A Multi-Scale Analysis of Jaguar (Panthera Onca) and Puma (Puma Concolor) Habitat Selection and Conservation in the Narrowest Section of Panama

Home , Jaguar, Puma (genus), San Blas Islands

Antioch University AURA - Antioch University Repository and Archive Student & Alumni Scholarship, including Dissertations & Theses Dissertations & Theses

2019 A Multi-Scale Analysis of Jaguar (Panthera onca) and Puma (Puma concolor) Habitat Selection and Conservation in the Narrowest Section of Panama. Kimberly A. Craighead Antioch University of New England

Follow this and additional works at: https://aura.antioch.edu/etds Part of the Animal Studies Commons, Biodiversity Commons, Environmental Studies Commons, Natural Resources and Conservation Commons, Other Animal Sciences Commons, Other Environmental Sciences Commons, and the Population Biology Commons

Recommended Citation Craighead, Kimberly A., "A Multi-Scale Analysis of Jaguar (Panthera onca) and Puma (Puma concolor) Habitat Selection and Conservation in the Narrowest Section of Panama." (2019). Dissertations & Theses. 474. https://aura.antioch.edu/etds/474

This Dissertation is brought to you for free and open access by the Student & Alumni Scholarship, including Dissertations & Theses at AURA - Antioch University Repository and Archive. It has been accepted for inclusion in Dissertations & Theses by an authorized administrator of AURA - Antioch University Repository and Archive. For more information, please contact [email protected], [email protected].

Department of Environmental Studies

DISSERTATION COMMITTEE PAGE

The undersigned have examined the dissertation entitled:

A MULTI-SCALE ANALYSIS OF JAGUAR (PANTHERA ONCA) AND PUMA (PUMA CONCOLOR) HABITAT SELECTION AND CONSERVATION IN THE NARROWEST SECTION OF PANAMÁ

Presented by Kimberly A. Craighead Candidate for the degree of Doctor of Philosophy, and hereby certify that it is accepted*

Committee Chair: Beth A. Kaplin, Ph.D. Professor, Department of Environmental Studies, Antioch University New England

Committee Member: Peter A. Palmiotto, D.F. Chair, Department of Environmental Studies, Antioch University New England

Committee Member: Marcella J. Kelly, Ph.D. Professor, Department of Fish & Wildlife Conservation, Virginia Tech

Defense Date: April 2019

*Signatures are on file with the Registrar’s Office, Antioch University New England.

A MULTI-SCALE ANALYSIS OF JAGUAR (PANTHERA ONCA) AND PUMA (PUMA CONCOLOR) HABITAT SELECTION AND CONSERVATION IN THE NARROWEST SECTION OF PANAMÁ

Kimberly A. Craighead

BSc. California State University, Hayward, CA, 2004

A dissertation submitted in partial fulfillment

of the requirements for the degree of

Doctor of Philosophy

Environmental Studies

Antioch University New England

Keene, New Hampshire, USA

2019

DEDICATION

For my dad, my biggest fan. I love you Pops.

ACKNOWLEDGEMENTS

This dissertation has been quite a journey. I have many people to thank, but first and foremost is my partner Milton Yacelga. Without your love and endless support, this dissertation would not have been possible. You’ve stuck by me through thick and thin, back and forth to Panamá and trekked umpteen miles over hill and dale checking camera traps with me, and for me…I can’t thank you enough!

My supportive committee has shown me how to advance my scholarship. I am grateful to my advisor Beth Kaplin, who has always had faith in my work and my process. Beth stood by my decisions to change research plans along the way and gave me the insight and confidence to carry on. Peter Palmiotto, your patience and compliments have been greatly appreciated!

Marcella Kelly, thank you for sharing your knowledge on camera traps, and for your flexibility.

Thank you all for providing me with the space to follow my passion.

I want to acknowledge all of my supportive friends. Sue Townsend you were a guiding light for me when I was on my puma track, still you are a wise mentor to me. Thank you for sharing your thoughts and opinions and for the tough love I needed along the way. I want to give special thanks to my friend and classmate Michele Losee who has been by my side from the start of this journey, I am so appreciative of your companionship. My good friend Lynne, thanks for driving me to campus and for lending me your car, because of you I was able to get to class! Dina, you have been my sounding board, our walks around the lake and your understanding of this process has been most appreciated. Amy Gotliffe, thanks for believing in my work.

I have sought the expertise of many who have helped guide me through this dissertation. Sam

Cushman, you’re a brilliant and collaborative scientist, thanks for letting me stay in your home while learning how to use Random Forests. Ho Yi Wan, your technical mind is something to envy! Thank you for pulling me through the analyses and for sharing your expertise. Bobby

Vogt, I am so thankful of your GIS skills, your help has been indispensable.

In Panamá, I thank the Mamoní Valley Preserve who graciously provided logistical support for my project. Nico, you are a star, always there to pick up the slack for me. I can’t thank my field team enough, Gabriel, Polo, Virgilio, Tomas and Modesto. These men are super-human and vital to the success of this research. Without their incredible knowledge of the forest and abilities to survive the harsh jungle environment, this project could not have existed. I want to thank you for keeping me alive and able to write this dissertation today!

This study was supported by Mamoní Valley Preserve, Tom Kaplan’s Panthera Foundation, the

Liz Claiborne and Art Ortenberg Foundation, The Laney Thornton Foundation and The

Shanbrom Foundation for their belief in my research. I want to thank Evan Shanbrom for his continued support after spending time with me in the jungle. I would also like to thank Idea Wild for contributing field gear, and I especially want to thank BioMaAs for their incredible generosity. Cullen Wilkerson, you made it possible to spend time away from my consulting work, thank you for having so much confidence in my project.

Immense gratitude goes out to my mom and dad. You have been endlessly supportive of me throughout my life. You’ve put up with my adventurousness and dedication to wildlife and travel; thank you for being there for me, no matter which path I chose.

iii

ABSTRACT

Over the past two centuries, large terrestrial carnivores have suffered extreme population declines and range contractions resulting from the synergistic anthropogenic threats of land-use change and indirect effects of climate change. In Panamá, rapid land use conversion coupled with climate change is predicted to negatively impact jaguar (Panthera onca) and puma (Puma concolor). This dissertation examined the environmental variables and scales influencing jaguar and puma habitat selection by season (annual, wet, and dry), using multi-scale optimized habitat suitability models and a machine-learning algorithm (Random Forests), in the narrowest section of Panamá. The models derived from the data of an intensive camera trapping effort (2016–

2018) captured a wide spectrum of ecological relationships for the sympatric felid species.

Jaguar habitat selection was limited by secondary forest at a broad scale (home range), suggesting that jaguars preferred primary forest. Therefore, the persistence of primary forest in the narrowest section of the country is key for the long-term survival of the species. Pumas incorporated primary forest at a fine scale (patch) and agropecuary (agriculture and livestock) at broad scales (home range), providing evidence of the plasticity and adaptability of the species to a diverse range of landscapes. Seasonal differences in habitat suitability were evident for both species which is most likely related to prey availability. The models also provided a set of seasonal habitat suitability maps (annual, wet season and dry season) from which spatial information on jaguar and puma distribution are presented. This study improves our understanding of species-environment relationships and habitat selection of jaguars and pumas in eastern Panamá, and contributes to the growing number of studies that demonstrate the strong conceptual and inferential advantages of a multi-scale approach. Further, I examined the

implications of the potential relocation of the Guna Yala Indigenous communities from the San

Blas Archipelago, Panamá, to a critical forest region on the mainland. My review highlights the interconnectedness of climate change, Indigenous people’s migration in response to climate change, and the implications for jaguar conservation in the region.

Keywords: distribution, habitat suitability, Random Forests, climate change, Guna Yala

This dissertation is available in open access at AURA, http://aura. antioch. edu/ and OhioLINK ETD Center, https:// etd.ohiolink.edu/etd.

TABLE OF CONTENTS

ABSTRACT ...... iv

LIST OF FIGURES ...... viii

LIST OF TABLES ...... x

LIST OF APPPENDICES...... xi

Chapter I: Introduction ...... 1

Dissertation overview ...... 5

Chapter II: Framed by the Forest: Scale-dependent Seasonal Habitat Selection by Jaguars

(Panthera onca) in Eastern Panamá ...... 12

Abstract ...... 12

Introduction ...... 13

Methods...... 17

Results ...... 30

Discussion ...... 41

Appendices ...... 59

Chapter III: Path of the Puma (Puma concolor): A Multi-Scale Analysis of Seasonal Habitat

Selection in the Narrowest Section of Panamá ...... 68

Abstract ...... 68

Introduction ...... 69

Methods...... 72

Results ...... 85

Discussion ...... 97

Appendices ...... 111

Chapter IV: Jaguars, Connectivity, and Climate Change in the Comarca Guna Yala, Panamá . 120

Abstract ...... 120

Introduction ...... 120

Biogeographic setting of The Republic of Panamá ...... 125

The Indigenous people of Guna Yala, past and present ...... 128

vii

LIST OF FIGURES

Figure

1. Location of study area …...………………………………………………………………21

2. Location of the study area indicating land cover types in eastern Panamá ….………….27

3. Variable importance plot for the predictor variables from Random Forests classifications used for predicting presence of jaguars in the study area ...……………………………..33

4. Partial dependency plots representing the marginal effect of each variable included in the Annual Random Forests jaguar model………..………………………………...………..34

5. Partial dependency plots representing the marginal effect of single variables included in the Wet Season Random Forests jaguar model …….……………………………………36

6. Partial dependency plots representing the marginal effect of single variables included in the Dry Season Random Forests jaguar model ………………………….………………37

7. Annual habitat suitability map showing the predicted occurrence of jaguars based on multi-scale habitat modeling in Panamá ………………………………………………...39

8. Wet season habitat suitability map showing the predicted occurrence of jaguars based on multi-scale habitat modeling in Panamá ………………………………………………...40

9. Dry season habitat suitability map showing the predicted occurrence of jaguars based on multi-scale habitat modeling in Panamá ………………………………………………...41

10. Location of study area …...………………………………………………….…………...76

11. Location of the study area indicating land cover types in eastern Panamá …………..…82

12. Variable importance plot for the predictor variables from Random Forests classifications used for predicting presence of pumas in the study area ...……………………..……….89

13. Partial dependency plots representing the marginal effect of each variable included in the Annual Random Forests puma model ………………………………………...…..……..90

14. Partial dependency plots representing the marginal effect of each variable included in the wet season Random Forests puma model ………………………………………...……..92

15. Partial dependency plots representing the marginal effect of each variable included in the dry season Random Forests puma model ………………………………………...……...93

viii

16. Annual habitat suitability map showing the predicted occurrence of pumas based on multi-scale habitat modeling in Panamá ………………………………………………...95

17. Wet season habitat suitability map showing the predicted occurrence of pumas based on multi-scale habitat modeling in Panamá ………………………………………………...96

18. Dry season habitat suitability map showing the predicted occurrence of pumas based on multi-scale habitat modeling in Panamá ………………………………………………...97

19. Location of the Comarca Guna Yala and the offshore islands of the San Blas Archipelago in Panamá ………………………………………………………………………………123

LIST OF TABLES Table

1. Description of variables used in the multi-scale jaguar habitat selection model developed for the study area ………………………………………………………………………...26

2. Optimal scales for the most important variables found in the annual, wet season, and dry season Random Forests jaguar models………………………………….……………….32

3. Description of variables used in the multi-scale puma habitat selection model developed for the study area ………………………………………………………………………...81

4. Optimal scales for the most important variables found in the annual, wet season, and dry season Random Forests puma models ….…………………………...…………………..87

LIST OF APPPENDICES Appendix

A. Examples of the three-land tenure areas Narganá Protected Wildland, Chagres National Park and the MVP ………………………………………………………………….59 – 61

B. Twenty-one predictor variables associated with the jaguar multi-scale habitat suitability models……………………………………………………………………………………62

C. Annual model OOB error rate table for jaguars ………………………………………....63

D. Wet season model OOB error rate table for jaguars……………………………………..64

E. Dry season model OOB error rate table for jaguars …………………………………….65

F. Photographs taken of jaguars by camera traps in the study area …………………..66 – 67

G. Examples of the three-land tenure areas Narganá Protected Wildland, Chagres National Park and the MVP ……………………………………………………………….111 – 113

H. Twenty-one predictor variables associated with the puma multi-scale habitat suitability models …..……………………………………………………………………………...114

I. Annual model OOB error rate table for pumas ………………………………………...115

J. Wet season model OOB error rate table for pumas…………………………………….116

K. Dry season model OOB error rate table for pumas …………………………………….117

L. Photographs taken of pumas by camera traps in the study area ……………...... 118 – 119

Chapter I: Introduction

Large terrestrial carnivores are some of the Earth’s most revered and iconic species.

Paradoxically, they are some of the most vulnerable taxa, 64% are threatened with extinction and

80% are in decline (Wolf & Ripple 2018). The rapid expansion of human populations and the conversion of natural habitats conflict with the biological needs of these charismatic species.

Worldwide, anthropogenic activities responsible for large terrestrial carnivore decline include habitat loss and fragmentation, reduction of prey, human persecution, illegal trade, and now, climate change.

Increasing human populations are linked to carnivore declines, geographic range reductions and changes in carnivore behavior (Woodroffe 2000; Ripple et al. 2014; Bruskotter & Wilson 2014).

The presence of humans coupled with changes in habitat structure often alters spatio-temporal habitat selection and the availability of resources (e.g., prey), which in turn, elicits changes in a species behavior to procure such resources (Wilmers et al. 2013; Smith et al. 2015; Weinberger et al. 2016; Suraci et al. 2019). In this context, a thorough understanding of the relationships between a species and its habitat is vital to large carnivore conservation (Tilman et al. 1994;

Haskell et al. 2013).

Animals are known to select habitats that maximize fitness by optimizing access to food, mates and other resources (Klaassen & Broekhuis 2018). It is commonly assumed that species occupy habitats that most suit their dietary and reproductive needs, if the benefits received outweigh the risks from competitors or other predators (Manly et al. 2007). In addition to prey abundance, the

2 physical characteristics of a habitat and the landscape context are also important correlates of habitat selection.

It is well recognized that habitat selection is an inherently scale-sensitive process (Rettie &

Messier 2000; Mayor et al. 2009). Since a multi-scale approach provides the most comprehensive characterization of habitat patterns, several studies have used multiple rather than single scales of analysis to determine at which scales an organism interacts with their environment (Wiens 1989; Rettie & Messier 2000; Mayor et al. 2009). To measure habitat selection at multiple scales, the orders of habitat selection as described by Wiens (1973) and

Johnson (1980) are often followed. In a hierarchical framework, the first order of selection is considered a population’s geographic range, the second order pertains to local site selection (i.e., a species home range), the third order addresses land use patterns within a home range (i.e., habitat patches), and the fourth order includes the resources selected within a habitat patch (e.g., food). A broad scale typically refers to the first and second order of habitat selection (geographic or home range) and a fine scale refers to the third or fourth orders (habitat patch and resources).

Understanding how habitat preferences by large carnivores shift across scales in a mosaic of human-altered and naturally occurring land cover is pivotal for characterizing the complexities of habitat selection. For instance, Apps et al. (2001) point out that broad-scale modeling can inform our understanding at a regional level, while fine-scale modeling may identify preferred topographic features or important habitat characteristics. Further, analyses of habitat selection at broad scales may reflect broader limiting factors (Rettie & Messier 2000; McLoughlin et al.

2002), while fine scale habitat selection analyses may reveal limiting resources within a habitat type (Hutto 1985).

Statistical modeling tools available to ecologists to model species’ distributions have increased at a rapid pace (e.g., Guisan & Thuiller 2005; Elith et al. 2006; Guillera-Ariota et al. 2015), as have the number of habitat suitability models (sometimes called species distribution models or SDM) published in the literature (e.g., Kanagaraj et al. 2013; Guillera-Ariota et al. 2015). In general,

SDM are empirical models that aim to capture the complexities and uncertainties underlying the biological mechanisms that drive species distribution and abundance (Kanagaraj et al. 2013).

Typically, these models estimate habitat preferences or predict distributions by correlating species locations with relevant environmental covariates (Guisan & Thuiller 2005; Guillera-

Ariota et al. 2015). The quality and distribution of species presence-absence records and the most relevant environmental predictors used, can influence the reliability of these models (Araújo &

Guisan 2006; Guisan et al. 2006).

Since habitat selection is inherently scale dependent, and most studies on wildlife habitat do not incorporate a multi-scale framework, which produces stronger inferences than single-scale models by identifying the operative scale. There is a need for a wider adoption of formal scale optimization of organism response to environmental variables (McGarigal et al. 2016).

Herein, I use a multi-scale optimized approach using the Random Forests method to investigate the habitat selection of two sympatric predators in the Americas, the jaguar (Panthera onca) and the puma (Puma concolor). The data for these analyses were taken from a two-year camera trap study focused on the collection of presence-absence locations of these species. My main goal is to gain ecological insight into the environmental factors and associated scales that determine species distribution, and identify areas where these species can occur across the human- dominated landscape in the narrowest section of eastern Panamá.

Similar to other Latin American countries, Panamá is rapidly succumbing to forest loss and fragmentation primarily due to human population growth and development. Panamá lost 14.3% of its forest cover between 1990 and 2010, leaving 43.7% still forested (FAO 2010). Over the last half century, the landscape on the east side of Panamá has experienced large-scale land use modification with expanding agriculture and deforestation around protected areas. The encroachment into wildlands has resulted in landscape fragmentation, which threatens large carnivore populations (i.e., jaguar and puma) by reducing the expansive landscapes and resources they require (Petracca et al. 2014). From that perspective, dispersal for these apex predators along the narrow Isthmus of Panamá may become problematic, as new human development creates further range contractions and consequential human-carnivore conflict.

As top predators in the Neotropics, jaguars often live in sympatry with pumas, sharing similar life history traits, behaviors and threats (Sunquist & Sunquist 2002; Palomares et al. 2016; de la

Torre et al. 2017; Gonzales-Borrajo et al. 2017). The jaguar is the largest cat in the Americas

(Nowell & Jackson 1996). Jaguars are a keystone species, considered near threatened throughout their range (Quigley et al. 2017). The two felids overlap in range from northern Mexico to northern Argentina. Similarly, jaguars and pumas have suffered substantial range reductions, but in many areas, pumas persist where jaguars have disappeared (i.e., United States, Chile)

(Sanderson et al. 2002; Sunquist & Sunquist 2002; De Angelo et al. 2011). Pumas, may therefore, be more resilient to human pressures and habitat modification than jaguars.

While jaguars have been extirpated from more than 50% of their historic geographic range

(Rabinowitz & Zeller 2010; Medellin et al. 2016; Quigley et al. 2017) and are considered near threatened (Quigley et al. 2017), the puma is listed as a species of Least Concern, with some subpopulations considered threatened or in decline (Nielsen et al. 2015). Yet, the puma is the

5 most wide-ranging terrestrial mammal in the Western Hemisphere, ranging from Canada through the Americas to the southernmost point of Chile (Sunquist & Sunquist 2002; Nielsen et al. 2015).

Pumas are considered more habitat generalists than jaguars, known to use open habitats such as converted landscapes (e.g., pastures or croplands) (Foster 2008; Fort 2016). The factors threatening these large sympatric felids make them species of conservation concern (Crooks

2002; Weber et al. 2010; De Angelo et al. 2011).

Dissertation overview

The goal of this study is to improve our understanding of jaguar and puma movement in eastern

Panamá to create more informed conservation management plans. As little is known about jaguar and puma habitat selection and their response to anthropogenic pressures in this region, I aim to identify and fill the gaps in the current knowledge by determining the environmental and anthropogenic variables and associated scales influencing jaguar and puma habitat selection. The distribution and habitat suitability for jaguars and puma are presented in separate chapters for each species to provide a thorough examination of the factors that determine the occurrence and ecological niche of each felid species. The core objectives of my study are to: (1) elucidate jaguar and puma habitat selection within this landscape to inform conservation management decisions, (2) identify the variables that most strongly influence habitat selection by jaguars and pumas, (3) identify the scale at which each of these variables is most important, (4) determine if there is a seasonal difference in habitat selection, (5) create seasonal (annual, wet and dry) habitat suitability maps for both the jaguar and puma in the study area, (6) posit the effects of climate change on the threatened jaguar population in eastern Panamá.

Common to Chapters II and III is the use of a multi-scale optimized approach to develop habitat suitability maps that model the probability of occurrence for each of these two guild members based on environmental and anthropogenic factors. I use presence-absence data taken from camera traps over a period of two years in eastern Panamá. The habitat suitability maps generated from these data include a predicted extent of jaguar and puma distribution of approximately 100 km outside the boundaries of the study area. The study area includes contiguous sections of three distinct areas under different tenure regimes, and therefore different dynamics of land cover and land use history: (1) Narganá Protected Wildland, (2) Chagres

National Park, and (3) The Mamoní Valley Preserve (Figure 1). Together, the study area comprises ~200 km2. The terrain in this section is rugged with altitudes ranging from 400 m to

1,000 m a.s.l. Annual average precipitation during the wet season is ~ 4000 mm and reduces to

~700 mm in the dry season.

Chapter II examines a set of predictor variables (e.g., land cover, topographic, hydrologic) across a gradient of 30 spatial scales (500 m– 15000 m) to identify the characteristic scale of each variable (Bellamy et al. 2013; McGarigal et al. 2016) at which jaguars select habitat. I predicted that using a multi-level, scale-optimized model approach would show that jaguars avoid human settlements and prefer primary forest habitat at broad scales. Further, considering that spatial seasonal distributions are species specific, I expected a reciprocal jaguar-specific distributional variation with predictor variables. Thus, I predicted that jaguars will exhibit a behavioral plasticity in terms of habitat requirements during the wet and dry season. Moreover, I predicted that jaguars would show differential habitat use at different spatial scales within each season, responding to hydrologic features at a fine scale in the dry season.

Chapter III investigates puma habitat selection using a set of predictor variables (e.g., land cover, topographic, hydrologic) across a gradient of 30 spatial scales (500 m– 15000 m) to identify the characteristic scale of each variable (Bellamy et al. 2013; McGarigal et al. 2016). I predicted that pumas would occur in areas of agropecuary (i.e., agriculture, livestock) in the wet season at a broad scale. Further, I predicted that pumas would exhibit a behavioral plasticity in terms of habitat requirements during the wet and dry season (i.e., differential habitat use at different spatial scales).

In Chapter IV, I explore the links between the slow-onset impacts of sea level rise, indigenous people’s displacement and the implications for jaguar conservation. I first describe the biogeographic setting of Panamá by highlighting its origin and impact on the biodiversity in the

Americas. I then review the history of the Guna Yala, who formerly lived in the tropical forest on the east side of the Isthmus of Panamá, before moving to the San Blas Islands along the

Caribbean coast. The socio-cultural and economic consequences of resettlement have set the stage for the impending migration of an estimated 40,000 people back to the mainland portion of the Comarca, one of the most extensive intact forests in Panamá. I end the chapter by outlining the potential impacts of this move on jaguars and discuss future recommendations and key challenges associated with development in a protected area and high biodiversity.

Chapter V includes a brief synthesis of findings, implications of the research, and considers future research needs.

Literature Cited

Apps, C. D., B. N. McLellan, T. A. Kinley and J. P. Flaa. 2001. Scale-Dependent Habitat Selection by Mountain Caribou, Columbia Mountains, British Columbia. The Journal of Wildlife Management 65:65–77.

Araújo, M. B. and A Guisan. 2006. Five (or so) challenges for species distribution modelling. Journal of Biogeography 33:1677–1688.

Bellamy, C., C. Scott and J. Altringham. 2013. Multiscale, presence‐only habitat suitability models: fine‐resolution maps for eight bat species. Journal of Applied Ecology 50:892–901.

Breiman, L., J. H. Friedman, R. A. Olshen and C. J. Stone. 1984. Classification and regression trees. Wadsworth and Brooks/Cole, Monterey, California, USA.

Bruskotter, J. T. and R. S. Wilson. 2014. Determining where the wild things will be: Using psychological theory to find tolerance for large carnivores. Conservation Letters 7:158–165.

