly 1997, Pauly 2006). Within bio- economic sciences, datasets from small-scale fisheries often result from rapid appraisals and diversity monitoring surveys conducted by non-profit organizations and regional offices

(Wiadnya et al. 2006). Assessments, constrained by funding and logistic impediments, lack the investigative depth necessary to provide useful information for behavioral modeling and

69 prediction (Pauly 2006, Zeller et al. 2007). Scarcity of information at a finer scale and paucity of longitudinal observations forces managers to rely on incomplete projections (Ingles et al. 2008,

Pauly 2006). As a consequence, policy design in small-scale fisheries still relies on a parametric and single-species understanding of the health of fishing stocks (the Maximum Sustainable Yield or MSY, McIlgorm et al. 2010, Acheson and Wilson 1996, McGoodwin 1990, Plaganyi 2007,

Wiadnya et al. 2006, Mous et al. 2005). This implies limitations on how the motivations behind decisions of resource use are understood.

A parametric model represents a population of elements through the study of a small set of its attributes (parameters) and their values (variation, mean). In ecology, it is possible to estimate the biomass of a fishery knowing the growth rate of the fish harvested and the total landings. As an example, the Maximum Sustainable Yield model (MSY) characterizes the associations between yields, monetary returns and fishing effort to project the moment when the maximum sustainable is approached (Sparre and Venema 1996, Conrad 2010). 14 On such bases,

MSY provides scientific support for the labels “overfishing” “extremely exploited” and “overly exploited fisheries” used by international organizations to define policy regulations and guide decision criteria (see IUCN). These regulations have concrete consequences for resource users, setting harvesting limits and quotas in a precautionary approach (Bene et al. 2010, Charles

2011). But, optimal exploitation level assessments suffer from problems of quantification and conceptualization.

14 An aspect of note is that fishing effort can be defined as the number of hours, trips or boats fishing over a certain period of time. In addition, it is important to observe that MSY presupposes the biological equilibrium of natural stocks. By keeping all conditions constant, the model assumes that changes in level of exploitation produce proportional responses in biomasses (Conrad 2010).

By constraining the kind of information utilized (fishing effort and yields) to estimate biomasses, MSY models do not mimic real socio-ecological scenarios (Wiadnya et al. 2006).

They are less adept at dealing with environmental and historical trends that explain population dynamics and resource use and confound causal agents with processes (Die and Caddy 1997,

Caddy 1996, 1999). In short, estimations of total catch might reflect intensity of fishing during a particular set of years rather than actual stocks. Inaccurate interpretations lead policy managers to the wrong conclusions and result in harvesting regulations that might endanger the livelihoods of thousands of fishermen and their families.

Limitations with parametric models in human behavior relate to perceiving resource exploitation as a problem of preventing the loss of returns in an open access environment

(Gordon 1953, 1954, Conrad 2010). Polices represent actors as narrow-minded self-interest maximizers (Feeny et al. 1990, McCay and Acheson 1990, Berkes et al. 1990) or firms that have the goal of increasing profits (Milner-Gulland 2011). Whereas this might be realistic in commercial settings, it does not work well when the units of analysis involve subsistence economies (Tucker 2007, 2011, Tucker and Taylor 2007). As underlined by Milner-Gulland

(2011:1): “Households act very differently to ﬁrms, because they often exploit the resource as one of a suite of productive activities, and their aim is not to maximize proﬁts but to maximize household welfare, or utility”.

Aside from ignoring the complexities and trade-offs within domestic economies, there is a tendency among fishery scientists to identify fishing with poverty (Bene 2003, Allison and

Ellis 2001). Lack of livelihood options encourages more people to take up fishing, thus exacerbating pressures on already depleted stocks (see Pauly 2006 for a Malthusian model; Ha et al. 2012). This brings managers and policy makers to the conclusion that resource degradation results from an absence of development initiatives and greedy fishermen.

As the main goal of this article, I suggest that restrictions in perspectives and approaches within fishery studies impair the comprehension of the role of anthropogenic forces in overharvesting and management failures. I argue that the most significant challenge in the research of “small-scale fishing communities” rests in the successful integration of the human dimensions of resource use in framesets like MSE, EBFM and SES (Perry et al. 2010, Miller et al. 2010, McIlgorm et al. 2010, Symes and Phillipson 2009, van Oostenbrugge et al. 2004,

Kittinger et al. 2012). This task requires the analyses of multiple sources of evidence that go beyond statistics and parametric studies on fishing effort (Bene and Tewfil 2001, Salas et al.

2004). It presupposes a systematic understanding of the ways in which resource policies and environmental transitions affect extractive behavior at the household level (Milner-Guland 2011,

2012).

Household level analysis: livelihoods and wellbeing

Livelihood, wellbeing and behavioral theories build from a household centered frameset to explore resource use practices and how different portfolios of economic activities enable the survival and regeneration of fishing families (Allison and Ellis 2001, Allison et al. 2009, Charles

2012, Allison and Horemans 2006, Coulthard et al. 2011, Britton and Coulthard 2012). They criticize narrow technical perspectives of fishery governance that do not consider the realities

72 behind subsistence decisions (van Oostenbrugge et al. 2004). The focus of livelihood and household level studies is on the cultural, socio-political and economic dimensions that explain exploitation strategies. The goal is to determine how such decisions transform into socio- ecological outcomes that might enhance or hinder the continuity of a household (da Silva and

Begossi 2009, Hanazaki and Begossi 2003). In some cases, these approaches benefit from an observational component that provides quantified valuations on the economic returns of different strategies (Oliveira and Begossi 2011, Lopes and Begossi 2011, van Oostenbrugge et al. 2004).

A household level analysis, because it focuses on the primary users of marine resources and extractive behaviors, can be of relevance to fishery models and policy design. First, it can inform managers and scientists on the impact of human activities on coastal ecosystems. It can provide observational data on the sustainability of certain stock populations by quantifying yields in artisanal and small-scale fisheries that are largely understudied (see Teh et al. 2007, Pauly and

Zeller 2014). It can also help anticipate the timelines of future extinctions and rates of extraction by providing information on the differential access to technology and aid, and on how socio- economic pressures translate into ecological pressures. Most importantly, observations can correct misconceptions of what data is representing. By capturing the local motivations of fishermen, it can help re-shift managers and scientists’ perceptions of resource users and facilitate the emergence of real co-participatory approaches.

The case study presented here takes on a socio-ecological systems approach (Charles

2012, Kittinger et al. 2012) and a household centered level of analysis (Milner-Gulland 2011,

2012) to: 1) explore the human dimensions of resource use in a small fishery in Ende, Flores,

Eastern Indonesia, 2) disaggregate the effect of human and environmental pressures in fish stocks. To that end, I focus on the changes experienced in non-industrial yields over the past

73 thirty years. I assume that said modifications in catches and household subsistence strategies can be discernible by exploring oscillations in demographics, environmental conditions and yields through a historical-ecology perspective (Al-Abdulrazzak et al. 2012, Saenz-Arroyo et al. 2005,

Butcher 2005, 2004, Jackson et al. 2001, Kurlansky 1997) and through the use of observational and ethnographic data.

It is my primary goal to challenge perceptions of fishermen as the main cause of the current decline of the fishery. I also suggest that households are spaces where social and ecological dimensions are integrated in decisions about resource use (Bene 2003). Thus, households should be considered a key unit of analysis in modeling harvesting strategies.

My research follows this rationale: First, I revise local information with provincial and

FAO Eastern Indian Ocean EZZ datasets to reconstruct inter and intra-annual yields according to a parametric approach. Then, I identify long-term trends in domestic fishing effort and in catches from late 1980s until present. I explore whether fish populations are decreasing, and if that is the case, whether decreases respond to historical ecological trends, reflect a recent transition to a semi-industrialized system along with other anthropogenic impacts, and show “fishing down the food web” effects. I also consider environmental factors that can explain the current state of the fishery. Finally, I rely on observational data to correct regional estimates and back cast a potential extraction level and domestic level catch per unit of effort. I use regression models, time series analysis and multivariate methods. Results underline the significance of mixed research approaches that operate at the household level to complement parametric models. In the last sections, I discuss the implications of these findings for fishery management.

4.3. Study area

The study area is located in the Regency of Ende (Kabupaten Ende), central Flores, Nusa

Tenggara Timur (Figures 3.1-9, 4.1). It comprises 10 subdistrics: Pulau Ende, Nangapanda,

Ende, Ende Utara, Ende Selatan, Ende Timur, Ndona, Ndori, Wolowaru and Wolojita, as well as an area of about 1670 km2 of fishing grounds in the Savu Sea. The main ethnic groups are Ende-

Lio (approximately 80.000 people), distributed in coastal and hinterland villages. Unlike other small-scale fishing communities with more stratified corporative groups (Barnes 1996), the primary unit of economic organization in Coastal Endenese villages is the extended household

(or family house, sa’o mere). To note is that anthropological accounts of kinship systems have emphasized the loose structure of clans in coastal regions of Ende (Needham 1968, Kennedy

1955, Van Wouden 1968) and the lack of a system of property and regulation. As of present time, there are no fishing organizations, cooperatives or customary institutions that can effectively regulate resource extraction or access to fishing areas. Finally, the most important individual in the household is the maternal uncle, with descent being traced patrilinealineally and residence being patrilocal.

Fishing grounds of the Savu are known for their catch of pelagic species (Scombridae,

Clupeidae, Carangidae families). Within a large bay, the grounds are characterized by shallow reefs near shore, and reef barriers that fringe deeper waters. Equipment includes gillnets, purse seines and ring nets, and hand-line fishing. There are about 2050 boats distributed in the ten counties studied, with the majority comprised of small plank boats (100-200 kg) and about 2 thirds of small engine motor boats (below 0.5 ton). About 8000 fishermen operate actively, with few seasonal sea weed farms and shark hunting activities (BPS Ende 2012: 275). Canoe fishermen stay within 2 miles off shore and fish over reef patches.

Figure 4.1 Gear and vessels used: Fish line for large pelagic fish, fish rod, plank boat (<1ton), canoe. All pictures by Victoria Ramenzoni.

Figure 4.2 Gear and vessels used 2: I mages from top to bottom: motorboat (<1ton) and large lampara boat (<5ton). All pictures by Victoria Ramenzoni.

Motorboats travel longer distances, and target upper water column pelagic fish (Spanish mackerel, Scomberomorus commerson). Fishing operations are at the level of subsistence, with large catches sold at regional markets. No industries are based in Ende. There are pole and line fishing fleets operating in the Savu and Timor seas based in Kupang (ACIAR 2001).

The average annual landing for the whole district of Ende from 1984 to 2011 is 6300.7 tons, about 2.6 % of the reported catch for the whole Eastern Indonesia Province (BPS Ende,

Dinas Perikanan Ende, FAO 2013). The smaller area focus of this study had a mean annual landing of 5674.3 tons in 2011 constituting a 75% of the marine production of the entire Ende regency.

The climate of Ende is dry in comparison to other areas of Indonesia and is strongly affected by the action of the North-West and South-East Monsoons. Fishing seasons are determined by the winds which prevent navigation. During the months of February and March, storms and cyclones constrain fishermen to stay at home. This is considered the low fishing season. In April-May conditions start to reverse and the peak season begins.

The landscape combines low precipitation, high intense solar radiation and strong winds.

It drastically varies depending on topography (Auffenberg 1981, Monk et al. 1997). The city of

Ende had in 2011 an annual total rainfall of 961 mm in the lowlands, while in Wolowaru, in the uplands, precipitation was 2169 mm (BPS Ende 2012: 38-39). Heavy rains occur during the monsoon season spanning from December to March, whereas July to August precipitation approaches zero mm a day. On average, annual temperatures are 28 °C at sea level (BPS Ende

2012:42), with heat indexes of 46 °C. Recent evidence indicates that since the 1980s the frequency of ENSO events appears to have increased (Annamalai and Liu 2005). Local authorities (personal communications, Kantor Pertanian, Kantor Bencana Alam Ende) report

78 changes on the onset of the rainy season, along with the occurrence of extreme events and the number of extreme days of heat. Sea surface level further threatens coastal communities through high levels of erosion and beach abrasion.

Ende is a relevant case study to explore changes in resource use occurring at the household level. First, as in many parts of the world, in Eastern Indonesia the quality of the landing information is of questionable validity (Stacey et al. 2011, Mous et al. 2005). The central government imposes a protocol for collecting information, but the level of implementation varies at the regency level (Wyadnia et al. 2006). Fishing auction offices are not functional. Most measurements are gathered in regional markets and then projected to calculate monthly and annual level figures. Estimations of yields suffer from underreporting or the conflation of results with other regencies. Observations are irregular and results vary among different offices. Along with serious problems in consistency, the statistics do not include domestic consumption or the amount that is shared with other households. Therefore, by implementing a household centered approach, findings from this case study are relevant to qualify and correct official statistics and can be used to derive recommendations on veracity of stock information.

Secondly, despite abundant historical literature on environmental relations in Southeast

Asia (Zerner 1994, 2003, Boomgard 2007, Fox 2005), ecological research is infrequent throughout the Eastern parts of Indonesia and Nusa Tenggara Timur (NTT). As a consequence, there is no reported census of fish species or data on the level of endemism of the region (Monk et al. 1997:177, Allen 2008, Allen et al. 2003, Allen and Adrim 2003). Conservation and transnational institutions have called attention to the high level of overexploitation and resource degradation in the Eastern Indian Ocean economic zones (UNEP 2008, Ingles et al. 2008). In

NTT, local fishermen have observed sharp reductions of populations, with benthic organisms and

79 coral species suffering a dramatic decline (Fox 2005, Munasik et al. 2011, Ahmad and Munasik

2013). There are also reports that species with rapid maturation and short life span such as squids and anchovies are increasing in numbers; while slow reproductive species like sharks, paddlefishes and rays are critically diminished (Tull 2009, Blaber et al. 2009). With scarce information on current the state of stocks, it is very difficult to assess claims of overfishing.

Anecdotal and ethnographic research indicates that numbers of fish have reduced notably in the last thirty years. But this conclusion has not been fully supported by quantification. Through a detailed study of household resource use, this case study can provide useful information for the design of models of human behavior and harvesting strategies that can complement stock assessments and biomass estimations.

Finally, in addition to intensive fishing, unreported catches and illegal practices such as non-permitted trawling gear (pukat harimau) are reported in the Eastern Indo Pacific region

(Stacey et al. 2011, UNEP 2010, Wagey et al. 2008, Heazley and Butcher 2007, Fox 2005). In the Arafura and Timor seas, Thai, Korean and Taiwanese vessels operate illegally under

Indonesian flags (Stacey et al. 2001, ACIAR 2001, Blaber et al. 2005). Indonesian long liners fish relatively close to Ende when monsoon conditions prevent the operation in other areas

(ACIAR 2001). These illegal fishing fleets, due to their larger tonnage and refrigeration capabilities, have a critical impact on depleted biomass that eclipses the effect of artisanal fishing. They remain largely unaccounted by regional governments, only mentioned by local, non-profit or media groups. Exploring how Endenese households rely on marine resources to secure their livelihoods can provide information on the level of extractive pressure exerted by local populations. It might be used to estimate the influence of illegal fleets in the region and design adequate monitoring and harvesting measures.

4.4. Data collection and sources

The combination of quantitative information, historical, ethnographic and observational material was one of the central goals of this project. The data were collected over multiple field seasons in different offices and archives. Interviews and archival research were carried out from

May until August 2009, from November 2010 until January 2011, and from June 2011 until

January 2013. I conducted archival research in Ende and Ndona at the Perpustakaan Daerah and at the Perpustakaan Umum Universitas Flores, Flores, and at the Perpustakaan Umum at the

University of Gadjah Mada and the Perpustakaan Kolese Ignatius, Yogyakarta, Java, at the

Candritya Seminar Library and the Ledalero Seminar Archive both in Maumere, Flores; and at the Royal Tropical Institute (Koninklijk Instituut voor de Tropen) in Amsterdam and at the

Catholic Archives at Radboud University in Nijmegen, The Netherlands. I obtained district level statistical information from the Statistical Bureau of Ende (Badan Pusat Statistik Ende or BPS), the Ende Regency Fishing Commission (Dinas Kelautan dan Perikanan or DKP), the volcanology and meteorological station (Stasiun Meteorologi dan Vulkanologi Gunung Iya), the environmental office (The Badan Lingkungan Hidup Daerah Ende or BLHD), the Agricultural office (Dinas Pertanian) and the Health Services Office (Dinas Kesehatan, Diskes).

From June 2011 until January 2013, I resided in Ende and Pulau Ende and conducted intensive ethnographic, observational and ecological research. In total 132 surveys and about 120 semi-structured interviews were carried out among active fishermen in Ende and Pulau Ende. In addition, 110 households (adults) were monitored for changes in weight in the dry and wet season and interviewed on food frequency and preferences in three villages: Ipy, Rendo Rate

Rua and Tanjung-Paupanda. Weighing of portions and special foods was done with a 3840

BLTBL™ Digital Nutritional Scale.

Primary demographic information was also obtained from the local village offices and sub-disctrict representatives, and from five partial censuses in Rendo Rate Rua, Ekoreko, Ipy I, II and IV, Koponggena II and Arubara. To monitor household fishing effort (the number of trips to capture fish), observations and diary logs were utilized at two villages in Pulau Ende from

September 2011 until July 2012 (9 months). A total of 1515 (766 canoe, 749 motorboat) fishing trips were recorded, including hour of departure, return, type of boat and gear, and species total number of fish caught. Fish could not be weighted as the catch would be sold at sea or right after return to port. In all cases, species were recorded in the local name and then identified with Reef

Fish Identification Guides during focus groups and interviews with fishermen (Allen et al. 2005,

Allen et al. 2003). Finally, 33 surveys were carried out at the local market of Mbogawani, Ende city, with fish vendors that assessed: 1) total catch sold per day, 2) number of trucks from other regencies, 3) average profits and other information.

4.5. Data analysis

4.5.1 Parametric model: Calculation of the Maximum sustainable yield (MSY) the

Maximum sustainable effort (fMSY) and potential productivity

I estimated the MSY and the fMSY for the fishery on the bases of effort data (number of boats, assuming an average trip of eight hours) and total yields from 1984 until 2011 for the entire Ende Regency. The Statistics Office of Ende (BPS Ende) and the District Fishing

Commission provided updated information on yields and effort. To calculate the CPUE (catch per unit of effort or yield per unit of effort), I corrected for differences in gear and tonnage of boats. Different weights were assigned to the number of boats depending on their tonnage and potential optimal catch (Wudianto et al. 2002). I assigned a weight of 0.50 to canoes, 0.8 to motorboats and 4 to boats with more than 5 tons. The weight represents the catchability of each

82 gear (Sparre and Venema 1998), with canoes equating 1/8th of the catch of medium size purse seiners. My estimation of catchability was very conservative and responded to observations of returns at the port of Ende. One unit of catchability corresponded to 100 kg. It is important to observe that the fishery has no heavily mechanized industry or fleet. Only 158 medium size purse seiners of more than 5 tons operate seasonally in the bay.

Because Ende has access to two different seas (Flores in the North, and Savu in the

South) but disaggregation of data for the two regions does not occur for all years, I calculated

MSY and fMSY for both areas combined. Further separation of data at the county level showed that landings and effort for the Savu region corresponds to 75% of the total reported per year. In addition, to insure the consistency of records, information on catches for the EZZ of Eastern

Indonesia (Eastern Indian Ocean, FAO Statistical Collections, FishStatJ 2013) was regressed against the annual landing data. This procedure permitted the extrapolation of missing values and the control of outliers.

The following model (Fox, state dependent) was used to generate a graphical representation of the relation between yield and effort (FAO/Venema and Sverra 1998, Conrad

2010).

Where Y is the yield, f the effort, c is the intercept and d the slope for the line ln(Y)/ ln(f) on f. This representation does not violate the assumption that CPUE declines with increasing effort.

Where B is the biomass and q is the catchability coefficient.

4.5.2 Seasonality and environmental factors affecting yield

To isolate environmental variability from effort in predicting yield, I conducted three kinds of analyses of the monthly catch and precipitation from 1994 until 2011: simple regressions, multiple regressions (Stepwise), and multivariate time series analysis. I could not include years before 1994 because I lacked consistent data on market operations that would allow me to establish monthly yields. Precipitation, wind and surface temperature data were obtained from the volcanology and meteorology office and the agricultural office in Ende city.

Measurements come from a single weather station that has been operating since the late 1970s at the foot of the Volcano Iya (W 121.63616, S 8.87278). The consistency of the data was checked by comparing other meteorological stations’ reports (Waingapu, Sumba; El Tari, Kupang; and

Darwin, Australia) through the Australian National Bureau of Meteorology and the NOAA databases for world climate. All climatic data points used were summarized monthly averages

(wind direction and wind speed) and totals (precipitation). In addition, I relied on a National

Weather Service NOAA database to assess ENSO/La Niña events from estimations of sea surface temperatures and their anomalies for El Niño 3.4 region (N 5-S 5, W 120-170) and the

Oceanic Niño Index (ONI). Based on the ONI (a three month running mean season of detected surface anomalies), the database classifies cold and warm episodes by season when values exceed a threshold of 0.5C for a minimum of five overlapping seasons. In accordance with this classification, I assigned a 2 for cold episodes (La Niña), a 1 for warm episodes (El Niño) and a zero for normal conditions for the months between 1994 and 2011.

I began the analysis exploring the existence of associations in the data with simple and multiple regressions on yield versus wind and precipitation. Initial statistical information on yields obtained from the Fishing commission and the Statistics Bureau in Ende was not

84 segregated by months. To disaggregate yields, I relied on statistics providing total number of tons sold by month at the local market for the same period. I estimated the percentage of the total yield that a particular month represented in the annual data of the tons that were sold, and extrapolated that percentage to the annual landings. This procedure was followed for each year of the series (1994-2011). When data was missing for particular years (3 cases) I calculated the average as a combination of the next and previous years for that particular month and then projected that percentage to the total annual landing data to assess the distribution per month.

To further explore the effect of environmental variables in predicting yield, I fitted different indicators in a Stepwise multiple regression model. I included monthly averages of precipitation, wind direction, low wind speed, high wind speed, effort, sea surface temperature anomaly, and fish mortality for the period of 1995-2011 as predictive variables. I also included

ENSO as a dummy variable. Mortality was calculated as the differences between CPUE by month (Conrad 2010). I chose the model based on the Akaike score and taking into account multicollinearity. Finally, because multiple regression models consider each observation as independent, I relied on multivariate time series analysis (Auto Regressive Integrated Moving

Average Models) to explore the autocorrelation of data points over time (Stram and Wei 1990,

Wilks 2011).

4.5.3 Correcting regional-level estimates for annual yields

To correct regional level statistics, I relied on surveys conducted at the local market. I estimated the amount of daily catch sold that originates in other regencies by multiplying the average reported number of foreign trucks by their mean tonnage (750 kg). This was scaled to the year by multiplying by 365 (total outside yields = mean average truck number x 750 kg x 365 days). Then I subtracted this number from the district statistics for the ten counties.

4.5.4 Estimating yields through observational and household level data

To formulate an informed estimation of resource extraction for the ten county area, I combined primary data obtained through household surveys and fishing effort observations. Only non-zero catches were used. Therefore, estimations of CPUE and other parameters might be positively biased.

Similarly to Teh et al. (2007), I assumed that fishing observations and survey results are representative of the whole fishing population (~8.000) in the area of study; and that fisheries were mostly artisanal and subsistence based. I calculated the annual yield per fisher by gear

(canoe and motorboat) by multiplying the mean CPUE (catch per unit of effort; table 4.1) by the reported number of days fishing (19.3 a month, 228 a year). Then I multiplied this number by the population estimate of active fishers in relation to the gear (BPS Ende 2011).

Because returns of fishing expeditions were in number of fish and not in kg, I disaggregated the catch per species and multiplied units by average weights of captured fish.

About 90% of the catch or more was constituted by kembung, also known as short or Indian mackerel. A scombrid that inhabits shallow areas (genus Rastrelliger, species Rastrelliger brachysoma), it has an average weight between 100 to 250 grams. Hence, I used 200 grams as the standard weight per unit of fish.

Table 4.1 Main characteristics of the fishery

Characteristics

 Coastal terrain surface area (km2): 131.92  Marine surface area (km2): 1669.95  Mean depth (m): -2345.82  Shelf area (km2): 425.916094  Mean depth of shelf (m): -89.88  Primary production (g Cm-2a-1): 220.3368 +/- 67.6815  Mean estimation of tons of pelagic and demersals per year: 3,100 tons (CI 5,900-1,800)  Mean number of days fishing (reported): 19.3  Mean number hours per unit of effort: 9

Perahu sampan kecil (jukung) or dugout canoe: 1-2 people.  Potential number of boats in active use: 200  Max Catch: 100-200 Kg16.  Small pelagics. Coral fishes.  Fishing gear: Hand line (monofilament no. 500-1200). Bottom long-line. Size of hooks 6,7, 16, 17. Bait made of chicken feathers, cloth, plastic.  Fishing areas: close to shore (max. distance 120 minutes) and near reef barriers.  CPUE: 27.18 fish (CI: 32.2-18.1). (9% other fish, 91% small pelagic): +/-7 kg (CI: 8.21Kg-4.6 Kg).  Extraction about: 316,118 kg per year (CI: 374,421-210,466 Kg) Small boats (1 ton): 1-4 people (2-3 crew).  Potential number of boats in active use: 1800  Max catch: 500 kga.  Small pelagics (tongkol, flyingfish, scad, mackerel) and tuna.  Fishing gear: gillnets (2.5-4.5 inches, multifilament, 5 to 7 pieces). Handline.  Fishing areas: Distances between 30 minutes to days.  CPUE: 49.8 fish (CI: 56-33). +/- 12 kg (CI: 14,2-8.4 Kg).  Extraction about: 5,116,200 kg per year (CI: 5,860,512-3,453,516 Kg).

15: See Annex for calculation. 16: Wudianto et al. 2002.b:c: http://www.iucnredlist.org/details/21857/0

4.5.5 Household consumption and back casting yields

I used population demographics and results from nutritional assessments to: 1) explore household livelihood conditions, 2) generate a level of consumption of marine products, and 3) to back cast changes in the level of resource extraction to the late 1880s. This calculation allowed me to speculate on how livelihoods returns might have changed through time.

To generate an approximation of the amount of marine resources consumed by the population of the studied area, I derived a weekly estimate of consumption of fish products by season and by type of household (fishing or farming) from nutritional surveys and food security reports (Reinhard 1997). I multiplied this estimate by the duration of the dry season and wet season (26 weeks each) and summed both to derive the total consumption per person per year.

To elevate this measurement to the population level, I multiplied then fish intake per person per year according to the proportion of farming and fishing households present in the area. Regency level data indicates that fishing households represent about 30% of the total population with the remaining 70% characterized by farming activities (Nakagawa 2006, Ende BPS). Therefore, I multiplied 30% of total population numbers by amount of annual intake per person in fishing households, and 70% of total population numbers by amount of annual intake per person in farming households. Adding both figures provided the total annual consumption.

To back cast prior values of resource extraction, I projected changes in CPUE based on

MSY calculations and historical data. These estimations are minimum projections and they assess the level of marine exploitation according mostly to non-numerical evidence. First, I assumed that the value of CPUE in 1880s was probably low with extraction only through the use of traditional gear (Weber 1899-1900, Roos 1872, 1877). Endenese were mostly devoted to inter- island trade and marine resources were harvested for subsistence. Roos mentions that fishing is

88 not as intensive as in other coastal societies (1877:493). Thus, I considered that if households comprised about 7 members in average, there should be about 1,428 fishermen from a population of 10,000 (Roos 187717). From these, I assumed that each represented a canoe or potential unit of effort. Then, I assigned a catch weight of 0,60 for each unit of effort (weighed unit of effort). The weight summarized the potential catch of a canoe (about 60 kg if one unit of catchability is 100 kg) given that stocks were probably in better conditions than in recent time. Total yields were calculated multiplying the unit of effort by the total of days fishing (180) and average weight of prey. I also corrected for average weight of prey, assigning a mean value of at least 400 grams for each unit of fish. I presupposed that trophic effects had scarcely occurred following considerations generated in other regions of Eastern Asia (Butcher 2004, 2005); therefore catches were bigger or included higher trophic species.

At the beginning of 1900s, CPUE probably increased due to Dutch administrative organization and the introduction of intensification practices (Dietrich 1983, 1989). With new jurisdictions and rajadoms created, consumption might have increased as new populations get access to fish. In addition, roads were built through extractive labor (Van Suchtelen 1921). In all, changes can be seen as demographic censuses became more comprehensive and population numbers went up. To account for these fluctuations, I doubled the CPUE from the late 1880s anticipating that reported increases in demography primarily reflected more efficiency in data collection and the creation of the Rajadom of Lio.

By late 1940s, CPUE was 4 times that of late 1800s in regions like Java (Butcher 2004).

In Ende, harvesting increased during the Japanese occupation due to the demand of fish from the military forces. In addition, the Japanese provided explosives to generate a fast easy way to

17: Roos mentioned that the total population including a about 4,000 slaves was 14,000 souls. He also mentioned that Endenese had about 6 to 7 children per family.

89 increment returns as fishermen mentioned in interviews. From 1950 until 1988, exploitation levels went up throughout Indonesia (about 10 times, Butcher 2004, Zeller et al. 2009).

Therefore, I captured this change by conservatively increasing CPUE to double from the late

1920s-1930s. However, since more boats were at sea and other practices to capture fish began to be used, I decreased the weight assigned to each unit of effort to 0.40.

Because during the 1970s it is likely the time when a transition to other motor boats began, I projected CPUE and Total yield values until 1975. I estimated values for 1983 considering information from regional measurements for 1984. The rest of the measurements also originate from this dataset. Starting 1989, CPUE went down critically. By 2011 it was 2.8 times lower than 1984.

It is important to observe the following conditions when interpreting these projections:

1) Lack of source data prevented me from including changes in consumption levels and in the number of active fishermen;

2) Aside from general population numbers, there was limited information that would allow me to investigate how household composition has changed through time so I supposed that households included 7 individuals until early 1970s following Roos (1872, 1877). This is when drastic political alterations and modernization occurred with Suharto;

3) I incorporated changes in CPUE and other measures in a linear way;

4) I considered that level of fishing effort determines captured yields without influence of environmental factors; and

5) Regarding estimations for the 1880s, I assumed that the whole population of available adults (about 1400 individuals) fished actively. In reality, it was probably closer to 1/3rd or ½ of the total. Therefore, CPUE is probably higher for this period. I based this decision on the lack of

90 a systematic assessment of population numbers before Dutch intervention. I supposed that the current population figure reported by anecdotal sources is underestimating total effort and, hence, impact on stocks as reported for other regions (Butcher 2004, 2005).

In all, projections are minimum estimates and should be read with caution.

4.6. Results

4.6.1 Parametric model: Maximum sustainable yield and effort

Based on the regression of the natural log of catch per unit of effort and effort measured as number of boats, I estimated the MSY in 5183.6 tons and the optimal effort (fMSY) in 1388 boats. The parameters for the regression model can be found in table 4.2.

Fishery scientists commonly refer to three developmental stages of a fishery: an initial moment of development that brings an increase in CPUE as mechanization and new equipment makes more species available and improve catch’s efficiency, the MSY situation when the optimal effort in terms of CPUE and costs is approached, and a third moment when yield and

CPUE fall with progressive overexploitation if proper sustainable measures are not adopted. In this case, the interpretation of the graphical relation between yield and effort indicates a linear decrease in the catch per unit of effort through time (Figures 4.3 and 4.4). The introduction of boats probably through government aid in the beginning of the 1990s created a substantial drop in the CPUE. Total catch went up in the following years, but the decrease in CPUE is about two and a half time times the original values. This might indicate that in the early 1990s, the degree of intensification exceeded the predicted optimal effort. As more fishermen gained access to boats and subtidal and mesopelagic fishing grounds, the catch suffered redistribution. Eventually, the increasing pressure on stocks affected growth and reproductive rates of the fishery, bringing yields down.

Table.4.2 Maximum Sustainable Yield and maximum sustainable effort

Predictive Equation: ln (Y/f) = 2.317 - 0.00054*f

Model Fit Parameters Values RSquare 0.81 RSquare Adj 0.80 Root Mean Square Error 0.17 Mean of Response 1.43 Observations (or Sum Wgts) 28

Fox Model Savu MSY: 5183.62 tons fMSY: 1388 boats

Maximum Sustainable Yield or MSY

Figures 4.3 and 4.4: Maximum Sustainable Yield and CPUE per year Interpretation of the graphical relation between yield and effort suggests a linear decrease in the catch per unit of effort as there is an increase in the number of boats, or with time.