Caso, A., C. Lopez-Gonzalez, E. Payan, E. Eizirik, T. de Oliveira, R. Leite-Pitman, M. Kelly and C. Valderrama. 2008. Puma concolor. The IUCN Red List of Threatened Species. Available at http://www.iucnredlist.org.

De Angelo, C., A. Paviolo and M. Di Bitetti. 2011. Differential impact of landscape transformation on pumas (Puma concolor) and jaguars (Panthera onca) in the Upper Paraná Atlantic Forest. Diversity and Distributions 17:422–436. de la Torre, J. A., J. M. Núñez and R. A. Medellín. 2017. Habitat availability and connectivity for jaguars (Panthera onca) in the Southern Mayan Forest: Conservation priorities for a fragmented landscape. Biological Conservation 206:270–282.

Elith, J., C. H. Graham and R. P. Anderson et al. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29:129–151.

Food and Agriculture Organization (FAO). 2010. Global Forest Resources Assessment. Food and Agricultural Organization of the United Nations, Rome, Italy.

Gonzalez-Borrajo, N., J. V. López-Bao and F. Palomares. 2017. Spatial ecology of jaguars, pumas, and ocelots: a review of the state of knowledge. Mammal Review 47:62–75.

Guillera‐Arroita, G. et al. 2015. Is my species distribution model fit for purpose? Matching data and models to applications. Global Ecology and Biogeography 24:276–292.

Guisan, A and W, Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993–1009.

Guisan, A. et al. 2006. Making better biogeographical predictions of species distributions. Journal of Applied Ecology 43:386–392.

Guisan, A., R. Tingley and J. B. Baumgartner et al. 2013. Predicting species distributions for conservation decisions. Ecology Letters 16:1424–1435.

Haskell, D. E., C. R. Webster, D. J. Flaspohler and M. W. Meyer. 2013. Relationship between Carnivore Distribution and Landscape Features in the Northern Highlands Ecological Landscape of Wisconsin. The American Midland Naturalist 169:1–16.

Hutto, R. L. 1985. Habitat selection by nonbreeding, migratory land birds. Pages 455–476 in Habitat selection in birds. M.L. Cody, editor. Academic Press, San Diego.

Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65–71.

Kanagaraj, R., T. Wiegand, A. Mohamed and S. Kramer-Schadt. 2013. Modelling species distributions to map the road towards carnivore conservation in the tropics. The Raffles Bulletin of Zoology 28:85–107.

Kelly, M. J., A. J. Noss, M. S. Di Bitetti, L. Maffei, R. L. Arispe, A. Paviolo, C. D. De Angelo and Y. E. Di Blanco. 2008. Estimating puma densities from camera trapping across three study sites: Bolivia, Argentina, and Belize. Journal of Mammalogy 89:408–418.

Klaassen, B. and F. Broekhuis. 2018. Living on the edge: Multiscale habitat selection by cheetahs in a human-wildlife landscape. Ecology and Evolution 8:7611–7623.

Laundré, J. W. and L. Hernández. 2010. What we know about pumas in Latin America. Pages 76–90 in Cougar: Ecology and Conservation. M. Hornocker and S. Negri editors. The University of Chicago Press, Chicago and London.

Manly, B. F. J., L. L. McDonald., D. L. Thomas, T. L. McDonald and W. P. Erickson. 2007. Resource Selection by Animals: Statistical Design and Analysis for Field Studies, second edition. Kluwer Academic Publishers, Boston, MA, USA.

Mayor, S. J., D. C. Schneider, J. A. Schaefer and S. P. Mahoney. 2009. Habitat selection at multiple scales. Ecoscience 16:238–247.

McGarigal, K., H. Y. Wan, K. A. Zeller, B. C. Timm and S. A. Cushman. 2016. Multi-scale habitat selection modeling: a review and outlook. Landscape Ecology 31:1161–1175.

McLoughlin, P. D., R. L. Case, R. J. Gau, Cluff HD, R. Mulders and F. Messier. 2002. Hierarchical habitat selection by barren- ground grizzly bears in the central Canadian Arctic. Oecologia 132:102–108.

Medellín, R. A., J. A. de la Torre, C. Chávez, H. Zarza and G. Ceballos. 2016. El Jaguar en el Siglo XXI: La Perspectica Continental. Fondo de Cultura Económica, Universidad Nacional Autónoma de México, Ciudad de México.

Nielsen, C., D. Thompson, M. Kelly and C. A. Lopez-Gonzalez. 2015. Puma concolor. The IUCN Red List of Threatened Species. Available at https://www.iucnredlist.org/species/18868/97216466.

Palomares, F., N. Fernández, S. Roques, C. Chávez, L. Silveira, C. Keller and B. Adrados. 2016. Fine-scale habitat segregation between two ecologically similar top predators. PloS One 11: e0155626.

Petracca, L. S., S. Hernández-Potosme, L. Obando-Sampson, R. Salom-Pérez, H. Quigley, H. S. Robinson. 2014. Agricultural encroachment and lack of enforcement threaten connectivity of range-wide jaguar (Panthera onca) corridor. Journal for Nature Conservation 22:436– 444.

Quigley, H., R. Foster, L. Petracca, E. Payan, R. Salom and B. Harmsen. 2017. Panthera onca (errata version published in 2018). The IUCN Red List of Threatened Species. Available at http://www.iucnredlist.org.

Rabinowitz, A., K. A. Zeller. 2010. A range-wide model of landscape connectivity and conservation for the jaguar, Panthera onca. Biological Conservation 143:939–945.

Rettie, W. J. and F. Messier. 2000. Hierarchical habitat selection by woodland caribou: its relationship to limiting factors. Ecography 23:466–478.

Ripple, W. J., et al. 2014. Status and ecological effects of the world’s largest carnivores. Science 343:1241484.

Sanderson, E. W., K. H. Redford, C. B. Chetkiewicz, A. Rodrigo, A. R. Rabinowitz, J. G. Robinson and A. B. Taber. 2002. Planning to save a species: the jaguar as a model. Conservation Biology 16:58–72.

Smith, J.A., Y. Wang and C. C. Wilmers. 2015. Top carnivores increase their kill rates on prey as a response to human-induced fear. Proceedings of the Royal Society B: Biological Sciences 282:20142711.

Sunquist, M. and F. Sunquist. 2002. Wild Cats of the World. The University of Chicago Press, Chicago and London.

Suraci, J. P., L. G. Frank, A. Oriol‐Cotterill, S. Ekwanga, T. M. Williams, C. C. Wilmers. 2019. Behavior‐specific habitat selection by African lions may promote their persistence in a human‐dominated landscape. Ecology:e02644.

Tilman, D., R. M. May, C. L. Lehman and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371:65–66.

Weinberger, I. C., S. Muff, A. de Jongh, A. Kranz and F. Bontadina. 2016. Flexible habitat selection paves the way for a recovery of otter populations in the European Alps. Biological Conservation 199:88–95.

Wiens, J. A. 1989. Spatial Scaling in Ecology. Functional Ecology 3:385.

Wilmers, C. C., et al. 2013. Scale dependent behavioral responses to human development by a large predator, the puma. PLoS One 8:e60590.

Wolf, C. and W. J. Ripple. 2018. Rewilding the world’s large carnivores. Royal Society Open Science 3:160252.

Woodroffe, R. 2000. Predators and people: using human density to interpret declines of large carnivores. Animal Conservation 1:165–173.

Chapter II: Framed by the Forest: Scale-dependent Seasonal Habitat Selection by Jaguars (Panthera onca) in Eastern Panamá

Abstract

As apex predators, jaguars (Panthera onca) occur at low population densities and have extensive

home ranges, making them susceptible to the high rates of deforestation occurring in Latin

American tropical forests. Thus, jaguars are a model species in forecasting species distribution

by examining species-environment relationships. Using camera trap presence-absence data

collected from 2016 – 2018 and a machine-learning algorithm (Random Forests), I developed

multi-scale optimized models to investigate the environmental factors and associated scales that

influence annual and seasonal (wet season and dry season) habitat selection of jaguars in

Panamá. Across all models, habitat selection for most variables was detected at fine scales

(68.7%), and few at broad scales (31.2%). In the annual model, secondary forest was the most

influential cover type at the home range level. Jaguar habitat suitability decreased as secondary

forest increased, suggesting that secondary forest is a key determinant of jaguar distribution in

the study area. However, the probability of jaguar occurrence increased with increasing primary

forest cover, highlighting the importance of primary forest for jaguar persistence. Seasonal

differences were apparent, most likely owing to resource availability within the ecosystem.

During the wet season, jaguar habitat suitability was highest with moderate amounts of

hydrologic features, a north-facing aspect, steep slopes and low elevations at fine scales. The dry

season model showed a decrease in habitat suitability as secondary forest and fallow land

increased, while jaguars preferred low and mid-elevations at fine scales. The jaguars limited use

of fallow land suggests that it provides shelter or food, thus may offer important ecological value

in land management strategies. This is the first study to highlight the importance of incorporating

multi-scale optimization to capture jaguar habitat suitability across scales. By addressing key

knowledge gaps in jaguar habitat selection in Panamá, this study offers both a framework for

predicting jaguar distribution throughout the country, and provides important insights for the

conservation of the species’ habitats in the narrowest section of Panamá.

Introduction

Species-habitat relationships are fundamentally scale dependent and considered an inherently

spatial process (Thomson & McGarigal 2002; Wasserman et al. 2012; Mateo-Sanchez et al.

2014; McGarigal et al. 2016). Because animals hierarchically select habitats, identifying how

habitat preferences shift across scales is paramount for characterizing the complexities of habitat

selection (Wiens 1989; Levin 1993; Leopold & Hess 2012). Given the ongoing understanding

that ecological patterns result from processes occurring at multiple spatial and temporal scales,

predictive modeling of habitat selection is best framed as a scale specific process (Wiens 1989;

McGarigal et al. 2016). Hence, incorporating explicit optimization of scales in identifying

species-habitat relationships produce results that are biologically more meaningful and

statistically often more powerful than a single scale framework (Mateo-Sanchez et al. 2014;

Timm et al. 2016; Wan et al. 2017).

Precise regional distributions of species are rarely known, mainly due to difficulties in obtaining empirical measures of wide-ranging species throughout a range (Torres et al. 2012; Kanagaraj et al. 2013; Jędrzejewski et al. 2018). The ecological nature of jaguars (Panthera onca), such as low population densities, high trophic level, extensive home ranges, and wide-ranging distribution, makes them a model species in forecasting species distribution by examining

14 species-environment relationship (Bellamy et al. 2013; Zeller et al. 2017). From a conservation standpoint, jaguars are ideal for predictive distribution models because habitat loss and fragmentation, depletion in prey, illegal poaching, and deficient protection are threatening the species’ long-term survival (Sanderson et al. 2002; Zanin et al. 2015; Jędrzejewski et al. 2017;

Arias-Alzate et al. 2017).

A growing body of research has revealed several aspects of the jaguar’s ecology such as diet, habitat use, density, demographics, abundance (Tobler et al. 2013; Morato et al. 2016; McBride

& Thompson 2018), and conservation efforts have been directed at solving the widespread and deleterious human-jaguar conflict with livestock (Zimmermann et al. 2005; Cavalcanti et al.

2010; Moyer & Shebell 2014; Fort et al. 2018; de Souza et al. 2018). More recently, species distribution modeling (SDM), has been adopted to reveal ecological patterns of jaguar distribution across its range. Torres et al. (2012) employed several SDM methods to test whether suitability measures derived from SDM (from presence-only data) can be a reliable surrogate for jaguar population densities, finding there was no straightforward positive relationship between density, abundance and habitat suitability. Later, Jędrzejewski et al. (2018) used available density and presence-absence data to reveal the spatial mechanisms (i.e., environmental variables related to primary productivity) that determine jaguar population density and distribution at a coarse spatial resolution throughout the Americas. However, habitat relationship studies on jaguars have not thoroughly considered scaling issues in habitat selection. To date, no studies in published literature have predicted habitat suitability for jaguars using a multi-scale analytical framework in Panamá.

Accelerating fragmentation of landscape and the frequent gap between available knowledge on the species and appropriate management actions (Giehl et al. 2017), has rendered the central jaguar population in Panamá critically endangered (de la Torre et al. 2018). The ecological aspects of jaguar populations in Panamá are scarcely known, and most studies have focused on human-wildlife conflict issues (Fort 2016; Brossard & Pritz 2013; Fort et al. 2018). Fort (2016), using camera trap records, reported for the first time, the impacts of anthropogenic land use and topographic features affecting jaguar occupancy in two populations in the region. But there is still substantial uncertainty about the status of the jaguar population, its distribution, and how accelerating changes in the landscape affect their movements. Attaining data for a wide-ranging and secretive species like the jaguar is difficult. Yet, identifying the combined effects of human impacts and environmental factors that determine jaguar distribution can reveal ecological patterns and processes to aid in conservation strategies. The objectives of my study were to:

1. Explore the relationship between jaguar distribution patterns and a set of predictor

variables, including environmental and anthropogenic variables, across a gradient of

scales.

2. Determine patterns and changes in jaguar spatial distribution due to seasonal variation. I

hypothesized.

3. Produce high-resolution habitat suitability maps for jaguars, incorporating seasonal

variations for conservation and management purposes.

Herein, I examine the role of environmental interactions and anthropogenic pressures in shaping the spatial distribution of jaguars, with a novel focus on multi-scale habitat modeling by incorporating a scale-optimized habitat selection analysis (McGarigal et al. 2016). That is, the

16 assessment of a set of predictor variables across a gradient of scales to identify the characteristic scale of each variable (Bellamy et al. 2013; McGarigal et al. 2016). To address this urgent need,

I use presence-absence data from camera traps to develop annual and seasonal habitat suitability models in the narrowest section of Panamá. In an attempt to predict jaguar habitat suitability beyond the study area, I extend the models to approximately 100 km outside the boundaries of my camera trapping area. The models utilized a machine-learning algorithm, Random Forests

(Breiman 2001), which consistently outperforms other methods (e.g., Linear discriminant analysis (LDA), logistic regression) (Cutler et al. 2007; Cushman et al. 2017; Cushman &

Wasserman 2018).

My first hypothesis is based on the long-time understanding that ecological patterns result from processes occurring at multiple spatial and temporal scales (Wiens 1989; McGarigal et al. 2016).

Yet, to date, factors reported as important determinants of jaguar-habitat occurrence have been based on analyses at fixed spatial scales (Cullen Jr 2006; Fort 2008). I predicted that using a multi-level, scale-optimized model approach will show that jaguars avoid human settlements and prefer primary forest habitat at broad scales. Jaguars have been reported to avoid open, exposed areas, favoring closed, forested areas further from anthropogenic land use (Schaller & Crawshaw

1980; Silveira 2004; Cullen Jr 2006; Fort 2008).

My second hypothesis relates to the compelling evidence that jaguar occurrence is influenced by the distribution of its primary prey species (Scognamillo et al. 2003; Cullen Jr 2006; Foster 2008;

Fort 2008; de la Torre 2018). Further, as spatial seasonal distributions are species specific, I expected a reciprocal jaguar-specific distributional variation with predictor variables. Thus, I predicted that jaguars will exhibit a behavioral plasticity in terms of habitat requirements during

17 the wet and dry season, selecting environmental attributes that provide ample resources within the ecosystem (Palomares et al. 1996). Moreover, the jaguar will show differential habitat use at different spatial scales within each season. This prediction is based on Johnson’s (1980) order of habitat selection, whereby, the jaguar’s second and third order selection (fine scale) should correspond to the selection of habitat parameters that relate to prey distribution. Since jaguars tend to be associated with water and dense cover to assess prey vulnerability (Nowell & Jackson

1996; Gese et al 2019), I predicted that jaguars would positively respond to hydrologic features at a fine scale in the dry season, as prey may concentrate near water at this time of year.

Methods

Study area – Panamá is the fourth smallest Central American nation of 3.6 million people (Davis

2010), spanning about 75,717 km2 on an east-west axis (Figure 1). Panamá occupies the

southernmost section of the Mesoamerican Biological Corridor (MBC), one of the largest

bioregional conservation programs in the world (Dettman 2006). Spanning over 767,990 km2,

the MBC extends from Selva Maya in Mexico’s South-East, to Darien in eastern Panamá

(Dettman 2006).

The physiographic features that characterize the country include the Talamanca Mountains that

abut the Serrania de Tabasara to the west, and the Cordillera de San Blas and the Serrania del

Darien to the east (Myers 1969). Together, the range of mountains form the Cordillera de San

Blas, from which the physical complexity and biological diversity originates. Elevational differences with associated temperature and precipitation patterns produce distinct vegetative regimes that further contribute to the country’s rich biodiversity (Myers 1969).

Humid tropical and premontane forests dominate the upper reaches of the mountains. These forests contain a mix of old-growth and secondary-growth stands that have canopy trees reaching

40 m in height, and an extremely rich epiphyte flora (for a comprehensive description of the region, see Holdridge et al. 1971; Condit et al. 2001). Panamá also boasts tropical montane cloud forests, one of the world’s most imperiled ecosystems (Aldrich et al. 1997). Frequently covered in cloud or mist (Stadtmuller 1987; Hamilton et al. 1995), cloud forests are exceptionally rich in biodiversity, sustaining numerous locally endemic species. Condensation over warm land produced by moisture-laden air from the Caribbean Sea, hits the mountains, producing the high humidity and precipitation characteristic of this forest type. Annual mean temperature ranges from 30°C in the valleys to 20°C in the mountains. The wet season occurs from May to

November and the dry season from December to April. Annual rainfall amounts range between

1,700 – 4,000 mm on the Pacific and Atlantic coasts respectively (Condit et al. 2001; Ibáñez et al. 2002).

Biogeographical importance of the study area – The study area was selected for its strategic location and ecological significance. The merging of the Isthmian Atlantic moist forest and

Choco-Darien moist forest Ecoregions makes it a biodiversity hot spot (Corrales et al. 2015). The location of the study area is also at the convergence of three important protected areas in eastern

Panamá: Chagres National Park, the Narganá Protected Wildland, one of the most representative areas with high probabilities of jaguar presence, and the Mamoní Valley Preserve, a large private land tenure that serves as a buffer and stepping-stone habitat where jaguars traverse through extant forested ridgelines. Together, the pervasive advance of agriculture and illegal poaching threaten this area. Beyond that, the study is located in one of the narrowest stretches (~15 km wide) of primary habitat remaining for jaguar dispersal in the country. The study area also lies

19 within the range of the Jaguar Corridor Initiative and is part of Jaguar Conservation Unit 206 -

Choco-Darien (Sanderson et al. 2002).

Study species – The jaguar is the largest cat in the Neotropics (Zeller et al. 2013), ranging from

Mexico to Argentina. To date, the jaguar has been extirpated from about 50% of its historic range (Rabinowitz & Zeller 2010; Medellin et al. 2016; Quigley et al. 2017). As apex predators, jaguars exert top-down control on the ecosystems where they are found, their removal can generate a negative cascade effect (Crooks 2002; Fort 2016). Jaguars inhabit a wide range of habitats encompassing lowlands, mountains and coastal landscapes, showing preference for wet habitats (Sanderson et al. 2002). Males have larger home ranges than females which may include the territories of 1 – 3 females. Jaguar home range size is variable throughout its range. For males, the smallest home range was registered in Belize (33 km2) (Rabinowitz 1986), and the

largest in the Brazilian Pantanal (263 km2) (Quigley et al. 2017). The variation is due to site-

specific factors such as prey availability, persecution from humans and other habitat variables

that affect its density. The density of jaguars also varies throughout their range. The highest has

been recorded in Belize (3.5 – 11/100 km2) (Quigley et al. 2017) and the lowest in Brazil and

Venezuela (2 – 3/100 km2) (Quigley et al. 2017; Jędrzejewski et al. 2017).

Camera-trap surveys – The survey area was selected for its biological significance and included contiguous sections of three distinct areas under different tenure regimes, and therefore different dynamics of land cover and land use history: (1) Narganá Protected Wildland, (2) Chagres

National Park, and (3) The Mamoní Valley Preserve (Figure 1, Appendix A). Together, the survey area comprised ~200 km2. The terrain in this section is rugged with altitudes ranging from

400 m to 1,000 m a.s.l. Annual average precipitation during the wet season is ~ 4000 mm which

reduces to ~700 mm in the dry season.

Narganá Protected Wildland, created in 1994, falls within the semi-autonomous territory of the

Indigenous Guna Yala people. It is co-managed by the Guna Yala Congress and the Ministry of the Environment (MiAmbiente) (Parker et al. 2004). Narganá Protected Wildland constitutes

51% of the study area and is situated within the general ridge system of the Cordillera de San

Blas. It includes the Caribbean side of the Cordillera de San Blas, which has a very wet climate owing to showers spawned by moisture-laden trade winds. Most of the forest within this area is largely undisturbed by human activities, but patches of fallow (locally known as rastrojo)

(pioneer forests 40 – 50 yrs. old fallow) and secondary mature forest are apparent. These patches are the result of settlers clearing the land for homesteads and agriculture in the late 1900s

(Chapin 2000) (Figure 1).

Figure 1. Location of study site and camera-trap stations distributed in a 2 x 2 km grid system used to determine jaguar habitat selection. It comprises three-land tenure areas Narganá Protected Wildland, Chagres National Park and the Mamoní Valley Preserve in the country of Panamá. Phase II data collection was not part of this study.

Chagres National Park, established in October 1984, is part of the Panamá Canal Watershed

(Condit et al. 2001) and borders the southwestern end of the Narganá Protected Wildland. It encompasses 13% of the study area. There are communities living within the park that depend on agriculture and subsistence hunting for their livelihoods. Old-growth forest dominates the upper ridges. Its topography is rugged with permanent and intermittent streams, steep rock walls along ridgelines and ravines with clear, fast-flowing, rocky streams, which serve as tributaries to one of the most ecologically important rivers in Panamá, the Chagres River (Minter 1948) (Figure 1).

The Mamoní Valley Preserve is on the leeward side of the Cordillera de San Blas facing the

Pacific slope. With most of the upper reaches still forested, the Mamoní Valley Preserve acts as a buffer zone, protecting the eastern border of Chagres National Park and the southwestern border of the Narganá Protected Wildland. The Mamoní Valley Preserve covers 36% of the study area.

It is comprised of 115 km2 of privately-owned lands. There are four villages within the preserve

with an estimated population of 400 people whose economy is largely agricultural, and ranching

based. The land cover in the preserve is composed of a mixed matrix that includes secondary

forest, fallow (with various stages of maturity), tree farms, pasture and agropecuary practices

(e.g., agriculture and livestock) (Figure 1).

Between July 2016 and September 2018, I used motion-activated camera traps to detect jaguar occurrence under a scientific research permit No. SE/A-14-17 obtained through the Ministerio de

Ambiente de Panamá. Camera traps were set throughout an area of approximately 200 km2 within the Panamá Province, Republic of Panamá (9°19’ 35.5261” E/-79°08’ 47.8356” N)

(Figure 1). Throughout the study period, I used Bushnell Trophy Cam Aggressor and Cuddeback

C-E cameras programmed for picture and video function.

Following Silver et al. (2004), a 2 x 2 km grid cell overlay was used to establish 48 camera stations. Using a hybrid design, I simultaneously deployed 36-single and 12 paired stations spaced approximately 2 km apart throughout the study grid. Camera spacing aimed to maximize photo capture probability, with two or more cameras per average home-range area (Kelly et al.

2008). Home range size was based on the smallest recorded home range size of jaguars in

Central America (10 km2; Rabinowitz & Nottingham 1986). Two cameras are favored over one

camera per station, as it increases the chances of capturing both flanks (left and right) of the

23 species for identification and reduces the risk of lost trap-nights from camera failure. However, the number of cameras were restricted by limited resources, hence the hybrid design. Based on site accessibility, camera stations were randomly established on pre-existing human trails, game trails, rivers, streams or creeks.

Cameras were placed approximately 30 – 50 cm above the ground and adjusted in response to local topography. Cameras were deployed year-round (wet season-May to November/dry season-

December to April) and programmed to operate 24h/day. During the dry season four camera stations were removed due to high levels of vandalism. Cameras were checked every three months to ensure functionality (e.g., replace memory cards, batteries, theft). Camera trap design decisions were based on attaining the largest grid feasible within logistical and financial constraints.

Data Analysis: Random Forests – Using models can be effective in evaluating habitat suitability for a species. Decisions regarding appropriate variables for the inclusion in candidate models is often guided by the available biological information on the species under study. However, the effect of a variable on habitat selection may occur at a range of spatial scales. Therefore, evaluating the optimality of the scales at which they represent the variable is essential

(Wasserman et al. 2012).

To create habitat suitability models for jaguars I used the ‘randomForest’ package in R statistical software (R Development Team 2017). Random Forests is a classification and regression tree based bootstrap method that provides well-supported predictions with large numbers of independent variables (Breiman 2001; Cutler et al. 2007). It is useful for modeling nonlinear

24 data, such as presence-absence data, with variables that likely interact in a hierarchical fashion

(De’ath & Fabricius 2000).

The Random Forests classifier includes two random processes that improve the predictive power of the classification. When building each decision tree in the forest (ensemble), at each tree branching node a subset of potential predictor variables is randomly selected on which the data is split (to create children nodes), thereby reducing the correlation between the trees, resulting in a lower error rate (Horning 2010). In addition, the Random Forests algorithm randomly selects a bootstrap sample from a subset of the total training data available to build each tree. A third of this subset is left out-of-bag (OOB) and not used to construct the tree. This OOB subset is then run through the constructed tree to cross-validate the classification, thereby deriving an unbiased estimate of the test-set error (Breiman & Cutler 2003). Random Forests models limit overfitting without increasing much error due to bias and are superior to most methods commonly used

(e.g., logistic regression and linear discriminant analysis (LDA)) (Cutler et al. 2007). I first conducted an error convergence pilot test using 10,000 trees. Based on the results, I determined that the error stabilized between 500 and 1500 trees. For all my models, I opted to use 2,000 trees to be conservative.

Predictor Variables – The predictor variables included in the models were selected based on a comprehensive review on previous studies of jaguar ecology (Zeller et al 2011; Fort 2016;

Morato et al. 2016; Astete et al. 2017; de la Torre et al. 2018). The variable classes used to fit the models included: topographic, landscape composition, land use, water body, and climate (Table

1, Appendix B).

Topographic covariates were derived from Aster Global DEM V002 (NASA) using a 30-m resolution digital elevation model (Table 1). The relative slope position measures the relative position of the focal pixel within a defined extent on a gradient from ground bottom to the top of a ridge, and the topographical roughness measures the topographical complexity of the landscape within a defined focal extent (Krishnamurthy et al. 2016). I calculated both the topographical roughness and relative slope position using the Geomorphometry & Gradient Metrix Toolbox in

ArcGIS (ESRI, Redlands, CA). The landscape composition variables were determined by the

interpretation of the Cobertura Boscosa y Uso de Tierra 2012 (land cover) map developed by the

Ministry of Environment of Panamá (MiAmbiente) (Table 1). Percent tree cover was derived by

the USGS Global Tree-Canopy Cover Circa 2010 (Table 1). Climate variables were derived from

Global Climate Data - WorldClim.org (annual average past 30 years) (Table 1). Finally, water

body and land use features were determined by interpreting Cobertura Boscosa y Uso de Tierra

2012 maps from MiAmbiente (Figure 2). All habitat variables were mapped at 30 m resolution,

which represented the finest resolution in the source data. The spatial grain of the analyses was

held constant at 30 m in the multi-scale analyses described below.

Table 1. Description of variables used in the multi-scale jaguar habitat selection model developed for the study area. Variables were classified into five groups: topographic, landscape composition, land use, hydrological, and climate.

Class Variable Variable Code Data source Topographic Aspect Aspect_cos NASA: Earth Data Elevation Dem NASA: Earth Data Slope Slope NASA: Earth Data Topographic Rou_new NASA: Earth Data Roughness

Landscape Agropecuary Agropec MiAmbiente* Composition Banana Plantation Banana MiAmbiente Broadleaf Forest Forlatif MiAmbiente Coniferous Forest Forconi MiAmbiente Herbaceous Vegetation Vegherb MiAmbiente Mixed Horticulture Hortmix MiAmbiente Pasture Pasto MiAmbiente Primary Forest Mix_forest MiAmbiente Fallow Rastrojo MiAmbiente Secondary Forest Foremix MiAmbiente Tree Cover Tree_r USGS Global Tree

Land use Human Settlement People MiAmbiente Infrastructure Infraest MiAmbiente

Hydrological Hydrologic Feature Hydro MiAmbiente

Climate Annual Precipitation Prec_all Global Climate Data Dry Season P_dry Global Climate Precipitation Data Wet Season P_wet Global Climate Precipitation Data

* Ministry of Environment: Mapa de cobertura boscosa y uso de la tierra, 2012 (land cover map)

Figure 2. Location of the study area indicating land cover types in eastern Panamá. The map depicts the agricultural frontier abutting the remaining stretch of forest within the narrowest section of Panamá. Taken from the Cobertura Boscosa y Uso de Tierra 2012 map from MiAmbiente.

Scale optimization and variable selection – To identify the optimized scale of each predictor variable that was most related to jaguar response, I calculated the focal mean of each predictor variable around each camera location across 30 spatial scales, ranging from 500 m – 15000 m radii at 500 m increments. To accomplish this, I first conducted a moving window analysis with the Focal Statistic tool in ArcGIS (ESRI, Redlands, CA), using a circular neighborhood and scales described above as search radii for each variable. This allowed me to obtain an output raster across the entire study area where the raster value at any given pixel was a mean function

28 of the values of all the input cells within the specified neighborhood (de la Torre et al. 2018).

Then, I extracted the raster values around each camera trap location for each scale and each variable. The full set of candidate variables was reduced to four topographic variables, seven landscape composition variables, three climate variables, two land use variables and one for water body (Appendix B).

Multi-scale optimized multivariate modeling – I developed a multi-scale model of jaguar occurrence as a function of the environmental predictor variables in three stages, following the recommendations of McGarigal et al. (2016). First, using the focal mean values from the moving window analysis described above, I conducted a univariate scaling analysis to identify the spatial scale at which each variable was most strongly related to jaguar occurrence. This was accomplished by testing one variable and one scale at a time using Random Forests. I selected the best-supported scale from each variable based on the model with the lowest OOB error rate

(Appendix C – E). Then, in the second stage, I optimized the number of variables for the final model in two subsequent steps. First, a variable was removed if it had p > 0.05 in the model and second, I used the multicollinear function in the rfUtilities R package to assess potential multicollinearity among all possible pairs of scale-optimized variables and removed variables that were highly correlated (p > 0.05). Finally, in the third stage, I ran Random Forests with the remaining variables to model the probability of jaguar occurrence. Using these procedures, I developed three final Random Forests models: annual, wet season, and dry season. With each of these models, I created a map predicting jaguar habitat suitability across the study area. The models also generated the scaled variable importance and partial dependency plots for each variable selected within the annual, wet, and dry season.

Model Validation – To assess the performance of the final models, I conducted random permutations, cross-validation using a resampling approach (Evans & Murphy 2018), whereby, one-tenth of the data was withheld as a validation set in each permutation. A total of 99 permutations were performed. The cross-validation produces a suite of performance metrics including OOB error rate (the proportion of OOB samples that are incorrectly classified), model error variance, and Kappa index of agreement, (“a measure of agreement between predicted presences and absences with actual presences and absences corrected for agreement that might be due to change alone” (Cutler et al. 2007)). Landis & Koch (1977) suggest the following ranges of agreement for the Kappa statistic: values < 0 indicate no agreement, 0 – 0.20 as slight,

0.21– 0.40 as fair, 0.41 – 0.60 as moderate, 0.61 – 0.80 as substantial, and 0.81 – 1 as almost perfect agreement.

Habitat selection and scale reference – To measure jaguar habitat selection at multiple levels, I followed the four hierarchical orders of habitat selection described by Johnson (1980) and Wiens

(1989). The first order refers to a species geographic range, the second order pertains to a species home range selection, the third order includes a species selection of habitat patches within its home range, and the fourth order refers to the selection of specific resources within a habitat patch (e.g., food items). Whilst second order modeling can provide an understanding of habitat selection at a regional level, fine-scale modeling may identify preferred topographic features or important habitat characteristics (Apps et al. 2001). To explain at what scale jaguars select their habitat, I divided the original 30 spatial scales (500 m – 15000 m), referring to responses at the lower half (500 m–7500 m) as fine scale habitat use, and those in the upper half (7500 m – 15000 m) as broad scale habitat use. Fine scale relates to the third and fourth orders of habitat selection, while a broad scale reflects the second order of habitat selection. The jaguar’s geographic range,

30 or first order selection, was beyond the scope of this study. Further, the division between fine and broad scale is based on the estimated home range size for male jaguars registered in the rainforest of Belize (28 km2 – 33 km2) (Rabinowitz & Nottingham1986). However, there is

variation in home range size due to site-specific factors such as prey availability, persecution

from humans and other habitat variables that affect density. Moreover, jaguars tend to increase home range size in the wet season (Nuñez-Perez & Miller 2019).

Results

Jaguar presence/absence data – The sampling effort accumulated during the two-year study period was 16,583 trapping nights from 48 camera stations, with the number of trap nights per station averaging 345.5 ± 23.30 SE (49 – 597). For the analysis, a single jaguar detection was considered one record per hour, per camera station (Appendix F). The sampling effort produced

153 records of jaguars with a capture success of 0.92 (153/16,583 x 100). Capture success for the wet season was 0.55 (92 jaguars), and for the dry season it was 0.35 (58 jaguars). In total, I detected jaguars at 32 camera stations (66.6%). During the wet season (May to November), cameras detected jaguars at 26 (54.2%) camera stations (1.9 ± 0.4 SE jaguars/camera). During the dry season (December to April), 19 (43.2%) cameras detected jaguars (1.3 ± 0.3 SE jaguars/camera) out of 44 active camera stations.

Scale optimization – The optimized multi-scale analyses displayed that the strength of the relationship between jaguar occurrence and predictor habitat variables was dependent on the spatial scale at which each variable was derived (Table 2). However, the analyses indicated that jaguar habitat selection was predominantly driven by habitat factors at fine scales in all three models (78.0%). In the annual model, jaguar presence was influenced by secondary, coniferous

31 and primary forests at broad scales and areas of human settlement, aspect and fallow at fine scales. The multi-scale annual model had a relatively high predictive performance with an OOB of 0.22. The Kappa index showed a moderate agreement at 0.52 (Table 2).

During the wet season, jaguar occurrence was associated with broadleaf forest and coniferous forest at relatively broad scales (Table 2). Jaguars were influenced by the set of topographic features (aspect, elevation, slope), hydrologic features (i.e., rivers, ponds, lakes) and the landscape composition variable fallow at fine scales. The multi-scale wet season model had a high predictive performance with an OOB of 0.18 and a Kappa index of 0.63 indicated a substantial agreement (Table 2).

The dry season scaling analyses did not show any variation in optimized scale preferences among variables (Table 2). In the dry season, jaguars responded to fallow, secondary forest and elevation at fine scales. The multi-scale dry season model had a high predictive performance with an OOB of 0.28 and a fair Kappa index of 0.40 (Table 2).

Table 2. Optimal scales for the most important variables found in the annual, wet season, and dry season Random Forests jaguar models. Code refers to the name of each variable taken from the Cobertura Boscosa y Uso de Tierra 2012 map, shortened for the purpose of analysis. Numbers in bold indicate broad scale. It includes Kappa, OOB and error variance.

Variables/Model Code Optimal Scale Kappa OOB Error Error Variance Jaguar Annual

Model accuracy 0.5283 0.2222 0.0007 Secondary Forest Foremix 10500 Coniferous Forest Forconi 14000 Human Settlement People 4000 Aspect Aspect_cos 6000 Fallow Rastrojo 3500 Primary Forest Mix_forest 10500

Jaguar Wet

Model accuracy 0.6303 0.181818 0.00107 Hydrologic Feature Hydro 4500 Aspect Aspect_cos 3500 Slope Slope 3000 Fallow Rastrojo 1500 Elevation Dem 3500 Broadleaf Forest Forlatif 11000 Coniferous Forest Forconi 12500

Jaguar Dry

Model accuracy 0.4008 0.28261 0.0008 Fallow Rastrojo 3500 Secondary Forest Foremix 5000 Elevation Dem 7000

Variable importance – As expected, the predictors deemed important by the Random Forests models are different for each of the developed models, and the ranking of predictors in order of importance is evident (Figure 3). Of the seventeen covariates used in the analyses, the final multi-scale annual model consisted of six variables. In decreasing order of importance, the predictors included in the model were secondary forest, coniferous forest, human settlement,

33 aspect, fallow and primary forest. The final wet season multi-scale model consisted of seven variables (Figure 3). The model revealed that hydrologic feature was the most important variable followed by aspect, slope, fallow, elevation, broadleaf forest, and coniferous forest. Finally, in the dry season analyses, three predictors contributed to jaguar habitat selection (Figure 3).

Fallow was the most important variable followed by secondary forest and elevation.

Figure 3. Variable importance plot for the predictor variables from Random Forests classifications used for predicting presence of jaguars in the study area. For each tree in the forest, there is a misclassification rate for the out-of-bag observations. To assess the importance of a specific predictor variable, the values of the variable are randomly permuted for the out-of- bag observations, and then the modified out-of-bag data are passed down the tree to get new predictions. The difference between the misclassification rate for the modified and original out- of-bag data, divided by the standard error, is a measure of the importance variable (Cutler et al. 2007). The variable Foremix refers to secondary forest, Forconi refers to coniferous forest, Rastrojo refers to fallow vegetation, Mix_forest refers to primary forest, Dem refers to elevation, Hydro refers to hydrologic features (i.e., rivers, ponds, lakes), Forlatif refers to broadleaf forest, and Agropec refers to agropecuary (i.e., agriculture and livestock).

Partial dependency plots – The annual partial dependency plots represent the marginal effect of

each of the most important predictors included in the annual model (Figure 4). Jaguar occurrence

34 had a non-linear relationship with secondary forest (foremix); the highest probability of occurrence was associated with lower amounts of this cover type (e.g., 13 – 15%). Human settlement (people) showed a strong negative relationship, indicating that the probability of jaguar occurrence drops drastically as the size of the inhabited area increases. Coniferous forest

(forconi) and fallow (rastrojo) showed negative correlations, indicating the jaguar’s lack of affinity for these two cover types as they increased in the landscape. Additionally, a northeastern aspect was preferred, while the probability of jaguar occurrence increased with more primary forest (mix_forest) (~70 – 75%), as shown by the positive unimodal peak.

Figure 4. Partial dependency plots representing the marginal effect of each variable included in the Annual Random Forests jaguar model. In a partial plot of marginal effects, only the range of values (and not the absolute values) can be compared between plots of different variables (Cutler et al. 2007). The gray area indicates the 95% confidence interval and the red line represents the mean. The variable mix_forest refers to primary forest, forconi refers to coniferous forest, dem refers to elevation, hydro refers to hydrologic features (rivers, ponds, lakes), agropec refers to agropecuary (agriculture and livestock), foremix refers to secondary forest, and aspect_cos refers to aspect.

The wet season partial dependency plots demonstrate that broadleaf forest (forlatif) was most positively associated with the presence of jaguars (Figure 5). Jaguar occurrence increased monotonically as the cover type increased on the landscape. Jaguar presence had a non-linear relationship with elevation; the probability of jaguar occurrence increased at low elevations (e.g.,

250 m) and was least likely at higher elevations (e.g., ~ 500 m). Alongside elevation, jaguars favored steep slopes with a north-facing aspect. The jaguar’s selection for hydrologic features

(i.e., rivers, ponds, lakes) was positively correlated with moderate amounts, but steeply declined as these features increased in the landscape. Lastly, jaguar occurrence decreased monotonically with increasing amounts of coniferous forest (forconi) and fallow (rastrojo) land.

Figure 5. Partial dependency plots representing the marginal effect of single variables included in the Wet Season Random Forests jaguar model. In a partial plot of marginal effects, only the range of values (and not the absolute values) can be compared between plots of different variables (Cutler et al. 2007). The gray area indicates the 95% confidence interval and the red line indicates the mean average. The variable mix_forest refers to primary forest, forconi refers to coniferous forest, dem refers to elevation, hydro refers to hydrologic features (rivers, ponds, lakes), agropec refers to agropecuary (agriculture and livestock), foremix refers to secondary forest, and aspect_cos refers to aspect.

The partial dependency plots for the dry season revealed that jaguar occurrence had a non-linear, negative relationship with fallow (rastrojo) and secondary forest (foremix). The probability of jaguar occurrence decreased as these cover types increased on the landscape (Figure 6).

Additionally, jaguar presence had a non-linear relationship with elevation; the probability of jaguar occurrence increased at a low elevation (e.g., 250 m), but then slowly declined before increasing again at a higher elevation (i.e., 550 m).

Figure 6. Partial dependency plots representing the marginal effect of single variables included in the Dry Season Random Forests jaguar model. In a partial plot of marginal effects, only the range of values (and not the absolute values) can be compared between plots of different variables (Cutler et al. 2007). The gray area indicates the 95% confidence interval and the red line indicates the mean average. The variable mix_forest refers to primary forest, foremix refers to secondary forest, and aspect_cos refers to aspect.

Habitat Suitability Models – The habitat suitability maps produced by Random Forests for the

jaguar included the annual, wet season, and dry season (Figure 7 – 9). Predictions for this

analysis were extrapolated to approximately 100 km outside the boundaries of my camera

trapping area. All three models displayed areas of high to low suitability represented in a

38 gradient from maximum probability for jaguars (in red) to the lowest (in blue). The maps derived by Random Forests indicated that the predictive accuracy (based on the Kappa index of agreement) varied between models and seasons (See results, Table 2). The annual model was moderate, the wet season model was substantial, and the dry season had a fair level of accuracy.

The annual model (Figure 7) indicated that the most suitable habitat is found in the northern reaches of the study area within the Narganá Protected Wildland, while areas of low habitat suitability were found where human disturbance is present south of the Cordillera de San Blas. In the wet season model (Figure 8) it is evident that highly suitable habitat was found in the northern portion of the study area, while a broad area of low habitat suitability is located south of the Cordillera de San Blas. The dry season model (Figure 9) shows a shift in habitat selection and a reduction of preferred habitat, most noticeably within the northern portion of the study area, which may indicate a change in seasonal elevation preferences. Additionally, a seasonal change of habitat use is apparent beyond the study area. Portions of moderate to high suitability in the Narganá Protected Wildland become less suitable in the dry season.

Figure 7. Annual habitat suitability map showing the predicted occurrence of jaguars based on multi-scale habitat modeling in Panamá. The map includes potential jaguar habitat suitability up to approximately 100 km outside the study area. The habitat suitability index ranges from high (in red) to low (in blue) with high representing more suitable habitat.

Figure 8. Wet season habitat suitability map showing the predicted occurrence of jaguars based on multi-scale habitat modeling in Panamá. The map includes potential jaguar habitat suitability up to approximately 100 km outside the study area. The habitat suitability index ranges from high (in red) to low (in blue) with high representing more suitable habitat.

Figure 9. Dry season habitat suitability map showing the predicted occurrence of jaguars based on multi-scale habitat modeling in Panamá. The map includes potential jaguar habitat suitability up to approximately 100 km outside the study area. The habitat suitability index ranges from high (in red) to low (in blue) with high representing more suitable habitat.

Discussion

Our ecological understanding of a species distribution often relies on habitat and environmental suitability modeling. However, limitations and decisions inherent in the construction of spatial distribution models remain (Guisan & Thuiller 2005; VanDerWal et al. 2009; Thuiller et al.

2010; Torres et al. 2012). Wide-ranging species like the jaguar are confronted with a challenging array of factors when making habitat choices, especially in highly anthropogenic-influenced

42 landscapes. In this study, such factors interacted dynamically, and with different strengths at multiple spatial scales. Still, fallow, secondary forest, coniferous forest, aspect and elevation — at different degrees of importance — were the most consistent predictors of habitat selection in each model.

Annual habitat selection – The results of the annual model indicated that jaguars responded to

secondary forest, coniferous forest and primary forest at broad scales (2nd order selection - home

range). Jaguars exhibited a low degree of tolerance for human settlement and fallow lands, and

preferred a northern aspect at fine scales (3rd order selection – e.g., foraging areas). Thus, my

main hypothesis that a multi-level, scale-optimized model approach on jaguar habitat selection

will show an association with primary forest at a broad scale was supported. However, primary

forest was ranked as the least important variable at broad scale. This finding may be due to the

study design, where the sampling units (camera stations) were not set at sufficient spatial

distances to include a broader spectrum of cover types, which did not allow for model

discrimination between sampling units. Boyce et al. (2003) point out that the heterogeneity of

resources is fundamental to the variation in selection amongst scales. Thus, the spatial

homogeneity of most of the sampling units in my study did not provide a wide-range of

conditions of potential jaguar habitat (i.e., different vegetation cover types). For example, most

stations fell within primary forest and patches of secondary forest. The few stations that did not

fall within those categories were located either in forest edge, small tree plantations or in areas

bordering stretches of mature forest (2 km from the forest edge on the Pacific slopes).

Additionally, the size of the study area coupled with jaguar occurrence records that were bias

towards primary forest, precluded the representation of other cover types (suitable or not) within

the area. Guisan & Thuiller (2005) suggested that for mobile organisms, various habitats may

43 need to be included in order to fulfill the different habitat requirements of the species. Moreover,

VanDerWal et al. (2009) point out that species’ dispersal capabilities and stochastic effects can result in species absence at sites with suitability potential, factors that are usually ignored by the models. The results support this view, since in my camera study area, jaguar dispersal may be framed by the Cordillera de San Blas, which may act as boundary for movement towards the

Pacific side, where most of the anthropogenic impacts and a lack of forested habitat are evident.

Therefore, dispersal beyond the Cordillera de San Blas may be uncommon in this jaguar population. That is, in my study area, jaguar range extension may be contingent on the approximately 15 km wide (North-South direction) strip of primary forest that is conspicuously framed by the Cordillera de San Blas to the south, and the Caribbean Sea to the north.

My data agrees with the theory that broad scales are the most relevant limiting factors among the hierarchal scales of habitat selection (Rettie & Messier 200). Because the spatial and temporal hierarchy of habitat selection reflects the hierarchy of factors potentially limiting individual fitness, species should avoid factors with greater potential to reduce individual fitness at broader spatial and temporal scales. Whereas, an individual will maximize fitness by avoiding the factors that are most limiting at each successive scale (i.e., fine) (Rettie & Messier 2000). The results are striking because they provide insight into the jaguars’ home range limitations, which in turn, reflect the distribution of suitable habitat.

As noted, portions of secondary forest along the southern limits of the study area appear to be key local determinants of jaguar distribution at the home-range level (i.e., broad scale). Beyond that, it highlights the importance of this habitat on the species’ dispersal limitations. However, in small amounts, secondary forest may provide a matrix of food sources at low energy costs, and

44 may also be used for territorial marking or dispersal. In previous studies, it has been suggested that carnivores select habitat at broad scales with attributes that supply ample resources and increase the probability of hunting success (Davidson et al. 2012). For example, collared peccary

(Pecari tajacu) are known to persist in disturbed habitats such as secondary forests near agricultural areas (Reyna-Hurtado & Tanner 2007). In addition, small mammal richness and diversity have been reported to increase in secondary forests (de Fonseca et al. 1990), while

Baird’s tapir (Tapirus bairdii) also tend to prefer secondary forest over primary forest due a greater availability of vegetation close to the forest floor (Foerster & Vaughan 2002).

Consistent with previous studies that have reported a positive association with jaguar selection for primary forest or high forest cover, the annual model results suggested that jaguars preferentially selected for habitats with complex structure (at broad scale) for use as secure refuge, reproduction, and to aid in the stalking and ambushing of prey (Crawshaw & Quigley

1991; Sunquist & Sunquist 2002; Silveira 2004; Cullen Jr 2006; Cullen et al. 2006; Davis et al.

2010; de la Torre et al. 2017).

The second prediction that jaguars would avoid human settlement at a broad scale was not supported, as jaguars exhibited a low tolerance for human settlements. Although jaguar’s showed fine scale tolerance based on the analysis, this can be explained by the scattered, and low human population density inhabiting the southern portion of the study area. Other studies have indicated the role of the human footprint as one of the main drivers of jaguar habitat use (Foster et al.