During 1999 and 2000, variation in CPUE per month was higher than in other years, maybe in relation to the strong La Niña event of 1999 that reduced all catches (Fegan 2001).

Throughout the mid-1990s and early 2000s, the level of effort oscillated given inconsistencies in plans and new decentralization policies. By 2008, it started to intensify again reaching 2021 boats in 2011. As a result, catches went up but for different species than twenty years before.

The scatterplots of yield per species per year indicate the different evolutions suffered by some groups (Figures 4.5, 4.6, and 4.7). Catches of big tuna and elasmobranches have decreased due to their particular life histories and spawning cycles (late age to reproductive maturity, low larval survival; CCSBT 2010). The decline in predator groups and potential fluctuations in the environment (increasing frequency of ENSO events) has favored smaller kinds of fish like sardines, fusiliers and macks with rapid maturation times. Overall, the fishery has become more productive (van Densen 2001). But this outcome might mask the effects of intensive exploitation as “fishing down the food web” occurs (Pauly et al. 1998) and accessible areas become depleted.

Figures 4.5 and 4.6 Trajectories of fish groups (1990s – 2010s)

Figure 4.7 Trajectories of fish groups (1990s – 2010s)

4.6.2 Environmental determinants of the catch

Because the changes in effort and yield are of a non-monotonic nature, the detection of an inter-annual trend becomes difficult (van Densen 2001, van Oostebrugge et al. 2002). Landings might respond to exceptional recruitment years given climatic conditions or changes in intensity of exploitation (Lehodey et al. 1997, 1998, 2003). To explore this association further, I conducted simple regression analyses of mean monthly precipitation and catches, and wind and catches. Results were highly significant in both cases (Adjusted R2: 0.78 yield by rain, and

Adjusted R2: 0.73 for wind by yield; p < 0.0002), indicating that fish availability and effort dynamics are influenced by the Monsoon seasons (Wiyono et al. 2006).

As ENSO events are becoming more frequent and affect the regimes of winds in the region, I included ENSO events as a dummy variable in a Stepwise multiple regression model

(OLS). I also included monthly averages of precipitation, wind direction, average low wind speed, high wind speed, effort, sea surface temperature anomaly, and fish mortality for the period of 1994-2011 as predictive factors. I expected that the model would drop some of the independent variables given multicollinearity effects. The following model minimized AIC

(25843.6); with Adjusted R2: 0.49 and p < 0.0001:

With indicating average precipitation per month, average low wind speed per month, average effort in number of boats per month, average wind direction per month, fish mortality per month, all weighted by month. The fit of the model suggests that about half of the variability in monthly yields is related to effort (.30) and climatic events (.19). For example, an increase of 1 millimeter of rain per month was associated with a decrease of 369 kg in yield (p

= 0.0023); and an increase in the average low wind speed per month was associated with a

97 decrease of 32670.9 kg in yield (p = 0.0004). This is not surprising as conversations with fishermen emphasized the importance of winds and precipitation as determining the numbers of days that can be fished across seasons (Table 4.3).

Table 4.3 Multiple regression yield per month

Summary of Fit Weight: Month SSE DFE RMSE RSquare RSquare Adj Cp p 3.067e+13 198 393583.38 0.50 0.49 6.20 6 AICc BIC 5843.692 5866.347

Current Estimates Parameter Estimate nDF SS "F Ratio" "Prob>F" Intercept 216686.75 1 0 0.000 1 Precipitation - 369.01 1 1.47e+12 9.474 0.00238* Low - 32670.91 1 1.95e+12 12.563 0.00049* Effort 304.03 1 1.87e+13 120.999 2.8e-22* Avdir - 682.95 1 1.39e+12 8.950 0.00313* Mortality - 699.57 1 4.93e+12 31.798 5.84e-8*

*: Indicates significance at : 0.05.

To further isolate environmental variability from effort in predicting yield, I conducted time-series analyses of catch (1995 to 2011) A total of 204 months of yields were analyzed. The study of the distribution of the yield over time showed a very mild quadratic trend with decreasing variability. Differentiation was not necessary. In addition, I applied a log transformation to the data but variability was not reduced. Therefore, analyses used non- transformed data points.

Through the Autocorrelations and Partial Correlations plots that examine associations among observations, I identified a seasonal ARIMA (2,0,0) (1,0,0)12 model. The model was

98 selected through AIC scores and considering the distribution of residuals (Table 4.4 and Figure

4.8). It accounts for about 66 percent of the variability in the data.

The multivariate time series model (Transfer Function Analysis) considered yield as an outcome variable of precipitation (predictive variable) from 1994 until 2011. I proceeded by exploring the best fit for precipitation through pre-whitening. Once I determined the parameters,

I selected the best model that could predict yield as a function of precipitation. The examination of autocorrelations and cross-correlations between the residuals of the model and outcome and predictive variables proved non-significant. Interpretation of the parameters of the model indicate that amount of precipitation predicts negatively yield on lag 1 and 2, with only lag 2 being statistically significant (p < .0001). That is, an increase in 1 millimeter in rain per month reduces catch for the subsequent month in 422 kg.

Table.4.4 Seasonal Arima and transfer function

Summary of Fit DF 212 Sum of Squared Errors 3.976296e+12 Variance Estimate 18756113745 Standard Deviation 136952.96 Akaike's 'A' Information Criterion 5728.68 RSquare 0.62 RSquare Adj 0.61 MAPE 19.23 MAE 87575.18 -2LogLikelihood 5720.68

Parameter Estimates Term Factor Lag Estimate Std Error t Ratio Prob>|t| AR1,1 1 1 0.66 0.07 9.25 <.0001 AR1,2 1 2 0.09 0.07 1.36 0.1756 AR2,12 2 12 0.34 0.06 5.06 <.0001 Intercept 1 0 520221.10 54982.25 9.46 <.0001 Constant Estimate 83691.33

Transfer Function Analysis Model Summary DF 200 Sum of Squared Errors 3.5528e+12 Variance Estimate 1.7764e+10 Standard Deviation 133281.76 Akaike's 'A' Information Criterion 5509.34 RSquare 0.66 RSquare Adj 0.65 MAPE 19.05 MAE 85899.66 -2LogLikelihood 5493.34

Parameter Estimates Variable Term Factor Lag Estimate Std Error t Ratio Prob>|t| PRECIP Num0,0 0 0 180.3 95.78 1.88 0.0612 PRECIP Num1,1 1 1 -126.9 97.82 -1.30 0.1960 PRECIP Num1,2 1 2 -422.0 98.16 -4.30 <.0001* PRECIP Den1,1 1 1 0.24 0.22 1.10 0.2727 YIELD AR1,1 1 1 0.63 0.07 8.81 <.0001* YIELD AR1,2 1 2 0.09 0.07 1.38 0.1679 YIELD AR2,12 2 12 0.35 0.07 5.06 <.0001* Intercept 0 0 423573.8 58895.39 7.19 <.0001*

100

Months from 1994- 2010

Figure 4.8: Seasonal Time Series

101

4. 6.3 Data corrections

Among serious problems with unreported catch, statistics do not include domestic consumption or the amount that is shared with other households. Official approximations of total yield are extrapolated through monthly visits to the regional market (Ende Regency Fishing

Commission, personal communication). But, my own observations and surveys conducted with

33 fish vendors at the market of Mbongawani, Ende, indicate that a significant proportion of the fish sold originates in the neighboring regencies of Sika (Sika and Maumere) and Nanga Keo

(Bajawa).

The market brings together about 125 vendors a day, a large fraction of which rely on partners in other towns to obtain fish. On average, five trucks a day (load of 350 kgs) come from

Maumere, Sika, and other fishing towns. This number doubles during the wet season when prices increase in several magnitudes. A conservative approach suggests that approximately 15 to 20 % of what is reported as sold might have originated in other regencies. Consequently, when corrected with my own estimations, total landings originated in the study area and sold at the market might have been closer to 3.314 tons.

4. 6.4 Yield estimates

Fishing effort observations and surveys carried out in Pulau Ende and Ende showed that if 19 days per month are considered as viable fishing days and all boats go fishing, the mean total maximum catch for line and small gillnets combined is about 5,432 tons per year (CI: 6,234-

3,663 tons). This seems like a reasonable estimation based on what I observed in the villages

(Table 4.1). The mean CPUE estimated for small canoes was about 27.1 units per fisher-1 day-1.

With 9% of the catch composed by other fish, and 91% small pelagic and coral reef individuals.

This brings about a mean estimate of 7 kg of catch per fisher-1 day-1. For motorboats, the mean

102

CPUE was 49.8 units of fish per fisher-1 day-1, or approximately a mean of 12 kg of catch per trip. Results are very similar to estimates reported for Banggi subsistence and artisanal fisheries of Banggi, Malaysia (Teh et al. 2007)

4. 6.5 Household level data and backcasting yields

Census and primary data for 2011 report that fishing households have at least 5.7 members, with 1.5 still at school. Information from the Statistics Bureau of Kupang (BPS

Provincial 1984) for the year of 1984 provides a household size of 5.8. This indicates little to no change in the number of family members over the past 27 years.

In relation to alternative sources of income, most of all the fishermen surveyed owned their house. However, only about half of them cultivated land in addition to fishing. Around one third of interviewees mentioned having migrated to other countries like Malaysia or other regions in Indonesia at some point and more than one third mentioned having at least one member of their family abroad. In very few cases remittances were received by those that stayed behind.

The self-reported mean weekly income is $ 380,000 rupiah in the dry season and

$125,000 rupiah in the wet season—these are about $35 and $10 U.S. dollars respectively. It must be noted that these estimates might be lower in reality. Fishing effort data suggests that returns from activities sometimes do not cover operational costs. These vary depending on the gear, with canoes spending about $ 1 to 2 dollars per trip and motorboats around $ 4 to 5 dollars per trip. Consequently, sharing constitutes a significant subsistence strategy with households redistributing their yields with four other families on average. There are virtually no other sources of income at the village but fishing and weaving, with the latter only marginally contributing to subsistence.

103

Results of nutritional observations and surveys place intake of marine products among fishing households at around 1.25 kg a week (5 times on average) during the dry season and

0.750 kg a week (3 times on average) during the wet season. Normal measured fish portions are

150-250 grams per individual per meal. The fish most frequently consumed belonged to the small group of tuna (kembung, genus Rastrelliger).

According to previous research, farming populations consume lower amounts of fish than their coastal counterparts. Households engage in fishing activities only one to three days a week in the non-harvest season and occasionally buy dry fish (Reinhard 1997:61). The frequency of marine products intake was about once a week, with the exception of Nangapanda (coastal village), in 1994.

If this dietary information is scaled to the whole area of study based on demographic reports, total consumption in 2011 for fishing households was about 3,047 tons (Table 3.1). The difference between what is sold at the market (~ 3,000 tonnes) and the landings estimated from observational data (~ 5,500 tonnes) might represent local consumption by fishing households.

However, if we take into account the total consumption for 2011 as estimated by surveys (~

4,824) there is a large proportion of demand that cannot be met by subsistence activities (~ 1,800 tonnes).

Changes in consumption have most likely occurred with modifications in catch per unit of effort along time (Table 4.5, Figure 4.9). For example, during the late 1880s, even when the projected minimum CPUE was comparatively lower than the values reported now, fish was probably consumed more frequently. In fact, anecdotal sources mention that exchange was common among coastal and hinterland villages (Roos 1872) and that large pelagic fishes were common in the area (Weber 1899). In the subsequent decades, CPUE changes might be

104

Table 4.5 Back cast of consumption values, total yield and CPUE

Minimum Minimum Fishing HH Farming HH Total Projected/ Total Projected/ Year Consumption Consumption consumption Real Total populationa Real in kg in kg in kg yields in CPUE kg 1872 14,000 728,000 728,000 0,072* 61,714*

1915 82,054 960,032 560,019 0,072* 1,520,050 132,927*

1924 66,000 772,200 450,450 0,144* 1,222,650 183,291*

1930 116,015 1,357,376 791,802 0,144* 2,149,178 322,190*

1952 141,620 1,656,954 966,557 0,288* 2,623,511 786,598*

1954 145,804 1,705,907 995,112 0,288* 2,701,019 809,837*

1975 179,331 2,098,173 1,223,934 3,322,107

1977 202,613 2,370,572 1,382,834 3,753,406 ~ 1983 211,851 2,478,657 1,445,883 ~ 7,4* 3,924,540 3,500,000*

2004 242,898 2,841,907 1,657,779 3,3 3,130,000 4,499,685

2011 260,428 3,047,008 1,777,421 3,03 5,674,000 4,824,429 a: Ende (Nakagawa 2006, Roos 1872:497, van Suchtelen 1921) *: Mínimum values projected from anecdotal sources.

105

Figure 4.9: Changes in monthly averaged CPUE per year (1990s-2010)

106

connected to external pressures and the introduction of new fishing gear. As the Dutch administration consolidated, the prohibition to continue inter-island trade consolidated and probably drove people to rely on fishing grounds. This might have affected returns having impacts in all kinds of subsistence activities. Households that relied on seafaring as a source of income saw their economic independence curtailed.

After the II world war, as other fishing devices became available, CPUE might have increased. How changes in technology affected stocks remains a mystery given the lack of any kind of information about the region. It can be mentioned, though, that ethnographic accounts indicate impacts in marine life in relation to the eruption of the Volcano Iya in 1969. Elderly fishermen mentioned that with increasing water temperatures coral death occurred, affecting reef stocks. In addition, there were numerous accounts of hunger occurring in the 1970s.

The most critical developments influencing fishing household economies seem to have occurred after the 1980s with important intensification in effort. CPUE values decreased in several magnitudes as population numbers continued to increase. In addition, it is reasonable to assume that more farming households began to incorporate fish into their diets as transportation means improved. But, this information cannot be demonstrated with the current dataset and merits further research. Ethnographic accounts point to a reduction of fish in the past thirty years, with critical modifications only occurring in the early 2000s. Such a perception matches what other researchers have reported in Banggi, Sabah, Malaysia (Teh et al. 2007). These issues, along with energetic requirement changes due to current activity levels and household composition, will be explored in future articles.

107

4.7. General Discussion and Conclusions

This article reconstructs current resource exploitation in a small scale tuna fishery in the

Savu Sea, Eastern Indonesia from a socio-ecological perspective. Despite limitations in the quality of the data, through analyses and projections based on observational, historical and ethnographic research at the household level, I was able to assess the importance of a parametric model in understanding the causes of resource overharvesting, and to explore livelihoods wellbeing.

All the different estimations (MSY and Observation based) approximate about 5000 +/-

1000 tons as the annual yield for the fishery. Results seem to indicate that stocks are under heavy exploitation and might have been overfished at the beginning of the 1990s. There is high variability in annual catches as constructed by observational data, with numbers of yields between 6,234-3,663 tons. In addition, effort numbers are beyond the parameters recommended by the fMSY. The fishery seems to be progressing to further intensification that might compromise sustainability of returns.

While drawing these final conclusions, it is difficult to say whether MSY calculations are reflecting intensity of effort instead of real biomass numbers (Wyadnia et al. 2006). Data does not permit to determine if the parametric model can offer reliable accounts of stock numbers because it does not include any biological estimations of growth or survival. Considering that caveat, recent increases in the yield suggest a mild positive trend for the fishery as a whole. But, the composition of the target species seems to have changed indicating overharvesting.

Previous research has shown four potentially different traits in coral resource clusters that denote overfishing (Carder et al. 2007:589): changes in mean sizes and body proportion of reef fishes, general recruitment numbers and abundance, abundance of snapper and grouper species to

108 other reef fishes (Lutjanidae and Serranidae families), and harvesting of pelagic fishes like tuna and sardines (Scombridae and Clupeidae families). Results from this study suggest some of these changes are happening in Ende. For example, Lutjanidae species have undergone declines in the early 2000s as well as the superorder Batoidea. In addition, fishermen are exploiting more intensively smaller pelagic species like mackerel tuna (Euthynnus affinis). Therefore, aside from resilient stocks that make the fishery more productive, two of the signs suggesting overfishing are met (Carder et al. 2007, Reitz et al. 2009, Reitz 2004).

Observational and household level results reveal that:

1) Fishing is conducted for subsistence and to satisfy local demand of fish products.

However, fish is neither consumed daily nor in large proportions among fishing or farming households. Ethnographic evidence indicated that fishermen prefer to sell big catches to the market to increase profit and sometimes buy smaller size fish to consume. A large proportion of diets is characterized by carbohydrates (cassava, rice) and might not satisfy the current energetic demands according to activity levels.

Although demographic information was not available to determine household composition changes through time, it is reasonable to assume that modifications in CPUE have presented additional stressors to fishing families given environmental and ecological alterations in sea surface level and extreme events. The oscillation over the last 25 years in monthly CPUE

(figure 3.7) might reflect environmental components affecting catches and might suggest increasing levels of variability in returns. Such findings coincide with what has been reported in other areas of Indonesia (Oostenbrugge et al. 2001, Teh et al. 2007, Wyadnia et al. 2006).

There are some indications that changes in yields have introduced additional risks to the health of fishing households for example the prevalence of diabetes has increased along with

109 other nutritional pathologies (Ende Regency Health Services personal communication, Ende BPS

2012). With instability in returns it is expected that fishermen will start to rely on other livelihood options. However, as there are no other sources of viable employment, fishing homes are experiencing high rates of migration. In terms of salaries, households barely reach the provincial poverty line (set at $ 20 a month). In fact, Flores and Nusa Tenggara Timur have poverty estimates of over 40% (Resosudarmo and Jotzo 2009), making this province the poorest in all Indonesia with the exception of Maluku (Barlow 2007).

2) Current demand for fish is not met by the fishery: surveys with fish vendors reported about 15 to 20% of all the fish sold at the market originates in other regencies. Market based measurements of yield are misleading as they do not account for the proportion of fish that is either imported or consumed at the household before reaching the point of sale. If relying in local government measurements alone, there is the potential to misrepresent the current health of stocks and the role of fishermen in overexploitation.

In all, extraction of resources might be, according to observations of fishing events, around the maximum number estimated for the ten county areas. Based on MSY projections and own observations, the exploitation of Ende bay fishing grounds is considered intense. The recent decrease in high trophic individuals, the ethnographic accounts mentioning reductions in CPUE, the variability in daily returns, and the impossibility of the fishery to meet the demand of fish seem to indicate that some of the marine species in Ende might be in a state of decline or thoroughly overfished.

These conclusions should be considered carefully. Environmental patterns and seasonality have a significant effect on fish recruitment processes and the level of effort, potentially affecting yields for several months and years. Pulses of effort intensification

110 interacting with particular ecological conditions might explain the current biomass and composition of the fishery in terms of a re-structuring the system. Further studies will be conducted to understand the thresholds affecting marine populations and how anthropogenic stressors (changes in fish consumption over time) might create potential tipping points. It is also anticipated that the changes in climatologic variables like precipitation and drought frequency will interact with anthropogenic pressures to augment the impact on stocks and affect exploitation patterns.

Taking all these points in consideration, findings do not clearly disaggregate fishermen’s behavior from other environmental and socio-economic factors in ways that permit an understanding of fishermen’s effect on the present state of the fishery. Results challenge institutional claims that fishermen are the main cause of decline and suggest that an increase of effort bolstered by economic policies might be responsible for the reduction of some high trophic level species.

4. 7.1 Data and policy management implications

Scholars have mentioned the importance of observational methods and livelihood studies to conservation practices (Tucker 2007, Colfer et al. 1998, Godoy et al. 2009). As discussed above, projections about fish stocks championed by parametric models and more recent EBFM,

MSE and SEE approaches rarely include information at the household level (Oostenbrugge and

Densen 2004, Allison and Ellis 2001, Charles 2012, Couldhart 2011). This exclusion can lead to policy failures and further socio-political stressors that might affect the local health of stocks as managers still rely on misrepresentations of what drives resource users and the causes of marine degradation.

111

For example, as it is the case of Ende, fishery programs might not be tailored to meet the local conditions. They rely on unrealistic assumptions about how fishing communities operate, lacking consistency, transparence and equity (Chozin 2008, Resosudarmo and Jotzo 2009). First, initiatives have increased fishing capacity through the introduction of bigger purse seiners accelerating the pace of exploitation of already endangered stocks. The increased number of small pelagic fish has generated the wrong perception that fishing stocks are undergoing a process of growth. As a result of these interventions, households do not report economic or nutritional indicators that suggest wellbeing and are below provincial poverty lines.

At the community level, government subsidies are provided in the form of equipment but only to small clusters of fishermen or preferential individuals. This arises from the belief that older traditional ways of fishing are not adequate as they foster self-interest and non- conservation behaviors. The new policy has the goal of forming fishing clusters or cooperatives.

However, while fishing is considered by Endenese an individual activity guided by personal luck and inscribed within a religious perspective, it is easy to see how policies can clash with local perceptions. The unequal distribution of gear has created conflicts and entrenched inequality structures. Policy failure has resulted in a return to damaging fishing practices as ways of extorting the local fishing commission to secure a fair distribution of aid.

112

Declining stocks, overfishing, and extinctions are not just the result of one event. They encompass the interaction of multiple variables and actors operating at different scales (West

2012, Lowe 2006, Brosius et al. 2005). Synergy between uneven technological modernization, environmental change, and development pressures can significantly stress biomass numbers and diversity. However, it is also the larger frame in which these interactions occur that can determine the outcome of ecosystem conservation policies. But this is also a level of abstraction that is not usually fostered in institutional settings.

Absence of private capital and industry in Ende does not mean that large scale vessels

(>10 tons) are not affecting its fish populations. Companies like PT Dharma Samudera

(Indonesia) have Bottom Long Liners boats (BLL) operating in the Savu and Timorese waters with Kupang (Timor) and Probolinggo (Java) as landing ports. In 2001, Brian Fegan estimated that about 350 to 400 BLL were fishing in Eastern Indonesia, with 70 BLL operating off Kupang

(ACIAR 2001: 9). “[E]ach of the BLL boat takes 6 to 8T/month of mixed demersals, of which

30- 40% is red snapper, 20-30% is P. multidens (=kunir=anggoli=kurisi bali), the rest mixed others.” The situation does not seem to have changed over the last ten years. BLL boats from

Benoa, Bali, are also operating in the eastern Sawu fishing for tuna (Bubler et al. 2009). As a consequence of illegal fishing, by-catch and unreported landings (UUI), elasmobranch, demersal and big pelagic fisheries are dwindling throughout the eastern Indonesian seas (Stacey et al 2011,

Ingles et al. 2008, Tull 2009).

Therefore, statistical data that is non-qualified by contextual and historical information at the household level obscures the cultural and political linkages that are driving the system. It can signify the loss of opportunities to achieve sustainable economic development and conservation regimes by making co-participation unachievable. When precautionary approaches like

113 zonification become impositions that are not backed up by popular consultation, managers risk turning fishermen into enemies. Unfortunately, this is already happening in Ende, where bad policy might be worse than no policy at all.

Over twenty years ago, McGoodwin wrote about the importance of humanizing fishing studies (1990). Ecosystem based fishery management inspired by socio-ecological approaches at the household level is a potential answer to this challenge. For that, better informed managers and policy makers are needed. This article is but one step into that direction.

4.7.2 Conclusion

Results from this case study indicate that the fishing grounds near Ende in the Sawu region are potentially overfished or under heavy exploitation. Because there is significant coupling between stocks and environmental conditions, the causes of degradation cannot be clearly established. Policies that plan to increase fishing effort should be discontinued before more assessments can be conducted in the area. Household-based approaches to the study of resource use need to inform future regulation as livelihoods are experiencing a significant stress.

The successful management of stocks in this area would benefit from solutions that consider long-term economic sustainability of the population and the control of UUI pressures.

114

CHAPTER 5

TEMPORAL PATTERNS IN FISHING EFFORT: USING MULTILEVEL METHODS

TO EXPLORE THE CHANGES IN TIME ALLOCATION ACCORDING TO THE LUNAR

CYCLE OF A SMALL-SCALE FISHERY IN EASTERN INDONESIA. IS NEW SOCIO-

ENVIRONMENTAL VARIABILITY REDEFINING FISHING PROFILES?18

18 Ramenzoni, V. C. 2014. To be submitted to Marine Policy.

115

Abstract

This study explored how time allocation to fishing activities changed according to the lunar cycle and seasonality among 25 tuna fishermen in Ende, Flores, Indonesia. It relied on a combination of standard statistical approaches, multilevel models, and discrimination analysis applied to long-term repeated observations of fishing events (n: 2633). Results permitted to generate three fishing profiles: generalist, conservative and opportunistic. Findings challenge previous research of the effect of the lunar cycle on effort, suggesting that new socio-ecological pressures have an impact on the time spent fishing during the full moon. This conclusion extends beyond equipment type (motorboat and canoes) and states the importance of studying single fishing profiles to design sound policy measures.

Key Words: small-scale fisheries, decision-making, traditional ecological knowledge, lunar cycle, climate change, multilevel models

5.1. Introduction

The study of fishing behavior is of critical importance to the design of conservation policies (Teh et al. 2012, Beitl 2012, Lopes and Begossi 2011, Aburto et al. 2009, Stelzenmüller et al. 2009, Hilborn 2007, Soulie and Thebaud 2006, Branch et al. 2006, Tewfik and Bene 2004,

Salas and Gaetner 2004, Babcock and Pikitch 2000, Johannes 1978). The Food and Agriculture

Organization (FAO) estimates that approximately 47.8 million people worldwide are full-time fishermen, with over 85% percent of fishing vessels being less than 12 m in length (2012). Small scale fishing fleets represent a significant contribution to world catches. However, their patterns of effort allocation remain largely underreported in official statistics and harvesting policies

(Guyader et al. 2013, Demaneche et al. 2008, Zeller et al. 2003, Pet-Soede 2000, Salas et al.

2004).

116

Over the last few years, scholars have tried to correct informational gaps through bio- economic, livelihoods and behavioral approaches (van Putten 2012, Branch et al. 2006, Lunn and

Dearden 2006). Across a variety of inland and coastal fisheries, researchers have investigated how fishermen allocate time to fishing by looking at temporal and spatial variability in catches

(Teh et al. 2012, Beitl 2012, Poos et al. 2010, Hamon 2009, Mahevas et al. 2008, Salas and

Charles 2007, Salas et al. 2004, Sosis 2002, Bene and Tewfik 2001, van Desen 2001, van

Oostenbrugge et al. 2001, 2002, Pet-Soede et al. 2001, Gragson 1992, Begossi 1992).

The exploration of the associations between seasonality, the lunar cycle, and fishing effort has offered insights into how temporal patterns affect subsistence decisions (Srisurichan et al, 2005, van Zwieten et al. 2002, van Densen 2001, van Oostenbrugge et al. 2001, 2002, Salas

2000, Pet-Soede et al. 2001, Di Natale and Mangano 1995). Findings have shown that yields experience substantial changes during the full moon and during different seasons (Teh et al.

2007, van Zwieten et al. 2002). Although there is a high level of heterogeneity in world catches given differences in techniques and target species, fluctuations in yields due to the lunar cycle can be of several orders of magnitude. As a consequence, understanding how behaviors are affected by temporal patterns is central to the effective governance of stocks.

Behavioral research on decision making has not been conducted at a scale that would allow for the identification of ideal-type strategies or detailed profiles in temporal effort allocation among subsistence fishermen (but see Salas and Gaetner 2004, Salas et al. 2004, van

Oostenbrugge 2001, Pet-Soede et al. 2001, 1999). In addition to this lack of specificity in fishing patterns, there seems to be an absence of information on how environmental fluctuations introduced by global climate change are already affecting harvesting decisions by changing the periodicity and predictability of seasons and tides. Assessments are conducted in a way that does

117 not allow for the identification of the full range of adaptation strategies (Dessai and Hulme

2004).

Ecological change endangers the continuity of fishing livelihoods around the world through multiple pathways. Within the next 10 to 20 years, acidification will affect survival of shellfish and finfish species. Stocks will not follow previous timelines of migration and recruitment, with significant relocations given temperature and circulatory changes in the ocean

(Cheung et al. 2009, 2010). The increasing level of uncertainty in yields will drive fishermen to adapt their level of effort. Projected responses involve changes in the intensity and the regional extension of fishing activities. These will introduce further pressures to conservation programs.

Yet, in small-island tropical ecosystems where impacts of climate change are already been experienced (IPCC 2013), responses are not projections but realities. If managers and scientists work under the assumption that adaptations are yet to come, they risk missing components of resiliency and flexibility (Dessai and Hulme 2004:109).

In the next sections, this article explores how the increase in socio-ecological uncertainty has introduced modifications in fishing patterns related to the moon phase and the seasons among small-scale fishermen in Ende, Flores, Indonesia. The purpose is to understand with high- level specificity how ecological and individual variability affect harvesting behaviors and generate a range of adaptive options. The goals are: 1) to propose ideal-type strategies or detailed profiles in temporal effort allocation that can be used for the design of policies in the fishery, 2) to provide a methodological roadmap of how to derive such profiles. To that end, I briefly discuss how temporal regularities have been studied in traditional fishing societies in section 2.

As a part of section 3, I characterize the limitations of foraging and behavioral approaches in their statistical treatment of observations and how that has restricted the understanding of fishing

118 effort and the formulation of profiles. Sections 4 and 5 present an ethnographic description and a detail of the methods employed. Section 6 and 7 include results and a general discussion that underlines the importance of behavioral profiles to conservation and harvesting policies.

5.2. Temporal patterns, regularities and decision-making in fisheries

Successful prediction of temporal regularities requires decision-makers to detect the behavior of the different components and events that characterize the informational structure of an environment (Simon 1947, 1957, Todd and Gigerenzer 2012, Gigerenzer 2008, Tucker 2007).

Climatic, atmospheric, physical, and biotic elements constitute a collection of units and processes that create the conditions for organisms to survive and reproduce in an ecosystem

(Cashdan 1990, 1992). In order to adapt to external and internal pressures, individuals must perceive change and regularity (van Densen 2001, Dall et al. 2005, Stephens 2008, Stephens and

Dunlap 2009). This is a psychological process that reflects the construction of constancies, contingencies and expectations from world events, and can be defined as variability discrimination (Wasserman et al. 2004).

In the domain of time, fishermen deal with variation at multiple levels (Howell and

Burnett 1978, Acheson 1981). Hence, it is expected that they will classify variability in different ways. For example, fishermen observe regular change in seasons and moon cycles. They can predict change by proxy indicators (i.e.: cold sea surface temperatures indicate less fish) or they can expect change without having precision on when it will actually happen (i.e.: the occurrence of winds in a particular season). Finally, there is change that, given its emergent nature from the combination of highly infrequent conditions, it is entirely random and unforeseen. One good example of that is the non-occurrence of winds in a certain season or the increased intensity of a tropical storm.

119

Ethnographic and anecdotal accounts have largely explored the effect of temporal patterns in regulating fishing activities (Gladwin 1970, Ammarell 2002a, 2002b, Acciaioli 2004,

Sopher 1965, Southon 1995, Chou 2003, Acheson 1981). This effect was described by

Malinowski back in 1918 in the Trobriand Islands (1918, 1921, 1978) and by Firth in Malaysia

(1965). Cordell (1974) detailed the extent to which traditional fishermen in Brazil rely on cyclical regularities of the tides and the moon to decide where to fish each day. He emphasized that tides are one of the most important ecological factors because they determine fish behavior and the positioning of gear, thus affecting fishing techniques and yields. The lunar month is another critical variable in explaining choice of fishing times and spots as it directly influences tides through gravitational forces and the duration and availability of light at night.

Cordell also suggested that fishermen’s traditional knowledge of the tidal-lunar cycle reflected an adaptive strategy to procure a greater catch. In this way he assumed that the calendar was a reflection of an ecological reality: the reproductive and migratory behaviors of fish respond to moonlight and tidal conditions. Fishermen developed a system that could trace and predict through repeated experiences and memory when it was more proficient to fish in certain patches in relation to others (1974:388).

In Ende, Flores, Eastern Indonesia, where the current research takes place, small-scale fishermen seem to closely adhere, at least in interviews and informal conversations, to the lunar cycle to allocate their monthly effort. While at sea, they make extensive use of equipment like nets, lures, and lamps to attract a catch. Fishing trips begin around 5 pm and extend into the early hours of the morning. On average, fishermen stay at sea for about 6 hours relocating to different places depending on the gear.

120

During the night, marine organisms migrate to depths between 0 to 85 meters to feed

(epipelagic region). This process is called diel vertical migration and congregates small and predator fish near the ocean’s surface. Lamps are used to capture fish as the light draws smaller organisms which are, in turn, predated by larger ones. Fishermen use baited hooks deployed at different depths and nets to trap the fish. When the moon approaches plenitude, the efficiency of artificial lights to attract a prey decreases and yields diminish. On these nights, fishermen are not likely to go fishing.