2010; Morato et al. 2016; Jędrzejewski et al. 2018). Jaguars are known to avoid human activity

(Cullen Jr et al. 2013; de la Torre et al. 2017), but jaguars in my study utilized forest edge where human activity persists in the form of agriculture or procurement of food (e.g., hunting) (K.C.,

45 unpublished data). In support of this finding, Dobbins et al. (2018) reported that jaguars utilized habitat with the highest availability of its preferred prey species, despite a high probability of interactions with humans.

Further, in the annual model, jaguars responded in complex ways to other variables. For example, the negative association with coniferous forest indicates that jaguars are sensitive to this landcover type, which acts as a limiting factor at a broad scale (i.e., home range). Coniferous forest is not likely to be an essential food source or cover for the jaguar due to the sparseness of the vegetation and lack of understory. Further, the limited association with the cover type fallow suggests that jaguars are unlikely to use much of this habitat type. Yet, when available in small amounts, it may provide shelter and perhaps a source of food. It may also play a role during seasonal shifts (further analysis below).

Wet season habitat selection – The annual model alone could not explain seasonal differences in resource selection by jaguars across Panamá. Several studies have established that seasonal fluctuation in spatial activity patterns and territory use has some impact on detection probability

(de la Torre & Medellin 2011; Harmsen et al. 2011; Tobler et al. 2013; Jędrzejewski et al. 2017).

The results corroborate my second hypothesis that jaguars select habitats seasonally. Habitats were selected for their environmental attributes that most likely supply ample resources within the ecosystem. The study illustrated the strong influence of fine scale habitat selection (i.e., hydrologic features, aspect, slope, elevation) during the lengthy wet season.

As shown by previous studies, jaguars were associated with hydrologic features (i.e., rivers, ponds, lakes) (Sollmann et al. 2012; Nuñez-Perez & Miller 2019). Nuñez-Perez and Miller

(2019) note that jaguars used streams to reach hilltops because they offer a clean travel path for

46 movement or dispersal. In addition, seasonal distribution and abundance of prey are known determinants for changes in home range patterns of wide-ranging carnivores (Crawshaw &

Quigley 1991; Davidson et al. 2012), including climatic factors such as rain (Quigley 1987;

Jędrzejewski et al. 2017). Further, plant productivity, distribution, and diversity in tropical rainforests correlate with water availability (Kursar et al. 2005) while aspect, slope and elevation factors control such productivity and distribution (Gentry 1988; Busing et al. 1992), creating climatic zones or habitats that harbor diverse faunas (Jankowski et al. 2008), in turn, providing a wide-range of opportunities for jaguars to efficiently stalk prey (Nuñez-Perez & Miller 2019).

Dry season habitat selection – To enhance fitness potential, jaguars will select a variety of habitats suitable for dispersal, optimal hunting patches, or to search for mates, if they offer some vegetation cover (Crawshaw & Quigley 2002; Nuñez-Perez & Miller 2019). This behavior is suggested by the dry season model results. However, my prediction that jaguars would select hydrologic features at a fine scale in the dry season was unsupported, as this model determined the most important variables were fallow, secondary forest and elevation at fine scales. De la

Torre & Rivero (2019) point out that jaguar movement was associated with a high percentage of forest cover at altitudes of 240 m. Similarly, in the dry season model, jaguars were associated with secondary forest and low or mid-elevations (~250 m and 550 m). Suggesting that secondary forest and fallow (at the height of maturity) may provide resources such as food and refuge, and even water sources that may be scarce in the primary forest during the short dry season.

Importance of Fallow – It is notable how fallow is associated with jaguar presence at a fine scale in all three models. Although this cover type was preferred in limited amounts, it warrants further discussion. The Panamanian Forestry Law defines fallow (i.e., rastrojo) as herbaceous, shrubby, and woody vegetation, including trees of 1 – 5 years of age, not higher than 5 m. In

Panamá, fallows are classified into young fallows of 1 – 2 years (i.e., rastrojo bajo) and older fallows of 10 – 15 years (i.e., rastrojo alto). As noted in the definition, fallow is a unique transitional vegetation cover that could represent a significant ecological value for the subsistence of jaguars due to its high prey productivity (i.e., peccaries, tapir) (K.C., unpublished data). Extensive sections of this habitat cover type are embedded in the primary forest up to 2 km from the Cordillera de San Blas in Narganá, the Guna Yala Territory on the Caribbean slope. It is found in lowlands, close to streams and rivers where 40 – 50 years ago, peasant farmers abandoned their lands. Subsequently, it developed into early successional mixed broad-leaf forest at lower elevations. Considering the value of this cover type to the larger ecosystem, this finding underscores the need to contemplate fallows in land management strategies (Jakovac et al. 2015). Further, it is valuable for planning more resilient landscapes for forest corridors, including forest patches deemed important for jaguar movement.

Conclusion – Simultaneous examination of the scales at which the variables in the models performed, provided hints to the marginal and outlying seasonal habitat use of jaguar in the study area. Sexton et al. (2009) points out that species range limits are essentially the expression of a species’ ecological niche in space. Thus, the annual model may reflect more of a biological reality rather than an artifact of sampling. While I observed an overall pattern of jaguar distribution limited by habitat suitability, the models predicted that the selection of suitable habitat for jaguars was strongest at fine scales (3rd – 4th order selection) accounting for 78.0% of

the most influential variables, and 21.9% for broad scale (2nd order selection). Next, the general

pattern of distribution predicted by the models conforms to findings from other studies, more

importantly; the models provide insight into jaguar home range limitations, which may shed light

on the mechanisms leading to observed patterns of the species distribution in the narrowest

48 section of Panamá. By incorporating a multi-level, scale-optimized habitat selection analysis approach (McGarigal et al. 2016; Zeller et al. 2017; Bauder et al. 2018), the spatially explicit information on the suitability of jaguar habitat in the complex tropical montane forest of Panamá has been revealed. Finally, my study contributes to the limited information regarding the extent of jaguar habitat selection in the region.

Limitations of the analyses – Models are an effective way to develop wildlife management plans

(Sunarto et al. 2012). However, they are a challenge to construct for rare, elusive and highly mobile species like the jaguar (Torres et al. 2012; Jędrzejewski et al. 2018), as data collection is a challenging task due to logistics and planning. Although Random Forests models often reach top predictive performances on ecological predictions (Prasad et al. 2006; Cutler et al. 2007), a weakness in my study is the limited geographic coverage of my data (~ 200 km2) compared to

the jaguar’s distribution across the regional landscape. Despite the size of my study area, and that

my findings were limited with respect to habitat types found outside the range of those

encountered in my sampling area; Panamá province, where the camera-traps were located, is a

good representation of the landscape used by jaguars. At the same time, it clearly represents the deleterious effects of fragmentation in the area. Hence, it will be meaningful to investigate further by placing additional camera traps at a broader extent across Panamá.

Owing to the powerful predictive ability of Random Forests, the habitat suitability maps produced by each model predicted reasonably well. The results, based on the existing strength of the relationships between the jaguars’ response and the predictors in the study area, confirm the findings of other studies. Still, caution must be granted regarding nationwide inferences from the habitat suitability models presented here, as they may not represent a comprehensive prediction for the entire country. In the meantime, as verification with additional camera trapping and GPS-

49 collaring is planned, the information provided by the interim models can be used as baseline predictions for management purposes in the area. The information can also drive future investigations to evaluate the accuracy of my predictions, especially in areas of high probability or potential corridors.

Conservation Implications – Habitat suitability models are not perfect representations of species distribution (Rondinini et al. 2011). Yet, the modeling efforts carried out in this study bolster our understanding of jaguar habitat selection in a one of the most critical regions of Panamá. It provides a valuable index for priority conservation actions and for other scientists interested in evaluating habitat suitability for other carnivores. Understanding the interplay between environmental and ecological factors, and the range of scales over which jaguars select habitats requires a joint effort and is fundamental to providing realistic scenarios of threats to the species–anthropogenic and climate change. Given that jaguars are an umbrella species and they are considered endangered in Central Panamá (de la Torre et al. 2017), successful jaguar conservation in this narrow section of the country will hinge upon land management practices that maintain the integrity of the remnants of primary forest, and the permeability of jaguar dispersal through secondary forest and old fallow to enable genetic flow throughout the region.

Literature Cited

Apps, C. D., B. N. McLellan, T. A. Kinley and J. P. Flaa. 2001. Scale-dependent habitat selection by mountain caribou, Columbia Mountains, British Columbia. The Journal of Wildlife Management 65:65.

Astete, S, J. Marinho-Filho, M. Kajin, G. Penido, B. Zimbres, R. Sollmann, A. T. A. Jácomo , N. M. Tôrres, and L. Silveira. 2017. Forced neighbours: coexistence between jaguars and pumas in a harsh environment. Journal of Arid Environments 146:27–34.

Bacon, C. S., D. Silvestro, C. Jaramillo, B. T. Smith, P. Chakrabarty and A. Antonelli. 2015. Biological evidence supports an early and complex emergence of the Isthmus of Panama. Proceedings of the National Academy of Sciences 112:6110–6115.

Bauder, J. M., D. R. Breininger, M. R. Bolt, M. L. Legare, C. L. Jenkins, B. B. Rothermel and K. McGarigal. 2018. Multi-level, multi-scale habitat selection by a wide-ranging, federally threatened snake. Landscape Ecology 33:743–763.

Bellamy, C., C. Scott and J. Altringham. 2013. Multiscale, presence‐only habitat suitability models: fine‐resolution maps for eight bat species. Journal of Applied Ecology 50:892– 901.

Breiman, L. 2001. Random forests. Machine learning 45:5–32.

Breiman, L. and A. Cutler. 2003. Setting up, using, and understanding random forests V4. 0. University of California, Department of Statistics.

Brossard, K. and J. A. Pritz. 2013. Human-Jaguar Conflict in the Alto Chagres National Park: A Socio-Ecological Study. McGill University.

Busing, R.T., P. S. White, MacKenzie MD. 1993. Gradient analysis of old spruce-fir forest of the Great Smokey Mountains circa 1935. Canadian Journal of Botany 71:951–958.

Cavalcanti, S., S. Marchini, A. Zimmermann, E. M. Gese and D. W. Macdonald. 2010. Jaguars, livestock, and people in Brazil: realities and perceptions behind the conflict. Pages 383–402 in The biology and conservation of wild felids. D. Macdonald and A. Loveridge, editors. Oxford University Press, Oxford, United Kingdom.

Chapin, M. 2000. Defending Kuna Yala: PEMASKY, the study project for the management of the wildlands of Kuna Yala, Panama. A case study for shifting the power: decentralization and biodiversity conservation (No. 333.95097287 C463). Biodiversity Support Program, Washington, DC (EUA).

Condit, R. et al. 2001.The status of the Panama Canal watershed and its biodiversity at the beginning of the 21st century. BioSciences 51:389–398.

Corrales, L., C. Bouroncle and J. C. Zamora. 2015. An overview of forest biomes and ecoregions of Central America. Pages 33–54 in Climate Change Impacts on Tropical Forests in Central America. Routledge.

Crawshaw, P. G. and H. B. Quigley HB. 1991. Jaguar spacing, activity, and habitat use in a seasonally flooded environment in Brazil. Journal of the Zoological Society of London 223:357–370.

Crawshaw Jr., P. G. and H. B. Quigley. 2002. Hábitos alimentarios del jaguar y el puma en el Pantanal, Brasil, con implicaciones para su manejo y conservación. Páginas 223–235. El jaguar en el nuevo milenio. R. A., C. Medellín, C. Equihua, C. L. B. Chetkiewicz, P. G. Crawshaw Jr., A. Rabinowitz, K. H. Redford, J. G. Robinson, E. W. Sanderson and A. B. Taber, editors. Universidad Nacional Autónoma de México, Wildlife Conservation Society y Fondo de Cultura Económica. México, DF.

Crooks, K. R. 2002. Relative sensitivities of mammalian carnivores to habitat fragmentation. Conservation Biology 16:488–502.

Cullen Jr., L. 2006. Jaguars as landscape detectives for conservation in the Atlantic Forest of Brazil. University of Kent.

Cullen Jr., L., D. A. Sana, F. Lima, K. C. de Abreu and A. Uezu. 2013. Selection of habitat by the jaguar, Panthera onca (Carnivora: Felidae), in the upper Paraná River, Brazil. Zoologia (Curitiba) 30:379–387.

Cushman, S. A., E. A. Macdonald, E. L. Landguth, Y. Malhi and D. W. Macdonald. 2017. Multiple-scale prediction of forest loss risk across Borneo. Landscape Ecology 32:1581– 1598.

Cushman, S. A. and T. N. Wasserman. 2018. Landscape Applications of Machine Learning: Comparing Random Forests and Logistic Regression in Multi-Scale Optimized Predictive Modeling of American Marten Occurrence in Northern Idaho, USA. Pages 185–203 in Machine Learning for Ecology and Sustainable Natural Resource Management. Springer, Cham.

Cutler, D. R., T. C. Edwards Jr., K. H. Beard, A. Cutler, K. T. Hess, J. Gibson and Lawler J. 2007. Random forests for classification in ecology. Ecology 88:2783–2792.

Davidson, Z., M. Valeix, A. J. Loveridge, J. E. Hunt, P. J. Johnson, H. Madzikanda and D. W. Macdonald. 2012. Environmental determinants of habitat and kill site selection in a large carnivore: scale matters. Journal of Mammalogy 93:677–685.

Davis, M. L., M. J. Kelly and D. F. Stauffer. 2011. Carnivore co‐existence and habitat use in the Mountain Pine Ridge Forest Reserve, Belize. Animal Conservation 14:56–65.

De´ath, G. and K. E Fabricius. 2000. Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 81:3178–3192. da Fonseca, G. A. and J. G. Robinson. 1990. Forest size and structure: competitive and predatory effects on small mammal communities. Biological Conservation 53:265–294. de la Torre, J. A. and R. A. Medellín. 2011. Jaguars Panthera onca in the Greater Lacandona Ecosystem, Chiapas, Mexico: population estimates and future prospects. Oryx 45:546–553. de la Torre, J. A., J. M. Núñez and R. A. Medellín. 2017. Habitat availability and connectivity for jaguars (Panthera onca) in the Southern Mayan Forest: Conservation priorities for a fragmented landscape. Biological Conservation 206:270–282. de la Torre, J.A., M. Rivero, G. Camacho and L. A. Álvarez-Márquez. 2018. Assessing occupancy and habitat connectivity for Baird’s tapir to establish conservation priorities in the Sierra Madre de Chiapas, Mexico. Journal for Nature Conservation 41:16–25. de la Torre, J. A. and M. Rivero. 2019. Insights of the Movements of the Jaguar in the Tropical Forests of Southern Mexico. Pages 217–241 in Movement Ecology of Neotropical Forest Mammals. Springer, Cham. de Souza, J. C., R. M. da Silva, M. P. R. Gonçalves, R. J. D. Jardim and S. H. Markwith. 2018. Habitat use, ranching, and human-wildlife conflict within a fragmented landscape in the Pantanal, Brazil. Biological Conservation 217:349–357.

Dettman, S. 2006. The mesoamerican biological corridor in Panama and Costa Rica: integrating bioregional planning and local initiatives. Journal of Sustainable Forestry 22:15–34.

Dickson, B. G. and P. Beier. 2002. Home-range and habitat selection by adult cougars in southern California. The Journal of Wildlife Management 66:1235–1245.

Dobbins, M. T., M. K. Steinberg, E. N. Broadbent and S. J. Ryan. 2018. Habitat use, activity patterns and human interactions with jaguars Panthera onca in southern Belize. Oryx 52:276–281.

Evans, J. S., Murphy MA. 2018. rfUtilities. R package version 2.1-3, https://cran.r- project.org/package=rfUtilities.

Foerster, C. R. and C. Vaughan. 2002. Home range, habitat use, and activity of Baird’s Tapir in Costa Rica. Biotropica 34:423–437.

Fort, J. L. 2016. Large Carnivore Occupancy and Human-Wildlife Conflict in Panama Large Carnivore Occupancy And Human-Wildlife Conflict In Panamá. Doctoral dissertation. Southern Illinois University Carbondale.

Fort, J. L., C. K. Nielsen, A. D. Carver, R. Moreno and N. F. Meyer. 2018. Factors influencing local attitudes and perceptions regarding jaguars Panthera onca and National Park conservation in Panamá. Oryx 52:282–291.

Foster, R. 2008. The ecology of jaguars (Panthera onca) in a human-influenced landscape Doctoral dissertation. University of Southampton.

Foster, R. J., B. J. Harmsen and C. P. Doncaster. 2010. Habitat use by sympatric jaguars and pumas across a gradient of human disturbance in Belize. Biotropica 42:724–731.

Gentry, A. H. 1988. Changes in plant community diversity and floristic composition on environmental and geographical gradients. Annals of the Missouri Botanical Garden 75:1– 34.

Gese, E. M., P. A. Terletzky, S. M. Cavalcanti and C. M. Neale. 2018. Influence of behavioral state, sex, and season on resource selection by jaguars (Panthera onca): Always on the prowl?. Ecosphere 9:e02341.

Guisan, A. and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993–1009.

Harmsen, B. J., R. J. Foster, S. C. Silver, L. E. T. Ostro and C. P. Doncaster. 2011. Jaguar and puma activity patterns in relation to their main prey. Mammalian. Biology 76:320–324.

Hernandez, P. A., C. H. Graham, L. L. Master and D. L. Albert. 2006. The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29:773–785.

Holdridge, L. R. 1967. Life Zone Ecology. Revised edition. Page 206. Tropical Science Center, San Jose, Costa Rica.

Holdridge, L. R., W. C. Grenke. 1971. Forest environments in tropical life zones: a pilot study. Forest environments in tropical life zones: a pilot study. Pergamon Press, Oxford, UK.

Horning, N. 2010. Random Forests: An algorithm for image classification and generation of continuous fields data sets. In Proceedings of the International Conference on Geoinformatics for Spatial Infrastructure Development in Earth and Allied Sciences. Volume 911. Osaka, Japan.

Ibáñez, R. et al. 2002. An ecosystem report on the Panama Canal: Monitoring the status of the forest communities and the watershed. Environmental Monitoring and Assessment 80:65– 95.

Jakovac, C. C., M. Peña-Claros, T. W. Kuyper, F. Bongers. 2015. Forest Resilience, Tipping Points and Global Change Processes: Loss of secondary-forest resilience by land-use intensification in the Amazon. Journal of Ecology 103:67–77.

Jankowski, J. E., A. L. Ciecka, N. Y. Meyer, K. N. Rabenold. 2009. Beta diversity along environmental gradients: implications of habitat specialization in tropical montane landscapes. Journal of Animal Ecology 78:315–327.

Jędrzejewski, W. et al. 2017. Density and population structure of the jaguar (Panthera onca) in a protected area of Los Llanos, Venezuela, from 1 year of camera trap monitoring. Mammal Research 62:9–19.

Jędrzejewski, W. et al. (2018). Estimating large carnivore populations at global scale based on spatial predictions of density and distribution–Application to the jaguar (Panthera onca). PloS One 13:e0194719.

Jin, X. M., Y. K. Zhang, M. E. Schaepman, J. G. P. W. Clevers and Z. Su. 2008. Impact of elevation and aspect on the spatial distribution of vegetation in the Qilian mountain area with remote sensing data. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences 37:1385–1390.

Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65–71.

Johnson, C. J., M. S. Boyce, R. Mulders, A. Gunn, R. J. Gau, H. D. Cluff and R. L. Case. 2004. Quantifying patch distribution at multiple spatial scales: applications to wildlife-habitat models. Landscape Ecology 19:869–882.

Krishnamurthy, R., S. A. Cushman, M. S. Sarkar, M. Malviya, M. Naveen, J. A. Johnson and S. Sen. 2016. Multi-scale prediction of landscape resistance for tiger dispersal in central India. Landscape Ecology 31:1355–1368.

Kursar, T. A., B. M. Engelbrecht and M. T. Tyree. 2005. A comparison of methods for determining soil water availability in two sites in Panama with similar rainfall but distinct tree communities. Journal of Tropical Ecology 21:297–305.

Leopold, C. R. and S. C. Hess SC. 2013. Multi-scale habitat selection of the endangered Hawaiian Goose. The Condor 115:17–27.

Levin, S. A. 1992. The problem of pattern and scale in ecology: the Robert H. MacArthur award lecture. Ecology 73:1943–1967.

Macdonald, D. W. et al. 2018. Multi-scale habitat selection modeling identifies threats and conservation opportunities for the Sunda clouded leopard (Neofelis diardi). Biological Conservation 227:92–103.

Mateo Sanchez, M. C., S. A. Cushman and S. Saura. 2014. Scale dependence in habitat selection: the case of the endangered brown bear (Ursus arctos) in the Cantabrian Range (NW Spain). International Journal of Geographical Information Science 28:1531–1546.

McBride, R. T. and J. J. Thompson. 2018. Space use and movement of jaguar (Panthera onca) in western Paraguay. Mammalia 82:540–549.

McGarigal, K., H. Y. Wan, K. A. Zeller, B. C. Timm and S. A. Cushman. 2016. Multi-scale habitat selection modeling: a review and outlook. Landscape Ecology 31:1161–1175.

Meyers, C. 1969. The Ecological Geography of Cloud Forest in Panama. American Museum Novitates, No 2396. The American Museum of Natural History, New York City, New York.

Minter, J. E. 1948. The Chagres: River of Westward Passage. Rinehart & Company, New York.

Morato, R. G. et al. 2016. Space use and movement of a neotropical top predator: The endangered jaguar. PLoS One 11:e0168176.

Moyer, J. and E. Shebell. 2014. Human-jaguar competition and conflict: a case study in the Colón Biological Corridor. McGill University, Canada.

Nuñez-Perez, R. and B. Miller. 2019. Movements and Home Range of Jaguars (Panthera onca) and Mountain Lions (Puma concolor) in a Tropical Dry Forest of Western Mexico. Pages 243–262 in Movement Ecology of Neotropical Forest Mammals. Springer, Cham.

Parker, T., J. Carrion, R. Samudio. 2004. Biodiversity and Tropical Forestry Assessment Of The Usaid/Panama Program: Environment, Biodiversity, Water, and Tropical Forest Conservation, Protection, and Management in Panama: Assessment and Recommendations. Chemonics International, Inc. Washington, D.C.

Prasad, A. M., L. R. Iverson and A. Liaw. 2006. Newer classification and regression tree techniques: bagging and random forests for ecological prediction. Ecosystems 9:181–199.

Quigley, H. B. 1987. Ecology and conservation of the jaguar in the Pantanal region, Mato Grosso do Sul, Brazil. Tesis de Doctorado, Universidad de Idaho, Moscow, Idaho, EUA.

Quigley, H., R. Foster, L. Petracca, E. Payan, R. Salom and B. Harmsen. 2017. Panthera onca (errata version published in 2018). The IUCN Red List of Threatened Species 2017. Available at http://www.iucnredlist.org.

Rabinowitz, A. R. and B. G. Nottingham. 1986. Ecology and behaviour of the Jaguar (Panthera onca) in Belize, Central America. Journal of Zoology 210:149–159.

Rabinowitz, A. and K. A. Zeller. 2010. A range-wide model of landscape connectivity and conservation for the jaguar, Panthera onca. Biological Conservation 143:939–945.

Rettie, W. J. and F. Messier. 2000. Hierarchical habitat selection by woodland caribou: its relationship to limiting factors. Ecography 23:466–478.

Reyna-Hurtado, R. and G. Tanner. 2007. Ungulate relative abun-dance in hunted and non-hunted sites in Calakmul forest (SouthernMexico).Biodiversity and Conservation 16:743–756.

R Development Team. 2017. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available at http://www.R- project.org.

Rondinini, C. et al. 2011. Global habitat suitability models of terrestrial mammals. Philosophical Transactions of the Royal Society B: Biological Sciences 366:2633–2641.

Scognamillo, D., I. E. Maxit, M. Sunquist and J. Polisar. 2003. Coexistence of jaguar (Panthera onca) and puma (Puma concolor) in a mosaic landscape in the Venezuelan llanos. Journal of Zoology 259:269–279.