Over the past decades, random ecological variability has increased in Eastern Indonesia with climate change (TROFFCA 2006). For example, extreme events such as droughts and s have incremented with more frequent El Niños. El Niño/La Niña Southern Oscillation (ENSO) refers to a coupled system of atmospheric and oceanic variations that produce extreme climatological events like excessive precipitation and dry conditions at a global scale. Along with ENSO, changes in currents and circulation patterns have also been observed. These have affected how fishermen perceive the structure of the environment and adjust their effort. When environments are highly unpredictable, knowledge is imperfect and uncertainty is high. Hence, fishers might be better off being non-systematic and non-cyclical in their effort allocation (van

Densen 2011, Dove 1993).

Researchers have also suggested that the stochastic nature of fish resources further constrains the optimization of strategies (see van Oostenbrugge et al. 2001; Pet-Soede et al.

2001) and can render non-sustainable practices more frequent (Lowe 2006, Mous et al. 2000,

Cesar et al. 2003, Pet-Soede et al. 2000). As the impact of climate fluctuations in marine stocks becomes more drastic (Cheung et al. 2009, 2010) and exacerbates socio-ecological pressures, the diversity and characteristics of allocation strategies might signify additional pressures on the

121 ecosystem. Fishermen might increase their exploitation of already intensively harvested species by fishing for longer periods of time, expand to other fishing grounds applying greater demand on stocks, rely on non-sustainable fishing strategies to increase yields and/or contribute to the extinction of threatened groups by non-adhering to governance regulations. Therefore, studies of fishing effort at a fine-grained resolution are central to devising sustainable management alternatives (Aburto et al. 2009, Salas et al. 2004, Branch et al. 2006, DeFeo and Castilla 2005,

Lopes and Begossi 2011).

Unfortunately, with the exception of a few cases, studies have not been conducted at a finer level of detail that would allow for the identification of ideal-type strategies in effort allocation (Salas and Gaetner 2004, Salas et al. 2004, van Oostenbrugge 2001, Pet-Soede et al.

2001). Ideal-type strategies are formal generalizations of stereotypical behaviors/actions that can provide the templates for modeling how decisions are made in terms of resource use (Weber

[1930] 2003). Maintaining specificity, they reflect common traits of a collection of phenomena without referring to a singular particular case. In the study of a fishery, the allocation of fishing events in time can be considered an ideal-type of strategy where individuals distribute their effort along the lunar month in accordance to their expectations of yields. Therefore, investigating their forms and characteristics, their profiles, is of great importance to managers and policy designers as they indicate potential paths of behavioral modification.

5.3. Merging individual data to create profiles: multilevel models and foraging studies

To create ideal-type profiles of temporal allocation a researcher has to: 1) collect longitudinal data (repeated measures); and, 2) analyze it in a way that acknowledges its variance while synthesizing its similarities. It is not surprising, then, that some of the fishing studies of foraging behavior have not been able to meet these conditions. First, it has to be considered that

122 fishermen are a highly heterogeneous group in terms of the technological characteristics of the gear they use and their preferences. At any particular time, a fishery exhibits multiple different arrangements of resource exploitation, with individuals fishing independently, in small corporate groups or even in associations. These arrangements are difficult to characterize in totality and in a consistent fashion. As a result, researchers tend to aggregate information of individual decisions to the group or community level, and lose resolution (Salas and Charles 2007).

Furthermore, variability in decision-making often results from individual preferences, characteristics, and skills that might not be explained by technical or ecological factors alone

(Salas and Charles 2007, King 2011, Ramenzoni 2013). This fact makes the generation of models or profiles difficult as the motivation behind strategies is not always the same.

Within fishery studies, researchers have long hypothesized the existence of a "skipper effect". The effect builds on the belief that skippers play a determinant role in fishing success because of personal attributes like individual luck and skills (Russell and Alexander 1996).

More skeptical researchers have argued that technical characteristics of the gear and randomness are the most significant determinants of yields (Durrenberger and Palsson 1986). As a result, a debate has ensued between “subjective” and “objective” sides. Scholars have tried to identify the causes of fishing success by exploring the practices, tactics, and knowledge of fishers (Ruttan and Tyedmers 2007, Garcia Quijano 2006, Thorlindsson 1988, Durrenberger and Palsson 1986,

Palsson and Durrenberger 1982, Acheson 1981, 1979, Cordell 1974).

123

Yet, the approach to fishing behavior has met the similar methodological difficulties of the more formalized foraging perspectives. Researchers have encountered many restrictions when considering the full ranges of variation (i.e.: how to separate random variation from other kinds of variation) as well as problems in operationalizing variables (i.e.: how to distinguish individual skill effects from technical considerations; Alvarez and Schmidt 2006, Ruttan and

Tyedmers 2007).

Probably the most significant limitation arises from the way that statistical analyses of behavior treat individual fishing events as independent data points (Lopes et al. 2011, Westneat and Fox 2010, Smith and Winterhalder 2000, Stephens 2008, Ruttan and Tyedmers 2007). For example, in optimal foraging theory, the units of analyses are consecutive measurements of time spent in subsistence activities. Researchers collect observations of fishing trips and estimates of catches within particular communities of fishermen (i.e.: Begossi 1992, Aswani 1998, Silvano and Begossi 2001, Sosis 2001, 2002, Lopes and Begossi 2011). To analyze the data, each single fishing event is aggregated to a dataset, which is then summarized in means and rates. Individual characteristics of the fishers only matter to weight the value of observations and are corrected for in the analysis. In the end, explanations are derived that refer to the average generic person or group (but see Koster et al. 2013 and McElreath and Koster 2013 for an application of mixed models in spot check data). Because fishing effort studies usually have a longitudinal design and participants are repeatedly sampled over consecutive days, months and seasons, data points show correlations that violate the assumptions of independence of observations presupposed in the statistical analysis. Correlations are due to individual elements that might bias measurements

(fishing skills), group effects (age), or to the closeness of measurements in time. They can be responsible for some of the variation that the explanatory variable is trying to account for. Most

124 significantly, if they are not included in the analysis they can lead to biased results and incorrect conclusions (Salas and Charles 2007:118, Hayes 2004, Snijder and Bosker 1999, 2011)19.

In the last ten years, mixed models, also known as multilevel or hierarchical models, have become more common in behavioral studies and biostatistics. They permit the simultaneous exploration of the different kinds of variability in the measurements (stochastic, within and between subjects variability) by considering random and fixed effects and by including multiple error terms20 (Cnaan et al. 1997:2349, Bolger and Laurenceau 2013, Bolger et al. 2003, Ott and

19: When applied to repeated serial observations, tests of fit between variables assume that variations in measurements that are a result of stochastic error or individual traits are neutralized over time. As more observations are included in the analysis by enlarging the sample size, idiosyncratic or contextual effects that affect measurements are evened out. Consequently, data points can be considered independent. However, these tests only analyze data at the level of measurements or observations (level 1, Mass and Snijders 2003, Hayes 2004). In truth, the tests do not take into account the different sources of variability (stochastic, within, and between subjects) or structure their analysis in different levels that account for the types of error in the data set. An analysis at level one considers all observations as independent without accounting for any effects that can be a result of the measurements belonging to a group or to a set of individuals that are repeatedly sampled. An analysis at level two implies looking at measurements as belonging to a first tier of analysis (within subjects) and including the effects of groups as defining a second tier of analysis (between subjects). In this way, the second level of analysis is “nested” within the first. If a statistical analysis does not consider the influences of these different levels of variability, it assumes the homogeneity among variances and covariances in the data set (Tabachnik and Fidell 2001). The probability of incurring into a type 1 error that is, proposing a trend when there is none, increases.

20: In mixed effect or multilevel models, fixed effects are usually interpreted as a population mean or the slope and intercept of the generic regression line (regression coefficients, Bolger and Laurenceau 2013). Therefore, they are considered an analysis at the level 1 (see previous footnote). In addition to incorporating fixed effects, multilevel models assume that differences across subjects can influence the values of the outcome variable and thus affect the functional relation with the IV. That is, analyses proceed at level one (observations) and at level two (subjects or groups). The random effect is characterized as the effect that originates from the random selection of a subject from the sample population. Hence, the inferences drawn from the experiment can be extended to the population of the study despite between-subjects variability (Ott and Lognecker 2010). The random component can be interpreted as representing two set of variances: the variance in the fit of an individual subject’s measurements to the general regression line (level 1) and the variance in the fit of an individual subject’s measurements to its own regression line (level 2). Therefore, mixed models can deal with unequal slopes in

125

Longnecker 2010). This gives them more flexibility and makes them a good alternative in datasets where individual profiles are of interest.

Complex approaches to fishing patterns have looked at cluster associations (Mahevas et al. 2008), principal component analysis and factorial techniques (Pelletier and Ferraris 2000,

Hamon 2009), dynamic studies of fleets (Branch et al. 2006), time series analysis (Marchal et al

2007, Srisurichan et al 2005, van Zwieten et al. 2002), and data envelopment analysis (Vazquez-

Rowe and Tyedmers 2013). However, it is only very recently that scholars are applying multilevel analysis to the consideration of fishermen’s behavior (Alvarez and Schmidt 2006,

Ruttan and Tyedmers 2007, see Millar et al. 1007, Millar and Willis 1999 and Stevenson and

Millar 2013 for limitations).

In this article, I rely both on classic statistical techniques and mixed methods to investigate whether fishermen in Ende, Flores, Indonesia, allocate their effort according to the lunar and seasonal cycles. I presuppose that if moon cycles and seasons have a demonstrable effect on catches, then fishermen would not go fishing on unfavorable days or seasons or would fish less than in more advantageous times. Following studies on the lunar calendar, seasonality, and ethnographic evidence presented above, I test these hypotheses:

Hypothesis 121: lunar phases affect probability of zero catch and catch rates, with lower catch rates and higher probabilities of zero catch per subject during the full moon 2 +/- days

(phase 1) in relation to other days of the month (phase 2 and 3).

regression lines of participants and assess the contribution of the IV to explaining variability (Bolger and Laurenceau 2013:45, Bolger et al. 2003).

21: Because of the absence of zero catch observations in the data set, the analysis might be positively biased. Consequently, changes in effort will be measured and perceived as modifications in catch rates (amount of catch per unit of time). See methods section for more details.

126

Hypothesis 2: seasonality affects catch rates, with fishermen having the lowest catch rates in the wet season (season 3) and the highest catch in the dry season (season 1).

Finally, I expect to see differences by type of gear, with motorboats reporting higher catches and lesser time spent fishing overall.

Through multilevel and discrimination analyses and by analyzing the progression of catch rates and time in the lunar month, I identify three types of fishing strategies. I also compare results from the discrimination analysis against a performance index (Millar et al. 1997, Salas and Charles 2007) to explore correlations with better catches per unit of effort, analyze coefficients of variation per subject and study the associations between catches and moon phases.

My goals are: 1) to generate ideal-type strategies or detailed profiles in temporal effort allocation that can be used for policy design, 2) to provide a methodological roadmap of how to derive such profiles. I conclude underlining how socio-ecological uncertainty is re-shaping behaviors and discussing the shortcomings of research and policies that do not incorporate the assessment of individual fishermen’s profiles.

5.4. Ethnographic description

Study site

Pulau Ende, with coordinates W 121.519460 and S 8.879934, is a small island located in the bay of Ende, Regency of Ende (Kabupaten Ende), central Flores, Nusa Tenggara Timur,

Indonesia (see figure 3.1). The population approximates 8000 people with about 1700 fishermen

(BPS Ende 2012) distributed in 7 villages. The main ethnic group is Endenese.

Fishing activities are mostly for subsistence, supporting small-scale trade at the local market. The list of harvested species comprises Scombridae, Clupeidae, Lutjanidae, Serranidae, and Carangidae families. Normal fishing equipment includes small gillnets (3.8 – 11.4 cm mesh,

127

5 to 7 pieces) and hand-line fishing (see Table 4.1). There are no industries based in Ende, nor private investment to develop the fishery. Pole and line fishing fleets operate in the Sawu, Timor and Indian oceans based in Kupang (ACIAR 2001) and Benoa, Bali (Stacey et al. 2011).

In the last twenty five years, Ende has experienced intensification of fishing activities. In

1986 the number of fishing boats was 196 canoes, 613 plank boats, and 72 motorboats. By 2012 these numbers had increased many-fold, while the catch per unit of effort has decreased about four times. Oscillations in the composition of the yields since late 1980s point to a substitution from bigger pelagic to smaller pelagic catches and potential modifications in the trophic levels of some species. The smaller area focus of this study had an annual landing of 918,640 kg in 2010 constituting 12.8 % of the marine production of the entire Ende regency (Dinas Perikanan Ende).

The marine ecosystem can be characterized as tropical, with sea surface temperatures approximating 28°C throughout the whole year. Primary productivity is about 561 mgC·m-

2·day-1 and is affected by the action of the monsoons (Potemra et al. 2003). Environmental patterns have an important effect in the timing and duration of fishing seasons (Wiyono et al.

2006, Pet-Soede et al. 2001). In fact, the eastern regions of the Indonesian archipelago are known for their extreme climatic conditions (D’Arrigo and Wilson 2008). Droughts and cyclones have increased with changes in climate, as well as the frequency of ENSO events (Annamalai and Liu

2005). With regional forecasting inaccessible (Pasaribu 2007), fishermen must rely on traditional fishing calendars of seasons and moon cycles. Unfortunately, these are losing effectiveness due to rapid climatic change (Ramenzoni 2013).

In addition to environmental uncertainty, to control overfishing and damaging fishing practices, local offices are implementing zonification precautionary solutions (i.e.: Marine

Protected Areas, BAPPEDA Ende 2012 and TNC Kupang 2012 personal communications). This

128 program constrains fishing livelihood strategies further for new rules imply changes in the access and use of fishing grounds and target the development of ecotourism options.

5.5. Data collection and methods

I collected observational and ethnographic data during 22 months of fieldwork.

Interviews were carried out from May until August 2009, from November 2010 until January

2011, and from June 2011 until January 2013 in Pulau Ende and Ende city, Flores, Indonesia. In total, I conducted about 150 surveys and about 120 semi-structured interviews. I obtained primary demographic information from the local village offices and sub-district representatives, and from five partial censi in Rendo Rate Rua, Ekoreko, Ipy I, II and IV, Koponggena II and

Arubara. I also relied on two Davis Instruments, Vantage Vue 6250 weather stations to gather meteorological data.

5.5.1 Participants

40 fishermen (20 motorboats and 20 canoes) completed diaries or were surveyed daily in relation to their fishing activities. The starting point was September 22, 2011 and some observations continued until the end of June 2012. Initial and final of inclusion in the study dates varied depending on the subject. Thus, the panel is unbalanced. Originally, I chose about 40 participants from a sampled population of 130 fishermen with mean age of 41.8 years (SD=

12.9). They belonged to the villages of Rendo Rate Rua and Ekoreko in Pulau Ende. Because rate of diary and survey completion varied, the total number of participants analyzed was 25 (12 motorboats and 13 canoes). The dataset consisted of 2633 observations.

5.5.2 Measures

Fishermen reported time of departure, time of return, catch by type and amount, and the reason for not fishing. Date of departure was matched against the lunar cycle by looking at the

129

US. Naval Observatory calendar and assigned to the corresponding lunar day and lunar phase.

Lunar day is a continuous ordinal variable that re-describes the current date on a 29-30 day moon cycle. Lunar phase is a categorical ordinal variable that indicates whether a particular day belongs to one of three classes according to the position of the moon and its relation to catch

(after Pet-Soede et al. 2001). Class one begins from the 11th day on a lunar moon and lasts until the 21st. It signals the full moon and the waxing and waning gibbous. Due to the fact that fishermen rely on lamps and bait to catch fish, this is the time when catches are the lowest and approaching zero. Class two comprises the days between the 21st and the 26th and the 6th and the

11th. This is when the moon is in its first and third quarters and approaching waxing and waning gibbous. Finally, class three includes days from the 26th until the 6th. This is when the moon is waning crescent or waxing crescent. Usually, the new moon day is not considered a good day for fishing.

I also added a categorical ordinal variable called season that described if the data belong to the dry season (April 16th to November 15th), transitional season (2 November 16th to

December 15th, March 15th to April 15th), or wet season (December 16th to March 14th). I relied on meteorological variables like atmospheric pressure and precipitation and combined them with information provided by the fishermen to construct the divisions.

For catch, I relied on total individual catch, catch rate or catch per unit of effort

(individual catch by unit of time for a day of effort), probability of catch being zero, and individual catch corrected by daily mean of total catch. I also estimated a Zscore for catches and a coefficient of variation (CV) per subject. The CV is the division between the standard deviation of catches and the mean of catches by subject.

130

In terms of catch composition and morphometrics, fish could not be measured or weighted as they would be sold at sea or right after return to port. In all cases, species were recorded in the local name and then identified with Reef Fish Identification Guides during focus groups and interviews with fishermen (Allen et al. 2005, Allen 2003). See figure 3.2-3.9 for details of catch species.

5.5.3 Time treatment

Days of observations were numbered consecutively. However, temporal effect was scaled depending on moon cycle and season. When I was looking at the effect of lunar phase in catch rates, fishing effort (time fishing), catch, and probabilities of catch I treated lunar day, the day of the month according to the position of the moon, as my time variable. This meant that I had more than one observation per annual day. When exploring seasonality, I used the annual scale for the time variable. Overall, this particular time treatment allowed the exploration of the relation between DV and IV comparatively over different subjects and over different temporal effects when conducting regression analyses.

5.5.4 Missing values and zero data events

The lack of an entry was treated as indicating one of two things: adverse conditions prevented fishing or missing data. I account for this difference by separating the cases where there was an explicit explanation for not going fishing on a particular day (full moon) from the cases where there was no information at all. I relied on fishermen’s observations and my own to derive a general calendar of days that were good for fishing. Later, I used this calendar to correct observations from subjects with missing values. I only considered missing value events for the months of December and January for that is when 3 diaries were incomplete (lacked proper justification for not going fishing). It must also be noted that there is a possibility that not all

131 fishermen might have recorded the days that they went fishing and catch was zero. There is no way to correct for this event which might introduce a positive bias in the analysis. However, observational research and surveys indicate that this situation is very rare. Finally, given that an important proportion of zero data events were recorded, I conducted a two-step analysis: one that included missing values and one that did not. Observing no significant differences, I decided to carry the analyses including the missing values when possible (Bolger and Laurenceau 2013).

5.6. Data Analysis

The analyses involved different steps. First, I explored the associations between significant variables through scatterplots, ANOVAS and regressions. Then, I carried out a nominal logistic regression to test the strength of associations. The second step involved determining the level of autocorrelation of the dataset and the generation of a multilevel model.

Finally, I relied on a discriminant analysis to identify membership groups and derive ideal-type profiles. I carried out all simple and multiple regressions, discriminant and mixed methods analyses with the software JMP Pro Version 10.0.1 Release 2 from SAS. See Table 5.1 for a summary of tests and results. The choice of statistical techniques was in accordance to the hypotheses to be tested.

132

Table 5.1. Chart detailing regression analyses and results

Regression DV IV Type of boat R2 Significance Least Squares Catch rate Season Canoe 0.04 <.0001

Boat 0.02 .0003

Least Squares Catch rate Lunar phase Canoe 0.01 .0089

Boat - ns

Least Squares Catch rate Lunar day Canoe 0.04 ns

Boat 0.02 ns

Least Squares Catch rate Lunar day Engine 0.01 <.0001 Engine*lunarday Analysis of effects tests of engine*lunarday and engine significant.. Binomial Probability Lunar phase Canoe 0.02 <.0001 Logistic of catch Boat 0.01 <.0001 Least Squares Time Season Canoe 0.03 <.0001 fishing Boat - .0052 Least Time Lunar phase Canoe 0.02 <.0001 Squares fishing Boat 0.04 <.0001 Least Time Lunar day Canoe 0.04 <.0001 Squares fishing Boat 0.04 <.0001

133

Table 5.1. Chart detailing regression analyses and results (cont.)

Regression DV IV Type of boat R2 Significance

Binomial Probability Lunar phase Canoe <.0001 Logistic of fishing Intercept: 0.22, <.0001 Phase 1: - 0.013, ns Phase 2: -0.31, <.0001

Boat <.0001 Intercept: 0.27, <.0001 Phase 1: -0.16, ns Phase 2: -0.23, <.0058

Multivariate Catch rate Mean hi temp, Canoe 0.01 AIC: -592.566 OLS mean dew, mean Stepwise heat index, mean barometric pressure.

Mean humidity, Boat 0.02 AIC: -459.034 mean wind direction, mean high wind direction.

Simple Probability Individual Zscore 0.10 <.0000 regression of class of mean catch (motorboat or canoe)

Simple Individual Individual mean 0.30 <.0023 regression mean CV proximity to full moon

134

5.6.1 Simple, Binomial Logistic and Multiple regressions

I started the analysis by constructing scatterplots and boxplots to explore the associations between the probability of catch on a given day, total catch, catch rates, and time spent fishing as dependent variables and seasonal, lunar day and lunar phase as independent variables correcting by type of fishing gear. To explore associations further, I carried out a binomial logistic regression with a logit link function fitting the probability of catch on a given day by lunar phase, given that lunar phase displays a sinusoid distribution (van Oostenbrugge et al. 2001, Pet-Soede et al. 2001). I also carried out ordinary least square regressions fitting catch rates and time spend fishing by lunar phase and day, and season (OLS). To assess the importance of other IVs, I conducted a multivariate Ordinary Least Square Stepwise analysis measuring the effect of environmental variables on probability of fishing on a particular day and time spent fishing. The new predictors were mean high and low temperatures, mean humidity, mean dew, mean wind speed, mean wind run, mean high wind speed, mean wind chill, mean heat index, mean barometric pressure, mean precipitation, mean precipitation rate, mean wind amplitude, mean wind direction1 and direction 2. Because meteorological information spans from September

2011 until May 24th 2012, I only included those dates in the regression.

5.6.2 Inter Class Correlation Analysis

Statistical classic methods like those tested in the previous section (i.e: multiple regression and ANOVAS) build on three assumptions: 1) observations are independent, 2) error terms are independent, and 3) error terms have equal variances (homoscedasticity; Maas and

Snijders 2003, Ott and Lognecker 2010). Such methods are not robust to deal with repeated observations, that being auto-correlated, might have unequal variances. For this reason, and to assess if measurements in the dataset were interdependent, I carried out Inter Class Correlation

135 test. This test allows for the determination of how much variation in a longitudinal panel study can be attributed to the effect of individual differences (between subjects) or to external differences (within subjects) in the observations.

5.6.3 Multilevel models

Considering the results of the ICC, I decided to carry out a mixed effect multilevel analysis. To explain the changes in catch rates in terms of lunar phases by having or not having an engine I proposed to test the following model:

Where is the catch ratio for subject j on a particular time point within a certain lunar phase; is the common (fixed) intercept for all subjects, the common (fixed) effect of engine (dual level) for population , the (fixed) effect of lunar phase for population ,

the (fixed) interaction of group by lunar phase, subject specific intercept deviation

(random effect), subject specific slope deviation by lunar phase (random effect), and as a residual within-subject error term. JMP Pro 10 relies on the Restricted Maximum Likelihood method for fixing mixed models.

5.6.4 Discriminant Analysis

To analyze the existence of profiles among fishermen, I relied on discriminant analysis and summary statistics. Discriminant analysis is considered as an inverse prediction from a multivariate analysis of variance (ANOVA, Sall et al. 2012:493). The prediction of membership of observations to a group (X) is based on the consideration of several continuous responses

(considered the Ys). I conducted a stepwise discriminant analysis with catch rate, total time spent fishing, total catch, and proximity to full moon as variables to predict membership to the groups motorboat or canoe. Proximity to the full moon was a weighted measure that assessed the

136 distance of the observation to the full moon depending on the lunar day. For example, if the day in the lunar calendar was 14, 15, or 16 I assigned a 1. Weights varied from 0.75, 0.5, and 0.25.

Variables were kept or discarded in relation to their significance levels. I used a quadratic discriminant analysis as the method (the covariates and means of the Y variables can be different across groups).

I also relied on summary statistics and graphics to construct different profiles

(generalistic, conservative and opportunistic). My classification was complemented by matching the results of the discriminant analysis to a within fisher normalized performance index adjusted after Millar et al. 1997 and Salas and Charles 2007 and a comparison between proximity of effort to full moon days and coefficients of variation (CV) in catches by subject.

5.7. Results

5.7.1 Exploratory findings

Preliminary findings indicate the existence of different patterns explaining the variability in the associations between moon phase and total catch by gear (figure 5.1). Lunar phase 1 (10 days surrounding the full moon) showed more concentrated boxplots in both types of gear for mean total catch, with some means being significantly different (One-way analyses and Students t, table and figure 5.2). Figure 5.3 displays the different progressions of catch rates by lunar days

(from 1 to 30) depending on the gear. It suggests that motorboats have higher mean catch rates than canoes when fixed by lunar days (t (30) = 4.78, p < .0001, One-way ANOVA, table 5.3). In addition, the time spent fishing varies depending on the season and depending on the gear (figure

5.4 and table 5.4). Motorboats might be spending on average more time fishing than canoes across all seasons, especially during the transition between the dry and the wet seasons. All these effects elicit interesting connections between temporal processes, time allocation and catch.

137

Figure 5.1 Boxplots mapping total catch according to gear and the Lunar Phase22

22 Lunar phase 1 corresponds to the days including and surrounding the full moon. Phase 2 includes the 21st and 26th and the 6th and 11th. Catches are the lowest at this time and approach zero. Phase 3 comprises days from the 27th until the 5th.

138

Table 5.2 / Figures 5.2: One-way ANOVA analyses and Student’s t of mean catches

for gear by lunar phase

Canoes

Connecting Letters Report Level Mean 3 A 23.352018 2 B 12.590909 1 B 9.412946

Motorboats

Connecting Letters Report Level Mean 3 A 30.945080 2 B 23.205069 1 B 20.168224

139

140

Figure 5.3. Mean catches trajectories by lunar day depending on gear

Table 5.3. One-way ANOVA of means of catch per lunar day

Canoes Boats Day Letter Letter Letter Mean Day Letter Letter Mean

27 A 11.852970 19 A 12.550170 23 A B 8.001227 5 A B 10.637830 26 B C 7.163492 16 A B 10.476852 21 B C 6.842923 24 A B 9.844287 4 B C 6.781857 3 A B 9.484679 28 B C 6.514857 25 A B 9.290867 3 B C 6.116955 15 A B 9.270632 8 B C 5.737175 8 A B 9.192831 14 B C 5.679048 7 A B 8.527125 29 B C 5.560673 10 A B 8.345903 1 B C 5.149330 11 A B 8.307332 2 B C 5.029292 18 A B 8.302579 24 B C 4.964753 30 A B 8.094439 20 B C 4.747205 17 A B 7.999885 15 B C 4.720282 6 A B 7.776082 30 B C 4.662024 14 A B 7.393043 7 B C 4.657647 22 A B 7.292753 19 B C 4.573942 9 A B 7.065901 6 B C 4.524679 12 A B 6.855832 18 B C 4.433766 4 A B 6.636115 9 B C 4.339286 26 A B 6.542405 25 B C 4.252639 1 A B 6.540329 17 B C 3.997427 23 A B 6.449005 5 B C 3.812476 28 A B 6.146825 22 B C 3.676775 13 A B 5.991689 13 B C 2.992962 29 B 5.857269 16 C 2.897116 20 A B 5.671154 12 B C 2.420635 27 B 5.623521 10 C 1.986111 2 B 5.164273 11 C 1.805351 21 B 4.942815

Levels not connected by the same letter are significantly different in the mean of catch rate per day.

141

Figure 5.4. Boxplot of total minutes spent fishing per season according to gear

142

Table 5.4. One-way ANOVA analyses for catches per season

Canoes

Connecting Letters Report

Level Mean 1 A 24.316406 3 B 12.379487 2 C 6.699074

Levels not connected by same letter are significantly different.

Motorboats

Connecting Letters Report

Level Mean 2 A 27.708333 3 A 26.041257 1 B 18.293548

Levels not connected by same letter are significantly different.

Simple, Binomial Logistic and Multivariate regressions

A binomial logistic regression analysis explaining probability of catching fish on a certain day (0,1) given the lunar phase (1-3) and fixed by type of boat (0,1), was conducted with the following model for the linear predictor :

Where and are unknown coefficients, and the effect of the lunar phase. Results were significant for some combinations of types of boat and lunar phases, with outcomes contradicting expectations (table 5.5). According to the first hypothesis testing the lunar phase effect on probability of catch, I was expecting a higher probability of experiencing no catch or catch zero during the full moon +/- days (phase 1) in relation to other days of the month (phase 2 and 3). In

143 the case of canoes, for example, lunar phase 2 reduces the probability of a successful catch about

29 % (recall that phase 1 represents the ten days surrounding and including the full moon).

However, transitioning from phase 2 to 3, both periods during the moon cycle when catches progressively increase and decrease, the probability of a successful catch increases 17%. In the case of motorboats, the probability of a successful catch is the lowest during the phase 2.

Similar results are obtained when considering the logistic regression for probability of going fishing, and the linear regressions for catch rates in terms of seasons, lunar cycles and days

(table 5.1). It is possible to observe different patterns in fishing preferences and effort, but choices seem to partially support the hypotheses proposed. The Adjusted R2 for the models are not relevant in terms of how much variation is accounted for. This might be a consequence of the selection of independent predictors. But when I included explanatory variables (environmental effects) in a Stepwise OLS regression, I was only able to explain about 0.02 percent of the total variation. Overall, findings seem to point to the inadequacy of the statistical technique used.

They suggest that between-subject variability might have a role in explaining catch rates and time allocation.

144

Table 5.5: Binomial Logistic Fit for probability of catch by type of boat

Generalized Linear Model Fit motor=0 Distribution: Binomial Link: Logit Estimation Method: Maximum Likelihood AIC: 1826.6862 Observations (or Sum Wgts) = 1334

Model -LogLikelihood L-R DF Prob>ChiSq ChiSquare Difference 9.27 18.5488 2 <.0001* Full 910.33 Reduced 919.60 Goodness Of Fit ChiSquare DF Prob>ChiSq Statistic Pearson 1334.000 1331 0.4717 Deviance 1820.668 1331 <.0001*

Effect Tests Source DF L-R Prob>ChiSq ChiSquare lunar phase 2 18.54 <.0001*

Parameter Estimates Term Estimate Std Error L-R Prob>ChiSq ChiSquare Intercept 0.175 0.055 10.08 0.0015* lunar 0.003 0.077 0.002 0.96 phase[1] lunar -0.293 0.078 14.18 0.0002* phase[2] *: Indicates significance at : 0.05.

145

Table 5.5: Binomial Logistic Fit for probability of catch by type of boat (cont.)

Generalized Linear Model Fit motor=1 Distribution: Binomial Link: Logit Estimation Method: Maximum Likelihood AIC: 1779.0061 Observations (or Sum Wgts) = 1299

Model -LogLikelihood L-R DF Prob>ChiSq ChiSquare Difference 11.73 23.47 2 <.0001* Full 886.49 Reduced 898.23 Goodness Of Fit ChiSquare DF Prob>ChiSq Statistic Pearson 1299.000 1296 0.4713 Deviance 1772.988 1296 <.0001*

Effect Tests Source DF L-R Prob>ChiSq ChiSquare lunar phase 2 23.47 <.0001*

Parameter Estimates Term Estimate Std Error L-R Prob>ChiSq ChiSquare Intercept -0.118 0.056 4.453 0.0348* lunar -0.173 0.079 4.777 0.0288* phase[1] lunar -0.207 0.079 6.850 0.0089* phase[2]

*: Indicates significance at : 0.05.

146

5.7.2 Intra Class Correlation Analyses

In consonance with the above results, I conducted an Intra Class Correlation analysis

(ICC) between catch rate and moon phases to investigate the independence of observations. The

ICC was 0.21 with bias and non bias. Hence, about 21 % of the variation in this dataset responds to variability in observations between the subjects. This is also reflected in individual One-way

ANOVAS of catch per lunar day and time spent fishing per lunar day.

5.7.3 Multilevel models

The multilevel analysis I tested was a mixed effect model regression between catch rate and lunar phase (table 5.6). In the models, ID was specified as a random effect variable, nested in engine (0,1). Results suggest the following points: First, it is important to observe, that when analyzing the random effects component of the model, the estimated variation between fishermen is 21% percent of the total variation. Second, lunar phase is non-significant as a predictor for individual catch rates (p = 0.90). This result is different from the results listed as parameter estimates (fixed effect model) in that the F test relies on a different variance component estimator in its F-ratio denominator (JMP 2012:118, table 5.6). In addition, engine was non-significant in explaining daily catch rates (p = 0.43).

But, the interaction between lunar phase and engine was highly significant (p = 0.0162).

For canoes, parameters revealed that catch rates are the lowest during the full moon. Estimates were diverse for motorboats.

In short, findings indicate that because there is a significant proportion of the variation of the sample (about 1/5th) that is explained as between subject differences, an overall mean would not be of interest. In terms of the IVs, the significance of the interaction between engines and lunar phase suggests that catch rates vary according to changes in this relation despite between

147 subjects variation. However, the way the catch rates vary is not the same for all subjects. This confirms preliminary assessments of scatterplots and regressions.

Table 5.6. Multilevel Mixed Model

Summary of Fit Response catch rate RSquare 0.18 RSquare Adj 0.18 Root Mean Square Error 8.82 Mean of Response 6.30 Observations (or Sum Wgts) 1392

Parameter Estimates Term Estimate Std Error DFDen t Ratio Prob>|t| Intercept 6.11 0.96 22.47 6.30 <.0001* lunar phase[1] -0.143 0.34 1369 -0.41 0.68 lunar phase[2] 0.02 0.35 1365 0.07 0.94 motor[0] -0.77 0.96 22.47 -0.80 0.43 lunar phase[1]*motor[0] -0.72 0.34 1369 -2.07 0.0387* lunar phase[2]*motor[0] -0.16 0.35 1365 -0.48 0.62

REML Variance Component Estimates Random Var Ratio Var Std Error 95% Lower 95% Upper Pct of Effect Component Total ID[motor] 0.2768758 21.55 6.99 7.84 35.26 21.68 Residual 77.84 2.98 72.31 84.03 78.31 Total 99.39 7.56 86.08 116.07 100.00 0 -2 LogLikelihood = 10070.656099 Note: Total is the sum of the positive variance components. Total including negative estimates = 99.399698

Fixed Effect Tests Source Nparm DF DFDen F Ratio Prob > F lunar phase 2 2 1367 0.1045 0.90 Motor 1 1 22.47 0.6379 0.43 lunar phase*motor 2 2 1367 4.1368 0.0162*

148

5.7.4 Discriminant analysis

The stepwise discriminant analysis allowed for the selection of a model predicting group membership that included the total time spent fishing, total catch, and proximity to full moon as the continuous variables (p <.00000). The model does not include days of not fishing. As a result of the implementation, a total of 30.76 % of observations were misclassified. This result is acceptable given the large number of observations and the high level of variability in the data.

The counts predicted by the model were the following: 589 canoe observations out of 742 were predicted as belonging to the canoe group, whereas 372 motorboat observations out of 646 were predicted as belonging to the motorboat group.

If aggregating the probabilities of being in one group or the other by subject, the model predicts correctly about 20 out of 25 cases. The level of fit of the systematization can also be deduced from the canonical plot by looking at the radius of the classifying circle.

It is interesting to note that the frequency of fishing observations close to the full moon or during the full moon is higher in canoes. This can be seen from the axe “proximity” being closer to canoe observations in the canonical plot. Motorboats also fish during the full moon, but considering group membership classification, not all subjects seem to be implementing that practice (figure 5.5). This contradicts some previous findings from the multilevel model but can be a result of the panel having unbalanced observations.

149

Figure 5.5 Canonical Plot

This plot shows the distribution of measurements according to whether they belong to the member classes motorboats (red) or canoes (blue). A circle is drawn around the measurement that represents the central value for the class. The three axes stand for the variables that are used to build the classes. Distances from the central values represent probability of belonging to the classifying class.

150

To assess the ecological fit of fishing during lunar phases, I compared the average normalized catches (Zscore of catch) obtained during days of full moon to the average normalized catches obtained during normal phase days. The comparison between means (One- way ANOVA, Table 5.7) only reported significant results between the first lunar phase and the third lunar phase, and between the second lunar phase and the third lunar phase. Consequently, fishing during the first moon phase seems to be as profitable as during the second moon phase or more.

151

Table 5.7 One-way ANOVA of zscore catch by lunar phase

Oneway Anova Means for Oneway Anova Level Number Mean Std Error Lower 95% Upper 95% 1 876 0.434036 0.00816 0.41803 0.45004 2 874 0.430129 0.00817 0.41410 0.44616 3 883 0.501748 0.00813 0.48580 0.51769

Means Comparisons Comparisons for each pair using Student's t Confidence Quantile t Alpha 1.96087 0.05 Positive values show pairs of means that are significantly different.

Connecting Letters Report Level Mean 3 A 0.50174847 1 B 0.43403560 2 B 0.43012939 Levels not connected by same letter are significantly different.

152

Finally, I also matched the probability of being classified as a motorboat to the performance index (mean Zscore of CPUE) and I obtained an Adjusted R2 of 0.10 (p<.0000).

This suggests that there is a mild connection between obtaining high variable catch rates to the type of gear used and the proximity to the full moon. To explore this relation better, I compared the coefficient of variation (CV) in catches between subjects and the average proximity to the full moon in fishing effort by fisherman (Table 5.8). I discovered that CV decreases as average proximity to the full moon in effort increases (R2: 0.30, p > 0.0023). I also compared the CV to gear types with a One Way ANOVA (Table 5.9). I discovered that CV does not seem to be related to gear type (t: 0.48, p: 0.63).

153

Table 5.8 Fit of mean CV by proximity to full moon by fisherman

Linear Fit CV = 2.8379518 - 3.0585973*mean prox full moon

Summary of Fit RSquare 0.33 RSquare Adj 0.30 Root Mean Square Error 0.405 Mean of Response 1.63 Observations (or Sum Wgts) 25 Parameter Estimates Term Estimate Std Error t Ratio Prob>|t| Intercept 2.83 0.35 7.91 <.0001* mean prox full moon -3.05 0.89 -3.43 0.0023* \

154

Table 5.9 One-way ANOVA of CV by type of gear

Oneway Anova Rsquare 0.010107 Adj Rsquare -0.03293 Root Mean Square Error 0.49851 Mean of Response 1.638924 Observations (or Sum Wgts) 25 t Test 1-0

Assuming equal variances Difference 0.09671 t Ratio 0.484604 Std Err Dif 0.19956 DF 23 Upper CL Dif 0.50954 Prob > |t| 0.6325 Lower CL Dif -0.31612 Prob > t 0.3163 Confidence 0.95 Prob < t 0.6837

155

5.8. General Discussion and Conclusion

5.8.1 Identification of fishing profiles

In this article, I explored how time allocation to fishing activities changes according to the lunar phase and the seasons among 25 fishermen. The first hypothesis tested whether the full moon increased the probability of zero catches and decreased catch rates in relation to other phases. I discovered, through linear and nominal logistic regressions, that the full moon had a negative effect only if compared to the new moon phase. But, catch rates and probability of catch were the lowest during the times when the moon was in the first and third quarters; that is, during the second phase. At that time, fishermen experience the highest oscillations in catch numbers, with rapid decreases and increases in yields as dates move to and from the full moon23. Withal, results only partially confirmed the hypothesis.

When I considered a multilevel model to explore this association, I discovered that the lunar phase had an effect on catch rates if the interaction with type of gear is considered. This might demonstrate the existence of different fishing profiles depending on the vessel used. For example, the analysis of the number of days spent fishing within the lunar calendar also showed gear type differences (figure 5.6-5.7). However, conclusions are equivocal. The high spikes in catch rates of canoes during the days 14, 15 and 16 of the lunar month, the multilevel analyses, the misclassification of motorboats in the discriminant analysis, and comparisons of CV to gear type, do not support a systematization based on vessel type or equipment alone.

23 Errors in the application of the logistic function might be related to how the different moon phases are divided and the rate of change in yields during the lunar month. The rate of fluctuations in catches might be an effect of the type of fish pursued, the aggregating behavior and other elements of its ecology. Hence, it is possible that changing the days included in each phase and using a multilevel model might provide a better fit. Future studies will look into this possibility.

156

Figure 5.6: Profiles of Canoe Users

157

Figure 5.7 Profiles of Motorboat Users

158

Drawing from previous results, and the comparisons of catch rates and time (figure 5.6

5.7) and the regression between CV and average proximity to full moon in fishing effort by subject (table 5.8) I identified the following ideal-type profiles that cut across membership groups (table 5.10). Profiles could be considered strategies to deal with the variability in returns

(Cordell 1974, Salas et al. 2004).

1. Generalistic/flexible TA: fishermen that prioritize the allocation of time to fishing,

regardless of the impact of the lunar cycle. These are fishermen that will fish with the

same probability on every single day of the month. They might tolerate high variance

returns per day, but in average have lower CVs than the conservative strategy (about 10

subjects).

2. Conservative TA: fishermen that minimize the number of days they spend fishing close

to the lunar cycle or reduce the number of hours spent fishing during the full moon.

These are fishermen that will show decreasing effort as days approach the full moon and

increasing effort as dates move further from the full moon. They will strive to minimize

the coefficient of variation in returns per day, but will have in average higher CVs (about

13 subjects).

3. Opportunistic TA: fishermen that will fish with a higher intensity on the full moon days

to benefit from lower competition and higher prices. They might tolerate high variance in

daily returns per day, but they have in average the lowest CVs and highest number of

days in proximity to the full moon of fishing effort (2 subjects).

159

Table 5.10 Detail of fishing profiles per subject

Prox Mean Mean SD Full Z ID Observations catch catch CV Gear Moon Catch Profile 1 204 16.76 16.99 1.01 canoe 0.55 0.47 3 2 77 2.05 2.10 1.02 canoe 0.48 0.47 3 3 76 3.26 5.72 1.75 canoe 0.46 0.45 1 4 75 23.95 43.74 1.83 canoe 0.34 0.46 2 5 69 7.10 7.39 1.04 canoe 0.43 0.48 1 6 126 16.86 28.32 1.68 canoe 0.40 0.45 2 7 131 18.86 43.28 2.29 canoe 0.33 0.44 2 8 126 40.24 65.06 1.62 canoe 0.40 0.45 2 9 125 16.59 39.52 2.38 canoe 0.19 0.44 1 10 126 12.43 25.83 2.08 canoe 0.31 0.45 2 11 61 2.92 4.49 1.54 canoe 0.40 0.45 2 12 60 2.55 3.73 1.46 canoe 0.45 0.45 1 13 208 17.64 35.44 2.01 motor 0.44 0.45 1 14 242 45.24 81.03 1.79 motor 0.54 0.46 1 15 217 13.63 38.17 2.80 motor 0.29 0.44 2 16 184 33.38 58.32 1.75 motor 0.50 0.45 1 17 62 48.61 55.08 1.13 motor 0.51 0.47 1 18 71 43.06 61.87 1.44 motor 0.31 0.46 2 19 69 8.68 20.37 2.35 motor 0.30 0.46 2 20 56 3.54 3.94 1.12 motor 0.35 0.48 2 21 51 8.61 14.93 1.73 motor 0.30 0.45 2 22 59 6.00 10.28 1.71 motor 0.25 0.45 2 23 40 11.08 14.00 1.26 motor 0.39 0.47 2 24 40 10.15 11.94 1.18 motor 0.47 0.47 1 25 78 5.45 5.43 1.00 canoe 0.42 0.48 1

160

The classification is particularly interesting to understand the impact of environmental and socio-economic uncertainty in behavioral patterns and risk tolerance. First of all, fishermen are willing to risk lower catches while maximizing the time spent fishing. Because returns are supposed to be lower according to the narratives and previous research (Cordell 1974), fishing during the full moon increases the likelihood of coming home empty handed and requires an investment in energy and time. But, fishing during the full moon, if successful, will report greater returns as prices of fish are higher than in other phases. Therefore, even if catches are predicted to be lower, fishermen might choose to go out on those nights and might end up bringing substantial yields that are higher than normal days. This is also validated in the comparison between means of normalized catches and lunar phases discussed in the previous sections.

Second, motorboats seem to have more stable catch rates throughout the lunar month in comparison to canoes. Indeed, having a motorboat allows fishermen to access areas which are further apart and provides more security against rough weather. In line with this, while testing the second hypothesis, I discovered that seasonality affected catch rates with fishermen having the lowest catch rates in the wet season and the highest catch in the dry season. I also discovered that effort allocation varied depending on the gear, with motorboats being able to fish during transitional periods when sailing conditions are non-favorable. However, motorboat’s costs are higher in terms of the returns (catches per individual are lower than canoes because they have to be redistributed among the crew). In addition, there is no difference in the CV in relation to type of gear. Consequently, the expectation that higher catches and highest profitability will be related to motorboats is not confirmed.

161

5.8.2 Conclusion

Considered as a whole, results contest the traditional perception of fishermen’s effort being determined by the lunar cycle and constrained by the full moon effect (i.e.: Cordell 1974, van Oostenbrugge et al. 2001, Pet-Soede et al. 1999, 2001, van Densen 2001). While conversations and interviews presented the lunar calendar as the main factor to allocate effort, behaviors and results contradict oral narratives. In addition, findings suggest that differences at the personal or individual level, like skills, traits or preferences, might be a better explanatory candidate for effort allocation than gear or the lunar calendar. They might endorse a “skipper effect” kind of argument (Acheson 1981, 1989, Thordlindsson 1988, King 2011, Ruttan and

Tyedmers 2007). Yet, additional evidence on skill levels needs to be considered before that statement can be made.

Researchers and policy managers should explore the determinants of time allocation in depth rather than assume the absence of change or that behavior is too erratic to present any pattern (van Oostenbrugge et al. 2001, 2004). This is a preliminary conclusion that requires further exploration but it bespeaks of socio-ecological uncertainty as a concrete driver of behavioral change. The causes of uncertainty can be traced back to the economic instability created by recent economic crises, subsidies, mechanization, increased competition from other fishermen and, most significantly, environmental variability.

Climatic change is creating a significant impact in the timeline of recruitment for some species (see Lehodey et al. 1997, Lehodey et al. 2003). Fishermen have consistently voiced their frustration that the older ways of scheduling time in calendars are no longer effective

(Ramenzoni 2013). In calendars, the allocation of annual fishing effort reflects the migratory and reproductive behaviors of fish as they occurred in certain months of the year. For example,

162 knowing that the fishing season for small scad happens in January, fishermen would strategize and parcel their activities in the previous months to prepare for that event. But, calendars of subsistence practices do not seem to be reliable anymore and might impede rapid responses to the new conditions.

In fact, with fish stock availability changing, flexibility is a key component of being able to procure a sustainable livelihood in terms of economic returns. Although this corollary is not new, it is extremely relevant at present times. Flexibility, defined beyond better gear, needs to encompass solutions that are just not of a technical order but that expand into the enhancement of households’ capabilities and livelihood options (Satria 2004, Resosudarmo and Jotzo 2009). It should be a precondition for successful management that matches local realities and needs.

This article argues that an analysis that does not take into account between-subjects variability cannot provide good explanations of the data (Alvarez and Schmidt 2006). The aggregation of observations can make the elicitation of fishing profiles a difficult task by obscuring the detection of behavioral patterns (Salas et al. 2004). In this case, it could mean that the analyses of the factors affecting catch rates and effort allocation might be outdated. Results show that responses are already incorporating changes in the use of temporal information and environmental cycles, and that such modifications in strategies can be systematized in new classifications of effort profiles.

Unfortunately, previous taxonomies of resource users have been functional to misinterpretations by approaches that are strictly economic, resulting on further policy failure

(Coulthard et al. 2011, Bene 2003). With constraints in technical and informational resources, regional governments have relied on rapid appraisal methods that, while providing snapshots of the situation, cannot capture its complexity or its dynamic rate of change (Bene and Tewfik

163

2001). As a result, there is a discord between highly specific qualitative and quantitative studies of fisheries and their instrumentation in policy. The complex narrative quality of many ethnographic studies had encountered obstacles in its transition to applied management solutions.

A multilevel approach to variability in fishermen’s behaviors and motivations allows for the generation of ideal-types of strategies that can be used to model behavior. It constitutes a good option to systematize and address the heterogeneity that characterizes fisheries, making the adaptation of results to governance guidelines easier.

Finally, it must be emphasized that the exploration of the different kinds of variability is crucial to designing sound policies and is not just a research challenge. In Lopes and Begossi’s words: “[d]esigning management initiatives without considering ﬁshermen’s individuality and personal goals that determine their behaviour means ignoring the most complex aspect of a chain that determines the degree of exploitation of a resource.” (2011:401). As fishermen adapt to new uncertain conditions, it is time human behavioral ecology perspectives take on the task of studying individuals in their terms and not in averages.

164

CHAPTER 6

ESTIMATIONS OF CATCHABILITY AND PROBABILITY OF CATCH AMONG

FISHERMEN IN ENDE, FLORES, INDONESIA: DO NARRATIVES AND SUBJECTIVE

PERCEPTIONS OF UNCERTAINTY DETERMINE FISHING EFFORTS AND RETURNS?24

24 Ramenzoni, V. C. 2014. To be submitted to Human Ecology.

165

Abstract

Technical efficiency, catchability and variation in returns have been matters of concern among fishery scholars in the last fifty years (Arreguin-Sanchez 1996, Mahevas et al. 2011,

Thornson and Berkson 2010a, 2010b). Despite the large number of studies within social and technical perspectives, there are gaps in the understanding of how subjective projections of productive efficiency, fishing strategies and estimations of returns are connected to fishing success and effort. In this article, I explore how fishermen rely on information of fish aggregations and gear efficiency to estimate probabilities of future returns. I propose a new measure of catchability and unpack local ideas of randomness through interviews, surveys and probability judgment protocols. I discover that subjective estimations and narratives of uncertainty match the variability in the environment and might shape effort. Catchability as a perception of fish aggreagations for the past fishing events explains total catch for that day.

Findings suggest important conclusions for fishery management and harvesting regulation.

Key Words: small-scale fisheries, luck, uncertainty representation, decision-making, catchability, traditional ecological knowledge, probabilistic judgment

6.1. Introduction

Over the past few decades, fishery scientists are making a renewed effort to gain knowledge about fishermen’s behavior. Despite the large number of studies within social and technical perspectives, there are gaps in the understanding of how subjective projections of productive efficiency, fishing strategies and estimations of returns are connected to fishing success. The literature approaches such topics by exploring skills, experience, and notions of luck (Acheson 1979, 1981, Alvarez and Schmidt 2006, Thordlindsson 1988, Durrenberger and

Palsson 1986, Ruttan and Tyedmers 2007, King 2011, Vazquez-Rowe and Tyedmers 2013).

166

However, two points remain controversial.

First, there are measurement limitations for technical efficiency, productivity of gear, catchability, and uncertainty in fishery economics. Catchability is the proportion of a fish population captured through a unit of effort. It is a probability that varies considerably depending on the fishery and the gear (Arreguin-Sanchez 1996, Wilberg and Bence 2006). But, until recently, scholars have treated it as a constant. Second, despite this coefficient being linked to individual factors and skills (Mahevas et al. 2011), studies of success rarely incorporate the human dimensions of catchability. These refer to fishermen’s perceptions and evaluation of fishing conditions and how they might affect the use of gear and fishing time allocation.

In this article, I address both shortcomings by exploring how individuals rely on information of fish aggregations and gear efficiency to estimate probabilities of success. To do so, I propose a new measure of catchability and unpack local ideas of randomness and chance through interviews and surveys. The article is motivated by the hypothesis that fishermen’s estimations of future returns and catchability capture the uncertainty in the environment by determining fishing effort and, ultimately, total catch.

In this article, randomness implies a lack of patterned variability in environmental circumstances. That is, changes cannot be predicted in their occurrence or outcome by either by scientists or decision-makers. Randomness is also described by the fishermen as chance or a 50% probability in a binary choice, a 33% probability in a ternary choice and so on. That is,in this estimation the likelihood of outcomes is equivalent so cues or probability assignments are non- informative. Within statistical sciences, randomness has been approached in a slightly different way. It refers to variability that cannot be explained by numerical parameters or models (Ott and

Longnecker 2010). Taking into consideration all of these definitions, I explore in this article

167 different ways in which randomness is perceived by fishermen and in relation to statistical methods to analyze measurements.

The next two sections discuss recent approaches to catchability and estimations from fishery sciences and cognitive anthropology perspectives. In section 4, I provide an environmental and ethnographic description of the site of the study. This is followed by a characterization of the interview and observational methods used, result sections and a general discussion. In short, I explore the validity of fishermen’sexpectations by measuring the role of local estimations of catchability in determining return rates and fishing effort allocation. I also present evidence that fishermen can verbally express probabilities regarding future catches which may not contradict empirical observations from fishing effort measurements. I conclude by discussing the possibility of designing adaptive strategies in highly uncertain environments.

6.2. Catchability and productive efficiency

Technical efficiency, catchability and variation in returns have been matters of concern among fishery scholars for over the past decades (Arreguin-Sanchez 1996, Mahevas et al. 2011,

Thornson and Berkson 2010a, 2010b). In addition to these issues, a multidisciplinary debate known as the “skipper effect” has looked at the connections between individual skills, technical and environmental characteristics, and random chance in determining fishing success (Acheson

1981, Bjarnason and Thorlindsson 1993, Durrenberger and Palsson 1983, Thordlindsson 1988,

Palsson and Durrenberger 1990, Ruttan and Tyedmers 2007, King 2011, Mahevas et al. 2011,

Vazquez-Rowe and Tyedmers 2013). The main scope of this polemic is to define whether fishing returns are highly correlated to differences in personal attributes or if catches are to be explained by other factors including luck.

168

As pointed out by King (2011:388), much of this debate has been of a terminological nature, failing to procure a consensual definition of what skill means or which variables need to be considered when testing a particular model. Technical efficiency has been sometimes equated to vessel size, time spent at sea and number of boats (fishing effort), or potential catch depending on the gear used (catchability).

In fishery sciences, catchability is defined as the proportion of total biomass that can be captured through a particular gear or fishing technique (Wilberg and Bence 2006). The parameter is a concrete measure of fishing equipment’s efficiency given a fixed unit of effort

(Sparre and Venema 1996:126, Arreguin-Sanchez 1996). For example, catchability is calculated as the division between the number of fish captured and the number of fishing units of a certain gear. It can be considered as a technical estimation of success, an anticipation of future returns if all conditions affecting a fishing event are kept constant (Arreguin-Sanchez 1996).

Within real life scenarios stability is rarely the case. The potentiality to catch fish can be influenced by environmental and cyclical factors, such as the periodicity of the moon phase or tidal regimes, by characteristics of the stocks, the type of bait employed or the quality of the equipment. Even when some of these attributes can be controlled for, the parameter is better defined as a probability or expectation over different sources of uncertainty (Ellis and Wang

2007). Uncertainty arises from the interaction among ecological, technical, human, and behavioral variables affecting gear efficiency. In fact, modelizations of stock biomass and harvesting behavior are increasingly acknowledging its multifaceted and dynamic nature

(Velazques-Abunader et al. 2013, Wilberg and Bence 2006, Ellis and Wang 2007). Much work remains to be done in applying these formalizations to small-scale fisheries.

169

Where it concerns the role of human capabilities and catchability, alternative approaches have emerged over the last few years that are testing the relation between technical efficiency, skills and luck. Studies have used, among many statistical tools, stochastic frontier models and data envelopment analysis (i.e.: Vazquez-Rowe and Tyedmers 2013). Nonetheless, for the large part, they often depend on static pre-defined parameters that refer to technical characteristics alone. That is, models do not include the behavioral or subjective dimension s behind catchability. This can be a byproduct of modeling limitations in the simulation of complex scenarios across multi-species fishing environments and the restrictions of dealing with long- term data (Alvarez and Schmidt 2006).

Although catchability seems like a highly technical construct, fishers have a deep knowledge of the efficacy of the gear they employ. They formulate their own judgments and assessment of efficiency; they rely on projections of potential catch to allocate their effort in spatial and temporal patterns. In some fisheries, the proper use of equipment according to species, environments and weather conditions is considered the single most significant skill

(Acheson 1981, 1989, Paolisso 2002). Thus, exploring how fishermen incorporate subjective estimations and probability judgments of future and past catchability when deciding how much labor to allocate to fishing can be fundamental in determining exploitation regimes and in understanding how preferences are formulated in terms of which species to pursue. Most significantly, it provides an understanding of how perceptions and expectations in terms of technical efficiency and environmental factors can directly influence behavior.

170

6.3. Prior expectations of catchability, perceptions of uncertainty and probability

judgments

Recent anthropological studies have shown the significance of prior expectations in determining behavioral choices in terms of resource use (Godoy et al. 2009). Within an Optimal

Foraging Theory approach, archaeological and ethnographic research has established that variables affecting efficiency of extractive behavior like cluster sizes of harvested species, encounter rates and mobility influence the selection of hunting techniques and return rates (Bird et al. 2009). They have also underlined the importance of other environmental variables like seasonality and short to mid-term climatic conditions in subsistence activities (Tucker 2007a,

Orlove et al. 2002, 2000).

On the other hand, studies of fisherman’s estimations of success and the role of expectations have encountered multiple obstacles. For example, Astuti underlined that the prevailing uncertainty in weather conditions does not allow fishermen to estimate future catches by short term prediction of climatic events (1995). Decisions to go fishing are made at the moment of departure, and do not necessary include any long term planning. This is indeed the case in many small scale fisheries where wind patterns and clouds affect sailing conditions in the order of minutes. Hence, it is not uncommon to find that many expectations of future returns are voiced as highly uncertain or depending on luck (Ramenzoni 2013).

However, this idea of randomness, chance or luck needs to be properly qualified or unpacked to elicit realistic judgments and estimations. Even when fishermen consider luck as the principal explanation for fishing variability and success (rezeki), they still utilize a system of climatic cues to anticipate sailing conditions. This system is not infallible as cues do not have a perfect predictive value, and the predictability of these cues has changed given climatic

171 alterations over the past 30 years. But, despite this rhetoric of unpredictability, it should not be assumed that fishermen are incapable of generating accurate predictions of upcoming returns.

Sosis (2000) showed that atoll fishermen select fishing patches according to expected productivity and that prior rates of success determine that decision in most cases. It is also not uncommon to find that skilled fishermen can deploy their gear to anticipate or follow changes in marine currents and sea surface temperatures. Wiyono et al. (2006) underscored the significance of climatic variables like precipitation in guiding gear allocation among western Javanese fisheries. Overall, results state that fishermen can formulate approximations about how their equipment will perform given previous and current conditions in terms of catches, weather, and maritime events. But do their returns reflect this predictive skillset? Do their judgments reflect objective assessments of the likelihood of obtaining a good catch? Unfortunately, this remains unclear.

In the 1970s, Gladwin (1971, 1980, 1989) and Quinn (1978) explored the decision mechanisms of fishermen and fish vendors in Ghana by looking at individual estimations of monetary returns when selling fish at the market. Gladwin (1976, 1979, 1980, 1989) proposed a hierarchical theory of choice through decision trees and flow charts, which is inspired by

Tversky’s (1972) elimination by aspects framework. Quinn relied on bounded rationality theories that emphasize the importance of ecological factors shaping the decision context when making a choice (Simon 1947, 1957, Nickerson 2012, Todd and Gigerenzer 2012). Among other things, ecological approaches to rationality build on the idea that heuristics or simple decision mechanisms explain behavior (Gigerenzer et al. 1999, Gigerenzer and Selten 2001). They challenge normative approaches, like optimization models or classical logic that assume complete knowledge of all the variables affecting the decision problem. In short, they propose

172 that individuals rely on single cues of an ecological nature to decide among multiple options and that these subjective judgments can generate satisfying returns.

Through her study, Quinn tried to identify potential cues that could explain observed behavior. She disputed the idea that fishermen were able to formulate conscious projections of future conditions or conditional probability judgments that could combine multiple sources of information to guide their activities. She suggested that the complex computations behind decisions, if present, were unconscious. Quinn concluded by casting doubts on whether we could fully describe fishermen’s expectations in light of the methodological limitations in eliciting cognitive processes.

The main caveats underlined by her approach are related to the difficulty of eliciting cognitive processes with oral protocols. Not only are subjects sometimes poorly cognizant of the steps and variables involved in deciding, but also their justifications for action might reflect hindsight biases or merely lack of knowledge. That is, some of the decision models that people use might embody post-hoc rationalizations making it hard to separate between cultural narratives, explanations and justifications that emerge when conducting the research (Boster

1984:347). Decision-making is a black box where anthropologists can only know the value of the alternatives and compare them to the outcomes. Such is the base for a statistical theory of decisions, where the researcher explores how inferences are produced from the relation between inputs and outputs about how the system works (Chibnik 1980).

In 1984, Alloy and Tabachnik conducted a thorough review of estimations about causality and covariation. Underscoring the importance of situational information provided by daily experience as the base of causal attribution and environmental prediction, they distinguished between data that is provided by contexts and experience and general expectations

173 or beliefs about the world. These distinctions allowed them to establish degrees of confidence in causal judgments and in perceptions of associations between cues and events (Nisbett et al. 1983,

Kunda 1999). For example, when environments are highly uncertain it is expected that individuals would assign lower levels of trust to their predictions or will have estimations that reflect an understanding of chance (Teigen et al. 1996, Hertwig and Gigerenzer 1999, Dove

1993). Put in other words, if the domain in which the probability judgment is conducted is close to random chance, people formulate expectations that reflect this uncertainty. This does not mean that their estimations would not be adequate, but that they would reflect their inherent lack of trust instead of their actual judgment (Duncan 1972, Howell and Burnett 1978). Such a difference is critical when comprehending narratives and estimations about uncertain events and questions recent explanations of cognitive biases as adaptive (Wilke and Barrett 2009, Haselton et al. 2009, Scheibehenne et al. 2011).

In this article, I address previous limitations in unpacking local notions of luck and uncertainty. I propose to explore the connections between behavior and estimations by separating between narratives of luck and numerical judgments of future outcomes based on recent experiences. Narratives of luck are expressions or explanations that emerge in spontaneous conversations when fishermen were prompted to discuss the likelihood of catches and their certainty (i.e.: “it depends on luck”, “it may happen or it may not happen”). Numerical judgments of probability are quantified responses about the probability of catching fish in the future and they were elicited with a structured questionnaire about prior catch success and fishing returns. All answers required a percentage estimation of probability as that was the preferred format used by fishermen.

174

To test the connections among judgments of probability, narratives and observed behaviors (fishing trips), I use a statistical approach (Boster 1984, Chibnik 1980). My objective is two-fold. Initially, I want to determine whether fishermen can formulate: 1) numerical estimations of their future performances that are consistent with what is expected according to a logical understanding of probability; and 2) match ecological variables. My second objective lies in determining whether narratives of uncertainty are representing real variability in the system that cannot be attributed to any other measurable factor. If that is the case, I propose that judgments and narratives of probability will help define adaptive patterns of behavior (Dove

1993).

The first research objective implies that judgments are not contradictory from and in terms of probability axioms. Hence, as more days are included in a prediction, it is expected that the probability of an event happening will be less certain or equal to estimations that include a lesser number of days (Ott and Longnecker 2010). In addition, I test whether these expectations have any real ecological value, that is, whether they can replicate environmental conditions. I assume after Cordell (1974), that estimations based in experience and current events are adaptive if they can provide ways of securing higher catches. I hypothetize that prior rates of success will provide information that will be used to allocate effort and secure higher returns (Sosis 2002).

In order to test this hypothesis, I have to establish what success means in terms of returns.

Because biomass numbers for the species harvested are unknown, I rely in fishing logs

(longitudinal repeated measurements of 25 subjects over 9 months) to assess the potential catch of a certain day. Yields, when considered in the context of total daily catches, provide a proxy for abundance which can be used to derive an index of overall success. To do so, and while controlling for the type of gear and fish genus, I operationalize a composite measure of

175 catchability. This measure relies on estimations about median cluster size of species and represents how successful a fisherman was when catching a particular genus of fish. I test the importance of prior returns by lagging this score of catchability and determining whether it can explain total yields for a day.

In relation to the second objective, I explore the value of other factors like weather conditions in accounting for the variability in yields. Following previous research in the skipper effect (Alvarez and Schmidt 2006, Palsson and Durrenberger 1990, Acheson 1979), I consider the residual variation left unexplained by the statistical models as randomness or luck. Then, I examine some of the core narratives of luck and if they contradict models and judgments.

Results allow me to draw the following conclusions: First, I propose that the selection of the cues from previous research that were used to test whether fishers formulate projections lacks a substantial ecological and behavioral component which makes the empirical assessment of expectations difficult. For this reason, I advocate for a more ecologically sound representation of decision environments and of the cognitive processes behind behavior. I consider that catchability is a suitable ecological cue for assessing fishermen’s expectations as it condenses information from multiple sources (biology, climatology, technology) that can determine success.