Silveira, L. 2004. Ecologia comparada de onça-pintada (Panthera onca) e onça-parda (Puma concolor) no cerrado do Brasil. PhD dissertation, Distrito Federal, Universidade de Brasília, Brasília, Brazil.

Silver, S. C., L. E. T. Ostro, L. K. Marsh, L. Maffei, A. J. Noss, M. J. Kelly, R. B. Wallace, H. Gómez and G. Ayala. 2004. The use of camera traps for estimating jaguar Panthera onca abundance and density using capture/recapture analysis. Oryx 38:148–154.

Sollmann, R., M. M. Furtado, H. Hofer, A. T. A. Jácomo, N. M. Tôrres and L. Silveira. 2012. Using occupancy models to investigate space partitioning between two sympatric large predators, the jaguar and puma in central Brazil. Mammalian Biology 77:41–46.

Sunarto, S., M. J. Kelly, K. Parakkasi, S. Klenzendorf, E. Septayuda and H. Kurniawan. 2012. Tigers need cover: multi-scale occupancy study of the big cat in Sumatran forest and plantation landscapes. PloS One 7:e30859.

Sunquist, M. and F. Sunquist. 2002. Wild Cats of the World. The University of Chicago Press, London.

Thompson, C. M. and K. McGarigal. 2002. The influence of research scale on bald eagle habitat selection along the lower Hudson River, New York (USA). Landscape Ecology 17:569– 586.

Thuiller, W., C. H. Albert, A. Dubuis, C. Randin and A. Guisan. 2010. Variation in habitat suitability does not always relate to variation in species’ plant functional traits. Biology Letters 23:120–123.

Timm, B. C., K. McGarigal, S. A. Cushman, and J. L. Ganey. 2016. Multi-scale Mexican spotted owl (Strix occidentalis lucida) nest/roost habitat selection in Arizona and a comparison with single-scale modeling results. Landscape Ecology 31:1209–1225.

Tobler, M. W. and G. V. Powell. 2013. Estimating jaguar densities with camera traps: problems with current designs and recommendations for future studies. Biological conservation 159: 109–118.

Tôrres, N. M., P. De Marco, T. Santos, L. Silveira, A. T. de Almeida Jácomo and J. A. Diniz‐Filho. 2012. Can species distribution modelling provide estimates of population densities? A case study with jaguars in the Neotropics. Diversity and Distributions 18:615– 627.

VanDerWal, J., L. P. Shoo, C. N. Johnson and S. E. Williams. 2009. Abundance and the environmental niche: environmental suitability estimated from niche models predicts the upper limit of local abundance. The American Naturalist 174:282–291.

Vergara, M., S. A. Cushman, F. Urra and A. Ruiz-González. 2016. Shaken but not stirred: multiscale habitat suitability modeling of sympatric marten species (Martes martes and Martes foina) in the northern Iberian Peninsula. Landscape Ecology 31:1241–1260.

Wan, H. Y., K. McGarigal, J. L. Ganey, V. Lauret, B. C. Timm and S. A. Cushman. 2017. Meta- replication reveals nonstationarity in multi-scale habitat selection of Mexican Spotted Owl. Condor 119:641–658.

Wasserman, T. N., S. A. Cushman and D. O. Wallin. 2012. Multi scale habitat relationships of Martes americana in northern Idaho, USA. Res. Pap. RMRS-RP-94. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station.

Wiens, J. A. 1989. Spatial Scaling in Ecology. Functional Ecology 3:385.

Zeller, K. A., S. Nijhawan, R. Salom-Pérez, S. H. Potosme and J. E. Hines. 2011. Integrating occupancy modeling and interview data for corridor identification: a case study for jaguars in Nicaragua. Biological Conservation 144:892–901.

Zeller, K. A., A. Rabinowitz, R. Salom-Perez and H. Quigley. 2013. The jaguar corridor initiative: a range-wide conservation strategy. Pages 629–59 in Molecular population genetics, evolutionary biology and biological conservation of Neotropical carnivores. New York, Nova Science Publishers, Inc.

Zeller, K. A., T. W. Vickers, H. B. Ernest and W. M. Boyce. 2017. Multi-level, multi-scale resource selection functions and resistance surfaces for conservation planning: Pumas as a case study. PloS One 12:e0179570.

Zimmermann, A., M. J. Walpole, N. Leader-Williams. 2005. Cattle ranchers' attitudes to conflicts with jaguar Panthera onca in the Pantanal of Brazil. Oryx 39:406–412.

Appendices

Appendix A: Examples of the three-land tenure areas Narganá Protected Wildland, Chagres National Park and the MVP.

Appendix B: Twenty-one predictor variables associated with the jaguar multi-scale habitat suitability models. Variable code name used for the analysis and the descriptive name used for discussion are included for reference.

*Indicates variables that were selected for the final analyses. The associated characteristics of the variable: bAgropec: Heterogeneous area of agricultural production, cHydro: Rivers, ponds, lakes, dInfraest: Major roads, airport, ePeople: Housing, buildings/urban area, fRastrojo: Shrubs and bushes.

Appendix C. Annual model OOB error rate table for jaguars.

Appendix D. Wet season model OOB error rate table for jaguars.

Appendix E. Dry season model OOB error rate table for jaguars.

Appendix F. Photographs taken of jaguars by camera traps in the study area.

Chapter III: Path of the Puma (Puma concolor): A Multi-Scale Analysis of Seasonal Habitat Selection in the Narrowest Section of Panamá

Abstract

Habitat loss is considered the most critical threat to puma (Puma concolor) populations, particularly in the Neotropics where landscape modification is happening rapidly. In Panamá, little is known about puma habitat use. Yet, understanding how pumas respond to the landscape across multiple spatial scales is an important metric for conservation endeavors of the species. In this study, I used presence-absence data from a two-year (2016 – 2018) camera-trap study to identify the environmental and anthropogenic variables and associated scales influencing puma habitat selection. By employing an optimized, multi-scale statistical approach and the non- parametric algorithm, Random Forests, I modeled the probability of puma occurrence in the narrowest section of eastern Panamá. According to the multi-scale models, primary forest and elevation were the most constant predictors of habitat selection at fine scales (patch selection).

Pumas responded to agropecuary (i.e., agriculture and livestock) and coniferous forest at broad scales (home range). Hydrologic features, secondary forest, and aspect were consistently selected at both fine and broad scales. Seasonal differences in habitat suitability were apparent. In the wet season, pumas incorporated the agropecuary landscape at a broad scale, likely due to prey availability and interspecific competition with jaguars (Panthera onca). Topographic and hydrologic features provided important resources (e.g., food and shelter) across multiple scales.

In the dry season model, habitat suitability increased with more primary forest in the landscape.

The models developed in this study using a multi-scale approach provide insight into puma habitat selection in a human-modified landscape and contribute to the limited knowledge of

69 puma occurrence in the region. The results can be considered a guide for priority conservation actions involving land management practices that conserve the integrity of critical habitat in

Panamá

Introduction

The inherent traits of large terrestrial carnivores, including high trophic level, extensive and continuous habitat requirements and low densities (Sunquist & Sunquist 2002; Ripple et al.

2014), make them highly susceptible to landscape level changes, habitat loss, and human disturbance (Ceballos 2005; Schipper et al. 2008; Zanin et al. 2015). Moreover, humans and large carnivores often have similar habitat requirements for land and resources, which often leads to human-carnivore competition (Treves & Karanth 2003; Broekhuis et al. 2017). In this context, understanding the relationship between the landscape and species spatial distribution, ecological processes, and species persistence, is vital to carnivore conservation (Tilman et al. 1994; Haskell et al. 2013).

The application of habitat suitability models show promise for understanding carnivore distribution and habitat use (Torres et al. 2012; Jędrzejewski et al. 2017; Hearn et al. 2018;

Macdonald et al. 2018). Gaining a better understanding of species-environment relationships of large, wide-ranging terrestrial carnivores is effective for predicting species distribution across human-dominated landscapes (Gonzalez-Borrajo et al. 2017). Such models play a critical role in conservation decision-making and ecological or biogeographical inference (Boyce & McDonald

1999; Guisan & Zimmermann 2000; Manly et al. 2002; Guisan & Thuiller 2005), especially in areas that harbor complex carnivore guilds, but at the same time witness severe declines.

One species that habitat suitability models can shed light on is the puma (Puma concolor) in

Panamá. Because of the puma’s ability to adapt, they are considered the most widespread felid in the Americas, ranging from Patagonia to Northern British Columbia (Sunquist & Sunquist 2002).

Puma ecology and behavior has been well documented in North America (Logan et al. 1999;

Gloyne et al. 2001; Zeller et al. 2014, 2017; Gustafson et al. 2018; Elbroch et al. 2018; Criffield et al. 2018). In the Neotropics, where puma is second in size to the sympatric jaguar (Panthera onca), much research has been generated regarding the temporal and spatial overlap of these two species (Scognamillo et al. 2003; de Angelo et al. 2011; Fort 2016; de la Torre et al. 2017; Rowe

2017). Research has also focused on competitive release, where pumas are known to persist in areas where jaguars have been eradicated (Moreno et al. 2006). However, detailed information on their regional distribution and habitat selection has been lacking in the Neotropics, hindering conservation efforts in the region (Nowell & Jackson 1996; Kelly et al. 2008; Foster et al. 2010;

Gonzalez-Borrajo et al. 2017).

Given the paucity of relevant information concerning pumas living in the tropical forest-human interface in Panamá, it is therefore, from an ecological and conservation perspective, important to assess the spatial distribution of this meso-predator. The objectives of my study were to:

1. Explore the relationships between puma distribution patterns and the set of environmental

and anthropogenic predictor variables across a gradient of scales.

2. Determine patterns and changes in puma spatial distribution due to seasonal variation.

3. Produce high-resolution, spatial distribution maps incorporating seasonal variation for

puma conservation and management purposes.

I used an optimized multi-scale habitat modeling approach to determine key environmental and anthropogenic variables influencing puma habitat selection in the narrowest section of Panamá. I use presence-absence data taken from camera traps to develop three habitat suitability models

(annual, wet season, and dry season). The models consider a formal scale optimization of an organism’s response to environmental variables (McGarigal et al. 2016), which can provide recommendations for the conservation of pumas in the region.

I selected the Random Forests algorithm (Breiman 2001) as my statistical method for habitat suitability modeling. Using Random Forests, I explored the relationship between the presence of pumas and the environmental factors deemed necessary for the occurrence of the species. Several studies have shown that Random Forests models, a classification tree-based model, often reach top predictive performances compared to other methodologies (e.g., LDA, logistic regression)

(Cutler et al. 2007; Cushman et al. 2017; Cushman & Wasserman 2018).

I hypothesized, that using a scale-optimized modeling approach on habitat selection would show that pumas respond at different scales to different environmental and anthropogenic features in the landscape. I predicted that pumas would avoid human settlements at broad scales. As described in previous studies, pumas may select against areas with high human activity (Sweanor et al. 2008; Burdett et al. 2010; Foster et al. 2010; Davis et al. 2011). This first hypothesis is based on the ongoing understanding that ecological patterns result from processes occurring at multiple spatial and temporal scales (Wiens 1989; McGarigal et al. 2016; Zeller et al. 2016). Yet, to date, factors reported as important determinants of puma-habitat occurrence have been based on analyses at fixed spatial scales (Wasserman et al. 2012).

My second hypothesis relates to the evidence that puma occurrence is influenced by the seasonal distribution of available prey species (Scognamillo et al. 2003; Foster 2008; Sollmann et al.

2012; Cullen Jr et al. 2013; Fort 2016; de la Torre et al. 2017). Thus, I hypothesized that pumas would exhibit behavioral plasticity in terms of habitat requirements during the wet and dry seasons and that pumas would differ in the spatiotemporal scales at which they select habitat during the wet and dry seasons. I predicted that pumas would select agropecuary (i.e., livestock and agriculture) in the wet season at a broad scale. This hypothesis follows the report that pumas in western Panamá exhibited an increase in detection in association with collared peccaries

(Pecari tajacu) during the wet season, suggesting a shift in prey distribution (Fort 2016).

Methods

Study area – Panamá is the fourth smallest Central American nation of 3.6 million people (Davis

2010), spanning about 75,717 km2 on an east-west axis (Figure 10). Panamá occupies the

southernmost section of the Mesoamerican Biological Corridor (MBC), one of the largest

bioregional conservation programs in the world (Dettman 2006). Spanning over 767,990 km2,

the MBC extends from Selva Maya in Mexico’s South-East, to Darien in eastern Panamá

(Dettman 2006).

The physiographic features that characterize the country include the Talamanca Mountains that

abut the Serrania de Tabasara to the west, and the Cordillera de San Blas and the Serrania del

Darien to the east (Myers 1969). Together, the range of mountains form the Cordillera de San

Humid tropical and premontane forests dominate the upper reaches of the mountains. These forests contain a mix of old-growth and secondary-growth stands that have canopy trees reaching

November and the dry season from December to April. Annual rainfall amounts range between

1,700 mm – 4,000 mm on the Pacific and Atlantic coasts respectively (Condit et al. 2001; Ibáñez et al. 2002).

Biogeographical importance of the study area – The study area was selected for its strategic

location and ecological significance. It is part of the largest remaining stretches of contiguous

rainforest in the Tumbes-Chocó-Magdalena eco-region, one of the top ecological hotspots on

Earth (World Bank 1998). The location of the study area is also at the convergence of three

important protected areas in eastern Panamá: Chagres National Park, the Guna Yala Indigenous

Territory, and the Mamoní Valley Preserve, a large private land tenure that serves as a buffer and

stepping-stone habitat. Together, the pervasive advance of agriculture and illegal poaching

threaten this area. Beyond that, the study is in one of the narrowest stretches (~15 km wide) of

primary habitat remaining for puma dispersal in the country.

Study species – The puma is listed as a species of Least Concern, but some subpopulations are considered threatened or declining due to anthropogenic pressure (Kelly et al. 2008; de Angelo et al. 2011; Nielsen et al. 2015). Though unconfirmed, the puma population is thought to be higher in Central and South America than in North America. During the early 1990s, the puma population was estimated at 3,500 – 5,000 in Canada and 10,000 in the western US (Nowell &

Jackson 1996; Nielsen et al. 2015). The puma is a solitary felid, typically associated with structurally complex habitats that enable them to remain inconspicuous, using habitat structure to their advantage to stalk and ambush prey (Lamprecht 1978; Elbroch & Wittmer 2012). They are opportunistic predators and will switch prey species depending on availability. Prey species range from white-nosed coati (Nasua narica) and armadillos (Dasypus novemcinctus) to deer

(Cervidae), feral horses (Equus caballus), and domestic animals (Logan & Sweanor 2001;

Monroy-Vilchis 2009; Smith et al. 2016; Ávila-Nájera 2018). Pumas tend to have a smaller body size in the tropics than in temperate areas and they select prey at least half of their body weight or smaller (Rowe 2018). They typically stalk their prey, and therefore, have an affinity for using dense understory as cover. Their predatory activities play a role in maintaining biodiversity and the structural integrity of the tropical forest ecosystems (Terborgh et al. 1999). Also, considered subordinate to the sympatric jaguar, pumas may suffer the costs of having to switch habitat use or be displaced at food resources (Elbroch & Kusler 2018). Pumas have the largest home ranges of felids in the Neotropics (Gonzalez-Borrajo et al. 2017), except for Brazil where they are equal to the jaguar (Silveira 2004). Density estimates in Central America are known from Belize where they are estimated at 1.42/100 km2 (Rich et al. 2014). In Western Mexico estimates are between

3–5/100 km2 (Nuñez et al. 1998).

Camera-trap surveys – The survey area was selected for its biological significance, it included contiguous sections of three distinct areas under different tenure regimes, and therefore different dynamics of land cover and land use history: (1) Narganá Protected Wildland, (2) Chagres

National Park, and (3) The Mamoní Valley Preserve (Figure 10, Appendix G). Together, the survey area comprised ~200 km2. The terrain in this section is rugged with altitudes ranging from

400 m to 1,000 m a.s.l. Annual average precipitation during the wet season is ~ 4000 mm which

reduces to ~700 mm in the dry season.

Narganá Protected Wildland, created in 1994, falls within the semi-autonomous territory of the

Indigenous Guna Yala people. It is co-managed by the Guna Yala Congress and the Ministry of

the Environment (MiAmbiente) (Parker et al. 2004). Narganá Protected Wildland constitutes

51% of the study area and is situated within the general ridge system of the Cordillera de San

Blas. It includes the Caribbean side of the Cordillera de San Blas, which has a very wet climate

owing to showers spawned by moisture-laden trade winds. Most of the forest within this area is

largely undisturbed by human activities, but patches of fallow (locally known as rastrojo)

(pioneer forests 40 – 50 yrs. old fallow) and secondary mature forest are apparent. These patches

are the result of settlers clearing the land for homesteads and agriculture in the late 1900s

(Chapin 2000) (Figure 10).

Figure 10. Location of study site and camera-trap stations distributed in a 2 x 2 km grid system used to determine puma habitat selection. It comprises three-land tenure areas Narganá Protected Wildland, Chagres National Park and the Mamoní Valley Preserve in the country of Panamá. Phase II data collection was not part of this study.

Chagres National Park, established in October 1984, is part of the Panamá Canal Watershed

The Mamoní Valley Preserve is on the leeward side of the Cordillera de San Blas facing the

It is comprised of 115 km2 of privately-owned lands. There are four villages within the preserve

with an estimated population of 400 people whose economy is largely agricultural, and ranching

based. The land cover in the preserve is composed of a mixed matrix that includes secondary

forest, fallow (with various stages of maturity), tree farms, pasture and agropecuary practices

(e.g., agriculture and livestock) (Figure 10).

Between July 2016 and September 2018, I used motion-activated camera traps to collect presence-absence data on puma occurrence under a scientific research permit No. SE/A-14-17

obtained through the Ministerio de Ambiente de Panamá. Camera traps were deployed

throughout an area of approximately 200 km2 within the Panamá Province, Republic of Panamá

(9°19’ 35.5261” E/-79°08’ 47.8356” N) (Figure 10). The original study was designed for jaguar

occupancy estimates, but pumas are similar in size and habits, so the study design is likely

appropriate for pumas (Harmsen et al. 2009).

Following Silver et al. (2004), a 2 x 2 km grid cell overlay was used to place 48 camera stations.

Using a hybrid design, I simultaneously deployed 36-single and 12 paired stations set

approximately 2 km apart throughout the study grid. Camera spacing aimed to maximize photo

capture probability, with two or more cameras per average home-range area (Kelly et al. 2008).

Based on previous studies in Mexico and South America, puma home range size in tropical and

subtropical habitats varies from 21 km2 to 60 km2 (Nuñez-Perez & Miller 2019). Two cameras

78 are favored over one camera per station, as it reduces the risk of lost trap-nights from camera failure. However, the number of cameras were restricted by limited resources.

Throughout the study period, I used Bushnell Trophy Cam Aggressor and Cuddeback C-E cameras, programmed for picture and video function. Cameras were placed approximately 30 cm

– 50 cm above the ground and adjusted in response to local topography. Cameras were deployed year-round (wet season-May to November/dry season-December to April) and programmed to operate 24h/day. During the dry season four camera stations were removed due to high levels of vandalism. Cameras were checked every three months to ensure functionality (e.g., replace memory cards, batteries, theft). Based on site accessibility, camera stations were randomly established on pre-existing human trails, game trails, rivers, streams or creeks. Larger water bodies such as lakes and ponds were not present in the study area. Grid cell size was based on attaining the largest grid feasible within logistical and financial constraints.

(Wasserman et al. 2012).

To create habitat suitability models for pumas I used the ‘randomForest’ package in R statistical software (R Development Team 2017). Random Forests is a classification and regression tree based bootstrap method that provides well-supported predictions with large numbers of independent variables (Breiman 2001; Cutler et al. 2007). It is useful for modeling nonlinear

79 data, such as presence-absence data, with variables that likely interact in a hierarchical fashion

(De’ath & Fabricius 2000).

The Random Forests classifier includes two random processes that improve the predictive power of the classification. When building each decision tree in the forest (ensemble), at each tree branching node a subset of potential predictor variables is randomly selected on which the data are split (to create children nodes), thereby reducing the correlation between the trees, resulting in a lower error rate (Horning 2010). In addition, the Random Forests algorithm randomly selects a bootstrap sample from a subset of the total training data available to build each tree. A third of this subset is left out-of-bag (OOB) and not used to construct the tree. This OOB subset is then run through the constructed tree to cross-validate the classification, thereby deriving an unbiased estimate of the test-set error (Breiman & Cutler 2003). Random Forests models limit overfitting without increasing much error due to bias, and for classification and regression, are superior to most methods commonly used (e.g., logistic regression and linear discriminant analysis (LDA))

(Cutler et al. 2007). I first conducted an error convergence pilot test using 10,000 trees. Based on the results, I determined that the error stabilized between 500 and 1500 trees. For all my models,

I opted to use 2,000 trees to be conservative.

Predictor Variables –The predictor variables included in the models were selected based on a comprehensive review on previous studies of puma ecology, and which I believed would be strongly related to puma occurrence (Zeller et al 2011; Fort 2016; Morato et al. 2016; Astete et al. 2017; de la Torre et al. 2018). The puma is an adaptable, solitary felid that exhibits behavioral plasticity, and inhabits a variety of forest types found throughout their range. Therefore, the variable classes used to fit the models included a wide-range of factors that are representative of

80 the behavioral strategies and resource use patterns thought to be important in influencing puma distribution: topographic, landscape composition, land use, water body and climate (Table 3,

Appendix H).

Topographic covariates were derived from Aster Global DEM V002 (NASA) using a 30-m resolution digital elevation model (Table 3). The relative slope position measures the relative position of the focal pixel within a defined extent on a gradient from ground bottom to the top of a ridge, and the topographical roughness measures the topographical complexity of the landscape within a defined focal extent (Krishnamurthy et al. 2016). I calculated both the topographical roughness and relative slope position using the Geomorphometry & Gradient Metrix Toolbox in

ArcGIS (ESRI, Redlands, CA). The landscape composition variables were determined by the

interpretation of the Cobertura Boscosa y Uso de Tierra 2012 (land cover) map developed by the

Ministry of Environment of Panamá (MiAmbiente) (Table 3). Percent tree cover was derived by

the USGS Global Tree-Canopy Cover Circa 2010 (Table 3). Climate variables were derived from

Global Climate Data - WorldClim.org (annual average past 30 years) (Table 3). Finally, water

body and land use features were determined by interpreting Cobertura Boscosa y Uso de Tierra

2012 maps from MiAmbiente (Figure 11). All habitat variables were mapped at 30 m resolution, which represented the finest resolution in the source data. The spatial grain of the analyses was held constant at 30 m in the multi-scale analyses described below.

Table 3. Description of variables used in the multi-scale puma habitat selection model developed for the study area. Variables were classified into five groups: topographic, landscape composition, land use, hydrological and climate.

Class Variable Variable Code Data source Topographic Aspect Aspect_cos NASA: Earth Data Elevation Dem NASA: Earth Data Slope Slope NASA: Earth Data Topographic Roughness Rou_new NASA: Earth Data

Landscape Composition Agropecuary Agropec MiAmbiente* Banana Plantation Banana MiAmbiente Broadleaf Forest Forlatif MiAmbiente Coniferous Forest Forconi MiAmbiente Herbaceous Vegetation Vegherb MiAmbiente Mixed Horticulture Hortmix MiAmbiente Pasture Pasto MiAmbiente Primary Forest Mix_forest MiAmbiente Fallow Rastrojo MiAmbiente Secondary Forest Foremix MiAmbiente Tree Cover Tree_r USGS Global Tree

Land use Human Settlement People MiAmbiente Infrastructure Infraest MiAmbiente

Hydrological Hydrologic Feature Hydro MiAmbiente

Climate Annual Precipitation Prec_all Global Climate Data Dry Season Precipitation P_dry Global Climate Data Wet Season Precipitation P_wet Global Climate Data

* Ministry of Environment: Mapa de cobertura boscosa y uso de la tierra, 2012 (land cover map)

Figure 11. Location of the study area indicating land cover types in eastern Panamá. The map depicts the agricultural frontier abutting the remaining stretch of forest within the narrowest section of Panamá. Taken from Cobertura Boscosa y Uso de Tierra 2012 map from MiAmbiente.