Second, findings show that fishermen not only have an accurate perception of the variability affecting catches, but that they allocate their effort accordingly. Therefore, in order to derive adequate harvesting recommendations that consider coping behavioral mechanisms to deal with uncertainty managers and policy makers need to unpack local notions about luck and randomness.

176

6.4. Ethnographic and environmental description

I collected observational and ethnographic data during 22 months of fieldwork.

Interviews were carried out from May until August 2009, from November 2010 until January

2011, and from June 2011 until January 2013 in Pulau Ende and Ende city, Flores, Indonesia. In total, I conducted about 50 surveys and about 122 semi-structured interviews. I obtained primary demographic information from the local village offices and sub-district representatives, and from five partial censuses in Rendo Rate Rua, Ekoreko, Ipy I, II and IV, Koponggena II and Arubara. I also relied on two Davis Instruments, Vantage Vue 6250 weather stations to gather meteorological data (temperature, precipitation, wind intensity, wind direction, and barometric pressure).

Pulau Ende is a small island with coordinates W 121.519460 and S 8.87993. It is located in the Regency of Ende (Kabupaten Ende), central Flores, Nusa Tenggara Timur, Indonesia (see figure 3.1). The population is mostly comprised by Endenese fishermen and farmers and approximates 8000 people, of whom 1700 are full time fishermen (BPS Ende 2012). Fishing activities support a small-scale trade at the local market of Mbogawani where local species include Scombridae, Clupeidae, Lutjanidae, Serranidae, and Carangidae families. There are no industries based in Ende and no private investment to develop the fishery. Pole and line fishing fleets operate in the Sawu, Timor and Indian oceans based in Kupang (ACIAR 2001) and Benoa,

Bali (Stacey et al. 2011).

Households in fishing villages meet their needs by fishing on average 9 hours a day.

Basic primary information on surveys indicated that they fish about 19 days a month and make a mean self-reported weekly income of $ 380,000 rupiah in the dry season and $125,000 rupiah in the wet season—these are about $35 and $10 u$s dollars respectively. Sharing constitutes a

177 significant subsistence strategy with households reporting redistributing their yields with 4 other families on average.

Ende is characterized by a tropical marine ecosystem with sea surface temperatures about

28°C year round. Seasonality is driven by the action of the monsoon system of winds that also affects surface currents, circulation, salinity, and turbidity patterns. Variability induced by atmospheric components is somewhat predictable through meteorological phenomena like precipitation. For example, fishermen know that the rainy season has begun when they experience three or more consecutive days of rain (Wheeler and McBride 2005). There is also a yearly cycle of winds that can be used to forecast transitions between dry and wet seasons. This is what Dove (1985, 1993) called cyclical resonance, a heuristic that allows people to make decisions based on the repetition of patterns.

Nevertheless, important changes seem to have occurred over the last thirty years with a higher frequency of El Nino Southern Oscillation events (Annamalai and Liu 2005). During El

Nino, the surface easterly winds are interrupted or weakened, and the thermocline in the Eastern

Pacific deepens creating abnormal sea surface temperatures in the area (Dayem et al. 2007).

There is a shift in the pressure zones that further modifies precipitation regimes specially in the eastern part of Indonesia and during the dry season (Van de Kaars et al. 2009:1, Hedon

2003:1782). As a result of all of these fluctuations, the predictability of climatic events has been altered. Environmental cues have become highly indeterminate in their capacity to anticipate the onset and duration of wet seasons which has affected yields. Environmental conditions have also an important effect in the timing and duration of fishing periods and thus in returns (Widoyono et al. 2006, Pet-Soede et al. 2001). Changes brought about by climatic fluctuations have coupled with the intensification of fishing activities. The catch per unit of effort has decreased to 25% of

178 the values in earlier 1980s. There have been also modifications in the trophic levels of some species, with bigger fish like tuna decreasing in numbers.

As in most fishing villages throughout the world, fishermen talk about the catch and the weather on a daily basis. Conversations and interviews in Ende underlined the fact that many unusual events were taking place, like winds not occurring when expected or numbers of certain fish dwindling dramatically in the past decades. When asked about the reason behind these changes, fishermen usually referred to the unpredictable nature of environmental processes. They attributed modifications to both supernatural and anthropogenic causes. They considered that natural events are not completely predictable as such because they decided by God; but, simultaneously, that human agency can be responsible for explaining the reduction in catches.

When probed about the factors predicting good catches, even old fishermen would downplay the significance of experience, net size, and hours spent fishing. Unlike other fisheries

(Cordell 1974, Morrill 1967, Paolisso 2002), my interviews showed that fishermen seemed to have little apparent knowledge of the behavior, preferences and reproductive stages of fish that could provide any advantages. They followed fishing calendars with skepticism, though they recognized the significance of migratory and recruitment patterns. My enquiries about identifying expert fishermen proved fruitless. It became clear in their responses that there was no single factor that could explain high returns but good luck.

179

Yet, some people consistently reported higher catches than others. They had different expectations (harapan) depending on the seasons and recognized individuals that were better at locating fish than others. Fishermen knew about schooling behaviors, that in some seasons it was possible to catch more than one fish at each trip, and that the moon cycle influenced the total catch. That led me to assume that, even when not voiced out loud, there should be an expectation of the future catch or a consideration of how future returns might be given prior performances.

6.5. Methods

6.5.1 Decision to go fishing and perceptions of stability

Probably one of the most important decisions a fisherman has to make is whether to stay at home or go fishing. Given the high level of risk associated with potential rough sailing conditions, there are possibly a multitude of factors that are considered before making that choice. During exploratory interviews, I asked 5 fishermen to free list the reasons or cues that were most usually followed to go fishing. From that list, I selected the 4 elements most frequently mentioned (weather conditions, need food, need money, luck in the previous days) and asked 132 participants to rank them in order of importance. As the effects of tides and the moon cycle were deemed obvious and fishermen did not include them in their explanations, I did not list them in the ranking. I also asked fishermen about their expectations regarding the amount of fish they could catch in the dry season and in the wet season, and whether the dry season was more stable in terms of returns than the wet season.

6.5.2 Perceptions of future catches and fish distributions

To determine whether fishermen perceived potential differences in future returns based on their previous success, I conducted 65 surveys. I asked fishermen to estimate their expectation of having a positive yield the following day if they had been successful in the past 1 to 4 days.

180

To retrieve a quantifiable answer, I tested multiple ways of asking about probabilities such as likelihood of occurrence, frequency of observing a certain event, certainty in judgment and other verbalized descriptions (Gigerenzer et al. 1999, Gigerenzer and Selten 2001, Todd and

Gigerenzer 2012). However, they were confusing when trying to elicit a response. After multiple trials, I discovered that fishermen were used to talking about percentages to discuss daily events like the proportion of people getting married before reaching adulthood and how it has changed in relation to time. Therefore, I relied on this form as a standard response. This led me to discard

15 cases for which fishing success was evaluated in incommensurate formats, bringing the number of surveys analyzed to 50. It should be noted that given the general degree of skewness in the distribution of results, I only considered medians in the analysis of responses. This can be a byproduct of using percentages because participants might feel more inclined to provide rounded estimates (for example, 75% instead of 72%). I conducted summary data and comparison between medians and means analyses through Stata 11 and JMP Pro 11.

I also showed fishermen three type of distributions (figure 6.1): (1) an even scattering of points that can be seen as random, (2) an aggregated concentration, and (3) a clumped distribution. Fishermen were asked to name the kinds of fish usually found in each distribution.

181

Figure 6.1 Distributions of fish species:

From top to bottom: even, aggregated, clumped.

182

6.5.3 Fishing diaries

40 fishermen (20 motorboats and 20 canoes) completed diaries or were surveyed daily in relation to their fishing activities, between September 22nd 2011 and June 16th 2012. Originally,

I chose about 40 participants from a sampled population of 130 fishermen with mean age of 41.8 years (SD= 12.9). They belonged to the villages of Rendo Rate Rua and Ekoreko in Pulau Ende.

Because rate of diary and survey completion varied, the total number of participants analyzed was 25 (12 motorboats and 13 canoes). The dataset consisted of 2457 observations. Fishermen reported time of departure, time of return, catch by type and amount, and the reason for not fishing.

6.5.4 Total catches and species identification

To analyze catches, I relied on total individual catch and catch rate or catch per unit of effort (individual catch by unit of time, considered the latter, a day of effort). Given the distribution of these variables (discrete and positive, varying according to tidal cycles) I relied on the log of total catch. In terms of catch composition and morphometrics, fish could not be measured or weighted as they would be sold at sea or right after return to port. In all cases, species were recorded in the local name and then identified with Reef Fish Identification Guides during focus groups and interviews with fishermen (Allen et al. 2005, Allen et al. 2003).

6.5.5 Gear types

Fishermen in Ende rely on two types of gear: hand line and small nets. Hand line fishing is often conducted over reef barriers or coral areas with canoes (100-200kg). Small nets are deployed in relatively deep areas (about 100 to 1000 mts deep) through the use of motor boats.

183

6.5.6 Catchability-Aggregation Index

I constructed a coefficient of catchability that considered the potential of catching fish according to the gear used and the species of fish pursued. To do so, I first stratified observations by type of gear. Then, I classified observations by the genus of the species caught and calculated the median for each group of fish. I relied on medians and not means because the data were highly skewed given that there were no negative events of fishing and returns might be affected by tidal patterns. Once I had the median for each genus per gear, I divided each single observation of catch per day by its corresponding mean. This allowed me to create an index or score that assessed the distance of individual daily yields from the median. For example, in the case of mackerel, I reclassified all the observations containing catches from any of the species belonging to this genus and calculated two medians, one for canoes and one for motorboats.

Later, if a fisherman captured n mackerel on day1, the n for day1 was divided by the mean of mackerel according to the gear used.

The assumption behind this process was that fish have different schooling or aggregation behaviors depending on their species and genus (Tables 6.1 and 6.2). For example, it is common for smaller pelagic fishes like anchovies and small tuna to swim in schools. Therefore, the probability of catching more than one fish at a given time is higher than among other solitary species (yellowfin tuna or sharks). It is important to observe, though, that this process did not necessarily reduce between-subject variation. In addition to this catchability-aggregation index, I also constructed a variable that indicated whether more than one species were caught per observation. This permitted the identification of different target species (coral vs. pelagic vs. large pelagic) by gear.

184

Table 6.1 Medians for canoes for all events

Type of fish Statistic Value

Coral fish Sum 937 Median 8 Mantas Sum 2 Median 1 Others Sum 1 Median 1 Sailfish Sum 1 Median 1 Sharks Sum 1 Median 1 Skips Sum 10224 Median 15 Scad , small tuna Sum 61 Median 3 Squirrel fish Sum 180 Median 40 Tuna Sum 11 Median 5.5 Squids Sum 44 Median 10 Octopus Sum 5 Median 1

185

Table 6.2 Medians for motorboats for all events

Type of fish Statistic Value Carangids Sum 2428 Median 21 Dolphin Sum 2 Median 1 Flying fish Sum 2987 Median 110 Mantas Sum 7 Median 1 Others Sum 79 Median 2 Rays Sum 1 Median 1 Sailfish Sum 4 Median 1 Sardines Sum 20 Median 20 Sharks Sum 10 Median 1 Skipjack Sum 9043 Median 20 Small tuna Sum 2541 Median 7 Squids Sum 8405 Median 65 Tuna Sum 151 Median 2

186

6.5.7 Data analyses

First, I conducted simple linear regressions for different types of gear (motor boats and canoes) with the log of total catch for a particular day as a dependent variable and previous (up to 4 days) catchability scores as independent variables. I also carried out Stepwise regressions to predict the log of total catch for a particular day considering results from the catchability index for the past four days, number of species pursued, and weather variables for the past day (wind direction, wind intensity, temperature, precipitation and barometric pressure). The sample size for the latter analyses is reduced as these analyses required cases where there were 5 consecutive days of fishing. Following previous researchers (Alvarez and Schmidt 2006, Palsson and

Durrenberger 1990, Acheson 1979), I consider the residual variation left unexplained by the regression models as randomness or, from the fishermen's point of view, luck.

To conclude, I carried out a mixed effect model regression by gear to explore the role of individual differences in yield (between subjects variation) in relation to catchability and randomness when accounting for the log of total yields (Alvarez and Schmidt 2006). The models controlled for seasonality and fishing effort by including two dummy variables (season and month) and total time spent fishing. Other environmental variables (meteorological measurements) were not included as they did not contribute to model fit in previous regressions.

Because this is a panel dataset, a mixed effect models separates the effect of time on catches, recognizing the dependence of repeated observations. This permits a teasing apart of attributes

(skills) embodied in the subject that might affect success by randomizing the individual as a factor. As individuals and their skills are time invariant within the time frame of the data, identifying their role in total yields while controlling for the effect of other variables in the model disaggregates the impact of chance.

187

6.6. Results

6.6.1 Decision to go fishing and perceptions of environmental stability

About 35% of canoe fishermen and 30% of motorboat fishermen ranked the need for money and food as the primary factor explaining their decision to go fishing on a particular day

(Table 6.3). Weather was ranked as the main factor by 27% of the fishermen while luck was only ranked first by 8% of the sample. Overall, luck was ranked fourth in order of importance by 69% of all the respondents, indicating that it is considered as a less significant reason to decide on the activities of a particular day.

Table 6.3. Ranking of factors when deciding to go fishing

Rank Weather Food Money Luck

1 27% 30% 35% 8%

2 27% 36% 26% 11%

3 28% 26% 27% 19%

4 18% 5% 8% 69%

About 78% fishermen considered that the dry season were catches were less variable than during the wet season, and 78% had higher expectations of catch in the dry season. When asked if fishermen could expect to catch at least one fish on every single day regardless of moon phase and season, 94% fishermen responded positively. 72% mentioned that they could expect to catch at least 10 fish with certainty in the dry season. However, only 68% of participants responded affirmatively that they could expect to catch at least 5 fish if going fishing on the wet season.

6.6.2 Perceptions of future catches and fish distributions

I started off by calculating the sample medians for each probability judgment. There were

188 seven questions that asked about the probability of having a successful catch the following day if previous catch was successful/ non successful in the past 1 to 4 days. Like when calculating the catchability index, I relied on the medians and not the means because the data was highly skewed

(see table 6.4 for standard error and means). A comparison of differences between medians (table

6.4) indicates that the probability of catching fish decreases as the number of previous successful days increases. However, it must be noted that some of the comparisons reported negligible differences. This is the case for estimations of success when previous days have not reported any yields.

Table 6.4. Perceptions of future catches and fish distributions

Previous days Median Mean SD

(1) 75 62.8 32.6

(1,1) 50 55.4 32.2

(1,1,1) 50 46.4 33.5

(1,1,1,1) 25 40.2 35.2

(0) 75 69.9 27.9

(0,0) 75 68.2 26.9

(0,0,0) 75 68.7 32.4

Results hint at different expectations based on previous yields. For example, the probability of having a successful catch is almost equal whether returns were positive or negative. However, this probability is higher when the previous three days reported no yields.

189

In relation to distributions of fish species, the frequency chart can be found in table 6.5.

Overall, big fish species like marlins, rays, sharks and large tuna were classified as randomly distributed and non-aggregated. Other species like small mackerel or skipjack were equally distributed as aggregated or clumped. Finally, smaller fish like sardines and coral fishes which are known to swim in schools were mostly classified as heavily clumped in distribution 3.

It is important to notice that types of distributions imply different encounter rates or probabilities of fishing more than one individual. Hence, equipment and fishing gear are selected on the base of the targeted species.

190

Table 6.5 Fish aggregations

Species D1 D2 D3 Total

7 0 0 Shark 7 100.0% 0.0% 0.0% 28 1 0 Ray 29 96.60% 3.40% 0.00% 13 9 13 Tuna 35 37.10% 25.70% 37.10% 0 16 25 Scad 41 0.00% 39.00% 61.00% 0 10 32 Sardines 42 0.00% 23.80% 76.20% 1 11 13 Small Tuna 25 4.00% 44.00% 52.00% 10 11 14 Skipjack 35 28.60% 31.40% 40.00% 17 0 0 Marlin 17 100.00% 0.00% 0.00% 0 2 5 Small fish 7 0.00% 28.60% 71.40% 0 1 7 Corals 8 0.00% 12.50% 87.50% 2 0 0 Big fish 2 100.00% 0.00% 0.00% Flying 1 4 3 8 fish 12.50% 50.00% 37.50% 1 0 2 Dolphin 3 33.30% 0.00% 66.70% TOTAL 259

191

The raters’ agreement showed high consistency in the ranking of the species (see figure

6.2). As an example, sharks, rays and marlins were mostly listed as belonging to a type 1 distribution whereas small mackerel and small tuna were almost evenly ranked in distributions 2 and 3.

Figure 6.2: Rater’s consistency

192

6.6.3 Catchability

To explore the predictive value of catchability as operationalized above, I carried out simple regressions to explain the catch of a particular day by considering results from the catchability index from up to four days before. Table 6.6 summarizes results. As indicated by responses to probability judgments, the effect of catchability in accounting for total catches decreases with time (R2 become progressively smaller even when the p values are within significance ranges). The value of estimations also seems to be different depending on the type of gear. However, simple regressions considered each single measurement as separate without requiring five consecutive days of data. That is, I could see the association between catchability for two days ago without having to consider catchability for yesterday.

Table 6.6 Simple regressions of total catch by catchability by gear

Gear DV IV R2 Significance Canoe Total catch Catchability-1 0.30 0.000

Total catch Catchability-2 0.15 0.000

Total catch Catchability-3 0.17 0.000

Total catch Catchability-4 0.07 0.000

Motorboat Total catch Catchability-1 0.09 0.000

Catchability-2 0.05 0.000

Catchability-3 0.01 0.023

Catchability-4 0.01 0.005

193

Multiple regressions allowed me to analyze only those cases for which there was at least five successive days of data. The final equation for the stepwise regression for canoes was the following:

With total minutes spent fishing, , , indicating different months of the year, catchability for the previous day, and catchability for three days before going fishing. The model fits about 35% of the variability in observations, with the change of one unit in the catchability of the previous day accounting for 40% increase of the catch in the following day

(Table 6.7). Catchability for the day before yesterday and four days before going fishing did not report any significant results. However, the catchability for three days before going fishing explained 20% of total catch for the present day.

194

Table 6.7. Multivariate regression of total catch by catchability for canoes

Summary of Fit RSquare 0.35 RSquare Adj 0.34 Root Mean Square Error 0.85 Mean of Response 2.60 Observations (or Sum Wgts) 297

Analysis of Variance Source DF Sum of Mean Square F Ratio Squares Model 6 116.09 19.34 26.7356 Error 290 209.87 0.72 Prob > F C. Total 296 325.96 <.0001*

Parameter Estimates Term Estimate Std Error t Ratio Prob>|t| Intercept 2.280 0.148 15.35 <.0001* Totalminutes 0.001 0.000 2.91 0.0039* month{1&2-3&4&5&10&11&12} -0.001 0.063 -0.02 0.98 month{3&4&5&10-11&12} -0.160 0.081 -1.97 0.0493* month{3-4&5&10} -0.120 0.068 -1.75 0.08 catchability -1 0.402 0.056 7.13 <.0001* catchability -3 0.197 0.057 3.43 0.0007*

195

For motorboats (Table 6.8), the regression equation was the following:

With different types of fish caught, total minutes spent fishing, , , , indicating different months of the year, barometric pressure for the last 24 hours, financial return for the day before, catchability for the day before going fishing, and the catchability for two days before going fishing. The model fits about 24% of the variability in observations, with the change of one unit in the catchability of the previous day and two days before accounting for 28% and 16% increases of the catch in the following day. Catchability for three days and four days before going fishing did not report any significant results.

196

Table 6.8. Multivariate regression of total catch by catchability for motorboats

Summary of Fit RSquare 0.26 RSquare Adj 0.24 Root Mean Square Error 1.20 Mean of Response 3.39 Observations (or Sum Wgts) 380

Analysis of Variance Source DF Sum of Mean Square F Ratio Squares Model 11 191.27 17.38 11.96 Error 368 535.01 1.45 Prob > F C. Total 379 726.28 <.0001*

Parameter Estimates Term Estimate Std Error t Ratio Prob>|t| Intercept -42.65745 30.48845 -1.40 0.1626 TYPESFI 0.6985191 0.117819 5.93 <.0001* Totalminutes 0.0010542 0.000429 2.46 0.0144* month{1&2-3&4&9&10&11&12} 0.2216932 0.091882 2.41 0.0163* month{3&4&9&10&11-12} -0.240106 0.123775 -1.94 0.0532 month{3&4-9&10&11} 0.3897059 0.108926 3.58 0.0004* month{9&10-11} -0.26791 0.195749 -1.37 0.1719 month{9-10} 0.0213777 0.27576 0.08 0.9382 BARPRESSREG 0.0443457 0.030327 1.46 0.1445 lag1total$ -3.05e-7 2.763e-7 -1.10 0.2704 catchability -1 0.2876784 0.07845 3.67 0.0003* catchability -2 0.166826 0.073014 2.28 0.0229*

197

It should be noted that in the case of canoes, catchability was the variable whose change reported the higher impact on total catch. For motorboats, catchability played a lesser role, with an increase of one in the number of species harvested signifying a positive change of 0.60 in the total catch for the following day. Environmental variables like temperature, pressure or wind direction did not have a significant explanatory role. Only the month of the year showed a significant effect indicating a trend that is associated with seasonality.

Finally, mixed effect models explored the value of individual differences in catches when quantifying the role of catchability and randomness in explaining yields. In the case of canoes, the variation between subjects accounts for 47% of the variability in observations. As a consequence, there is a strong individual effect that might suggest important differences in skills or personal attributes. Despite this individual effect, seasonality, fishing effort, month and yesterday’s catchability reported significant coefficients. Yields are about 43% higher in the dry season, and 85% higher in November in relation to October. They decrease about 50% as the month changes from February to March. In addition, a change of one unit in yesterday’s catchability creates an increase of 23% in the total yield of today. Catchability for three days before going fishing also has a positive effect on yields, but it is non-statistically significant (p =

0.0581). Regarding the role of randomness and luck, the model predicts about 54% of the variability (see table 6.9), leaving 46% to chance. The effect of random variation is high implying that a large part of fluctuations in catch are not due to seasonal, time dependent, individual or technical characteristics.

198

Table 6.9. Mixed effect Canoes

Summary of Fit RSquare 0.571302 RSquare Adj 0.542585 Root Mean Square Error 0.700161 Mean of Response 2.674098 Observations (or Sum Wgts) 224

Parameter Estimates Term Estimate Std Error DFDen t Ratio Prob>|t| Intercept 2.4233985 0.395982 57.28 6.12 <.0001* catchability -1 0.2339983 0.070504 205.1 3.32 0.0011* catchability -2 -0.048519 0.067044 204 -0.72 0.4701 catchability -3 0.1363532 0.071542 205 1.91 0.0581 catchability -4 -0.017465 0.064033 202.3 -0.27 0.7853 season[DRY] 0.4303313 0.15509 206.5 2.77 0.0060* season[TRANSITION] -0.149114 0.099092 204.9 -1.50 0.1339 Totalminutes 0.001267 0.000549 206.5 2.31 0.0220* month[2-1] 0.2421594 0.214285 203 1.13 0.2598 month[3-2] -0.482155 0.232665 209 -2.07 0.0395* month[4-3] -0.455311 0.262907 208.8 -1.73 0.0848 month[5-4] 0.0369344 0.308824 208.1 0.12 0.9049 month[10-5] -0.647846 0.367428 206.9 -1.76 0.0793 month[11-10] 0.8595081 0.298249 204.3 2.88 0.0044* month[12-11] 0.4960495 0.299589 201.3 1.66 0.0993

REML Variance Component Estimates Random Var Ratio Var Std Error 95% Lower 95% Upper Pct of Total Effect Component ID 0.913868 0.4480007 0.2624089 -0.066311 0.9623126 47.750 Residual 0.4902248 0.0492198 0.4064368 0.6030349 52.250 Total 0.9382255 0.2640994 0.578244 1.7814153 100.000

-2 LogLikelihood = 531.23205909. Note: Total is the sum of the positive variance components. Total including negative estimates = 0.938

199

For motorboats, there is also an effect by month, indicating decreases when switching between dry and wet season (Table 6.10). Catchability for yesterday and two days ago report significant coefficients predicting increases of 20% and 17% respectively. The individual component only accounts for 25% of total variability in the model. Hence, other factors might be affecting catches. However, the R2 is 31% which suggests a high level of randomness given the large proportion of data that is not represented by the model.

200

Table 6.10. Mixed effect Motorboats

Summary of Fit RSquare 0.352997 RSquare Adj 0.318159 Root Mean Square Error 1.148094 Mean of Response 3.555795 Observations (or Sum Wgts) 275

Parameter Estimates Term Estimate Std Error DFDen t Ratio Prob>|t| Intercept 3.1891088 0.510096 145.4 6.25 <.0001* catchability -1 0.2021013 0.079757 257 2.53 0.0119* catchability -2 0.1774209 0.083823 253.5 2.12 0.0353* catchability -3 -0.172448 0.088244 252.9 -1.95 0.0518 catchability -4 -0.060584 0.091573 254 -0.66 0.5088 season[DRY] 0.1792858 0.225355 259.6 0.80 0.4270 season[TRANSITION] 0.2184106 0.129044 254 1.69 0.0918 Totalminutes 0.0011087 0.000559 259.9 1.98 0.0486* month[2-1] 0.18447 0.294275 255.2 0.63 0.5313 month[3-2] -0.390488 0.303455 254.5 -1.29 0.1993 month[4-3] -0.381444 0.282475 258.9 -1.35 0.1781 month[9-4] -2.852119 1.221853 254.4 -2.33 0.0204* month[10-9] 1.829869 1.193539 250.1 1.53 0.1265 month[11-10] 0.4708754 0.433436 254.7 1.09 0.2783 month[12-11] 1.4548365 0.431769 256.6 3.37 0.0009*

REML Variance Component Estimates Random Var Ratio Var Std Error 95% Lower 95% Upper Pct of Effect Component Total ID 0.3354305 0.4421375 0.2529617 -0.053658 0.9379333 25.118 Residual 1.3181194 0.1180683 1.1141874 1.5839775 74.882 Total 1.7602569 0.2736883 1.3265406 2.4490156 100.000

-2 LogLikelihood = 896.49131877 Note: Total is the sum of the positive variance components. Total including negative estimates = 1.7602569

Fixed Effect Tests Source Nparm DF DFDen F Ratio Prob > F catchability -1 1 1 257 6.4209 0.0119* catchability -2 1 1 253.5 4.4800 0.0353* catchability -3 1 1 252.9 3.8190 0.0518 catchability -4 1 1 254 0.4377 0.5088

201

Source Nparm DF DFDen F Ratio Prob > F Season 2 2 257.5 2.7817 0.0638 totalminutes 1 1 259.9 3.9268 0.0486* Month 7 7 256.3 2.8790 0.0065*

202

Overall, for both canoe and motorboat fishers, the same conclusion can be reached: there is an important part of the daily catch that cannot be accounted by any of the variables considered or controlled for in the different models. There is a high level of unpredictability in returns. It might be the case that “luck is more important than skill” (Alvarez and Schmidt

2006:22). Of the variables that showed significant estimates, seasonality and catchability for the first two days were the only ones that provided good meaningful predictors of returns. As it will be discussed next, these three variables condense ecological information at different scales

(short-term, medium term) that might help provide adaptive strategies to the unpredictability of the environment.

6.7. General Discussion and Conclusions

In this article I explored how narratives and estimations about luck and catchability affect future outcomes in fishermen’s behavior. Are their narratives and estimations misrepresenting reality? There are good reasons to think they are not. First, I discovered that despite high uncertainty in returns, fishermen still relied on expectations about their future catch that depended on the seasons and types of species pursued; and that such expectations did not contradict the idea of luck. For example, they recognized that every single fisherman if he tries hard enough can at least get one fish in every trip. However, the possibility of getting more than a certain number varies depending on the season.

Secondly, unpacking the uncertainty on their estimations through questions about probability of catch, I was able to establish that there is logical consistency in their expectations in relation to empirical observations. For example, the role of prior catchability in accounting for future returns diminishes as previous days are taken into consideration. Nonetheless, such estimations were not far from chance (50%) as they oscillated between 75 to 50%. This is also

203 reflected in the different regression models tested as there is an important part of the daily catch that cannot be accounted by any of the variables proposed. Consequently, there is a high level of unpredictability in returns that is constantly reflected in narratives about chance and luck.

Findings from this article do not match anthropological research on the role of magic and prescriptions in dealing with anxiety and uncertainty. In 1918, Malinowski proposed a theory to explain differences in religious behavior and fishing taboos in relation to type of gear and fishing location. He discovered that in areas where returns were more uncertain like the open seas, people would rely on magic and prescriptions. Whereas, in other spots where catches were more secured, he virtually discovered no religious behavior associated to fishing. This was the base for his anxiety-reduction explanation of ritual taboos and magic beliefs. Similar findings were mentioned by Cordell (1974), with other examples in the literature (Firth 1965, Sother 1965,

Felson and Gmelch 1979). However, the explanation is challenged by Palmer (1989) and King

(2011) who underscored the role of magic as a mechanism to generate cooperation and the socialization of critical fishing information.

Results from this research do not seem to indicate that magic and rhetorics of luck are more frequent in areas of high unpredictability of returns or where cooperation is critical. The emphasis on the inadequacy of magic and ways of manipulating luck, uncertainty of predictions and the lack of importance of personal skills, as discussed next, seem to suggest that randomness is a central component of the cycle. Indeed, my interviews showed that randomness is embodied in the perception of events, from fishing returns to daily life like sickness and happiness. Luck is the driving principle in one’s existence as it is through the blessings and trials imposed by God that a path is set in life.

204

Within this explanatory frameset, luck as an idiom of causation sets limitations in how important situational and experiential knowledge are in obtaining a good outcome (Tucker et al.

2011). Because the mere essence of luck is unpredictability, indeterminacy and non-patterned variability, this constrains the options that someone has to manipulate or anticipate their destiny.

Luck cannot be controlled by any possible human means. Ritual practices associated with spells or magic are considered sin (dirty luck, Ramenzoni 2013) and are only employed at a significant spiritual and economic cost.

Furthermore, interviews demonstrated that ideas associated with higher success in other fisheries such as being an expert fisherman or having better equipment (i.e.: Poggie 1979) are not necessarily among the factors that Endenese fishermen rely on to explain high yields. There were no rankings or distinctions among the fishermen that would allow the identification of a single quality that would make one successful. Only hard work and persistence make the difference in obtaining a higher yield.

The emphasis on randomness and unpredictability does not imply that environmental or behavioral cues are not relevant at all to the task of being a fisherman. During interviews it became known that fishermen recognize a core set of knowledge that every single person regularly venturing at sea must possess. Such lore refers to indicators about the behavior of the currents and winds like lightning and water color. But, this body of information is implicit and taken for granted. It only emerged as a topic of conversation when discussing the unpredictability associated with current sailing conditions and how indicators were not reliable anymore. The problem lies in the new uncertainty introduced by climate change and how it determines which cues are important to know given the situation.

205

In this respect, fishermen consistently recognized that different kinds of fish have different clustering behaviors. Thus, aggregation patterns may be among the few ecological indicators that can have a value in predicting returns. This conclusion is reflected in the high level of inter-subject agreement manifested by fishermen when free-listing species according to their schooling behavior.

There is probably one other cue that is somewhat valuable in anticipating returns: seasonality. Many studies in the history of Southeast Asia have indicated its importance in shaping sailing and trade patterns (Ammarell 2002, Acciaioli 2004, Reid 1984, Butcher 2004).

Fishermen know that during the wet months of the monsoon season, seas can be rough and fishing can be dangerous. They also know that high pressure winds reinforcing the easterlies create upwelling conditions in some regions, increasing the catch. Nevertheless, the transitional periods between seasons and the actual date of onset of a season remain highly unpredictable given alterations in climatic patterns. In short, fishermen can only rely on general assertions about which season is supposed to come next or how the current season is like in comparison to prior seasons to respond to uncertainty. The value of seasonality as a cue emerges through the aggregation of daily conditions over a span of time. Its predictability on a situational basis is, even though acknowledged, restricted.

Finally, in relation to weather prediction, responses to which factors can better predict decisions to go fishing showed changes in the value of environmental information. Weather events only mattered to the degree that they determined immediate conditions that might affect currents and safety while at sea. As I discovered when running stepwise regressions on total catch, meteorological variables could hardly explain returns or the time spent fishing. Once again, this does not imply that fishermen do not value weather cues at all. They pay attention to

206 the formation of clouds, the duration of storms, they can tell differences in water temperature and what that means in terms of which species can be caught. But given the highly variable value of these cues it is somewhat adaptive that they would only be considered on a situational basis.