Scale optimization and variable selection – Since there is no way to know the spatial extent surrounding a sample point at which a variable is most related to puma presence, it is best to measure each environmental variable across a range of radii surrounding a sampled point to determine the scale at which each predictor variable is most related to the puma’s response. Scale extent selection was based on the estimated home range size of the species, and was restricted by the narrow geographic parameters of Panamá. To identify the optimized scale of each predictor variable that was most related to puma response, I calculated the focal mean of each predictor

83 variable around each camera location across 30 spatial scales, ranging from 500 m – 15000 m radii at 500 m increments. To accomplish this, I first conducted a moving window analysis with the Focal Statistic tool in ArcGIS (ESRI, Redlands, CA), using a circular neighborhood and

scales described above as search radii for each variable. This allowed me to obtain an output

raster across the entire study area where the raster value at any given pixel was a mean function

of the values of all the input cells within the specified neighborhood (de la Torre et al. 2018).

Multi-scale optimized multivariate modeling – I developed multi-scale optimized habitat suitability models of puma occurrence as a function of the environmental predictor variables in three stages, following the recommendations of McGarigal et al. (2016). First, using the focal mean values from the moving window analysis described above, I conducted a univariate scaling analysis to identify the spatial scale at which each habitat variable was most strongly related to puma occurrence. This was accomplished by testing one variable and one scale at a time using

Random Forests. I selected the best-supported scale from each variable based on the model with the lowest OOB error rate (Appendix I – K). Then, in the second stage, I reduced the number of variables for the final model in two subsequent steps. First, a variable was removed if it had p >

0.05 in the model and second, I used the multicollinear function in the rfUtilities R package to assess potential multicollinearity among all possible pairs of scale-optimized variables and removed variables that were highly correlated p > 0.05. Finally, in the third stage, I ran Random

Forests with the remaining variables to model the probability of puma occurrence. Using these

84 procedures, I developed three final Random Forests models: annual, wet season, and dry season.

With each of these models, I created a map predicting puma habitat suitability across the study area. The models also generated the scaled variable importance and partial dependency plots for each variable selected within the annual, wet, and dry seasons.

Model Validation – To assess the performance of the final models, I conducted random permutations, cross-validation using a resampling approach (Evans & Murphy 2018), whereby, one-tenth of the data was withheld as a validation set in each permutation. A total of 99 permutations were performed. The cross-validation produces a suite of performance metrics including OOB error rate (the proportion of OOB samples that are incorrectly classified), model error variance, and Kappa index of agreement, “a measure of agreement between predicted presences and absences with actual presences and absences corrected for agreement that might be due to change alone” (Cutler et al. 2007). Landis & Koch (1977) suggest the following ranges of agreement for the Kappa statistic: values < 0 indicate no agreement, 0 – 0.20 as slight, 0.21 –

0.40 as fair, 0.41 – 0.60 as moderate, 0.61 – 0.80 as substantial, and 0.81 – 1 as almost perfect agreement.

Habitat selection and scale reference – To measure puma habitat selection at multiple levels, I followed the four hierarchical orders of habitat selection described by Johnson (1980) and Wiens

85 features or important habitat characteristics (Apps et al. 2001). To explain at what scale pumas, select their habitat, I divided the original 30 spatial scales (500 m – 15000 m), referring to responses at the lower half (500 m – 7500 m) as fine scale habitat use, and those in the upper half

(7500 m – 15000 m) as broad scale habitat use. Fine scale relates to the third order of habitat election, while a broad scale reflects the second order of habitat selection. The puma’s geographic range, or first order selection, was beyond the scope of this study. The division between fine and broad scale is further based on the nearest estimated home range for male pumas in the tropical dry forest of western Mexico (~83 km2) (Nuñez-Perez & Miller 2019).

Results

Puma presence-absence data – The sampling effort accumulated during the two-year study

period from 48 camera stations was 16,583 trapping nights. With the number of trap nights per

station averaging 345.5 ± 23.30 SE (49 – 597). For the analysis, a single puma detection was considered one record per hour, per camera station (Appendix L). The sampling effort produced

218 records of pumas, 145 individual records during wet season (0 – 16 pumas/station) and 73 during dry season (0 – 8 pumas/station). Total capture success (218/16,583 x 100) was 1.3.

Capture success for the wet season was 0.87 (145 pumas) and for the dry season it was 0.44 (73 pumas). I detected pumas at 38 of the 48 camera stations (79.2%). During the wet season (May

to November), cameras detected pumas at 33 (68.8%) stations (3.02 ± 0.58 SE pumas/camera).

During the dry season (December to April), cameras detected pumas at 27 (61.4%) stations (1.66

± 0.29 SE pumas/camera) out of 44 active camera stations.

Scale optimization – For all optimized multi-scale models, the strength of the relationship

between puma occurrence and predictor habitat variables was dependent on the spatial scale at

86 which each variable was derived (Table 4). The annual multi-scale analysis indicated that pumas were most closely associated with areas of primary forest, elevation, hydrologic features (i.e., rivers, ponds, lakes) and secondary forest at fine scales, while, pumas responded to coniferous forest, aspect, and agropecuary (i.e., agriculture and livestock) at broad scales.

During the wet season, puma occurrence was influenced by primary forest, elevation, and slope at fine scales. Variables such as agropecuary (i.e., agriculture and livestock), hydrologic features

(i.e., rivers, ponds, lakes), aspect, and coniferous forest were responded to at broad scales. The dry season optimized multi-scale analyses indicated that pumas responded to both primary forest and aspect at fine scales, while pumas were associated with secondary forest at a broad scale.

In general, the prediction performance of the models as indicated by the Kappa index of agreement, showed a relatively good level of accuracy (performance). According to Landis &

Koch (1977), the annual model showed substantial agreement (0.61), the dry season model showed moderate (0.56), while the wet season model showed just below the threshold for moderate agreement (0.38) (Table 4).

Table 4. Optimal scales for each most important variable for the annual, wet and dry Random Forests puma models. Code refers to the name of each variable taken from the Cobertura Boscosa y Uso de Tierra 2012 map, shortened for the purpose of analysis. Numbers in bold indicate broad scale. The table also includes Kappa, OOB and error variance.

Optimal OOB Error Variables/Model Code Kappa Scale Error Variance Puma Annual

Model accuracy 0.6102 0.10869 0.000312 Primary forest Mix_forest 500 Coniferous forest Forconi 13500 Elevation Dem 5500 Aspect Aspect_cos 11000 Hydrologic feature Hydro 500 Agropecuary Agropec 10500 Secondary forest Foremix 2000

Puma Wet Season

Model accuracy 0.3852 0.22222 0.000719 Primary forest Mix_forest 500 Agropecuary Agropec 14000 Elevation Dem 6000 Hydrologic feature Hydro 15000 Aspect Aspect_cos 11000 Coniferous forest Forconi 13500 Slope Slope 1500

Puma Dry Season

Model accuracy 0.5616 0.20000 0.000791 Primary forest Mix_forest 500 Aspect Aspect_cos 1000 Secondary forest Foremix 11000

Variable importance – Variable importance measures of the annual puma model revealed that

the most influential predictor variable was primary forest followed by coniferous forest and

elevation. In decreasing order of importance, the other predictors included in the model were

88 aspect, hydrologic features (i.e., rivers, ponds, lakes), agropecuary, (i.e., agriculture and livestock) and secondary forest (Figure 12).

In the wet season, seven variables were selected for the final model and were ranked in order of importance. Primary forest, agropecuary (i.e., agriculture and livestock), and elevation were the three most influential variables. Hydrologic features, aspect, coniferous forest, and slope were ranked lower in their relative variable importance, respectively (Figure 12).

In the dry season, from the entire set of variables used in the analyses, the most influential variables were substantially reduced to three variables. In order of importance, the variables included primary forest, aspect, and secondary forest (Figure 12).

0.1

0.09

0.08

0.07 (%IncMSE)

0.06

0.05 Importance

0.04

Variable 0.03

0.02

0.01

0 Dem Dem Slope Hydro Hydro Aspect Aspect Aspect Forconi Forconi Foremix Foremix Agropec Agropec Mix_forest Mix_forest Mix_forest ANNUAL WET SEASON DRY SEASON

Figure 12. Variable importance plot for the predictor variables from Random Forests classifications used for predicting presence of pumas in the study area. For each tree in the forest, there is a misclassification rate for the out-of-bag observations. To assess the importance of a specific predictor variable, the values of the variable are randomly permuted for the out-of-bag observations, and then the modified out-of-bag data are passed down the tree to get new predictions. The difference between the misclassification rate for the modified and original out- of-bag data, divided by the standard error, is a measure of the importance variable (Cutler et al. 2007). The variable Mix_forest refers to primary forest, Forconi refers to coniferous forest, Dem refers to elevation, Hydro refers to hydrologic features (rivers, ponds, lakes), Agropec refers to agropecuary (agriculture and livestock), and Foremix refers to secondary forest.

Partial dependency plots – The partial dependency plots representing the most important predictors for the puma annual model (Figure 13) showed non-linear relationships. Most noteworthy is the positive relationship between pumas and primary forest (mix_forest); the probability of puma occurrence increased with higher amounts of this cover type. On the other hand, the highest probability of puma occurrence was associated with low amounts of secondary forest (foremix). Pumas showed a positive relationship with agropecuary (agropec) (i.e.,

90 agriculture and livestock), the probability of puma occurrence increased as agropecuary lands increased on the landscape. Pumas showed a negative relationship with high elevations; the probability of puma occurrence dropped drastically as elevation increases over ~250 m. An aspect ranging from north to northeast was preferred. Coniferous forest (forconi) was negatively related to puma presence as it increased on the landscape, while puma occurrence decreased as hydrologic features (e.g., rivers) increased in the landscape.

Figure 13. Partial dependent plots represent the marginal effect of single variables included in the Annual Random Forest puma model. In a partial plot of marginal effects, only the range of values (and not the absolute values) can be compared between plots of different variables (Cutler et al. 2007). The gray area indicates the 95% confidence interval and the red line indicates the mean average. The variable mix_forest refers to primary forest, forconi refers to coniferous forest, dem refers to elevation, hydro refers to hydrologic features (rivers, ponds, lakes), agropec refers to agropecuary (agriculture and livestock), foremix refers to secondary forest, and aspect_cos refers to aspect.

The partial dependency plot for the wet season (Figure 14) indicated a strong positive relationship between pumas and primary forest (mix_forest); the probability of puma occurrence increased with higher amounts of this landcover type. Additionally, the probability of puma occurrence decreased above ~250 m. Alongside elevation, pumas favored steep slopes (~20%) with a northern aspect. The puma’s selection for hydrologic features (i.e., rivers, ponds, lakes) was associated with low amounts. Finally, agropecuary (agropec) (i.e., agriculture and livestock) lands and coniferous forest (forconi) were negatively correlated with puma occurrence as they increased on the landscape.

Figure 14. Partial dependency plots representing the marginal effect of single variables included in the Wet Season Random Forest puma model. In a partial plot of marginal effects, only the range of values (and not the absolute values) can be compared between plots of different variables (Cutler et al. 2007). The gray area indicates the 95% confidence interval and the red line indicates the mean average. The variable mix_forest refers to primary forest, forconi refers to coniferous forest, dem refers to elevation, hydro refers to hydrologic features (rivers, ponds, lakes), agropec refers to agropecuary (agriculture and livestock), foremix refers to secondary forest, and aspect_cos refers to aspect.

Lastly, the partial dependency plots for the dry season (Figure 15) showed a positive association between pumas and primary forest (mix_forest); with the highest probability of occurrence associated with higher amounts of this cover type on the landscape (~90 – 100%). On the other

hand, the probability of puma occurrence in secondary forest (foremix) was low, decreasing as

93 secondary forest increased on the landscape. Finally, pumas showed a preference for north- facing aspects in the dry season.

Figure 15. Partial dependent plots representing the marginal effect of single variables included in the Dry Season Random Forest puma model. In a partial plot of marginal effects, only the range of values (and not the absolute values) can be compared between plots of different variables (Cutler et al. 2007). The gray area indicates the 95% confidence interval and the red line indicates the mean average. The variable mix_forest refers to primary forest, foremix refers to secondary forest, and aspect_cos refers to aspect.

Habitat Suitability models – The habitat suitability maps produced by Random Forests for the puma included the annual, wet and dry seasons (Figure 16 – 18). All three models displayed areas of high to low suitability represented in a gradient from the highest probability of puma occurrence (red) to the lowest (blue). The maps derived by Random Forests indicated that the predictive accuracy (based on the Kappa index of agreement) varied between models and

94 seasons (See results, Table 4). The annual model showed substantial agreement, the dry season showed moderate, while the wet season model showed just below the threshold for a moderate agreement. As indicated in the models, areas of high suitability were found within Chagres

National Park and the Narganá Protected Wildland where primary forest is prevalent, representing the northern portion of the study area. Compared to the annual model (Figure 16), the wet season model (Figure 17) showed an increase in moderately suitable habitat in the southern portion of the study area. The dry season model (Figure 18) pointed to a solid band of highly suitable habitat along the Narganá Protected Wildland, while a significant increase in low habitat suitability is evident south of the Cordillera de San Blas, within the inhabited areas of the valley. Beyond the study area, a seasonal difference was apparent, as significant portions of the region were reduced to low habitat suitability in the dry season, likely due to a shift in habitat preferences related to prey availability.

Figure 16. Annual season habitat suitability map showing the predicted occurrence of pumas based on multi-scale habitat modeling in Panamá. The map includes potential puma habitat suitability up to approximately 100 km outside the study area. The habitat suitability index ranges from high (in red) to low (in blue) with high representing more suitable habitat.

Figure 17. Wet season habitat suitability map showing the predicted occurrence of pumas based on multi-scale habitat modeling in Panamá. The map includes potential puma habitat suitability up to approximately 100 km outside the study area. The habitat suitability index ranges from high (in red) to low (in blue) with high representing more suitable habitat.

Figure 18. Dry season habitat suitability map showing the predicted occurrence of pumas based on multi-scale habitat modeling in Panamá. The map includes potential puma habitat suitability up to approximately 100 km outside the study area. The habitat suitability index ranges from high (in red) to low (in blue) with high representing more suitable habitat.

Discussion

Multi-level, multi-scale models capture a wide spectrum of ecological relationships for a

population (Zeller et al. 2017). From that perspective, and through multi-scale optimized

analyses (McGarigal et al. 2016), I found support for my first hypothesis regarding

spatiotemporal differences in habitat use by pumas at multiple scales. In all three models, the

most influential and consistent predictors of habitat selection at fine scale were primary forest

and elevation. At broad scale, agropecuary (i.e., agriculture and livestock) and coniferous forest

98 were selected, while hydrologic features, secondary forest, and aspect were consistent at both scales.

The annual model results agree with other studies that have found pumas prefer native protected forests and those that provide resources and cover for stalking prey, such as primary forest (fine scale) (DeAngelo 2009; Foster 2010; LaRue & Nielsen 2011; Angelieri et al. 2016; Rowe 2018).

Agropecuary and coniferous forest were influential for home range establishment (broad scale).

The selection of primary forest and its attributes, lower elevations and hydrologic features at fine scales, points to a preference for complex habitats, which may offer more niches to attract prey, than simpler habitats (Klopfer & MacArthur 1960; Levins 1968; August 1983). Therefore, primary forest may provide high-quality habitat because of its prey richness, topography; coupled with hydrologic features, it may confer advantages for pumas to stalk and ambush prey

(Murphy & Ruth 2010; Guerisoli et al. 2019; Nuñez-Perez & Miller 2019). Thus, in the tropical montane cloud forest of Panamá, pumas appear to prefer the structurally complex habitat

(Lamprecht 1978; Elbroch & Wittmer 2012).

In all puma models, primary forest was the most important predictor variable linked to habitat suitability at the same fine scale (500 m). The results show that puma occurrence increased greatly as the proportion of primary forest increased (80 – 100%). Similarly, pumas responded at fine scales (annual and wet season) to elevational gradients. In the annual model for instance, pumas selected low elevations (~250). While in the wet season model, pumas showed a preference for low elevations, but were not averse to mid-elevations (Figure 14). The slight difference in elevational gradients used at fine scales and between models, may be related to puma habitat preferences for food acquisition, mates, or dispersal (Burdett et al. 2010; Nuñez-

Perez & Miller 2019). For example, Nuñez-Perez & Miller (2019) reported that puma movement increased during the wet season, which may reflect a pursuit of resources. Additionally, pumas incorporated a broad scale preference for an aspect ranging from north to northeast at low to mid-elevations, which is known to favor forest productivity and species richness on the

Caribbean side of the study area. The results suggest that the determinants of puma habitat in the tropical montane cloud forest includes a complexity of steep slopes, and the interacting effects of aspect and elevation.

The variable human settlement was not included in the final models. Therefore, my first prediction that pumas would avoid human settlements at a broad scale was not supported.

However, if puma selection for agropecuary (i.e., agriculture and livestock) areas is considered a proxy for human settlement, then my prediction was upheld. In the analysis, pumas exhibited a broad scale response to agropecuary (annual and wet season models), being the second most important variable during wet season. Given that the southern end of the study area is an agricultural frontier abutting the remaining stretch of forest within the narrowest section of

Panamá, human modified-habitats appear to have some relevance to puma habitat suitability

(Scognamillo et al. 2003; De Angelo et al. 2011; Elbroch & Wittmer 2012).

Further, the scale and the importance at which the variable agropecuary (i.e., agriculture and livestock) was selected for in the wet season model, and its absence from the dry season model, strongly suggests seasonal changes in habitat selection by pumas. This finding corroborates my second hypothesis that pumas would select habitats differently between seasons, specifically my prediction that pumas would select agropecuary during the wet season at a broad scale.

100

Traditionally, it is assumed that seasonal changes in the home range of carnivores, and therefore, habitat selection, is determined by changes in the distribution and abundance of prey (Sunquist

1981; Fort 2016; Azevedo et al. 2018). It is reasoned that wet seasons play important roles in patterns of species distribution and abundance (August 1983; Judas & Henry 1999). For example, in a study using remote cameras to investigate patterns of site occupancy, Fort (2008) found that collared peccary detection increased at the onset of the wet season. Furthermore, pumas have been reported to follow collared peccaries during the wet season (Mendez Ponte &

Chivers 2007). Within a prey dependency framework, pumas may have to incorporate anthropogenic lands at the interface (edge) between primary forests and open areas, where prey availability tends to increase (Guerisoli et al. 2019). Nuñez-Perez & Miller (2019), point out that water during the wet season creates a wider dispersion of prey, which in my study, may be reflected by the increased importance of the agropecuary variable in the wet season model. It is possible that certain agricultural practices or crops within the year correlate with prey distribution and availability that attract pumas. Smith (2005), for example, showed that in neotropical rainforest regions (including Panamá), heterogeneous human landscapes (i.e., orchards, fallows, and gardens) provide highly suitable habitat for many species like nine-banded armadillo (Dasypus novemcintus), white-nosed coati (Nasua narica), pacas (Agouti paca), and collared peccary (Tayassu tajacu), which are known to be potential prey for pumas (Harmsen et al. 2010; Foster et al. 2010; Laundré & Hernández 2010).

Interspecific competition may offer another explanation for pumas incorporating agropecuary

(i.e., agriculture and livestock) into their home range (broad scale) during the wet season.

Elbroch & Kusler (2017) found that when pumas are the subordinate species, they suffer displacement at food sources. It has also been suggested that differences in habitat segregation

101 allow jaguars and pumas to coexist and decrease competition to avoid injury, death, or prey depletion (Foster et al. 2010; Sollmann et al. 2012; Palomares et al. 2016). Thus, the avoidance of competitive interactions between these two carnivores in Panamá may be reflected by the puma’s incorporation of human-altered landscapes.

Regarding the puma’s selection for hydrologic features (e.g., rivers), my annual and wet season models suggest limited associations. This finding agrees with previous studies in Central and

South America (Sollmann et al. 2012; Angelieri et al. 2016; Rowe 2018). Yet, according to

Nuñez-Perez & Miller (2019), riparian features allow for ease of access into hillsides, offering cover and hunting opportunities due to the dense vegetation. These results also concur with behavior observed in GPS-collared pumas in the mountains of California, where pumas used riparian zones as cover and escape pathways (Burdett et al. 2010; Wilmers 2013).

Lastly, coniferous forest was considered an important variable in both the annual and wet season models at a broad scale. Pumas were associated with limited amounts of coniferous forest in the wet season. However, in the annual model, the negative relationship indicated that coniferous forest is a limiting factor at the home range level. This land cover type does not likely provide essential prey, nor would it provide cover for puma due to the sparseness of the vegetation and lack of understory.

Conclusion – Pumas have suffered range contraction throughout their range caused by habitat loss and fragmentation, human persecution and alteration of natural habitat. Therefore, understanding how pumas select habitat, and at what scale, is important in species-habitat relationship analysis. Evaluating habitat selection at a range of scales reveals the true grain at which pumas respond within the landscape (Wasserman et al. 2012). In this work, I have

102 identified the environmental and anthropogenic factors affecting puma habitat use at multiple scales. My results are mostly consistent with previous research in the Neotropics, with additional insights related to the relative importance, and scale at which each habitat variable had the largest influence on puma occurrence. The results emphasized the intrinsic complexity of decision-making in the puma’s response to landscape changes in the study area, and potentially countrywide. My results indicated that pumas selected primary forests and relevant topographic qualities (e.g., elevation, aspect, hydrologic features) at fine scales. At broad scales, pumas incorporated anthropogenic features like agropecuary (i.e., agriculture and livestock) landscapes.

Additionally, the multi-scale analyses indicated a noticeable seasonal change in habitat use.

Because animals often select different habitat components at different scales, and species vary in their scales of selection (Wiens 1995; Mayor et al. 2009), the multi-scale, optimized approach

(McGarigal et al. 2016) is proving to have inferential advantages over single-scale models

(Zeller et al. 2017; Bauder et al. 2018). My study was the first to develop a multi-scale, optimized method for determining puma habitat suitability in Panamá, providing insight into patterns of scale-dependent habitat selection for the region.

Although pumas are widespread across the Americas, their populations are in decline, and their status is poorly known in Central America (Caso et al. 2008; Kelly et al. 2008; Laundré &

Hernández 2009; Foster et al. 2016). The results of this study bolster our understanding of puma habitat selection and can be considered a guide for priority conservation actions involving land management practices that conserve the integrity of critical habitat in Panamá. The results can also be helpful for revising management policies or for the mitigation of negative effects to critical patches or corridors where future development may occur. Pumas play an important role

103 in the natural dynamics of the ecosystems throughout their range. They help to control the number of herbivores, which decreases the probability that herbivore populations will become so large as to have negative impacts on plant productivity. Without pumas, ecosystems may experience changes in community structure. Therefore, research efforts must be continued and expanded upon to improve our knowledge of puma distribution and habitat selection, and to refine conservation strategies. Zanin et al. (2015) noted that although local felid research is important for species conservation, felid research also needs a broader scale approach to understand the effects of habitat fragmentation and improve conservation strategies. Future efforts should focus on surveying areas to confirm the presence or absence of pumas nationwide.

Further, nationwide surveys are needed to enhance habitat suitability models to best evaluate the feasibility of establishing or maintaining viable corridors for pumas in the region.

104

Literature Cited

Angelieri, C. C. S., C. Adams-Hosking, K. M. P. M. de Barros, M. P. de Souza and C. A. McAlpine. 2016. Using species distribution models to predict potential landscape restoration effects on puma conservation. PLoS One 11:e0145232.

August, P. V. 1983. The role of habitat complexity and heterogeneity in structuring tropical mammal communities. Ecology 64:1495–1507.

Ávila-Nájera, D. M., C. Chávez, S. Pérez-Elizalde, R. A. Guzmán-Plazola, G. D. Mendoza and M. A. Lazcano-Barrero. 2018. Ecology of Puma concolor (Carnivora: Felidae) in a Mexican tropical forest: adaptation to environmental disturbances. Revista de Biología Tropical 66:78–90.