In all, it is crucial to emphasize that luck implies a certain attitude toward nature that shapes the perception of ecological patterns. But luck also defines suitable rules on how to interact with an environment and which expectations are valid. This, in turn, constrains decision making processes and resource use practices. Similarly to what Dove discovered among the

Kantu in West Kalimantan (1993), the notion of luck as a cultural construct might enhance the indeterminacy of the environment in local perceptions allowing for the identification of adaptive strategies and the value of particular ecological cues. By reinforcing ideas of randomness and chance, ideas of luck prevent biased causal attributions or overconfident judgments. Because luck is based on arbitrariness, fishermen know that their estimations have only a limited predictive value that is tied to a concrete configuration of ecological events.

Exploring the relationship between catchability and fishing strategies, the importance of prior estimations of catchability needs to be contextualized considering the needs that inspire a fisherman to go out to sea, the kinds of fish and the type of gear preferred. Catchability as a cue permits the integration of prior returns with technical efficiency, environmental conditions, and prey behavior. This can delimit particular fishing styles and strategies.

In the case of canoes, fishermen usually exploit reef areas where there is a high diversity of species, whereas motorboats frequently rely on kinds of gear (nets) that require them to fish in open waters. Their catches are mostly comprised of pelagic species like tuna or skipjack alone.

Pelagic species display highly migratory patterns. Thus, when a school is caught it is most likely that there is a high probability of having subsequent days of high returns. With the exception of a

207 few families, coral fishes are localized species, displaying highly stationary behaviors. This would indicate that the probability of catching more than one fish should be high disregarding the day. However, in this particular case, the type of gear employed restricts the total amount of fish that can be caught at a given time.

In other words, the decision to fish in particular locations not only affects the diversity of catch but also restricts the kinds of equipment that can be used, and vice-versa. For example, motorboats rely on nets that can capture more than one individual, but they cannot be deployed in reef areas without incurring substantial damages to the mesh. On the other hand, canoes are more flexible in pursuing a wider spectrum of species for they rely on hand line but can only get a handful of fish at a time depending on the number of hooks in the water.

Going back to the theoretical questions that inspired this research it can be said that abstract parameters that consider the technical aspects of catchability alone cannot provide good descriptions of behavior or the potentiality of resources in a system. Because decisions to allocate the gear include a myriad of other components that oscillate among a perception of environmental conditions, cultural preferences, and ecological knowledge it is necessary to provide a human pragmatic dimension in modeling catchability. I strove to do so by selecting an index based on individual performances and fish aggregations. There are probably other ways to include a socio-ecological component, but this is just one attempt in that direction.

In terms of estimations, my findings show that judgments of probability can grasp the uncertainty that prevails in the environment. Eliciting such judgments was an arduous process that involved the discussion of alternative ways of perceiving uncertainty, luck and the value of ecological cues. Unlike prior research, I think that this process was aided by at least two factors: first, the significance of luck in religion and daily life in the community; and second, by the

208 numerous environmental fluctuations that are being experienced recently. I believe that these two elements made the perception of randomness more prevalent and will be tested in future articles.

In brief, my results indicate that perceptions of uncertainty cannot be discounted as ecologically unsound or non-consistent with reality without prior evaluation; and that the selection of ecological cues is critical when assessing decision making mechanisms about natural resources.

Much has been written on the value of experience, systems of knowledge and adaptive strategies in the context of climate change (Johannes 1993, Berkes et al. 2000, 2007, Folke 2004,

Berkes 2009). This article is grounded in the conviction that there is value in identifying how communities are re-shaping their behavioral and cognitive patterns in dealing with uncertainty.

However, as communities are heterogeneous and are facing dissimilar situations it is difficult to derive comprehensive recommendations. Moreover, with changes in configurations of uncertainty it is expected that attitudes towards resource management will also change. New problematics and adaptation responses will arise sometimes at a faster pace than they can be identified by scholars. As Dove suggested (1993), in these highly dynamic ecosystems strategically managed solutions are ill advised.

Therefore, policy makers and fishery scientists will have to face the challenge of providing alternatives that are highly flexible but also tuned to the local contexts. With the goal of facilitating these efforts, findings presented here suggest the importance of carefully determining the drivers of uncertainty in the environment and how local systems are failing or succeeding at predicting them. It is only by identifying similarities in the informational trade-offs that communities are experiencing and enabling structural responses beyond technical solutions that flexibility might be attained (Satria 2004).

209

CHAPTER 7

CONCLUSION

In this concluding chapter I discuss the principal findings of the dissertation. I address the significance of results in anthropology, followed by some direct implications for management and policy design in fisheries. I end the chapter mentioning the limitations, challenges and opportunities to the behavioral study of decision in uncertainty.

7.1 Summary of Main Results

I relied on a combination of ethnographic, ecological and experimental tools to explore how environmental uncertainty and variability is affecting extractive behavior in small-scale fisheries in Ende, Eastern Indonesia. In chapter three I generated a parametric model of the fishery which I corrected with observational data from household and market surveys. Results revealed substantial exploitation of stocks and “fishing down the food web” effects. By 2011, catches per unit of effort were three times lower in comparison to the values estimated in 1980s.

The causes of decline combine the interaction between intensification practices and environmental variation.

Environmental variability has a central role in explaining the current state of the fishery.

Temporal covariance between monthly yields and rainfall is significant. In the past thirty years, the Savu region has seen modifications in precipitation regimes (see Annex 2). Changes may be indicative of fluctuations in other climatological and atmospheric components that have probably

210 affected fish stocks. Unfortunately, data sets are incomplete so it is difficult to assess the magnitude of such changes with precision.

Intensification has responded to state-level modernization policies applied to fishery and agriculture sectors in the late 1980s. Devastating effects of an El Niño in 1997 and further decentralization policies in 1999 probably contributed to exacerbate the demand on stocks. As more people entered the fishery, bigger species were overharvested and reduced drastically

(Batoidae family, Thunnus albacares and sharks). As predatory relations are altered, medium to small size pelagic species proliferated. Nowadays, the fishery is mostly concentrated in exploiting mackerel and small to mid size pelagic fishes (Rastrelliger genus, Katsuwonus pelamis, Clupeidae family).

Fishermen have responded to catch reductions and environmental uncertainty by modifying exploitation behaviors and consumption of marine products. Secondary data, ethnographic testimonies and household-level observations showed that new fishing profiles guide activities. For example, in terms of daily variability, fishermen seem to consider fishing returns and catchability from previous days to assign their time to productive activities.

In addition to behavioral changes, non-patterned environmental variability has led fishermen to question old ways of predicting phenomena and to distrust fishing calendars and weather cues. Fishermen still rely on some ecological indicators like seasons and fish aggregations, and manifest assessments of likelihood of future catches that match the probabilistic structure of the environment. However, they mention that effort allocation is synchronized with the lunar cycle which contradicts observations. This dissonance between behaviors and rationales merits further research. It can be interpreted as an acknowledgement that fishing efficiency is lower while fish prices are higher during days of full moon.

211

Taken as a whole, findings indicate that fishermen can perceive environmental variability in ways that match the structure of the environment and have adapted their behavior to these new challenges. The emphasis on the inadequacy of magic and unreliability of predictions, skills and experience to explain returns indicates that randomness is a central component of the cycle.

Despite local awareness of changes in environmental conditions, there is no system to conceptualize modifications or how to respond to new challenges. Institutional policy continues to operate under ideas of intensification and is selectively subsidizing the expansion of the fishing fleet to bigger purse seiners In the short to long-term future, the fishery may continue to reduce species richness as temperature increases and modifications to tidal dynamics will further constrain stocks.Policy makers and governance mechanisms should consider the introduction of sustainable livelihood alternatives that can enhance local capacities.

Table 7.1 Operationalization of findings

Decision  Go fishing? Rest? Work in the garden? Work on equipment?

Decision environment: set of informational structures that represents the behavior and the relations between the component units of the decision setting.

Uncertainty Findings Conclusion

Subjective  Estimations of future returns seem to Judgments of probability Uncertainty match empirical observations on catches match the structure of Estimations of (decreasing probability assigned to the environment. uncertainty successful catches as number of previous Judgments of days with positive returns increase). probability  Prior catchability seems to determine Narratives of chance fishing effort and total catches.

212

Table 7.1 Operationalization of findings (cont.)

Uncertainty Findings Conclusion

Objective Uncertainty  Parametric models indicate High levels of Quantifiable overexploitation of the fishery in uncertainty on a day to estimations of 1990s. CPUE has decreased 4 times day basis, patterned variability through from early 1980s. seasonal variability. However, there are parametric, non-  Colwell Index shows high levels of changes in patterned parametric and complex predictability for precipitation events, methods variability. Future SPI shows change in drought analyses will look at conditions starting 1990 (Annex 2). changes in starting date  Differing levels of variation (CV) in of the season and yields determined by personal significance of attributes (relevance of mixed effect barometric pressure. models) and not by gear type or time spent fishing.  Regression, transfer functions and ARIMA models indicate significant covariation between yields and precipitation rates at a monthly and interannual scale (70%). Also significant effect of seasonality in yields.  Lunar cycle (moon phases), do not seem to explain time allocation or catches.  Daily variability in catches is explained by fish aggregation (catchability for the previous day). Weather variables like temperature and precipitation for the last 24 hours are not significant. Only wind direction and intensity reported significant results.  However, the best fit for most of regression models leaves about 50% of variability unexplained even when including between subject differences. Thus, chance and unpredictable variation is high in daily yields.

213

Table 7.1 Operationalization of findings (cont.)

Heuristics Fishing rules

 Multiple cues for  Fishing profiles: allocation of time to fishing might be explained decisions: by tolerance of risk in catches (individual CV in yields) and Generalistic TA, preferences in terms of effect of lunar cycle and fish prices. Conservative TA,  Probability of fishing today might be explained by catchability Opportunistic TA. for the previous days.  One cue decision: look at prior catchability

214

7.2 Theoretical and methodological implications for behavioral ecology, cognitive

sciences and anthropology

1) Behavioral studies

Within Human Behavioral Ecology there has been a long-standing critique about the incapacity of the theory to account for mechanisms of decision processes. This has been known as the phenotypic gambit (Laland and Brown 2002). By exploring informational structures of the decision settings and understanding how uncertainty can combine a subjective and objective component, this dissertation exemplifies how the informational mechanisms behind decisions can be identified and proposed. For example, in Chapter 5 the careful study of individual behavior through statistical and ethnographic methods and its matching to environmental structures (lunar cycle) permits to elicit potential behavioral rules at the level of information inferences or heuristics. Chapter 6 shows how behaviors might reflect a consideration of probabilistic scenarios in terms of catch and suggest a one-cue heuristic rule for decisions (see

Table 7.1). Future modeling of strategies of decision making through simulations can help determine whether such mechanisms are both psychological and ecological plausible.

Another limitation that has been attributed to behavioral studies is its lack of treatment of socio-cultural and psychological variables in decision processes. This has been a consequence of models equating subjective preferences in choice to an energetic or utility based rationale. In addition, it is a by product of difficulties in methodologies and statistical techniques used when analyzing data sets. For example, models rely on statistical analyses that are unable to tease apart the effect of individual traits in accounting for behavior. By combining a multilevel kind of analyses with ethnographic methods, this dissertation shows that decisions can be to a substantial part explained by individual characteristics and random environmental uncertainty. Therefore, a more careful consideration of social and cultural variables needs to be incorporated.

215

Finally, HBE models have not expanded to include the different kinds of variability or uncertainty that can be found in an ecosystem. If not assuming perfect knowledge, models have relied on operationalizations of uncertainty as risk. However, ecological scenarios encompass a myriad of processes and, in many of these cases; their complexity impedes their conceptualization in terms of known probability distributions. Modelers can produce ecological realistic simulations if they focus on the informational structure of the environment and derive behavioral patterns that are tuned to those settings. Moreover, a distinction between epistemic

(subjective) and ontological (objective) uncertainty can help operationalize costs and constraints in individual choices. The perfect information requirement can be relaxed and rules of behavior can be derived that represent adaptive processes. Chapter 5 and 6 offer examples of how environmental structures shape individual inferences and mold behavioral preferences.

2) Cognitive sciences and anthropology

Findings from this dissertation add to previous debates in cognitive anthropology and psychology regarding: 1) the value of subjective assessments of uncertainty and probability, 2) the formats in which assessments are made, and 3) methodological constraints in exploring psychological processes outside controlled settings (Quinn 1978, Chibnik 1980, Dove 1993).

In terms of the value of subjective estimations, I discovered that fishermen can generate adequate representations mimicking probability distributions under conditions of high uncertainty. While previous researchers have been very skeptical on whether subjects can consciously flesh out probability estimations that are correct (Kahneman et al. 1982, Gigerenzer

2008, Quinn 1978), I found out that such assessments are not unconscious, that they reflect ecological distributions of stocks, and that they seem to be logically consistent. Moreover, fishermen were able to produce judgments in percentages (normalized formats of probability); casting doubts on whether percentages should not be used in evaluating probabilistic reasoning

216 as suggested by other researchers (Gigerenzer et al. 1999, Heinrich and Gigerenzer 1999,

Cosmides and Tooby 1996).

I also discovered that estimations can be ascertained better if judgments are elicited in ecologically relevant domains. That is, inferences seem to be in consonance with the environmental structure of the environment (Wilke 2006, Wilke and Barrett 2009). This calls for a study of decision mechanisms that pays equal attention to the characterization of the setting as well as the cognitive components that take part in probability judgments. To this matter, researchers should recognize that ecological factors are not just restricted to biological or abiotic processes but that they include socio-cultural considerations. Culture and narratives about the predictability of events, what Tucker (2011) calls the idiom of causation, reinforce an accurate perception of the structure of the environment (Dove 1993). Cultural and ecological domains are highly relevant to psychological processes. Their influence is not limited to providing context but also to the shaping of performances.

One way to include ecological and socio-cultural domains of information lies in better defining the environment of the decision maker. Unfortunately, bounded rationality approaches have only glided at the surface when it comes to providing a consistent methodology to operationalize heuristics (i.e.: Todd and Gigerenzer 2012, Gigerenzer et al. 1999). They have fallen prey to the same criticism of generalism they have posed to Tversky’s and Kahneman’s

Heuristic and Biases program (1974). By being ambiguous and inspecific on the attributes of environments, they fail to define concrete steps to test how well heuristics tune to ecological settings. Therefore, they restrict the opportunities a researcher has to replicate their empirical findings.

217

In this dissertation, I show that by including environmental and ecological variables decision settings can be better defined. This speaks both to psychologists and anthropologists.

First of all, a model of decision environments must be realistic and thus it must include socio- ecological dimensions, not just information (see i.e.: Todd and Gigerenzer 2012). This does not necessarily imply that a researcher has to cover all kinds of attributes and phenomena in the world of the decision-maker to capture the essence behind a decision problem25. It is true that some social events defy apprenhension and even modelization, but subsistence decisions lend themselves to analysis by being both repeated over time, and tied to observable behaviors and materials. As a consequence, a pattern of behavioral preferences can be derived. In this case, the focus on certain kinds of structures in terms of how their behavior (recurrence, repetition, constancy, predictability, etc.) changes or perdures provides constraints that solve the computational problem of capturing complexity. Also a perspective that looks into adaptability and constraints in terms of cognitive limitations can guide a modeler in selecting which structures need to be considered (Sperber and Wilson 1984, Cummins 2002).

25 Modelers, when faced with real life situations, tend to think that the complexity and contingent nature of behavior cannot be represented formally (Elio 2002). Likewise, anthropologists tend to assume that because situations are shaped by values and cultural beliefs they are not deemed formalizabled.

218

Secondly, environments are not only defined as biophysical or monolithyc structures. By considering both subjective and objective dimensions of uncertainty, a scholar can include cultural variables and their effect on the individual in decision models. This is so because the characterization of the decision setting as portrayed in chapter 2 permits the inclusion of individual assessments and certainties as well as behavioral rules that can affect the outcome of a decision process. In all, anthropologists can benefit by maintaining a certain level of cultural particularism without compromising generabilizability or dynamism.

7.3 Implications for anthropologists

Research needs to answer a policy question!

Despite the importance of understanding adaptation in uncertainty, with the exception of ethnoecologists and ethnobiologists, not many anthropologists are pursuing research in comparative cognition or decision making. Because of such gap, we lack good models about how decisions are made and updated information on how societies are responding to new ecological uncertainties by re-shifting their decision patterns. Economic interpretations of what Simon called Models of Man that is conceptualizations of what drives human behavior, are dominant in institutional settings (Simon 1957).

Given large failures in governance actions, a wider range of scientists acknowledge the importance of including social scientists in interdisciplinary teams. Because of the preminence of economic rationales, anthropologists spend substantial time in the criticism of models and frameworks that do not acknowledge the plurality behind decision motivations. However, unable to offer concise solutions, the role of anthropologists becomes passive, and more often than not, regarded with suspicion.

219

Why this is so may be a by product of our inability to communicate and articulate what anthropology is about. After more than a hundred years of anthropological research, we have not been able to transmit effectively the role of culture as a driver of behavior to other disciplines dealing with human action. Most significantly, we have not been able to articulate what our contribution should be as social scientists to current efforts in climate change adaptation and mitigation.

I believe that one way in which we can start addressing our shortcomings in communicating what anthropologists can do is through more policy relevant research and applied work. Ninety years ago Malinowski began his introduction to the Argonauts by underscoring how his work might help answer problems in dealing with local populations (1978). He was ostracized and criticized by recent generations of anthropologists because of his apparent lack of political commitment and ethical integrity. Disregarding the validity of such accusations, it is by reuniting anthropology with practice that I see a way forward. It is also my conviction that we should devote more time to translating and operationalizing our findings in easily understandable models or framesets, and pursuing more engagement in institutional settings.

7.4 Implications for policy and governance

Results indicate that conservation policies need to:

1) Understand the role of socio economic and environmental uncertainty in fishing effort and how it impacts household patterns of resource use before assuming that an area is overfished.

Labels like “overfishing” or “extreme exploitation” have important consequences for stakeholders. They set the strategy for regulatory framesets that may constrain stressed local economies from accessing resources. They may stigmatize local communities and blame them

220 for ecosystem degradation. Managers and policy makers need to conduct assessments of societal impacts before making claims on the state of a fishery. The US Magnusson and Steven’s Act is a good example of how a responsible and pluralistic policy frame can balance societal needs with ecological priorities.

2) Aproach individual decision making modeling in terms of the economics of resource extraction and incorporate socio-environmental uncertainty to explain current level of fishing effort.

The use of parametric models in poor data fisheries may be a solution to management and informational gaps if such estimates are corrected and qualified by observations of the socio- economic conditions of households. Policy makers and modelers should consider that motivations behind the use of resources are varied and do not necessarily respond to maximization principles. This is a challenge in that resource users may embody a myriad of different preferences. Furthermore, extractive behavior may respond to complex socio-political conditions. Social scientists are equipped to capture such complexity by ethnographic, observational and analytical techniques. Most significantly, they can provide guidance in the design of policies that are better targeted to address societal needs.

3) Study the process of adaptation.

Adaptation responses have emphasized over the last decades the value of traditional knowledge and local practices. Many agencies and organisms are including local responses as benchmarks for the design of solutions to vulnerability and mitigation problems (see UNEP or

IPCC). However, the assumption that adaptation is a process that will happen or has already occurred is prevalent across institutional discourses (Dessai and Hulme 2004). Research is, thus, directed to the identification of solutions to the detriment of the process of adaptation. By

221 adopting such a static approach to change, managers lack the flexibility to identify possibilities and opportunities to assist local communities and enhance resilience.

4) Better informed responses. Recognize that flexibility in strategies is a result of enabling structural responses in local communities that go beyond technical solutions.

Adaptation solutions more often than not include unrealistic assessments of community needs and gaps. Therefore, some of the solutions proposed, like ecotourism, are not viable.

Flexibility in adapting to future change requires enhancing the local capacities of vulnerable communities so the response is not just a one time event. That is, managers should strive to generate solutions that encompass structural changes in well-being and material indexes. With this purpose, alternative livelihood programs need to explore potential shifts in subsistence practices that offer sustainable options in socio-cultural and ecological ways.

7.5 Limitations, challenges and opportunities

Many limitations and challenges can be indentified in this dissertation. First of all, regional datasets are incomplete. To correct measurements, observations were complemented with information from other regions like Northern Australia. However, the fidelity of the data affects the validity of the conclusions drawn in chapter 3 and 4. In addition, there are substantial gaps on oceanographic and fishery parameters that impede a comprehensive exploration of the status of stocks.

Projections in chapter 4 and in Annex 1 are only estimates and were conducted through parametric approaches. In the future, simulations that have the potential of including stochastic parameters and non-parametric techniques will be explored.

Regarding other instruments that were analyzed in chapter 5 and 6 limitations are of sample size and replicability of procedures and results. Many of the analyses carried out

222 responded to the particular structure of the environment in which decisions were studied, and assume a high level of specificity (i.e.: they deal with individual traits). As a consequence, there are important constraints to the generalization of findings (external validity) and of the conclusions drawn. I paid careful attention, when possible, to detailing the steps on the analysis and transformations of the variables. These shortcomings might be solved if more research is pursued in the area (increase inferential power through bigger samples sizes) and if a comprehensive model of decision making is proposed.

Finally, I see the translation of the principlies inspiring this research to other systems of resources like forests and pastures as an opportunity. I believe that an environmental structure based approach to decision making can report benefits to the governance of open access systems and traditional tenure regimes. In the future, I will continue working in formalizing a decision model through simulations and multilevel analyses, and in addressing the different levels in which environmental uncertainty shapes behavior.

223

REFERENCES

Aburto, J., M. Thiel, and W. Stotz. 2009. Allocation of effort in artisanal fisheries: The

importance of migration and temporary fishing camps. Ocean & Coastal Management 52

(12) (12): 646-54.

Acciaioli, G. 2004. From Economic Actor to Moral Agent: Knowledge, Fate and Hierarchy

among the Bugis of Sulawesi. Indonesia 78: 147-179.

Acharya, S. 2010. Presage Biology: Lessons from nature in weather forecasting. Indian Journal

of Traditional Knowledge, 10(1): 114-124.

Acheson, J. M. & Wilson J.A. 1996. Order out of chaos: the case for parametric fisheries

management. American Anthropologist 98:579-594.

Acheson, J. M. 1979. Variations in traditional inshore fishing rights in maine lobstering

communities. North Atlantic Maritime Cultures: Anthropological Essays on Changing

Adaptations.

Acheson, J. M. 1981. Anthropology of fishing. Annual Review of Anthropology: 275-316.

Achmad, A., M. Munasik, and D.P. Wijayanti. 2013. Kondisi ekosistem terumbu karang di rote

timur, kabupaten rote ndao, taman nasional perairan laut Sawu menggunakan metode manta

tow. Journal of Marine Research 2 (3): 211-9.

Al-Abdulrazzak, D., R. Naidoo, M. L. D. Palomares, and D. Pauly. 2012. Gaining perspective on

what we've lost: The reliability of encoded anecdotes in historical ecology. PloS One 7 (8):

e43386.

Aldrian, E. &Susanto, R.D. 2003. Identification of three dominant rainfall regions within

224

Indonesia and their relationship to sea surface temperature. International Journal of

Climatology, 23(12):1435 1452.

Alfaro-Shigueto, J., J. C. Mangel, M. Pajuelo, P.H. Dutton, J. A. Seminoff, and B. J. Godley.

2010. Where small can have a large impact: Structure and characterization of small-scale

fisheries in Peru. Fisheries Research 106 (1): 8-17.

Allen, G. R. 2008. Conservation hotspots of biodiversity and endemism for Indo‐Pacific coral

reef fishes. Aquatic Conservation: Marine and Freshwater Ecosystems 18 (5): 541-56.

Allen, G.R., and M. Adrim. 2003. Coral reef fishes of Indonesia. Zoological Studies-Taipei- 42

(1): 1-72.

Allen, G. R., R. Steene, P. Humann, and Ned DeLoach. 2003. Reef fish identification: Tropical

pacificNew World Pubns Inc.

Allison, E.H. and F. Ellis. 2001. The Livelihoods Approach and Management of Small-Scale

Fisheries, Marine Policy, 25(5): 377-388.

Allison, E. H., A. L. Perry, M.‐C. Badjeck, W. N. Adger, K. Brown, D. Conway, A. S. Halls,

Graham M. Pilling, J.D. Reynolds, and N. L. Andrew. 2009. Vulnerability of national

economies to the impacts of climate change on fisheries. Fish and Fisheries 10 (2): 173-96.

Allison, E. H., and B. Horemans. 2006. Putting the principles of the sustainable livelihoods

approach into fisheries development policy and practice. Marine Policy 30 (6): 757-66.

Allison, E. H., and I.Kelling. 2009. Fishy crimes: The societal costs of poorly governed marine

fisheries. Paper presented at Third Annual Convention of the Consortium of Non-

Traditional Security Studies in Asia, Singapore.

Allison, E. H., F. Ellis, P. M. Mvula, and L. F. Mathieu. 2001. Fisheries management and

uncertainty: The causes and consequences of variability in inland fisheries in africa, with

225

special reference to malawi edward H. Allison1, frank Ellis1, peter M. Mvula2 and laurence

F. Mathieu1. Paper presented at Lake Malawi Fisheries Management Symposium.

Alloy, L. B., and N. Tabachnik. 1984. Assessment of covariation by humans and animals: The

joint influence of prior expectations and current situational information. Psychological

Review 91 (1): 112.

Alvarez, A., and P.Schmidt. 2006. Is skill more important than luck in explaining fish catches?

Journal of Productivity Analysis 26 (1): 15-25.

Ammarell, G. 2002 a. Bugis migration and modes of adaptation to local situations. Ethnology 41

(1): 51-68.

Ammarell, G. 2002 b. Knowing when to set sail: Practical knowledge and simple heuristics in

bugis navigational strategies. Bijdragen Tot De Taal-, Land-En Volkenkunde 158 (2): 191-

223.

Anderson, J. R. 1991b. Is human cognition adaptive? Behavioral and Brain Sciences 14 (03):

471-85.

Anderson, J. R. 1991a. The adaptive nature of human categorization. Psychological Review 98

(3): 409.

Annamalai, H., and P. Liu. 2005. Response of the Asian summer monsoon to changes in El Niño

properties. Quarterly Journal of the Royal Meteorological Society 131 (607): 805-31.

Arreguín-Sánchez, F. 1996. Catchability: A key parameter for fish stock assessment. Reviews in

Fish Biology and Fisheries 6 (2): 221-42.

Astuti, R. 1995. People of the sea: Identity and descent among the vezo of Madagascar.

Cambridge University Press.

Aswani, S. 1998. Patterns of marine harvest effort in southwestern new georgia, solomon

226

islands: Resource management or optimal foraging? Ocean & Coastal Management 40 (2):

207-35.

Auffenberg, W. 1981. The behavioral ecology of the komodo monitor. University Press of

Florida.

Babcock, E. A., and E. K. Pikitch. 2000. A dynamic programming model of fishing strategy

choice in a multispecies trawl fishery with trip limits. Canadian Journal of Fisheries and

Aquatic Sciences 57 (2): 357-70.

Badan Pusat Statistik Kabupaten Ende .1985 to 2011. Ende Dalam Angka. Badan Pusat Statistik

Kabupaten Ende. Ende, Flores, Indonesia.

Badjeck M-C, et al. 2009. Impacts of climate variability and change on fishery-based livelihoods.

Marine Policy , doi:10.1016/j.marpol.2009.08.007.

Barlow, C. 2007. Socio-economic conditions and poverty alleviation in nusa tenggara timur.

Bateson, M., and A. Kacelnik. 1998. Risk-sensitive foraging: Decision making in variable

environments. Cognitive Ecology: 297-341.

Beckerman, S. 1991. Risk sensitive foraging behavior among the barí. Human Ecology(4): 509.

Begossi, A., M. Clauzet, N. Hanazaki, P.F.M. Lopes, M. Ramires, and R. AM Silvano. 2009.

Fishers' decision making, optimal foraging and management. Proceedings of III Seminário

De Gestão Socioambiental Para o Desenvolvimento Sustentável Da Aqüicultura e Da Pesca

no Brasil: 1-3.

Beitl, C. M. 2012. Beyond Collective Action: A Multi-Scale Analysis of Sustainability in the

Mangrove Fishery Commons of Coastal Ecuador. Dissertation. The University of Georgia.

Béné, C., and A. Tewfik. 2001. Fishing effort allocation and fishermen's decision making

process in a multi-species small-scale fishery: Analysis of the conch and lobster fishery in

227

turks and caicos islands. Human Ecology 29 (2) (Jun.): 157-86.

Béné, C., B. Hersoug, and E. H. Allison. 2010. Not by rent alone: Analysing the Pro‐Poor

functions of Small‐Scale fisheries in developing countries. Development Policy Review 28

(3): 325-58.

Béné, C. 2003. When fishery rhymes with poverty: A first step beyond the old paradigm on

poverty in small-scale fisheries. World Development 31 (6): 949-75.

Berkes, F., J. Colding, and C.Folke. 2000. Rediscovery of traditional ecological knowledge as

adaptive management. Ecological Applications 10 (5): 1251-62.

Berkes, F., M. Kislalioglu Berkes, and H. Fast. 2007. Collaborative integrated management in

Canada's north: The role of local and traditional knowledge and community-based

monitoring. Coastal Management 35 (1): 143-62.

Berkes, F. 2006. From community-based resource management to complex systems: The scale

issue and marine commons. Ecology and Society 11 (1): 45.

Berkes, F. 2009. Evolution of co-management: Role of knowledge generation, bridging

organizations and social learning. Journal of Environmental Management 90 (5): 1692-702.

Bernstein, P. L. 1996. Against the gods: The remarkable story of risk.

Bettinger, R. L. 1991. Middle-range theory and hunter-gatherers. In Hunter-gatherers., 61-

82Springer.

Bird, D. W., R. Bliege Bird, and B. F. Codding. 2009. In pursuit of mobile prey: Martu hunting

strategies and archaeofaunal interpretation. American Antiquity: 3-29.

Bjarnason, T. and T. Thorlindsson. 1993. In defense of a folk model: the ‘skipper effect’ in the

Icelandic cod fishery, American Anthropology. 95:371–394.

Blaber, S.J.M., C.M. Dichmont, R.C. Buckworth, B. Sumiono, S. Nurhakim, B. Iskandar, B.

228

Fegan, DC Ramm, and JP Salini. 2005. Shared stocks of snapper (lutjanidae) in australia

and Indonesia: Integrating biology, population dynamics and socio-economics to examine

management scenarios. Reviews in Fish Biology and Fisheries 15 (1-2): 111-27.

Blaber, S.J.M., C.M. Dichmont, W. White, R. Buckworth, L. Sadiyah, B. Iskandar, S. Nurhakim,

R. Pillans, and R. Andamari. 2009. Elasmobranchs in southern Indonesian fisheries: The

fisheries, the status of the stocks and management options. Reviews in Fish Biology and

Fisheries 19 (3): 367-91.

Bockstael, N. E., and J. J. Opaluch. 1983. Discrete modelling of supply response under

uncertainty: The case of the fishery. Journal of Environmental Economics and Management

10 (2): 125-37.

Bolger, N., and J.-P. Laurenceau. 2013. Intensive longitudinal methods: An introduction to diary

and experience sampling research. Guilford Press.

Bolger, N., A. Davis, and E. Rafaeli. 2003. Diary methods: Capturing life as it is lived. Annual

Review of Psychology 54 (1): 579-616.

Boomgaard, P. 2005. Resources and People of the Sea in and around the Indonesian

Archipelago; 900-1900. In Boomgaard, Peter, David Henley and Manon Osseweijer

(eds.) Muddied Waters: Historical and Contemporary Perspectives on the Management of

Forests and Fisheries in Island Southeast Asia. Leiden: KITLV Press. pp.97-120.

Boomgaard, P. 2007. A world of water: Rain, rivers and seas in Southeast Asian histories, Brill.

The Netherlands.

Boster, J. 1984. Inferring decision making from preferences and behavior: An analysis of

aguaruna jivaro manioc selection. Human Ecology 12 (4): 343-58.

Boyd, R., and P. J. Richerson. 2002. Group beneficial norms can spread rapidly in a structured

229

population. Journal of Theoretical Biology 215 (3): 287-96.