Azevedo, F. C., F. G. Lemos, M. C. Freitas‐Junior, D. G. Rocha and F. C. C. Azevedo. 2018. Puma activity patterns and temporal overlap with prey in a human‐modified landscape at Southeastern Brazil. Journal of Zoology 305:246–255.

Boyce, M. S. and L. L. McDonald. 1999. Relating populations to habitats using resource selection functions. Trends in Ecology & Evolution 14:268–272.

Breiman, L. 2001. Random forests. Machine Learning 45:5–32.

Broekhuis, F., S. A. Cushman and N. B. Elliot. 2017. Identification of human–carnivore conflict hotspots to prioritize mitigation efforts. Ecology and Evolution 7:10630–10639.

Burdett, C. L., et al. 2010. Interfacing models of wildlife habitat and human development to predict the future distribution of puma habitat. Ecosphere 1:1–21.

Caso, A., C. Lopez-Gonzalez, E. Payan, E. Eizirik, T. de Oliveira, R. Leite-Pitman, M. Kelly, C. Valderrama and M. Lucherini. 2008. Puma concolor. IUCN Red List of Threatened Species. Available at http://www.iucnredlist.org.

Ceballos, G. 2005. Global Mammal Conservation: What Must We Manage? Science 309:603– 607.

Cushman, S. A., E. A. Macdonald, E. L. Landguth, Y. Malhi and D. W. Macdonald. 2017. Multiple-scale prediction of forest loss risk across Borneo. Landscape Ecology 32:1581– 1598.

105

Cushman, S. A. and T. N. Wasserman. 2018. Landscape Applications of Machine Learning: Comparing Random Forests and Logistic Regression Multi-Scale Optimized Predictive Modeling of American Marten Occurrence in Northern Idaho, USA. Pages 185–203 in Machine Learning for Ecology and Sustainable Natural Resource Management. Springer, Cham.

Cutler, D. R., T. C. Edwards, K. H. Beard, A. Cutler, K. T. Hess, J. Gibson and J. J. Lawler. 2007. Random Forests for Classification In Ecology. Ecology 88:2783–2792.

Davis, M. L., M. J. Kelly and D. F. Stauffer. 2011. Carnivore co-existence and habitat use in the Mountain Pine Ridge Forest Reserve, Belize. Animal Conservation 14:56–65. de Angelo, C., A. Paviolo and M. Di Bitetti. 2011. Differential impact of landscape transformation on pumas ( Puma concolor ) and jaguars ( Panthera onca ) in the Upper Paraná Atlantic Forest: Impact of landscape transformation on pumas and jaguars. Diversity and Distributions 17:422–436. de la Torre, J. A., J. M. Núñez and R. A. Medellín. 2017. Spatial requirements of jaguars and pumas in Southern Mexico. Mammalian Biology 84:52–60.

Elbroch, L. M. and H. U. Wittmer. 2012. Puma spatial ecology in open habitats with aggregate prey. Mammalian Biology-Zeitschrift für Säugetierkunde 77:377–384.

Elbroch, L. M. and A. Kusler. 2018. Are pumas subordinate carnivores, and does it matter? PeerJ 6:e4293.

Evans, J. S. and M. A. Murphy. 2018. rfUtilities. R package version 2.1–3, https://cran.r- project.org/package=rfUtilities.

Fort, J. L. 2016. Large Carnivore Occupancy And Human-Wildlife Conflict In Panamá. Doctoral dissertation. Southern Illinois University Carbondale.

Foster, R. 2008. The ecology of jaguars (Panthera onca) in a human-influenced landscape Doctoral dissertation. University of Southampton.

Foster, R. J., B. J. Harmsen and C. P. Doncaster. 2010. Habitat use by sympatric jaguars and pumas across a gradient of human disturbance in Belize. Biotropica 42:724–731.

Gonzalez-Borrajo, N., J. V. López-Bao and F. Palomares. 2017. Spatial ecology of jaguars, pumas, and ocelots: a review of the state of knowledge. Mammal Review 47:62–75.

Guerisoli, M. D. L. M., N. Caruso, E. M. Luengos Vidal and M. Lucherini. 2019. Habitat use and activity patterns of Puma concolor in a human-dominated landscape of central Argentina. Journal of Mammalogy.

106

Guisan, A. and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147–186.

Guisan, A. and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology letters 8:993–1009.

Harmsen, B. J., R. J. Foster, S. Silver, L. Ostro and C. P. Doncaster. 2010. Differential Use of Trails by Forest Mammals and the Implications for Camera-Trap Studies: A Case Study from Belize: Trail Use by Neotropical Forest Mammals. Biotropica 42:126–133.

Hearn, A. J., S. A. Cushman, J. Ross, B. Goossens, L. T. B. Hunter, and D. W. Macdonald. 2018. Spatio-temporal ecology of sympatric felids on Borneo. Evidence for resource partitioning? PLoS One 13:e0200828.

Hirzel, A. H., G. Le Lay, V. Helfer, C. Randin and A. Guisan. 2006. Evaluating the ability of habitat suitability models to predict species presences. Ecological Modelling 199:142–152.

Jędrzejewski, W., et al. 2017. Density and population structure of the jaguar (Panthera onca) in a protected area of Los Llanos, Venezuela, from 1 year of camera trap monitoring. Mammal Research 62:9–19.

Judas, J. and O. Henry. 1999. Seasonal variation of home range of collared peccaries in tropical rain forest of French Guiana. The Journal of Wildlife Management 63:546–552.

Kie, J. G, R. T. Bowyer and K. M. Stewart. 2003. Ungulates in western coniferous forests: habitat relationships, population dynamics, and ecosystem processes. Pages 296–340 in Mammal Community Dynamics. Management and Conservation in the Coniferous Forests of Western North America. C. J. Zabel and R. G. Anthony, editors. Cambridge University Press, Cambridge, UK.

Klopfer, P. H. and MacArthur, R. H. 1960. Niche size and faunal diversity. The American Naturalist 94:293–300.

107

Lamprecht, J. 1978. The relationship between food competition and foraging group size in some larger carnivores: a hypothesis. Tierpsychology 46:337–343.

Landis, J. R and G. G. Koch. 1977. An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers. Biometrics 33:363–374.

LaRue, M. and C. Nielsen. 2011. Modelling potential habitat for cougars in Midwest North America. Ecological Modelling 222:897–900.

Laundré, J. W. and L. Hernández. 2010. What we know about pumas in Latin America. Pages 76–90 in Cougar: ecology and conservation. M. Hornocker and S. Negri, editors. The University Press, Chicago and London.

Levins, R. 1968. Evolution in changing environments: some theoretical explorations (No. 2). Princeton University Press.

Logan, K. A. and L. L. Sweanor. 2001. Desert puma: evolutionary ecology and conservation of an enduring carnivore. Island Press.

Macdonald, E. A., S. A. Cushman, E. L. Landguth, A. J. Hearn, Y. Malhi and D. W. Macdonald. 2018. Simulating impacts of rapid forest loss on population size, connectivity and genetic diversity of Sunda clouded leopards (Neofelis diardi) in Borneo. PLoS One 13:e0196974.

Macdonald, D. W., et al. 2018. Multi-scale habitat selection modeling identifies threats and conservation opportunities for the Sunda clouded leopard (Neofelis diardi). Biological Conservation 227:92–103.

Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald and W. P. Erickson. 2002. Resource Selection by Animals: Statistical Design and Analysis for Field Studies, second edition. Kluwer Academic Publishers, Boston, MA. USA.

McGarigal, K., H. Y. Wan, K. A. Zeller, B. C. Timm and S. A. Cushman. 2016. Multi-scale habitat selection modeling: a review and outlook. Landscape Ecology 31:1161–1175.

Mendes Pontes, A. R. and D. J. Chivers. 2007. Peccary movements as determinants of the movements of large cats in Brazilian Amazonia. Journal of Zoology 273:257–265.

Monroy-Vilchis, O., Y. Gómez, M. Janczur and V. Urios. 2009. Food niche of Puma concolor in central Mexico. Wildlife Biology 15:97–106.

Moreno, R. S., R. W. Kays and R. Samudio. 2006. Competetive release in diets of ocelot (Leopardus pardalis) and puma (Puma concolor) after jaguar (Panthera onca) decline. Journal of Mammology 87:808–816.

108

Murphy, K. and T. Ruth. 2010. Diet and prey selection of a perfect predator. Pages 118–137 in M. Hornocker and S. Negri, editors. Cougar: Ecology and Conservation. University of Chicago Press, Chicago.

Nielsen, C., D. Thompson, M. Kelly and C. A. Lopez-Gonzalez. 2015. Puma concolor. The IUCN Red List of Threatened Species. Available at https://www.iucnredlist.org/species/18868/97216466.

Novack, A. J., M. B. Main, M. E. Sunquist and R. F. Labisky. 2005. Foraging ecology of jaguar (Panthera onca) and puma (Puma concolor) in hunted and non-hunted sites within the Maya Biosphere Reserve, Guatemala. Journal of Zoology 267:167–178.

Nowell, K. and P. Jackson. 1996. Status survey and conservation action plan. Wildcats. IUCN/SSC Cat Specialist Group, IUCN. Gland, Switzerland and Cambridge, UK.

Palomares, F., N. Fernández, S. Roques, C. Chávez, L. Silveira, C. Keller and B. Adrados. 2016. Fine-scale habitat segregation between two ecologically similar top predators. PloS One 11: e0155626.

Paviolo, A., et al. 2016. A biodiversity hotspot losing its top predator: The challenge of jaguar conservation in the Atlantic Forest of South America. Scientific Reports 6:37147.

Rich, L. N., et al. 2014. Comparing capture-recapture, mark-resight, and spatial mark-resight models for estimating puma densities via camera traps. Journal of Mammalogy 95:382–391.

Ripple, W. J., et al. 2014. Status and ecological effects of the world’s largest carnivores. Science 343:1241484.

Rowe, C. B. 2017. The influence of habitat features and co-occurring species on puma (Puma concolor) occupancy across eight sites in Belize, Central America. Doctoral dissertation. Virginia Tech.

Schipper, J., et al. 2008. The status of the world’s land and marine mammals: diversity, threat, and knowledge. Science 322:225–230.

Smith, D. A. 2005. Garden game: shifting cultivation, indigenous hunting and wildlife ecology in western Panama. Human Ecology 33:505–537.

109

Smith, J. A., Y. Wang, and C. C. Wilmers. 2016. Spatial characteristics of residential development shift large carnivore prey habits. The Journal of Wildlife Management 80:1040–1048.

Sollmann, R., M. M. Furtado, H. Hofer, A. T. A. Jácomo, N. M. Tôrres, L. Silveira. 2012. Using occupancy models to investigate space partitioning between two sympatric large predators, the jaguar and puma in central Brazil. Mammalian Biology 77:41–46.

M. E. Sunquist. 1981. The social organization of tigers (Panthera tigris) in Royal Chitawan National park, Nepal. Smithsonian Contributions to Zoology. Smithsonian Institution Press, Washington D.C.

Sunquist, M. and F. Sunquist. 2002. Wild Cats of the World. The University of Chicago Press, Chicago and London.

Tilman, D., R. M. May, C. L. Lehman and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371:65–66.

Treves, A. and K. U. Karanth. 2003. Human-Carnivore Conflict and Perspectives on Carnivore Management Worldwide. Conservation Biology 17:1491–1499.

Tritsch, I. and D. Arvor. 2016. Transition in environmental governance in the Brazilian Amazon: emergence of a new pattern of socio-economic development and deforestation. Land Use Policy 59:446–455.

Wiens, J. A. 1989. Spatial scaling in ecology. Functional ecology 3:385–397.

110

Wilmers, C. C., Y. Wang, B. Nickel, P. Houghtaling, Y. Shakeri, M. L. Allen, J. Kermish-Wells, V. Yovovich and T. Williams. 2013. Scale dependent behavioral responses to human development by a large predator, the puma. PLoS One 8:e60590.

Zanin, M., F. Palomares and D. Brito. 2015. What we (don't) know about the effects of habitat loss and fragmentation on felids. Oryx 49:96–106.

Zeller, K. A., K. McGarigal P. Beier, S. A. Cushman, T. W. Vickers and W. M. Boyce. 2014. Sensitivity of landscape resistance estimates based on point selection functions to scale and behavioral state: pumas as a case study. Landscape Ecology 29:541–557.

Zeller, K. A., K. McGarigal, S. A. Cushman, P. Beier, T. W. Vickers, W. M. Boyce. 2016. Using step and path selection functions for estimating resistance to movement: pumas as a case study. Landscape ecology 31:1319–1335.

111

Appendices

Appendix G: Examples of the three-land tenure areas Narganá Protected Wildland, Chagres National Park and the Mamoní Valley Preserve, Panamá.

112

113

114

Appendix H. Twenty-one predictor variables associated with the puma multi-scale habitat suitability models. Variable code name and descriptive name used for discussion are included for reference.

*Indicates variables that were selected for the final analyses. aAgropec: Heterogeneous area of agricultural production, bHydro: Rivers, ponds, lakes, cInfraest: Major roads, airport, dPeople: Housing, buildings/urban area, eRastrojo: Shrubs and bushes.

115

Appendix I. Annual model OOB table for pumas.

116

Appendix J. Wet season model OOB table for pumas.

117

Appendix K. Dry season model OOB table for pumas.

118

Appendix L. Photographs taken of pumas by camera traps in the study area in Panamá.

119

120

Chapter IV: Jaguars, Connectivity, and Climate Change in the Comarca Guna Yala, Panamá Abstract

Among large terrestrial carnivores, jaguars (Panthera onca) are suffering from significant population reductions throughout their range. The causes may not be restricted to the overexploitation of resources (i.e., hunting, grazing, deforestation frontiers) that have compromised forest ecosystems, but also to the advent of climate change. The effects of climate change are forcing the island communities of the Guna Yala Indigenous people to relocate to one of the most extensive, intact forests remaining for jaguars in Panamá. This review explores, for the first time, the links between the impacts of climate change, indigenous people’s displacement, and the implications for jaguar conservation. I examine the consequences of the proposed resettlement location, as it may exacerbate the susceptibility of the jaguar population in the region. Then, I put into perspective one of the most challenging obstacles to resettlement as an adaptive strategy to climate change, securing a location where people’s livelihoods, traditions, and cultures are not significantly altered. Finally, I conclude by discussing future recommendations and key challenges associated with human development in an area of high biodiversity. Few studies have focused on the synergies between climate change and habitat loss for felids, yet the potential impacts warrant consideration for the future of these species.

Introduction

Globally, ecosystems have been radically compromised by increased natural resource exploitation (e.g., agriculture, grazing, hunting, fishing, timber removal) that has disrupted ecosystem processes and diminished ecosystem services (i.e., the benefits provided by ecosystems that contribute to maintaining human life) (Mooney et al. 2009; Mace 2011). This

121 includes services provided by apex predators (Ripple 2014), as they help maintain biodiversity and ecological processes by exerting top down control in ecosystems they inhabit (Estes et al.

2011). A decline of these species has been exacerbated by human induced habitat encroachment.

Hence, human-carnivore coexistence in shared landscapes will depend upon the tolerance of the people to wildlife.

Among the apex predators, jaguars (Panthera onca), are suffering from significant population reductions throughout their range (Ray et al. 2005; Pitman et al. 2017). Their high trophic level, extensive home range, and low population density makes them highly vulnerable to landscape change. Although deforestation and fragmentation, prey depletion, and human persecution are the main threats to their survival, an extensive body of research points to climate change as another anthropogenic driver of carnivore decline (Ripple et al. 2014).

Studies on the linkages between jaguars and human-altered ecosystems have provided substantial insight into the human-jaguar dynamic (i.e., prey competition, retaliatory killing) (Polisar et al.

2003; Cavalcanti et al. 2010; Zanin et al. 2015). Yet, environmental change, and the impacts it will have, such as the displacement of island communities to forested areas, generates a new challenge to jaguar persistence. As suggested by Pitman et al. (2017), jaguars may not persist in human-dominated landscapes without connectivity or permeable matrices between natural or protected areas. Therefore, strategies that facilitate human coexistence with apex species (i.e. jaguars) across the landscape are essential to their long-term survival, and for maintaining the high levels of biodiversity, characteristic of tropical forests. Thus, a functional regional ecosystem will increase species resilience in the face of climate change, thereby, increasing the capacity to deliver ecosystem services for human well-being (Mooney et al. 2009).

122

The effects of climate change have become evident in Panamá, particularly in the Indigenous territory (i.e., Comarca) of the Guna Yala, which encompasses a thin strip of land along the

Caribbean coast, and the offshore islands of the San Blas Archipelago (San Blas) (Arenas 2014,

2016) (Figure 19). The slow-onset impacts from rising sea levels is particularly grave for these island communities, where climate change displacement is inevitable. The Guna Yala are already experiencing recurrent floods, waterlogging, salinity intrusions, and depletion of marine resources due to coral bleaching (Arenas 2016).

123

Figure 19. Location of the Comarca Guna Yala and the offshore islands of the San Blas Archipelago in eastern Panamá. The potential resettlement location along the Caribbean coast and the Carti road are highlighted.

124

Though the environmental changes are threatening, it will be the breakdown of ecosystem- dependent livelihoods that will likely drive human displacement in the next two to three decades

(Raleigh & Jordan 2010). Underdevelopment, high population densities, and poverty simultaneously play a dynamic role in prompting human displacement (Heltberg et al. 2010).

Therefore, in the face of climate change, practical and holistic approaches are needed not only to address the severe risk of losing ecosystem services (e.g., iconic species, biodiversity) (Ellis

2011; Pedrono et al. 2016), but also to promote climate-resilient livelihoods for the rising number of displaced people (Raleigh & Jordan 2010), and prevent the loss of cultural assets that communities value (Adger et al. 2013).

The proposed resettlement of Guna Yala from San Blas (islands) back to the forest (mainland), brings a new set of concerns regarding its effects on both a tropical forest ecosystem (Arenas

2014), and the persistence of the iconic, near threatened jaguar. While maintenance of forest ecosystem services is part of the solution to building resiliency and adaptation to climate change, ecosystem services are also affected by changing climatic conditions (Locatelli 2016). Therefore, human resettlements to areas of high biological significance should take into account (1) a geoecological approach–the relationships between the abiotic, biotic and anthropogenic impacts to the forest ecosystem (Caviglia-Harris & Harris 2011; Manoiu 2015) and (2) a biocultural conservation approach that recognizes the dynamic interconnectedness of biophysical and sociocultural components of conservation actions (Gavin et al. 2015; Sterling et al. 2017;

McCarter et al. 2018), as frameworks for confronting the loss of biological and cultural diversity in the region. This review explores, for the first time, the links between the slow-onset impacts of climate change, indigenous people’s displacement, and the implications for jaguar conservation.

I first describe the biogeographic setting of Panamá by highlighting its origin and impact on the

125 biodiversity in the Americas. Then, I review the history of the Guna Yala, who formerly lived in the tropical forest in eastern Panamá, before moving to San Blas along the Caribbean coast. After analyzing the consequences of climate change on island ecosystems, the vulnerability of the

Guna Indigenous people is discussed. Further, the socio-cultural and economic consequences of resettlement set the stage for the impending migration of an estimated 40,000 people back to the mainland portion of the Comarca, one of the most extensive intact forests in Panamá. After establishing the roles of biodiversity, specifically apex predators, for sustaining ecosystems, the impacts of habitat fragmentation on jaguar are outlined. I conclude by discussing future recommendations and key challenges associated with development in protected areas with high biodiversity.

Biogeographic setting of The Republic of Panamá

The Republic of Panamá (hereafter Panamá) is the fourth smallest Central American nation of

3.6 million people (Davis 2010), spanning about 75,717 km2 on an east-west axis (Figure 19).

There are 3,525 million hectares of forest covering ~47% of Panamá, including some of the most biodiverse habitats in the world, harboring about 1,298 endemic species (Midler et al. 2014).

Panamá shares two biodiversity hotspots, the Tumbes-Choco Magdalena hotspot in the east, and

the southern end of the Mesoamerica hotspot to the west; which merge east of the Canal Area

(Myers 1969). Hotspots are areas with an extremely high amount of the world’s biodiversity in a relatively small proportion of area (Brummitt & Lughadha 2003). Hotspots are typically defined by areas of exceptional concentrations of biodiversity, measured by species richness, high

endemism, number of threatened species, and degree of habitat loss.

126

The origins of this extraordinary diversity in the Panamanian Isthmus goes back about 3 million years (Woodburne et al. 2006; Weir 2009; Cione et al. 2009, 2015; Leigh et al. 2013). When the land bridge formed between North and South America, it initiated one of the most significant and rapid episodes of intercontinental interchange of tropical forest faunas (Woodburne et al. 2006 and references therein; Weir 2009). Through a series of taxa invasions and extinctions, the

“Great American Biotic Interchange” (GABI) (Stehli & Webb 1985) transformed the tropical fauna of the New World, creating the strongest biogeographic link between North and South

America (Woodburne et al. 2006).

Today, Panamá boasts tropical montane cloud forests (TMCFs), one of the Earth’s most imperiled ecosystems; much of what remains globally are believed to be small or remnant fragments (Aldrich et al. 1997). The total cover has been estimated to be 1.4% of the total area of all tropical forests (Oliveira et al. 2015). TMCFs defined as, “forests that are frequently covered in clouds or mist” (Stadtmuller 1987; Hamilton et al. 1995), are exceptionally rich in biodiversity, and sustain numerous locally endemic species. Furthermore, the ecosystem services provided by cloud forests (e.g., fresh water supply to communities, habitat maintenance for biodiversity, carbon storage, and climate regulation) are of paramount importance for human well-being, both locally and globally (Aldrich et al. 1997; Bonan 2008 in Spracklen, Oliveira

2014). But cloud forest ecosystems are poorly understood, and their fate remains uncertain

(Myers 1969; Aldrich et al. 1997).

The total forest cover in Panamá has decreased over the last 20 years (CATHALAC 2008). The

Global Forest Resources Assessment 2010 (FAO 2010) reports that the annual deforestation rate between 2005 and 2010 was 0.36%. In 2010, forest cover comprised 44% of the national

127 territory (FAO 2010). The Panamanian forests, biodiversity, and ecosystem services are threatened by a range of anthropogenic pressures including the expansion of logging, agriculture, cattle ranching, infrastructure, illegal trafficking of species and their products, and more recently, the alteration of cloud formation patterns due to climate change (Parker et al. 2004). Seeking to reverse deforestation, Panamá began working towards a more sustainable future for development by integrating environmental conservation into decision-making and economic planning

(Vergara-Asenjo & Potvin 2014). Two programs were implemented, the establishment of national protected areas for in situ forest conservation in 1966; and in 2008, the incorporation of the United Nations Collaborative Program on Reducing Emissions from Deforestation and Forest

Degradation in Developing Countries (REDD+) described briefly below.

In the face of rapid loss and degradation of natural habitat, the formation of protected areas

(PAs) that act as repositories of native biodiversity and facilitate natural ecosystem processes, has been the global norm. In this regard, Panamá is no exception (Nelson et al. 2001; Oestreicher et al. 2009; Haruna 2010). In 2004, Panamá boasted 51 protected areas, accounting for about

25% of the country, equal to roughly 2 million ha, of which 60% comprised terrestrial environments (Parker et al. 2004). Although the effectiveness of PAs varies within and between regions and countries (Spracklen et al. 2015), and success generally depends upon the location, governance, and budgets (Nelson & Chomitz 2011), Panamá is one of five countries with the best performing PAs based on effectiveness in reducing forest loss (Spracklen et al. 2015).

Vergara-Asenjo & Potvin (2014) stated that PAs and the indigenous territories of Panamá have been the most effective land-tenure regimes for avoiding deforestation. Later, Vergara-Asenjo

(2016) pointed out that together, PAs and indigenous territories held 77% of total mature forest

128 area in the country, highlighting the importance of indigenous territories for the conservation of biodiversity.

Panamá joined REDD+ to move towards a goal of developing a national strategy that has the potential to simultaneously contribute to climate change mitigation and poverty alleviation whilst conserving biodiversity and sustaining vital ecosystem services (World Bank 2011; UNDP 2012;

Vergara-Asenjo et al. 2017). However, a successful venture between REDD+ and Indigenous people relies on the recognition of traditional lands and indigenous rights, capacity building, and respect for indigenous cosmology (Vergara-Asenjo et al. 2017). The REDD+ program has been met with resistance by some indigenous groups that prefer not to commodify the forest and its ecosystem services, nor change their traditions or forfeit their rights to accommodate the program. For example, the Guna Yala rejected the program, believing REDD+ might threaten traditional land uses and rights (Potvin & Mateo-Vega 2013).