Branch, T. A., R. Hilborn, A. C. Haynie, G. Fay, L. Flynn, J. Griffiths, K. N. Marshall, Jeffrey

K. Randall, J.M. Scheuerell, and E. J. Ward. 2006. Fleet dynamics and fishermen behavior:

Lessons for fisheries managers. Canadian Journal of Fisheries and Aquatic Sciences 63 (7):

1647-68.

Britton, E., and S. Coulthard. 2013. Assessing the social wellbeing of Northern Ireland's fishing

society using a three-dimensional approach. Marine Policy 37: 28-36.

Brosius, P. J., A.Lowenhaupt Tsing, and C. Zerner. 2005. Communities and conservation:

Histories and politics of community-based natural resource management. Rowman

Altamira.

Brunswik, E. 1943. Organismic achievement and environmental probability. Psychological

Review 50 (3): 255.

Butcher, J.G. 2005. The marine animals of South East Asia: towards a demographic history 1850-

2000. In Boomgaard, Peter, David Henley and Manon Osseweijer (eds.) Muddied Waters:

Historical and Contemporary Perspectives on the Management of Forests and Fisheries

in Island Southeast Asia. Leiden: KITLV Press. pp.63-96.

Butcher, J.G. 2004.The Closing of the Frontier: A History of the marine fisheries of Southeast

Asia c.1850-2000. Institute of Southeast Asian Studies (ISEAS): Singapore.

Caddy, JF, and R. McGarvey. 1996. Targets or limits for management of fisheries? North

American Journal of Fisheries Management 16 (3): 479-87.

Caddy, JF. 1999. Fisheries management in the twenty-first century: Will new paradigms

apply? Reviews in Fish Biology and Fisheries 9 (1): 1-43.

Cancian, F. 1980. Risk and uncertainty in agricultural decision-making. In Agricultural decision-

230

making: Anthropological contributions to rural development,(ed.) Peggy F. Barlett, 161-

176. Studies in Anthropology Series. Orlando, Fla., USA: Academic Press.

Caraco, T. 1981. Risk-sensitivity and foraging groups. Ecology: 527-31.

Carder, N., E.J. Reitz, and J.G. Crock. 2007. Fish communities and populations during the post-

Saladoid period (AD 600/800–1500), anguilla, Lesser Antilles. Journal of Archaeological

Science 34 (4): 588-99.

Cashdan, E. A. 1990. Risk and uncertainty in tribal and peasant economiesWestview Pr.

Cashdan, E. A. 1992. Spatial organization and habitat use. Evolutionary Ecology and Human

Behavior: 23-7.

Cesar, H., L.Burke, and L.Pet-Soede. 2003. The economics of worldwide coral reef degradation.

Chakravarty-Kaul, M. 1998. Transhumance and customary pastoral rights in himachal pradesh:

Claiming the high pastures for gaddis. Mountain Research and Development: 5-17.

Charles, A. 2011. Good practices for governance of small-scale fisheries. World Small-Scale

Fisheries Contemporary Visions.Delft: Eburon: 285-99.

Charles, A. 2012. People, oceans and scale: Governance, livelihoods and climate change

adaptation in marine social–ecological systems. Current Opinion in Environmental

Sustainability 4 (3): 351-7.

Cheung, W. W.L., V. W.Y. Lam, J. L. Sarmiento, K. Kearney, R. Watson, and D. Pauly. 2009.

Projecting global marine biodiversity impacts under climate change scenarios. Fish and

Fisheries 10 (3): 235-51.

Cheung, W. W.L., V. WY Lam, J. L. Sarmiento, K. Kearney, R.E.G Watson, D. Zeller, and D.

Pauly. 2010. Large‐scale redistribution of maximum fisheries catch potential in the global

ocean under climate change. Global Change Biology 16 (1): 24-35.

231

Chou, C. 2003. Indonesian sea nomads; Money, magic, and fear of the Orang Suku Laut.

London: Routledge Curzon.

Chozin, M. 2008. Illegal but Common: Life of Blast Fishermen in the Spermonde Archipelago,

South Sulawesi, Indonesia.

Cnaan, A., N.M .Laird, and P. Slasor. 1997. Tutorial in biostatistics: Using the general linear

mixed model to analyse unbalanced repeated measures and longitudinal data. Statistics in

Medicine 16 (20): 2349-80.

Colfer, C.P., Wadley, R. and Venkateswarlu, P. 1999. Understanding local people’s use of time:

A pre-condition for good co management, Environmental Conservation 26(1): 41-52.

Conrad, J. M. 2010. Resource economics. Cambridge University Press.

Cordell, J. 1974. The lunar-tide fishing cycle in Northeast Brazil. Ethnology 13:379-392.

Cosmides, L., and J.Tooby. 1996. Are humans good intuitive statisticians after all? rethinking

some conclusions from the literature on judgment under uncertainty. Cognition 58 (1): 1-73.

Coulthard, S. 2011. Poverty, sustainability and human wellbeing: A social wellbeing approach to

the global fisheries crisis. Global Environmental Change 21 (2): 453.

Cummins, D. D. 2002. The evolutionary roots of intelligence and rationality. Common Sense,

Reasoning and Rationality: 132-47.

da Silva, A.L., and A.Begossi. 2009. Biodiversity, food consumption and ecological niche

dimension: A study case of the riverine populations from the Rio Negro, Amazonia, Brazil.

Environment, Development and Sustainability 11 (3): 489-507.

Dall, S. RX, L.-A. Giraldeau, O. Olsson, J. M. McNamara, and D. W. Stephens. 2005.

Information and its use by animals in evolutionary ecology. Trends in Ecology & Evolution

20 (4): 187-93.

232

D'Andrade, R. G. 1989. Cultural cognition. Paper presented at Foundations of cognitive science.

D'Arrigo, R., and R. Wilson. 2008. El nino and Indian Ocean influences on Indonesian drought:

Implications for forecasting rainfall and crop productivity. International Journal of

Climatology 28 (5): 611-6.

Dayem, K. E., D. C. Noone, and P. Molnar. 2007. Tropical western pacific warm pool and

maritime continent precipitation rates and their contrasting relationships with the walker

circulation. Journal of Geophysical Research: Atmospheres (1984–2012) 112 (D6). de Oliveira, L. E. Chimello, and A. Begossi. 2011. Last trip return rate influence patch choice

decisions of small-scale shrimp trawlers: Optimal foraging in são francisco, coastal

brazil. Human Ecology 39 (3): 323-32.

DeFeo, O., and J. C. Castilla. 2005. More than one bag for the world fishery crisis and keys for

co-management successes in selected artisanal Latin American shellfisheries. Reviews in

Fish Biology and Fisheries 15 (3): 265-83.

Demanèche, S., C. Merrien, J. Vigneau, O. Guyader, P. Berthou, P. Lespagnol, E. Leblond, and

F. Daurès. 2008. A new approach to estimate catches and fishing effort of small scale

fisheries by sampling fishing trips on-site. Paper presented at Proceedings of the ICES

Annual Science Conference.

Dessai, S., M. Hulme, R. Lempert, and R. Pielke Jr. 2009. Climate prediction: A limit to

adaptation. Adapting to Climate Change: Thresholds, Values, Governance: 64-78.

Di Natale, A., and A. Mangano. 1995. Moon phases influence on CPUE: A first analysis of

swordfish driftnet catch data from the italian fleet between 1990 and 1991.

Collect.Vol.Sci.Pap.ICCAT 44 (1): 264-7.

Die, D. J., and J. F. Caddy. 1997. Sustainable yield indicators from biomass: Are there

233

appropriate reference points for use in tropical fisheries? Fisheries Research 32 (1): 69-79.

Dietrich, S. 1983. Flores in the nineteenth century: Aspects of Dutch colonialism on a

non‐profitable island. Indonesia Circle 11 (31): 39-58.

Dietrich, S. 1989. Kolonialismus und mission auf flores (ca. 1900-1942). Vol. 1K. Renner.

Dove, M. R. 1985. The Kantu system of land tenure: The evolution of tribal land rights in

Borneo. Studies in Third World Societies 33 : 159-82.

Dove, M. R. 1993. Uncertainty, humility, and adaptation in the tropical forest: The agricultural

augury of the Kantu'. Ethnology: 145-67.

Duncan, R. B. 1972. Characteristics of organizational environments and perceived environmental

uncertainty. Administrative Science Quarterly 17 (3).

Dunlap, A. S., and D. W. Stephens. 2009. Components of change in the evolution of learning and

unlearned preference. Proceedings.Biological Sciences / the Royal Society 276 (1670) (Sep

7): 3201-8.

Durrenberger, E. P. and G. Pálsson. 1983. Riddles of herring and rhetorics of success. Journal of

Anthropological Research: 323-35.

Durrenberger, E. P. and G.i Pálsson. 1986. Finding fish: The tactics of icelandic skippers.

American Ethnologist 13 (2): 213-29.

Edjid, H. 1977. Tiga Ratus Tahun Kerajaan Ende. 1638-1962. Fakultas Keguruan Universitas

Negeri Nusa Cendana, Ende.

Eggert, H., and P. Martinsson. 2004. Are commercial fishers risk-lovers? Land Economics 80

(4): 550-60.

Elio, R. 2002. Common sense, reasoning, and rationalityOxford University Press.

Ellis, N., and Y.-G. Wang. 2007. Effects of fish density distribution and effort distribution on

234

catchability. ICES Journal of Marine Science: Journal Du Conseil 64 (1): 178-91.

Ellis, R. 2009. Tuna: love, death, and mercury. Vintage Books, Random house Books: New York.

Fauzi, A., and Z. Anna. 2010. Social resilience and uncertainties. the case of small-scale fishing

households in the north coast of central java. Journal of Maritime Studies (MAST) 9 (2): 55-

64.

Feeny, D., F. Berkes, B. McCay and J. M. Acheson 1990. The tragedy of the commons: twenty-

two years later. Human Ecology 18 (1): 1-19.

Felson, R. B., and G.e Gmelch. 1979. Uncertainty and the use of magic. Current Anthropology:

587-9.

Firth, R. 2006. Malay fishermen. Routledge.

Foley, R.A. 1985. Optimality theory in anthropology. Man 20:222-242.

Folke, C. 2004. Traditional knowledge in social-ecological systems. Ecology and Society 9 (3):

Forth, G. 1998. Beneath the volcano: Religion, cosmology and spirit classification among the

Nage of eastern Indonesia. Leiden: The Royal Institute (KITLV) Press.

Fox, J. 1977. Harvest of the Palm: Ecological Change in Eastern Indonesia, Harvard University

Press, Cambridge, Mass.

Fox, J. 2005. In a Single Generation: A Lament for the Forests and Seas of Indonesia. In

Boomgaard, Peter, David Henley and Manon Osseweijer (eds.) Muddied Waters:

Historical and Contemporary Perspectives on the Management of Forests and Fisheries

in Island Southeast Asia, Leiden: KITLV Press. pp. 43-60.

Fulton, E. A., A. D.M. Smith, D.C. Smith, and I. E. van Putten. 2011. Human behaviour: The

key source of uncertainty in fisheries management. Fish and Fisheries 12 (1): 2-17.

235

Garciá-Quijano, C. G. 2006. Resisting Extinction: The Value of Local Ecological Knowledge for

Small-Scale Fishers in Southeastern Puerto Rico. Dissertation. The University of Georgia.

Garro, L. C. 1998. On the rationality of Decision‐Making studies: Part 1: Decision models of

treatment choice. Medical Anthropology Quarterly 12 (3): 319-40.

Gigerenzer, G. 2008. Rationality for Mortals. How people cope with uncertainty. New

York:Oxford University Press.

Gigerenzer, G. and H. Brighton. 2009. Homo heuristicus: Why biased minds make better

inferences. Topics in Cognitive Science 1(1:)107-144.

Gigerenzer, G. and W. Gaissmaier. 2011. Heuristic Decision Making. Annual Review of

Psychology 62: 451-482.

Gigerenzer, G., P. M. Todd, and the ABC Group. 1999. Simple heuristics that make us smart.

New York: Oxford University Press.

Gigerenzer, G. and D. J. Murray. 1987. Cognition as intuitive statistics.Lawrence Erlbaum

Associates, Inc.

Gigerenzer, G. and R. Selten. 2002. Bounded rationality: The adaptive toolboxMit Press.

Gladwin, C. H. 1980. A theory of real-life choice: applications to agricultural decisions. In

Agricultural Decision Making, Peggy Barlett (ed). New York: Academic Press, pp: 45-

83.

Gladwin, C. H. 1976. A view of the plan puebla: An application of hierarchical decision models.

American Journal of Agricultural Economics 58 (5): 881-7.

Gladwin, C. H. 1979. Cognitive strategies and adoption decisions: A case study of nonadoption

of an agronomic recommendation. Economic Development and Cultural Change: 155-73.

Gladwin, C. H. 1983. Contributions of decision-tree methodology to a farming systems program.

236

Human Organization 42 (2): 146-57.

Gladwin, C. H. 1989. Ethnographic decision tree modeling. Vol. 19Sage.

Gladwin, H.. 1970. Decision making in the Cape Coast (Fante) fishing and fish marketing

system. Dissertation. Department of Anthropology, Stanford University.

Godoy, R.A., V. Reyes-García, V. Vadez, O. Magvanjav, W. R. Leonard, T. McDade, S. Kumar,

J. Iqbal, D. Wilkie, S. Tanner, T. Huanca, and TAPS Bolivia Study Team. 2009. Does

culture affect the present? The effect of future weather on the current collection of planted

crops and wildlife in a native Amazonian society of Bolivia. Human Ecology 37(5):613-

628.

Gordon, H. S. 1953. An economic approach to the optimum utilization of fishery

resources. Journal of the Fisheries Board of Canada 10 (7): 442-57.

Gordon, H. S. 1954. The economic theory of a common-property resource: The fishery. The

Journal of Political Economy: 124-42.

Gowdy, J. 2008. Behavioral economics and climate change policy. Journal of Economic

Behavior and Organization, 68:632-644.

Gragson, T. L. 1992. Fishing the waters of Amazonia: Native subsistence economies in a tropical

rain forest. American Anthropologist 94 (2): 428-40.

Guyader, O., P. Berthou, C. Koutsikopoulos, F. Alban, S. Demaneche, M.B. Gaspar, R.

Eschbaum, E. Fahy, O. Tully, and L. Reynal. 2013. Small scale fisheries in Europe: A

comparative analysis based on a selection of case studies. Fisheries Research 140 : 1-13.

Ha, T. T.T.u, H.van Dijk, and S. R. Bush. 2012. Mangrove conservation or shrimp farmer's

livelihood? the devolution of forest management and benefit sharing in the mekong delta,

vietnam.Ocean & Coastal Management 69 : 185-93.

237

Hacking, I. 1990. The taming of chance. Vol. 17Cambridge University Press.

Halim, A. 2002. Adoption of cyanide fishing practice in Indonesia. Ocean and Coastal

management 45:313-323.

Hall, K. R. 1985. Maritime trade and state development in early Southeast Asia. University of

Hawaii Press Honolulu.

Hall, Kenneth R. 2011. A history of early Southeast Asia: Maritime trade and societal

development, 100-1500. Lanham, Md.: Rowman & Littlefield.

Hamada, J. , M.D. Yamanaka, J. Matsumoto, S. Fukao , P.A. Winarso and T. Sribimawati T.

2002. Spatial and Temporal Variations of the Rainy Season Over Indonesia and Their

Link to ENSO. Journal of the Meteorological Society of Japan 80:285-310.

Hamon, K. 2009. Typology and dynamics of a spatially and temporally heterogenous fishery:

The tasmanian rock lobster fishery. Paper presented at 18th World Imacs Congress And

Modsim09 International Congress On Modelling And Simulation-Interfacing Modelling

And Simulation With Mathematical And Computational Sciences, .

Hanazaki, N. and A. Begossi. 2003. Does fish still matter? changes in the diet of two brazilian

fishing communities. Ecology of Food and Nutrition 42 (4-5): 279-301.

Haselton, M. G., G. A. Bryant, A. Wilke, D. A. Frederick, A. Galperin, W. E. Frankenhuis, and

T. Moore. 2009. Adaptive rationality: An evolutionary perspective on cognitive bias. Social

Cognition 27 (5): 733-63.

Hastie, R. 2001. Problems for judgment and decision making. Annual Review of Psychology 52

(1): 653-83.

Hayes, A. F. 2006. A primer on multilevel modeling. Human Communication Research 32 (4):

385-410.

238

Heazle, M. and J.G. Butcher. 2007. Fisheries depletion and the state in Indonesia: Towards a

regional regulatory regime. Marine Policy 31(3): 276-286.

Helfman, G. 2007. Fish Conservation. Island Press, Washington, DC.

Henley, D. and M. Osseweijer 2005. Introduction; Forests and fisheries in island Southeast Asia;

Histories of natural resource management and mismanagement. In Boomgaard, Peter,

David Henley and Manon Osseweijer (eds.) Muddied Waters: Historical and

Contemporary Perspectives on the Management of Forests and Fisheries in Island

Southeast Asia. Leiden: KITLV Press. pp.1-42.

Henrich, J., S. J. Heine, and A. Norenzayan. 2010. The weirdest people in the world? Behavioral

and Brain Sciences 33 (2-3): 61-83.

Hertwig, R. and G. Gigerenzer. 1999. Theconjunction fallacy'revisited: How intelligent

inferences look like reasoning errors. Journal of Behavioral Decision Making 12 : 275-306.

Hilborn R and M. Mangel. 1997. The Ecological Detective: Confronting Models with Data.

Princeton, New Jersey:Princeton University Press.

Hilborn, R., and M. Mangel. The ecological detective: Confronting models with data. 1997.

Princeton New Jersey.

Hilborn, R. 2007. Moving to sustainability by learning from successful fisheries. AMBIO: A

Journal of the Human Environment 36 (4): 296-303.

Holland, D. S. 2008. Are fishermen rational? A fishing expedition. Marine Resource Economics

23 (3).

Houston, A., C. W. Clark, J. M. McNamara, and M. Mangel. 1988. Dynamic models in

behavioural and evolutionary ecology. Nature 332:29-34

Howell, W. C. and S. A. Burnett. 1978. Uncertainty measurement: A cognitive taxonomy.

239

Organizational Behavior and Human Performance 22 (1): 45-68.

Hutchinson, J. and G. Gigerenzer. 2005. Simple heuristics and rules of thumb: Where

psychologists and behavioural biologists might meet. Behavioural Processes 69 (2): 97-124.

Ingles, J., J. Flores and I. Musthofa. 2008. Getting of the hook. Reforming tuna fisheries of

Indonesia. Coral Triangle Initiative.World Wildlife Foundation.

Jackson, J. B., M. X. Kirby, W. H. Berger, K. A. Bjorndal, L. W. Botsford, B. J. Bourque, R. H.

Bradbury, et al. 2001. Historical overfishing and the recent collapse of coastal

ecosystems. Science (New York, N.Y.) 293 (5530) (Jul 27): 629-37.

Jacquet, J. and D. Pauly. 2008. Funding priorities: Big barriers to Small‐Scale

fisheries. Conservation Biology 22 (4): 832-5.

Johannes, R. E. 1978. Traditional marine conservation methods in oceania and their demise.

Annual Review of Ecology and Systematics 9 (1): 349-64.

Johannes, R. E. 1993. Integrating traditional ecological knowledge and management with

environmental impact assessment. Traditional Ecological Knowledge: Concepts and Cases:

33-9.

Kahneman, D. , P. Slovic. and A. Tverski. 1982. Judgement under uncertainty: Heuristics and

biases. CUP, Cambridge.

Kahneman, D. and A. Tversky. 1974. Subjective probability: A judgment of representativeness.

In The concept of probability in psychological experiments., 25-48Springer.

Kahneman, D. and A.Tversky. 1984. Choices, values, and frames. American Psychologist 39

(4): 341.

Kahneman, D. and A. Tversky. 1996. On the reality of cognitive illusions. .

Kahneman, D. and A. Tversky. 2000. Choices, values, and framesCambridge University Press.

240

Kahneman, D. 2003. Maps of bounded rationality: Psychology for behavioral economics.

American Economic Review: 1449-75.

Kaplan, H., and K. Hill. 1992. The evolutionary ecology of food acquisition. Evolutionary

Ecology and Human Behavior: 167-201.

King, T.J. 2011. The" skipper effect": Riddles of luck and rhetorics of individualism. Human

Organization 70 (4): 387-96.

Kittinger, J. N., Elena M. Finkbeiner, Edward W. Glazier, and Larry B. Crowder. 2012. Human

dimensions of coral reef social-ecological systems. Ecology & Society 17 (4).

Knaap, G. and H. Sutherland. 2004. Monsoon Traders: Ships, Skippers and Commodities in

Eighteenth-Century Makassar. Leiden, The Netherlands: KITLV Press.

Koster, J. M., M. N. Grote, and B. Winterhalder. 2013. Effects on household labor of temporary

out-migration by male household heads in nicaragua and peru: An analysis of spot-check

time allocation data using mixed-effects models. Human Ecology 41 (2): 221-37.

Kujala, H., M. A. Burgman, and A. Moilanen. 2013. Treatment of uncertainty in conservation

under climate change. Conservation Letters 6 (2): 73-85.

Kunda, Z. 1999. Social cognition: Making sense of people. MIT press.

Lavapie-Gonzales, F. 2005. Growth, mortality and recruitment of Decapterus Kurroides in

Davao Gulf, Philippines. Fishbyte 9 (2): 6-9.

Lehodey, P., F. Chai, and J. Hampton. 2003. Modelling climate‐related variability of tuna

populations from a coupled ocean–biogeochemical‐populations dynamics model. Fisheries

Oceanography12 (4‐5): 483-94.

Lehodey, P., M. Bertignac, J. Hampton, A. Lewis, and J. Picaut. 1997. El niño southern

oscillation and tuna in the western pacific. Nature 389 (6652): 715-8.

241

Lehodey, P., M. Bertignac, J. Hampton, .A Stoens, C. Menkès, L. Memery, and N. Grima. 1998.

Predicting skipjack tuna forage distributions in the equatorial pacific using a coupled

dynamical bio‐geochemical model. Fisheries Oceanography 7 (3‐4): 317-25.

Link, J. S. 2005. Translating ecosystem indicators into decision criteria. ICES Journal of Marine

Science: Journal Du Conseil 62 (3): 569-76.

Lopes, P.F.M. and A. Begossi. 2011. Decision‐making processes by small‐scale fishermen on the

southeast coast of brazil. Fisheries Management and Ecology 18 (5): 400-10.

Lopes, P. F. M., M. Clauzet, N. Hanazaki, M. Ramires, R. A. M. Silvano, and A. Begossi. 2011.

Foraging behaviour of brazilian riverine and coastal fishers: How much is explained by the

optimal foraging theory? Conservation & Society 9 (3) (10): 236-46.

Low, B. S. 1990. Human responses to environmental extremeness and uncertainty: A cross-

cultural perspective. Risk and Uncertainty in Tribal and Peasant Economies: 229-55.

Lowe, C. 2006. Wild Profusion. Princeton University Press, Princeton and Oxford.

Luce, R.D. and H. Raiffa. 1957. Games and decisions: Introduction and critical survey. Courier

Dover Publications.

Ludvico, L. R., I. M. Bennett, and S. Beckerman. 1991. Risk sensitive foraging behavior among

the Barí. Human Ecology 19 (4): 509-16.

Ludwig, D., R. Hilborn, and C. Walters. 1993. Uncertainty, resource exploitation, and

conservation—lessons from history. Science 260:17-36.

Lunn, K. E., and P. Dearden. 2006. Monitoring small-scale marine fisheries: An example from

Thailand's Ko Chang archipelago. Fisheries Research 77 (1): 60-71.

Mahévas, S., L. Bellanger, and V. M. Trenkel. 2008. Cluster analysis of linear model coefficients

under contiguity constraints for identifying spatial and temporal fishing effort patterns.

242

Fisheries Research 93 (1): 29-38.

Mahévas, S., Y. Vermard, T. Hutton, A. Iriondo, A. Jadaud, C. D. Maravelias, A. Punzón, J.

Sacchi, A. Tidd, and E. Tsitsika. 2011. An investigation of human vs. technology-induced

variation in catchability for a selection of European fishing fleets. ICES Journal of Marine

Science: Journal Du Conseil 68 (10): 2252-63.

Malinowski, B.. 1918. 53. fishing in the trobriand islands. Man 18 : 87-92.

Malinowski, B.. 1921. The primitive economics of the trobriand islanders. The Economic

Journal: 1-16.

Malinowski, B.. 1978. Argonauts of the Western Pacific: An account of native enterprise and

adventure in the archipelagoes of Melanesian New GuineaRoutledge.

Mangel, M. 1990. Dynamic information in uncertain and changing worlds, Journal of

Theoretical Biology, 146:317-332.

Mangel, M. and C.W. Clark. 1983. Uncertainty, search, and information in fisheries. Journal of

the International Council for the Exploration of the Seas 43:93-103.

Mangel, M. and C.W. Clark. 1986. Towards a unified foraging theory. Ecology 67:1127-1138.

Mangel, Marc. 1990. Dynamic information in uncertain and changing worlds. Journal of

Theoretical Biology 146 (3): 317-32.

Marchal, P., J.J. Poos, and F. Quirijns. 2007. Linkage between fishers’ foraging, market and fish

stocks density: Examples from some North Sea fisheries. Fisheries Research 83 (1): 33-43.

McCay, B. J. 1981. Optimal Foragers or Political Actors? Ecological Analyses of a New Jersey

Fishery. American Ethnologist 8:356-382.

McElreath, R., and J. Koster. 2013. Using multilevel models to estimate variation in foraging

returns: Effects of failure rate, harvest size, age, and individual heterogeneity. Human

243

Nature.

McGoodwin, J.R. 1990. Crisis in the World’s Fisheries: People, Problems, and Policies.

Stanford: Stanford University Press.

McIlgorm A., S. Hanna, G. Knapp, H. P. Le Floc’, F. Millerd and M. Pan. 2010. How will

climate change alter fishery governance: insights from seven international case studies.

Marine Policy. 34 (1): 170-7. doi:10.1016/j.marpol.2009.06.004.

McKee, T. B., N. J. Doesken, and J. Kleist. 1993. The relationship of drought frequency and

duration of time scales. Eighth Conference on Applied Climatology, American

Meteorological Society, Jan17-23, 1993, Anaheim CA, pp.179-186.

Millar, R.B, J.E McKenzie, J. D Bell, and LD Tierney. 1997. Evaluation of an indigenous fishing

calendar using recreational catch rates of snapper Pagrus Auratus in the north island of New

Zealand.

Millar, R.B., and T. J. Willis. 1999. Applications: Estimating the relative density of snapper in

and around a marine reserve using a Log‐Linear Mixed‐Effects model. Australian & New

Zealand Journal of Statistics 41 (4): 383-94.

Miller, K., A. Charles, M. Barange, K. Brander, V. F. Gallucci, M. A. Gasalla, A. Khan, G.

Munro, R. Murtugudde, and R . E. Ommer. 2010. Climate change, uncertainty, and resilient

fisheries: Institutional responses through integrative science. Progress in Oceanography 87

(1): 338-46.

Milner-Gulland, E. J. 2011. Integrating fisheries approaches and household utility models for

improved resource management. Proceedings of the National Academy of Sciences of the

United States of America 108 (4) (Jan 25): 1741-6.

Milner-Gulland, E. J. 2012. Interactions between human behaviour and ecological

244

systems. Philosophical Transactions of the Royal Society of London.Series B, Biological

Sciences 367 (1586) (Jan 19): 270-8.

Mistiaen, J. A., and I. E. Strand. 2000. Location choice of commercial fishermen with

heterogeneous risk preferences. American Journal of Agricultural Economics: 1184-90.

Mithen, S. J. 1990. Thoughtful foragers: a study of prehistoric decision making

Cambridge[England]; New York: Cambridge University Press.

Mithen, Steven J. 1988. Modeling hunter-gatherer decision making: Complementing optimal

foraging theory. Human Ecology 17 (1): 59-83.

Monk, K. A., Y. de Fretes, and G. Reksodiharjo-Lilley. 1997. The Ecology of Indonesia. Volume

V: The ecology of Nusa Tenggara and Maluku. Hong Kong: Periplus Editions.

Morrill, W. T. 1967. Ethnoicthyology of the cha-cha. Ethnology: 405-16.

Mous, P., L. Pet-Soede, M. Erdmann, H. Cesar, Y. Sadovy, and J. Pet. 2000. Cyanide fishing on

Indonesian coral reefs for the live food fish market-what is the problem. Collected Essays

on the Economics of Coral Reefs.Kalmar, Sweden: CORDIO, Kalmar University: 69-76.

Mous, P. J., J.S Pet, Z. Arifin, R. Djohani, M.V. Erdmann, A. Halim, M. Knight, L. Pet‐Soede,

and G. Wiadnya. 2005. Policy needs to improve marine capture fisheries management and

to define a role for marine protected areas in Indonesia. Fisheries Management and

Ecology 12 (4): 259-68.

Muller, J., J. F. McManus, D. W. Oppo, and R. Francois. 2012. Strengthening of the northeast

monsoon over the Flores Sea, Indonesia, at the time of Heinrich event 1. Geology 40 (7):

635-8.

Munasik, H. Adri, ATP. Wibowo, R. Kiswantoro, Y. Fajariyanto, and H. Sofyanto. 2011.

Kondisi Terumbu Karang di Taman Nasional Perairan Laut Sawu. Provinsi Nusa

245

Tenggara Timur Workshop Sosialisasi Peraturan Gubernur tentang Pengelolaan Terumbu

Karang COREMAP II Provinsi NTT, Kupang 25 Juli 2011.

Murtaugh, M., and H. Gladwin. 1980. A hierarchical decision-process model for forecasting

automobile type-choice. Transportation Research Part A: General 14 (5): 337-47.

Nakagawa, S. 1984.Endenese. In Muslim Peoples: A World Ethnographic Survey, Richard V.

Weekes (ed.), 251-253. Westport, Conn.: Greenwood Press.

Nakagawa, S. 1996. Endenese. In Encyclopedia of World Cultures. Retrieved March 22, 2010

from Encyclopedia.com:http://www.encyclopedia.com/doc/1G2-3458000787.html

Nakagawa, Satoshi. 2006. As things stood.

Needham, R. 1968. Endeh: Terminology, alliance, and analysis. Bijdragen Tot De Taal-, Land-

En Volkenkunde: 305-35.

Needham, R. 1980. Principles and variations in the structure of the Sumbanese society. In The

Flow of life: essays on eastern Indonesia. James J. Fox, Monni Adams (eds.) Harvard

Press, Boston. pp. 21-47.

Nickerson, R. S. 2012. Aspects of rationality: Reflections on what it means to be rational and

whether we arePsychology Press.

Nisbett, R. E., and Y. Miyamoto. 2005. The influence of culture: Holistic versus analytic

perception. Trends in Cognitive Sciences 9 (10): 467-73.

Nisbett, R. E., D. H. Krantz, C. Jepson, and Z.Kunda. 1983. The use of statistical heuristics in

everyday inductive reasoning. Psychological Review 90 (4): 339.

Nisbett, R. and L. Ross., L.1980. Human inference: Strategies and shortcomings of social judg-

ment. Englewood Cliffs, N]: Prentice-Hall.

Orlove, B.S., J.C.H. Chiang, and M. A. Cane. 2002. Ethnoclimatology in the Andes. American

246

Scientist, 90:428–435.

Orlove, B. S., John CH Chiang, and Mark A. Cane. 2000. Forecasting Andean rainfall and crop

yield from the influence of El Niño on Pleiades visibility. Nature 403 (6765): 68-71.

Ota, Y., and R. Just. 2008. Fleet sizes, fishing effort and the ‘hidden’factors behind statistics: An

anthropological study of small-scale fisheries in UK. Marine Policy 32 (3): 301-8.

Ott, L. and MT Longnecker. 2010. An introduction to statistical methods and data analysis,

Brooks/Cole cengage. Learning, Belmont CA.

Palmer, C. T. 1989. The ritual taboos of fishermen: An alternative explanation. MAST.Maritime

Anthropological Studies 2 (1): 59-68.

Palomares, M. L. D., and J.J. Heymans. 2006. Historical ecology of the Raja Ampat archipelago,

Papua province, Indonesia. .

Palsson, G., and E. P. Durrenberger. 1990. Systems of production and social discourse: The

skipper effect revisited. American Anthropologist 92 (1): 130-41.

Pálsson, G. and P. Durrenberger. 1982. To dream of fish: The causes of icelandic skippers

fishing success. Journal of Anthropological Research Anc Southwestern Journal of

Anthropology Albuquerque, NM 38 (2): 227-42.

Panaĭotov, T. and D. Panayotou. 1986. Occupational and geographical mobility in and out of

Thai fisheries. Food & Agriculture Org.

Paolisso, M. 2002. Blue Crabs and Controversy on the Chesapeake Bay: A Cultural Model for

Understanding Watermen’s Reasoning about Blue Crab Management. Human

Organization 61(3):226 239.

Parrek, A. and P. C. Trivedi. 2010. Cultural values and indigenous knowledge of climate change

and disaster prediction in Rajasthan, India. Indian Journal of Traditional Knowledge,

247

10(1): 183-189.

Pasaribu, S. M. 2007. Revitalizing institutions to enhance climate forecast application in East

Nusa Tenggara province, Indonesia. Anal Kebijak Pertan 5 (3): 239-53.

Pauly, D., V. V. Christensen, J. Dalsgaard, R. Froese, and F. Torres Jr. 1998. Fishing down

marine food webs. Science (New York, N.Y.) 279 (5352) (Feb 6): 860-3.

Pauly, D. and D. Zeller. 2014. Accurate catches and the sustainability of coral reef

fisheries. Current Opinion in Environmental Sustainability 7 : 44-51.

Pauly, D. 2006. Major trends in small-scale marine fisheries, with emphasis on developing

countries, and some implications for the social sciences. Maritime Studies 4 (2): 7-22.

Pelletier, D.and J. Ferraris. 2000. A multivariate approach for defining fishing tactics from

commercial catch and effort data. Canadian Journal of Fisheries and Aquatic Sciences 57

(1): 51-65.

Pelras, C. 1996. The Bugis. Oxford: Blackwell.

Perry, A. L., P. J. Low, J. R. Ellis, and J. D. Reynolds. 2005. Climate change and distribution

shifts in marine fishes. Science (New York, N.Y.) 308 (5730) (Jun 24): 1912-5.

Perry, R. I., R. E. Ommer, M. Barange, and F. Werner. 2010. The challenge of adapting marine

social–ecological systems to the additional stress of climate change. Current Opinion in

Environmental Sustainability 2 (5): 356-63.

Pet-Soede, C., W. L. T. Van Densen, J. G. Hiddink, S. Kuyl, and M. A. M. Machiels. 2001. Can

fishermen allocate their fishing effort in space and time on the basis of their catch rates? an

example from spermonde archipelago, SW sulawesi, indonesia. Fisheries Management &

Ecology 8 (1) (02): 15-36.

Pet-Soede, L., H. Cesar, and J. Pet. 2000. Blasting away: The economics of blast fishing on

248

indonesian coral reefs. Collected Essays on the Economics of Coral Reefs.CORDIO,

Kalmar: 77-84.

Plagányi, É. E. 2007. Models for an ecosystem approach to fisheries. Food & Agriculture Org.

Pollnac, R., M. Bavinck, and I. Monnereau. 2012. Job satisfaction in fisheries compared. Social

Indicators Research 109 (1) (10/01): 119-33.

Poos, JJ, JA Bogaards, FJ Quirijns, DM Gillis, and AD Rijnsdorp. 2010. Individual quotas,

fishing effort allocation, and over-quota discarding in mixed fisheries. ICES Journal of

Marine Science: Journal Du Conseil 67 (2): 323-33.

Potemra, J. T., S. L. Hautala, and J. Sprintall. 2003. Vertical structure of indonesian throughflow

in a large-scale model. Deep Sea Research Part II: Topical Studies in Oceanography 50

(12): 2143-61.

Quinn, N. 1978. Do Mfantse fish sellers calculate probabilities in their heads? American

Ethnologist 5:206-226.

Ramenzoni, V. C. 2013. Endenese fisheries: Exploratory findings on environmental perceptions,

fish effort, and overfishing in eastern indonesia. Ethnobiology Letters 4 : 39-51.

Ranyard, R., W. R. Crozier, and O. Svenson. 1997. Decision making: Cognitive models and

explanations. Vol. 1Psychology Press.

Rappaport, R. A. 1968. Pigs for the Ancestors: Ritual in the Ecology of a New Guinea People.

New Haven.

Reid, A. 1984. The pre-colonial economy of indonesia. Bulletin of Indonesian Economic Studies

20 (2): 151-67.

Reinhard, I. 1997. Food security, nutrition security, and poverty in central flores, Indonesia:

Situation after four years implementation of an integrated food security project (IFSP)Lit.

249

Reitz, E. J. 2004. "Fishing down the food web": A case study from St. Augustine, Florida,

USA. American Antiquity 69 (1): 63-84.

Reitz, E. J., I. R. Quitmyer, and R. A. Marrinan. 2009. What are we measuring in the

zooarchaeological record of prehispanic fishing strategies in the Georgia Bight, USA? The

Journal of Island and Coastal Archaeology 4 (1): 2-36.

Resosudarmo, B. P., and F. Jotzo. 2009. Development, resources, and environment in eastern

Indonesia. Working with Nature Against Poverty: Development, Resources and the

Environment in Eastern Indonesia.Singapore: ISEAS.

Roncoli, C., Ingram, K. and Kirsben, P. 2002. Reading the Rains: Local Knowledge and Rainfall

Forecasting Among Farmers of Burkina Faso. Society and Natural Resources 15:411-

430.

Roos, S. 1877. Iets over Endeh. Tijdschrift voor Indische Taal-, Land- en Volkenkunde, 24: 481-

582.

Roos, S. 1872. Bijdrage tot de kennis van taal, land en volk op het eiland soemba. Bruining &

Wijt.

Russell, S. D., and R. T. Alexander. 1996. The skipper effect debate: Views from a Philippine

fishery. Journal of Anthropological Research 52 (4): 433-60.

Ruttan, L. M., and P. H. Tyedmers. 2007. Skippers, spotters and seiners: Analysis of the “skipper

effect” in US menhaden (< i> Brevoortia spp.) purse-seine fisheries. Fisheries Research

83 (1): 73-80.

Sáenz–Arroyo, A., C. M. Roberts, J.Torre, and M. Cariño‐Olvera. 2005. Using fishers’

anecdotes, naturalists’ observations and grey literature to reassess marine species at risk:

The case of the gulf grouper in the gulf of california, mexico. Fish and Fisheries 6 (2): 121-

250

33.

Salas, S. and A. Charles. 2007. Are small-scale fishers profit maximizers?: Exploring fishing

performance of small-scale fishers and factors determining catch rates. Proceedings of the

60th Gulf and Caribbean Fisheries Institute.

Salas, S. and D. Gaertner. 2004. The behavioural dynamics of fishers: Management implications.

Fish and Fisheries 5 (2): 153-67.

Salas, S., U. R. Sumaila, and T. Pitcher. 2004. Short-term decisions of small-scale fishers

selecting alternative target species: A choice model. Canadian Journal of Fisheries &

Aquatic Sciences 61 (3) (03): 374-83.

Salas, S. 2000. Fishing Strategies of Small-Scale Fishers and their Implications. Dissertation.

Sall, J., A.Lehman, M.Stephens, and L.Creighton. 2012. JMP start statistics: A guide to statistics

and data analysis using JMPSAS Institute Inc.

Sareng Orin Bao, P. 1969. Nusa Nipa nama pribumi Nusa Flores: warisan purba. Nusa Indah,

Ende, Flores.

Satria, A. and Y. Matsuda. 2004. Decentralization of fisheries management in Indonesia, Marine

Policy 28:437–450

Satria, A. 2009. Ekologi politik nelayanPT LKiS Pelangi Aksara.

Scheibehenne, B., A. Wilke, and P. M. Todd. 2011. Expectations of clumpy resources influence

predictions of sequential events. Evolution and Human Behavior 32 (5): 326-33.

Semedi, P. 2001. Close to the Stone, Far from the Throne: the Story of a Javanese Fishing

Community, 1820s-1990s. Yogyakarta, Indonesia: Benang merah.

Shackeroff, J. M., L.M. Campbell, and L. B. Crowder. 2011. Social-ecological guilds: Putting

people into marine historical ecology. Ecology & Society 16 (1).

251

Shafir, E. and A. Tversky. 1995. Decision making.The MIT Press.

Silvano, R. AM, and A. Begossi. 2001. Seasonal dynamics of fishery at the piracicaba river

(brazil). Fisheries Research 51 (1): 69-86.

Silvestre, G., and D.Pauly. 1997. Management of tropical coastal fisheries in asia: An overview

of key challenges and opportunities. Paper presented at Workshop on Sustainable

Exploitation of Tropical Coastal Fish Stocks in Asia. ICLARM, Manila, Philippines.

Simon, H.A. 1957. Models of man – social and rational. New York: John Wiley and Sons.

Simon, H.A. 1955. A behavioral model of rational choice. The Quarterly Journal of Economics

69 (1): 99-118.

Simon, H. A. 1956. Rational choice and the structure of the environment. Psychological Review

63 (2): 129.

Simon, H.A. 1987. Satisficing. The New Palgrave: A Dictionary of Economics 4 : 243-5.

Simon, H. A. 1947. El Comportamiento Administrativo, Aguilar, Buenos Aires.

Smith, E.A. 1983. Anthropological applications of optimal foraging theory: A critical review.

Current Anthropology 24:625-651.

Smith, E.Al. and B. Winterhalder. 1992. Evolutionary ecology and human behaviorTransaction

Publishers.

Smith, E. A., R. L. Bettinger, C. A. Bishop, V. Blundell, E. Cashdan, M. J. Casimir, A. L.

Christenson, B. Cox, R. Dyson-Hudson, and B. Hayden. 1983. Anthropological applications

of optimal foraging theory: A critical review [and comments and reply]. Current

Anthropology: 625-51.

Snijders, T. A. B. 2011. Multilevel Analysis. Springer.

Snijders, T. A. B., and R. J. Bosker. 1999. Multilevel analysis : An introduction to basic and

252

advanced multilevel modeling. London ; Thousand Oaks, Calif. : Sage Publications, 1999.

Sopher, D. 1965. The sea nomads: a study of the maritime boat people of Southeast Asia.

Memoirs, No. 5. Singapore: National Museum.

Sosis, R.. 2001. Sharing, consumption, and patch choice on ifaluk atoll. Human Nature 12 (3):

221-45.

Sosis, R.. 2002. Patch choice decisions among Ifaluk fishers. American Anthropologist 104 (2):

583-98.

Soulié, J.-C., and O.Thébaud. 2006. Modeling fleet response in regulated fisheries: An agent-

based approach. Mathematical & Computer Modelling 44 (5) (09): 553-64.

Southon, M. 1995. The navel of the perahu. Australian Centre for Agricultural Research. The

Australian National University Canberra, ACT, Australia.

Sparre, P. and S. C Venema. 1992 Introduction to tropical fish stock assessment. part 1. Manual.

FAO Fisheries Technical Paper(306.1).

Srisurichan, S., N. Caputi, and J.Cross. 2005. Impact of lunar cycle and swell on the daily catch

rate of western rock lobster (panulirus cygnus) using time series modelling. New Zealand

Journal of Marine and Freshwater Research 39 (3): 749-64.

Stacey, Ne, S. Nurhakim, D. Nugroho, H. Soselisa, B. Resosudarmo, O. Kalis, J. Monteiro, J.

Prescott, J. Martin, and J. Karam. 2011. Socio-economic profile of the arafura and timor

seas. Report Prepared for the Arafura Timor Seas Ecosystem Action (ATSEA)

Program.Jakarta.

Staples, D., B. Satia, and PR Gardiner. 2004. A research agenda for small-scale fisheriesFAO,

Regional Office for Asia and the Pacific.

Stelzenmüller, V., F. Maynou, G. Bernard, G. Cadiou, M. Camilleri, R. Crec’hriou, G.Criquet,

253

M.Dimech, O. Esparza, and R. Higgins. 2008. Spatial assessment of fishing effort around

European marine reserves: Implications for successful fisheries management. Marine

Pollution Bulletin 56 (12): 2018-26.

Stephens, D. W. 2008. Decision ecology: Foraging and the ecology of animal decision making.

Cognitive, Affective, & Behavioral Neuroscience 8 (4): 475-84.

Stephens, D.W, and J.R Krebs. 1986. Foraging theory. Princeton: Princeton University Press 1

(10): 100.

Stevenson, B. C., and R. B. Millar. 2013. Promising the moon? Evaluation of indigenous and

lunar fishing calendars using semiparametric generalized mixed models of recreational catch

data. Environmental and Ecological Statistics 20 (4): 591-608.

Stocker, TF, D. Qin, GK Plattner, M. Tignor, SK Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex,

and PM Midgley. 2013. IPCC, 2013: Climate change 2013: The physical science basis.

contribution of working group I to the fifth assessment report of the intergovernmental panel

on climate change.

Symes, D., and J. Phillipson. 2009. Whatever became of social objectives in fisheries

policy? Fisheries Research 95 (1): 1-5.

Tabachnick, B. G., and LS Fidell. 2001. Multivariate statistics. Needham Heights, MA: Allyn &.

Teh, L. S. L, D. Zeller, A. Cabanban, LCL Teh, and U. R. Sumaila. 2007. Seasonality and

historic trends in the reef fisheries of Pulau Banggi, Sabah, Malaysia. Coral Reefs 26 (2):

251-63.

Teh, L. S. L, L. C. L. Teh, and U. R. Sumaila. 2011. Quantifying the overlooked socio-economic

contribution of small-scale fisheries in Sabah, Malaysia. Fisheries Research 110 (3): 450-8.

Teh, L., L. Teh, and M. Meitner. 2012. Preferred resource spaces and fisher flexibility:

254

Implications for spatial management of small-scale fisheries. Springer Science & Business

Media B.V, 40.

Teigen, K. H., M. Martinussen, and T. Lund. 1996. Conjunction errors in the prediction of

referendum outcomes: Effects of attitude and realism. Acta Psychologica 93 (1): 91-105.

Tewfik, A., and C. Béné. 2004. “The big grab”: Non-compliance with regulations, skewed

fishing effort allocation and implications for a spiny lobster fishery. Fisheries Research 69

(1) (8): 21-33.

Thorlindsson, T. 1988. The skipper effect in the Icelandic herring fishery. Human Organization

47 (3): 199-212.

Thorson, J. T., and J. Berkson. 2010. Evaluating single-and multi-species procedures to estimate

time-varying catchability functional parameters. Fisheries Research 101 (1): 38-49.

Todd, P. M. and G. Gigerenzer, G. 2003. Bounding rationality to the world. Journal of Economic

Psychology 24:143-165.

Todd, P. M., and G. Gigerenzer. 2012. Ecological rationality: Intelligence in the world. Oxford

University Press.

Tucker, B. 2007a. Applying behavioral ecology and behavioral economics to conservation and

development planning: An example from the Mikea Forest, Madagascar. Human Nature

18:190-208.

Tucker, B. 2007b. Perception of Interannual Covariation and Strategies for Risk Reduction

among Mikea of Madagascar. Human Nature 18:162-180.

Tucker, B., A. Huff, J. Tombo, P. Hajasoa, and C. Nagnisaha. 2011. When the wealthy are poor:

Poverty explanations and local perspectives in southwestern madagascar.American

Anthropologist 113 (2): 291-305.

255

Tule, P. 2004. Longing for the House of God, Dwelling in the House of the Ancestors. Local

Belief, Christianity, and Islam among the Kéo of Central Flores (Studia Instituti

Anthropos, Vol. 50). Fribourg: Academic Press.

Tule, P. 1998. The indigenous muslim minority group in Ma’undai (keo) of central flores:

Between the house of islam and the house of culture. Anthropologi Indonesia 56 : 68-85.

Tull, M. 2009. The History of Shark Fishing in Indonesia: A HMAP Asia Project Paper.

Tversky, A., and D. Kahneman. 1981. The framing of decisions and the psychology of choice.

Science (New York, N.Y.) 211 (4481) (Jan 30): 453-8.

Tversky, A. and D. Kahneman. 1986. Rational choice and the framing of decisions. Journal of

Business: S251-78.

Tversky, A. 1972. Elimination by aspects: A theory of choice. Psychological Review 79 (4): 281.

UNEP. 2008. Nellemann, C., Hain, S., and Alder, J. (Eds). February 2008. In Dead Water –

Merging of climate change with pollution, over-harvest, and infestations in the world’s

fishing grounds. United Nations Environment Programme, GRID-Arendal, Norway,

www.grida.no.

Uskul, A. K., S. Kitayama, and R. E. Nisbett. 2008. Ecocultural basis of cognition: Farmers and

fishermen are more holistic than herders. Proceedings of the National Academy of Sciences

of the United States of America 105 (25) (Jun 24): 8552-6.

van Densen, W. L. Theodorus. 2001. On the perception of time trends in resource outcome: Its

importance in fisheries co-management, agriculture and whalingUniversity of Twente.

van Oostenbrugge, H. J.A.E., J.P. Powell, J. P.G Smit, J.-J. Poos, S. Kraak, and E. F.C Buisman.

2008. Linking catchability and fisher behaviour under effort management. Aquatic Living

Resources 21 (03): 265-73.

256

van Oostenbrugge, J. A. E., E.J. Bakker, W. L.T. van Densen, M. A.M. Machiels, and P. A.M.

van Zwieten. 2002. Characterizing catch variability in a multispecies fishery: Implications

for fishery management. Canadian Journal of Fisheries and Aquatic Sciences 59 (6): 1032-

43.

van Oostenbrugge, J. A. E., J. J. Poos, W. L. T. van Densen, and M. A. M. Machiels. 2002. In

search of a better unit of effort in the coastal liftnet fishery with lights for small pelagics in

Indonesia. Fisheries Research 59 (1) (12/30): 43.

van Oostenbrugge, J. A. E., W. L. T. Densen, and M. A. M. Machiels. 2001. Risk aversion in

allocating fishing effort in a highly uncertain coastal fishery for pelagic fish, Moluccas,

Indonesia. Canadian Journal of Fisheries and Aquatic Sciences 58 (8): 1683-91.

van Oostenbrugge, J.A. E., W.L.T. Van Densen, and M.A.M Machiels. 2004. How the uncertain

outcomes associated with aquatic and land resource use affect livelihood strategies in

coastal communities in the central moluccas, indonesia. Agricultural Systems 82 (1): 57-91.

van Oostenbrugge, J.A.E. 2001. Uncertainty in daily catch rate in the light fisheries around

ambon and the lease islands: Characterisation, causes and consequences[Sl: sn].

Dissertation.

van Putten, I. E., S. Kulmala, O. Thébaud, N. Dowling, K. G. Hamon, T. Hutton, and Sean

Pascoe. 2012. Theories and behavioural drivers underlying fleet dynamics models. Fish and

Fisheries 13 (2): 216-35.

Van Suchtelen, B. C. C. M. M. 1921. Endeh (Flores). Mededeelingen van het Bureau voor de

Bestuurszaken der Buitengewesten, bewerkt door het Encyclopaedisch Bureau, Aflevering

26. Weltevreden: Uitgeverij Papyrus.

van Wouden, F. A. E. 1968. Types of social structure in Eastern Indonesia. Vol. 11.Nijhoff.

257

Van Zwieten, P.A.M, F.C. Roest, M.A.M. Machiels, and W.L.T .Van Densen. 2002. Effects of

inter-annual variability, seasonality and persistence on the perception of long-term trends in

catch rates of the industrial pelagic purse-seine fishery of northern Lake Tanganyika

(Burundi). Fisheries Research 54 (3): 329-48.

Varkey, D. A., C. H. Ainsworth, T. J. Pitcher, Yohanis Goram, and Rashid Sumaila. 2010.

Illegal, unreported and unregulated fisheries catch in Raja Ampat regency, Eastern

Indonesia. Marine Policy 34 (2): 228-36.

Vázquez-Rowe, I., and P. Tyedmers. 2013. Identifying the importance of the “skipper effect”

within sources of measured inefficiency in fisheries through data envelopment analysis

(DEA). Marine Policy 38 : 387-96.

Velázquez-Abunader, Iván, Silvia Salas, and Miguel A. Cabrera. 2013. Differential catchability

by zone, fleet, and size: The case of the red octopus (octopus maya) and common octopus

(octopus vulgaris) fishery in yucatan, mexico. Journal of Shellfish Research 32 (3): 845-54.

Wasserman, E. A., M. E. Young, and R.G. Cook. 2004. Variability discrimination in humans and

animals: Implications for adaptive action. American Psychologist 59 (9): 879.

Weber, E. U., and E. J. Johnson. 2009. Mindful judgment and decision making. Annual Review

of Psychology 60 : 53-85.

Weber, M. 1899. Introduction et description de l'expedition. Siboga-Expeditie , V. 1, E.J. Brill,

Leide, p. 152.

Weber, M. 2002. The protestant ethic and the spirit of capitalism: And other writingsPenguin.

Wei, W. W.S, and D. O. Stram. 1990. Disaggregation of time series models. Journal of the Royal

Statistical Society. Series B (Methodological): 453-67.

West, P. 2012. From modern production to imagined primitive: The social world of coffee from

258

Papua New Guinea. Duke University Press.

Westneat, D. and C. W. Fox. 2010. Evolutionary behavioral ecology. Oxford University Press.

Wheeler, M.C. and JL McBride. 2005. Australian-indonesian monsoon. Intraseasonal

Variability in the Atmosphere-Ocean Climate System: 125-73.

Wiadnya, D.G.R, P.J. Mous, R. Djohani, M.V. Erdmann, A. Halim, M. Knight, L. Pet-Soede,

and J.S Pet. 2005. Marine capture fisheries policy formulation and the role of marine

protected areas as tool for fisheries management in Indonesia. Mar.Res.Indonesia 30 : 33-

45.

Wiber, M., F. Berkes, A. Charles, and J. Kearney. 2004. Participatory research supporting

community-based fishery management. Marine Policy 28 (6): 459-68.

Wilberg, M. J., and J. R. Bence. 2006. Performance of time-varying catchability estimators in

statistical catch-at-age analysis. Canadian Journal of Fisheries and Aquatic Sciences 63

(10): 2275-85.

Wilke, A. and H. Clark Barrett. 2009. The hot hand phenomenon as a cognitive adaptation to

clumped resources. Evolution and Human Behavior 30 (3): 161-9.

Wilke, A. and P. M. Todd. 2010. Past and present environments: The evolution of decision

making. Psicothema 22 (1).

Wilke, A. 2006. Evolved Responses to an Uncertain World. Doctoral Dissertation.

Wilks, D. S. 2011. Statistical methods in the atmospheric sciences. Vol. 100Academic press.

Williams, A. J., A. C. Ballagh, G. A. Begg, C. D. Murchie, and L. M. Currey. 2008. Harvest

patterns and effort dynamics of indigenous and non-indigenous commercial sectors of the

Eastern Torres Strait reef line fishery. Continental Shelf Research 28 (16): 2117-28.

Williams, M.J. and D. Staples. 2010. Southeast Asian fisheries. In: Handbook of marine fisheries

259

conservation and management. Grafton, R.Q., Hillborn, R., Squires, D., Tait, M.,

Williams, M. (eds.). Oxford (United Kingdom), Oxford Univ. Press, pp. 243-257.

Winterhalder, B. 1981. Foraging strategies in the boreal forest: An analysis of Cree hunting and

gathering. In Hunter-Gatherer Foraging Strategies: Ethnographic and Archaeological

Analyses. B. Winterhalder and E. A. Smith (eds). Pp. 66-98. Chicago: University of

Chicago Press.

Winterhalder, B. 1996. A marginal model of tolerated theft. Ethology and Sociobiology 17: 37-

53.

Winterhalder, B. 2001. The behavioural ecology of hunter-gatherers. Hunter-Gatherers: An

Interdisciplinary Perspective 13 : 12.

Winterhalder, B. 2007. Risk and decision-making. Oxford Handbook of Evolutionary

Psychology: 433-46.

Winterhalder, B. and E. A. Smith. 2000. Analyzing adaptive strategies: Human behavioral

ecology at twenty-five. Evolutionary Anthropology 9:51-72.

Winterhalder, B., F. Lu, and B. Tucker. 1999. Risk-senstive adaptive tactics: Models and

evidence from subsistence studies in biology and anthropology. Journal of Archaeological

Research 7 (4): 301-48.

Wiyono, E. S., S. Yamada, E. Tanaka, T. Arimoto, and T. Kitakado. 2006. Dynamics of fishing

gear allocation by fishers in small-scale coastal fisheries of pelabuhanratu bay, indonesia.

Fisheries Management & Ecology 13 (3) (06): 185-95.

Yates, J. Frank, Ju Wei Lee, Hiromi Shinotsuka, Andrea L. Patalano, and Winston Sieck. 1998.

Cross-cultural variations in probability judgment accuracy: Beyond general knowledge

overconfidence? Organizational Behavior and Human Decision Processes 74 : 89-117.

260

Zeller, D., S. Booth, and D. Pauly. 2006. Fisheries contributions to the gross domestic product:

Underestimating small-scale fisheries in the pacific. Marine Resource Economics 21 (4).

Zeller, D., S.Booth, G. Davis, and D. Pauly. 2007. Re-estimation of small-scale fishery catches

for US flag-associated island areas in the western pacific: The last 50 years. Fishery

Bulletin105 (2): 266-77.

Zeller, D., W. Cheung, C. Close, and D. Pauly. 2009. Trends in global marine fisheries–a critical

view. Fisheries, Sustainability and Development. Royal Swedish Academy of Agriculture

and Forestry, Stockholm: 87-107.

Zerner, C. 1994. Through a green lens: The construction of customary environmental law and

community in Indonesia's Maluku islands. Law & Soc'y Rev. 28 : 1079.

Zerner, C. 2003. Culture and the question of rights: Forests, coasts, and seas in Southeast Asia

Duke University Press.

261

ANNEX A: Primary production

8.1. Spatial analyses and primary production

To estimate the potential productivity in tons per square kilometer of neritic and demersal shelf fisheries of the research area, Dalzell and Pauly’s (1989) log-linear models were applied.

The equations state:

and

2 with Py being the potential yield in tons km of shelf, PP the annual primary productivity in gCm2a-1, and D the average shelf depth in m.

Bathymetric data was obtained from the The General Bathymetric Chart of the Oceans

(GEBCO, GEBCO_08 Grid, version 20100927, http://www.gebco.net) housed online at the

British Oceanographic Data Centre. A portion of the file with coordinates was exported to a .cvs format file using The Gebco Grid Display Software Version 2.13. Through spatial analyses

(ArcGIS 10), the different depths were reclassified and converted into polygons representing depth classes. I used these depth polygons to calculate areas distant 45 km from the point of origin or departure port for fishing expeditions in Pulau Ende (coordinates: W 121.519460 S

8.879934). The threshold of 45 km was selected as the maximum distance fishing boats

(motorized) might travel in one single trip according to observations (see Figure 8.1).

The first depth class contained non bathymetric information for above surface level geographic features and was not considered in the analysis. The second class included depths

262 between 0.1 m and 50 m comprising the intertidal zone (neritic). The third class had depths between 50 m and 100 m (first subtidal zone) and the fourth between 100 and 200 m (second subtidal zone). Finally the fifth, sixth and seventh classes included depths between 200 m and

1000 m (mesopelagic), 1000 and 2000 m (bathypelagic), and 2000 m to maximum (hadal) respectively. Information on the water column stratification of the oceans of Eastern Indonesia, and particularly the Sawu was derived from relevant oceanographic sources (Wyrtki 1961,

Longhurst and Pauly 1987, Monk et al 1997). Classes (II, III and IV) were merged into the one to establish the depths of the continental shelf. The mean annual primary production for the area was estimated from satellite imagery MODIS r2013 as provided by Ocean Productivity Oregon

State University Website and available through the SeaDAS database from NASA within a

Carbon Based Production Model (CbPM; Westberry et al. 2008) for 2011 and the first ten months of 2012 by employing SeaDAS 7.0 software. The scale resolution was 4 km pixels.

8.2. Potential production and resource extraction Satellite data analysis showed that the area produces estimates around 0.603 +/- 185.43 gC·m-2·day-1. Table 1 includes the profile of depths for the region studied. Estimations predict that the annual potential for the region is about 5,900 tons in its best primary productivity.

Projections of low and mean productivity are 1800 tons and 3100 tons respectively. According to official statistics, the annual landing in 2011 was 3909 tons for pelagic and demersal species only

(total landing of 5212.3 tons). This indicates that reported yields are approximately 2000 tons below the best productive estimate of 5900 tons of fish. However, they exceed by 800 tons the mean estimate and by almost 2100 tons the low estimate. It is possible to conclude, then, that fish resources are at least intensively exploited or even overexploited in terms of potential production. Findings must be considered with caution as many of these species are highly migratory (Carangidae family), originating in spawning grounds outside the study area.

263

Moreover, there are some species that might reproduce more than once per year (Clupeidae family) or reach mature reproductive age in just a few months.

Figure 8.1: Reclassification of Depths

264

ANNEX B: Environmental Structures

9.1. Colwel Index of Predictability, Constancy and Contingency (Low 1990)

I calculated, following Low (1990) the Colwel index of predictability, constancy and contingency.

Table 9.1 Colwel Index for Ende

Predictability 0.529 Contingency -0.42 Constancy 0.95

This allowed me to observe that environmental variability in terms of monthly precipitation is patterned. Future studies will look at changes in a more fine grained approach by considering the onset of the rainy season.

9.2. Calculation of the Standard Precipitation Index (SPI)

The SPI calculation is based on the 30 year long-term precipitation record for Ende. I calculated the cumulative mean of precipitation trimonthly from 1980 until 2011 and then I converted those values to a gamma distribution (McKee et al. 1993, 1995). Because events of precipitation are positive (there cannot be negative values) and heavily skewed, a Pearson III distribution (or Gamma) reliably fits the data. I then transformed this value to an inverse normal distribution so that the mean and standard deviation of the dataset is zero and one, respectively.

Positive SPI values indicate wet condition greater than median precipitation, whereas negative values the dry condition less than median precipitation.

265

9.3. Results

I discovered that there is a change in the drought conditions starting 1990. It can be observed in the plotting of SPI values for the time series (figure 9.1). I also calculated a matched samples T Test (table/figure 9.3) that returned significant results for a comparison between the periods 1980-1990 and 1991-2011.

Standard Precipitation Index 3.00

2.00

1.00

0.00

-1.00

-2.00

-3.00

-4.00

Figure. 9.1: Standard Precipitation Index (1981-2011)

266

Table 9.2 Cumulative probabilities for various SPI values and possible interpretation of wet (or dry).

SPI Cumulative Probability Interpretation -3.0 0.0014 extremely dry -2.5 0.0062 extremely dry -2.0 0.0228 extremely dry (SPI < -2.0) -1.5 0.0668 severely dry (-2.0 < SPI < -1.5) -1.0 0.1587 moderately dry (-1.5 < SPI < -1.0) -0.5 0.3085 near normal 0.0 0.5000 near normal 0.5 0.6915 near normal 1.0 0.8413 moderately wet (1.0 < SPI < 1.5) 1.5 0.9332 very wet (1.5 < SPI < 2.0) 2.0 0.9772 extremely wet (2.0 < SPI) 2.5 0.9938 extremely wet 3.0 0.9986 extremely wet

267

Table/Figure 9.3 Comparison between matched means

Matched Pairs Difference: Column 2-Column 9

Column 2 0.33956 t-Ratio 6.08047 Column 9 -0.2378 DF 119 Mean Difference 0.57733 Prob > |t| <.0001* Std Error 0.09495 Prob > t <.0001* Upper 95% 0.76533 Prob < t 1.0000 Lower 95% 0.38932 N 120 Correlation 0.47635

268

ANNEX C

Month Rain Wind Fish Activities Name 1-2 Intense (40 West-North. Pelagic, Night and Hunger Barat days). Angin Mbohu, Sardines, noon fishing. events. Floods. Amburugazi, anchovies, Insects. Zera. coral fishes. Tongkol 3 Begins to East Flying fish. Harvest Pancaroba decrease maize and beans. 4-8 Southeast – Small tunas, Planting of Best fishing Tenggara South. skipjacks, manioc season. Angin wete sailfishes, big gardens and Living big. (timur) tunas, other crops mantas, etc. (maize). Kombong.

9 East Sardines, Pancaroba sailfishes. 10 Should begin East Sardines, Afternoon Barat Daya transitioning sailfishes. and night fishing. 11 12 Floods Squids.

Figure 10.1 Fishing Calendar

269

Date Hours depart- return Fish name and quantity Reason for not fishing

Figure 10.2: Fishing logs

270

Figure 10.3 Kulavu or magical objects

271

Figures 10.4 and 5: Endenese Landscapes. Wet season in hinterlands, Village of Braai. Pictures by Victoria Ramenzoni.

272

Figures 10. 6 and 7: Endenese Landscapes. Corals around Pulau Ende and Meti Numba Village during neap tide. Pictures by Victoria Ramenzoni