The Indigenous people of Guna Yala, past and present

Panamá is a multi-ethnic country with five legally established indigenous territories referred to as ‘Comarcas’: Guna Yala, Emberá-Wounaan, Guna de Madungandi, Ngäbe Buglé, Kuna de

Wargandi, and eight clearly defined indigenous groups: Guna, Ngäbe, Buglé, Teribe/Naso,

Bokota, Emberá, Wounaan, and Bri Bri (Vergara-Asenjo & Potvin 2014). According to the 2010

Panamanian national census, the indigenous population was estimated at 417,559, which represents 12.3% of the total population (Davis 2010). The census revealed that 23.9% have left their Comarcas, representing a trend of urban migration of indigenous people observed throughout Latin America. According to the World Bank report (2008), 49% of the indigenous population currently lives in urban areas.

129

The Comarca Guna Yala (hereafter Comarca) region was organized by Law 16 on February 16,

1953. The Guna (formerly called ‘Kuna’ until 2011) is the second largest indigenous society in

Panamá after the Ngäbe. Per the 2010 census, the total population was estimated at 80,526 people, representing 19.3% of the indigenous societies. By 2010, 40.6% lived in urban areas

(Davis 2010; Arenas 2014). Within the San Blas Archipelago (hereafter San Blas), the number of inhabitants was 30,458. The Gunas are hailed as pioneers in the autonomy and management of their nation, and as forward-thinking conservationists (PEMESKY 1986). Their societal and political structure has been a model for the other Panamanian and foreign indigenous groups

(PRONAT 2008).

The Comarca comprises a long, thin strip of land along the Atlantic coast of eastern Panamá

(Apgar 2010), and a seaward portion that consists of more than 365 small coral offshore islands, together constituting ~5,400 km2 (CGK 2008, Apgar 2010). The mainland portion of the

Comarca (~3260 km2) stretches northwest from Colombia to Parque Nacional Chagres, forming a band of forest between 10 – 20 km wide, framed by the Caribbean Sea to the north and the

Cordillera Serrania de San Blas to the south. Historically, the Guna lived across the Isthmus of

Panamá in what is now Darien Province and Northern Colombia. Although it is not clear when

and why the Guna began moving westward towards what is now the Comarca, the culture

developed in the cloud forest. But by the middle of the nineteenth century, they began moving to

islands off the Atlantic coast, which were relatively free of the insects and diseases found on the

mainland (Chapin 2000; Howe 2002).

In the early 1900s, the newly formed Panamanian government tried, unsuccessfully, to control

and “civilize” the Guna through commercial and missionary activities (Martinez Mauri 2007).

130

With help from the United States, the Guna rebelled against the new government and after a deadly uprising in 1925 (Tule Revolution), they achieved their independence and a treaty was signed, acknowledging their cultural and political autonomy (Chapin 2000). In 1938, San Blas was recognized as a Guna territory after the Panamanian government granted the Guna legal dominion over their tribal land (Apgar 2010). However, in the early 1970s, threats to the integrity of the Comarca began with the “expansion of the agriculture frontier” (Peterson St-

Laurent et al. 2013). During this period, the Panamanian government was promoting the colonization of remote areas. This enticed farmers or landless individuals to search for new opportunities and land ownership (Hernandez 1984; Peterson St-Laurent et al. 2013). At the same time, roads were constructed in eastern Panamá allowing farmers to establish themselves more easily in previously inaccessible areas.

Since the Guna were strategically settled on the coastline and offshore islands for subsistence, they had no physical presence along the southern mountain border of their Comarca. Thus, the

Guna were unable to protect their forest mainland territory from non-Guna squatters.

Consequently, both sides of the Cordillera de San Blas were cleared by non-indigenous settlers for homesteads, creating negative effects on the environment, culture, and society (Chapin 2000).

At the same time, the Panamanian government was granting a form of land ownership to individuals who could justify their “functional use” of the land; which involved the practice of slash-and-burn agriculture for several years until soil fertility was depleted and they were forced to move to another location (Wali 1993; Peterson St-Laurent et al. 2013). Thus, the Gunas’ desire to protect the forest from colonization conflicted with the government policies that incentivized colonization in this region.

131

In 1983, as the Guna had increasingly focused on protecting their land and managing their natural resources, a group of Guna leaders conceived and implemented the first indigenous project in Latin America to establish and manage a protected area (Study Project for the

Management of the Wildlands of Kuna Yala—PEMASKY) (PEMASKY 1986; Chapin 2000).

The goal was to defend their territory, and the task was to define the physical limits of the

Comarca, while at the same time, defend Guna lands against the encroachment of outsiders. Yet,

Chapin (2000) describes that it was seen by conservationists as an inadvertently significant move to conserve biodiversity at the southern border of their territory. The legacy of PEMASKY’s short-term initiative was perhaps the Guna’s appreciation for the fragile state of their own forest environment, and a collective movement towards seeking strategies to protect and restore the ecosystem from outside forces, which continues today.

Climate change: implications for Guna Yala traditional land

Whilst coastal and island communities are on the front line of environmental changes associated with sea level rise, the effect of climate change on the cloud forest constitutes another significant environmental concern (Bruijnzeel 2001; Oliveira 2014). Because of the intrinsic relationship between the Indigenous people of the Guna Yala, the cloud forest, and the San Blas islands in which they live, climate change is expected to have a wide range of detrimental effects, including a loss of ecosystem services and life-sustaining marine and coastal resources (Murray

& Oullet-DeCoste 2008; Arenas 2014, 2016).

The International Panel for Climate Change (IPCC) (1997) notes that small islands and atolls, such as those found in San Blas, which rarely exceed 3 – 4 m above sea level, will be threatened by rising seas, as they are most susceptible to flooding and inundation (Murray & Oullet-

DeCoste 2008). Between 1901 and 2010, the global mean sea level rose 0.19 (0.17 – 0.21) m,

132 while the global average combined land and ocean surface temperatures showed a warming of

0.85 [0.65 to 1.06] C ° between 1880 and 2012 (IPCC 2015).

The effect of climate change on ocean ecosystems has been more pronounced. Globally, rising sea surface temperature has been linked to phytoplankton productivity, which is the foundation of the aquatic food web (Hoegh-Guldberg 1999). Ocean acidification has resulted in the bleaching of coral reefs (IPCC 2007), which account for one-third of marine biodiversity around the world. Coral reefs also protect island communities from storm damage (Murray & Oullet-

DeCoste 2008). At the current rate of warming, scientists have predicted that 90% of coral reefs in the Pacific Ocean may bleach or degrade by 2050. Further, the coral reefs of the Caribbean are some of the most degraded on the planet, having been reduced by 80% in just three decades

(Pandolfi et al. 2003). Yet, the reef systems of San Blas have provided a wide range of services for almost the entire Guna population (within and outside of the Comarca). Without abundant marine life, Guna livelihoods, economic progress in the form of tourism, food security, cultural expression, and identity, are at risk (Elliott & Tanguay 2006; Jackson 2014).

Over a period of 30 years, rising sea levels have reduced the surface area by up to 50,363m2 on uninhabited islands in San Blas (Guzmán et al. 2003). The increasing food insecurity due to declining fish populations (overexploitation) and coral bleaching (Guzmán 2003; IPCC 2007), combined with the overpopulation on many islands (Arenas 2014, 2016; Davis 2010) makes the future of the Guna living on offshore islands seem bleak. Natural disasters and weather-related events have already affected the islands to the point that inhabitants of the most affected island,

Gardi Sugdub, have initiated the process of relocation back to the mainland. The inhabitants of

133 neighboring islands (Playon Chico, Gardi Maladup, Digir, and Yandub) have also expressed the desire to be relocated (Arenas 2014, 2016).

Climate change is having significant and often negative social, cultural, and economic consequences on the Guna Yala community, which may lack the adaptive capacity to maintain their way of life on the islands in the face of change. However, there is still uncertainty about the extent and timeframe of the relocation of the Gunas from their small island-states, which is shifting from an “option to a necessity” (Arenas 2016). Relocation mitigation measures will require many considerations and assessments, particularly regarding impacts to the environment and the biodiversity in proposed resettlement areas (Murray & Oullet-DeCoste 2008). If successfully organized and implemented, the Guna Yala relocation of the first island community

(Gardi Sugdub), could become a model for neighboring islands to follow. However, according to

Arenas (2014, 2016), the Panamanian government does not have a strategy for assisting the affected island communities, nor a policy for the relocation process to the mainland.

Displacement of the Guna Yala: an option or necessity?

According to the World Bank (2008), indigenous peoples account for just 5% of the world’s population, but they protect an estimated 22% of the Earth’s surface, 80% of remaining biodiversity and 90% of cultural diversity on the planet. Globally, they are among the poorest and most socially marginalized people (Mearns & Norton 2009; Kronik & Verner 2010; World

Bank 2015). Owing to their heavy (sometimes sole) dependence on their local ecosystems (e.g., tropical forests, high mountains, small islands) as a source of livelihood, they tend to be disproportionately affected by climate change (Heltberg et al. 2010). Further, the UN

International Migration Agency (2018) highlights the link between climate change impacts and

134 the migration of indigenous people from ancestral lands, stating there is insufficient knowledge on how climate change is affecting indigenous migration patterns worldwide.

Yet, the slow-onset effects of climate change such as sea level rise, are expected to bring about

significant displacement patterns for the Guna people, within and outside of San Blas. Barnett

(2001) explains, to address this issue, the most likely adaptive strategy is a progressive and

permanent relocation. However, Raleigh & Jordan (2010) suggest, that although resettlement

reduces people’s physical vulnerability to disaster risk, it may conversely decrease living

standards, thereby increasing the economic and social vulnerability of resettled populations.

Moreover, the socio-ecological thresholds surrounding sea level rise may trigger the resettlement

process more than the physical impacts of climate change itself (Barnett & Adger 2003). Thus,

the primary factors to likely force the exodus of the Gunas from the islands include high

population densities and low levels of available resources for adaptive measures.

Over the past ten years, several severe weather-related events and natural disasters have

underscored the threat of climate change and rising sea levels. Moreover, population growth has exacerbated the issues. As a result, relocating to the mainland has become an increasingly important item on the agenda for the Guna General Congress and the local Sailas – known as wise men or community leaders (Arenas 2014). One of the most challenging obstacles to

resettlement as an adaptive strategy to climate change, is perhaps securing a location where

people’s livelihoods, traditions, and cultures are not significantly altered (Gavin et al. 2015). In

this regard, the Guna Yala face an interesting scenario. After living for decades in San Blas, they

are now confronting the return to their ancestral land, the cloud forest, which they left more than

200 years ago due to environmental hostility (e.g., malaria, yellow fever) (Howe 2002; Apgar

2010).

135

In years to come, it is estimated that approximately 28,000 people from 48 Guna communities will eventually have to be relocated from the islands to the mainland. Additionally, another

12,000 people that already live in Panamá City, may want to return to their home province to rejoin their families (Arenas 2014). Yet, according to Raleigh & Jordan (2010), such large-scale community relocation stemming from sea level rise is unlikely to happen within the next 20 years. This assertion appears to be true for the communities of Gardi Sugdub and Playon Chico, partly because the government has not taken action to assist the Guna in the relocation process.

Further, the government’s attempts to address environmental issues have a very poor reputation, as inadequate planning and facilities combine with the politicized nature of the resettlement process (Arenas 2016). In addition, participatory planning is not easy, as it requires the involvement of governments and planners and the voice of those who are relocating.

Nonetheless, inadequate planning (e.g., housing development, infrastructure, agriculture) comes with consequences, in this case, potential degradation of the forest and its biodiversity (Arenas

2014).

Large carnivores, connectivity and protected areas

It has been demonstrated that one of the most significant anthropogenic impacts on nature is climate change and the extirpation of large carnivores (Estes 2010; Estes et al. 2011). Estes et al.

(2011) refer to trophic downgrading, as the process of removing large apex predators from nature. Downgrading effects the structure and dynamics of global ecosystems, as it acts synergistically with climate change, land use change, and habitat loss. Moreover, habitat loss and fragmentation, human persecution, and depletion of prey can affect the social structure and influence the ecological role of large carnivores (Ripple 2014). The ecological argument is that despite existing in low densities, large carnivores can exert strong regulatory effects on

136 ecosystems (Estes 2010; Ripple 2014). Further, Estes et al. (2011) proposed that many of the ecological issues (e.g., pandemics, shifts in ecosystems, increases of unwanted species) were ultimately the result of the loss of apex predators. As apex predators, they regulate the relative abundance of smaller carnivores and herbivores (Crooks & Soule 1999; Haskell et al. 2013).

Thus, they enhance biodiversity, reestablish native vegetation (ensuring CO2 absorption), and

even regulate diseases (Hebblewhite et al. 2005; Prugh et al. 2009; Berger et al. 2012; Ripple

2014). In addition, apex carnivores provide not only ecosystem services, but also economic

services (Ripple 2014). Because of their iconic status and charismatic nature, large carnivores

deliver economic benefit to communities associated with tourism and conservation (e.g., jaguars

in South America, lions in Africa, tigers in India).

Yet, large carnivores are prone to extinction because of their slow growth rates, large habitat

requirements, and naturally low population densities (Purvis et al. 2000). Petracca et al. (2014)

argue that conservation strategies for large carnivores should stress the maintenance of range-

wide connectivity that includes protected areas (PAs) and dispersal corridors that facilitate

movement and ensure connectivity between populations. Maintaining forest connectivity

prevents inbreeding by allowing for genetic exchange, and offers greater resilience to

disturbances, which ultimately leads to higher species fitness and persistence (Olds et al. 2012).

However, Pitman et al. (2017) suggest that measuring connectivity levels is difficult due to

limited datasets, resulting in a lack of integration of connectivity in conservation planning and

wildlife management. Additionally, an essential component to connectivity conservation is the

identification of resources and factors within the habitat that facilitate species movement (Pitman

et al. 2017). Without this information, it is difficult to provide the habitat required for

connectivity or permeable matrices between non-protected areas and PAs.

137

Large carnivores face challenges in human-dominated landscapes, as their predatory behavior and large habitat requirements often demand more landscape than what most PAs provide.

Sometimes, PAs are insufficient to maintain ecological and population processes or movement of organisms within their boundaries (Hansen & Defries 2007). PAs cover more than 12% of the

Earth’s land area and they are recognized as the most important core units for in situ conservation (Chape et al. 2005), yet, they do not always function as planned (Hansen & Defries

2007). Despite their intended role, increasing human populations and intensified land use surrounding PAs causes changes to the ecological functions and biodiversity within. Thus, maintaining buffer areas surrounding PAs enhances the capacity to conserve species richness by increasing the size, which in turn helps to maintain ecological processes (Hansen & Defries

2007).

Anthropogenic activities such as road construction, agriculture, and extraction of natural resources, lead to the clearing of primary forests around PAs in the tropics (Mustard et al. 2004).

Subsequently, edge effects can lead to habitat change due to increased disturbance rates and forest mortality within the PA, or other less obvious but influential effects, such as greater access to hunting and poaching of predators and prey (Hansen & Defries 2007). Beyond that, most protected areas are embedded within matrices of heterogeneous land uses that are dominated by human-related activities (Hansen & Defries 2007). The future of biodiversity within a PA and the surrounding matrices are linked, making the surrounding landscapes around protected areas equally as important for the conservation of large, wide-ranging carnivores (Harvey et al. 2009).

Consequently, strategies to facilitate human coexistence with large carnivores across the landscape are essential to their long-term survival. Hansen & Defries (2007) determined that the

138 management of human needs and long-term conservation of ecosystems is a complex balance, with the goal to “identify management solutions that satisfy human needs while maintaining ecological function.” Yet, the need for large carnivore conservation near human populations is centered on whether the negative impacts humans and large carnivores have on each other can be minimized (Carter & Linnell 2016). Large, predatory felids are one of the most susceptible to landscape level change and habitat loss (Ceballos et al. 2005; Schipper et al. 2008), since these carnivores generally require expansive areas with ample prey to persist (Crooks 2002). Human- felid conflict and overexploitation of felids and prey are the leading causes of their decline and extirpation today (Macdonald & Loveridge 2010). Thus, most large felids are considered species of conservation concern (Weber & Rabinowitz 1996; Crooks 2002; De Angelo et al. 2011).

Reconciling the needs of both humans and felids is a complex but essential management task considering the decreasing numbers of wild cats worldwide (Inskip & Zimmermann 2009).

Ripple et al. (2014) reported that 9 out of 10 of the largest felid species (body mass range 161–18 kg) were described as declining per the IUCN Red List (IUCN). Of those ten felid species, 10% are considered endangered, 10% are near threatened, 60% are considered vulnerable, and 20% are listed as least concern (IUCN 2015, 2016, 2017).

Implications of climate change on felids through time

Felids are a phylogenetically and ecologically homogeneous group comprised of 38 wild species

(Schneider et al. 2015). Thirteen felid species inhabit the Americas. The jaguar is the largest and occupies a broad range of habitats from northern Mexico to Argentina. Their high trophic level, extensive home range, and low population density make them highly vulnerable to habitat loss and fragmentation (Sunquist & Sunquist 2002). Though felids are among the most studied mammal groups (Amori & Gippoliti 2000), the understanding of the effects of habitat loss and

139 fragmentation is limited, and present knowledge varies considerably among species (Zanin et al.

2015). Further, few studies have focused on the synergies between climate change and habitat loss for felids and their prey. Yet, the potential impacts warrant serious consideration for the future of these species (Bellard et al. 2012).

Although the evidence of current extinctions from the effects of climate change is limited

(Bellard et al. 2012), a significant range shift for most species across their prehistoric distributions, driven by large scale environmental change has been identified (Arias-Alzate et al.

2016). For example, prehistoric records of jaguars found in a diversity of habitats indicated a high degree of environmental plasticity (Rodrigues et al. 2018). Over recent decades, climate change has led to heritable, genetic changes in populations of animals with short life cycles, allowing them to persist (Walther et al. 2002; Bradshaw & Holzapfel 2006; Fuller 2010). On the other hand, populations of large animals with long lifespans and smaller population sizes

(including jaguars), may decline in population size or may be displaced by other species

(Walther et al. 2002; Bradshaw & Holzapfel 2006).

A pressing issue in the face of global climate change is to better understand the adaptive responses of species to avoid extinction. Two possible responses are (1) micro-evolution – the heritable shifts in allele frequency in a population (see Bradshaw & Holzapfel 2006) and (2) phenotypic plasticity – the ability of individuals to modify their behavior, morphology, or physiology in response to altered environmental conditions (Fuller et al. 2010; Bellard et al.

2012). The former is the most likely response available for long-lived species like the jaguar.

Besides phenology (i.e., timing of the biological events in species) and morphology (i.e., color patterns, body shape and size), Fuller et al. (2010) postulate that the physiological capacity (i.e.,

140 plasticity) of organisms to respond to climatic variability is a key component in predicting the effects of climate change (see Fuller et al. 2010 and references therein; Bellard et al. 2012).

Additionally, the ecological niche concept has been used to explain species distribution patterns, which are governed by different factors that include dispersal abilities, biotic interactions and climatic conditions (see Arias-Alzate et al. 2016 and references therein). Since climate is a key factor in determining a species ecological niche, it can help to understand the implications of range shifts. Historical patterns of species responses can shed light on how they survived climatic changes, which have generated shifts in the availability of suitable habitat and provide insight on the increased or decreased probabilities of their extinction. As large carnivores are more susceptible to range reduction given their area requirements and biological characteristics, they require large distributional ranges to persist, avoid bottlenecks, and protect against extinction (see Arias-Alzate et al. 2016 and references therein).

The link between forest loss and climate change should raise urgency for forest conservation

(Haruna et al. 2014). Since habitat loss and fragmentation, prey loss, and poaching are the biggest threats to felids in the Americas, operating synergistically with climate change; in conservation planning, it is of utmost importance to protect diverse habitats where a species lives and genetic diversity within a species, to provide the capacity for a species to adapt (Bellard et al. 2012). Protection should focus on forest regions, since they serve as climate change regulators and biodiversity refuges. Based on prehistoric records of environmental change, the current synergy of cumulative threats, and species biological requirements, one of the most vulnerable big cats in the Americas is the jaguar (Arias-Alzate et al. 2016).

Minimizing the human footprint and jaguar research needs

141

The pending resettlement of the Guna back to the forest will not only initiate social, cultural, and economic change for people, but also may trigger the onset of ecological changes to the forest in the Comarca. The mainland Comarca has one of the highest potentials for jaguar conservation in

Panamá because it is one of the largest strongholds of intact forest, mostly free of development.

Further, it falls within the path of the jaguar corridor initiative, a range-wide initiative that has identified priority areas for jaguar connectivity between populations (Sanderson et al. 2002;

Zeller 2007; Rabinowitz & Zeller 2010). Plans to address long-term environmental challenges resulting from the impending resettlement of thousands of people to the most well-preserved forest area, at the narrowest section of the country (Arenas 2014), are essential.

Rapid changes to the Guna culture and development in the Comarca are inevitable due to the proposed relocation. The projected development and associated infrastructure would promote the formation of more settlements, increasing rates of colonization from Guna and outside colonists; acting as catalysts for land cover change and biodiversity loss (Perz 2014; Espinosa et al. 2018).

The potential for the Comarca to provide suitable habitat for jaguars will be compromised if it becomes more accessible by roads, allowing deeper access into the forest, likely resulting in increased hunting opportunities and habitat degradation (Chomitz & Gray 1999; Campbell et al.

2017). The implications of development in a biodiversity hotspot and the synergism of threats could impede jaguar survival by creating a potential bottleneck for species dispersal and migration.

The cumulative effects of human development, through agricultural expansion and the creation of infrastructure (i.e., houses, roads, and services), have been inexorably linked to deforestation and landscape deterioration (Caviglia-Harris & Harris 2011). Though it is imperative to

142 acknowledge the rights of indigenous people’s self-determination and autonomy, it is also important to identify the ecological impact of any settlement design on one of the most biodiverse ecosystems in Panamá. Sustainable development within the forest must be a policy goal for the Guna struggling to survive the effects of climate change, while addressing environmental disturbances related to new settlements, and the impacts of human population growth and biodiversity loss.

Sustainable development has become a policy goal for some governments facing population growth, loss of biodiversity, and climate change. Determining a settlement design that incorporates environmental protection and social needs is the objective, but no human settlement can be implemented without deforestation (Caviglia-Harris & Harris 2011). Arenas (2014) recommended that both local Panamanian governments and the Guna leaders coordinate with development programs to minimize the ecological effects on the ecosystem, reduce conflict with conservation objectives, and develop community–based strategies for wildlife management.

Espinosa et al. (2018) suggest establishing policies specifically for road construction, since roads perpetuate pervasive encroachment into the forest. This is of importance considering the high vulnerability of jaguars to habitat loss and development. Additionally, policies must seek to identify priority areas for conservation (Crooks et al. 2011), and enforcement of conservation actions should aim to prevent future loss of jaguars and other threatened species (Di Minin et al.

2016). Conservation actions need to consider the fact that social-ecological systems are always changing (Gavin et al. 2015). Hence, incorporation of adaptive approaches for cultural systems is needed. Finally, plans must assure that settlement designs meet social goals (e.g., maintain a united community, contact between families, access to services), but not at the expense of environmental principals (Caviglia-Harris & Harris 2011).

143

There are several considerations for the design and implementation of a new settlement in a biodiversity hotspot (Caviglia-Harris & Harris (2011). First, selection of settlement sites should consider the constraints of the biophysical landscape and the subsequent environmental impacts, such as access to water and surface topography. Second, housing designs should maintain the traditions of the Guna’s, tailored to the environments that their communities have known intimately for generations, which will maintain a sense of Guna cultural aesthetics in the landscape. Third, an eventual shift in the economy from subsistence-based to service-based (e.g., intensive commercial fishing to commercial agriculture) is expected, and considerations should be made to minimize the ecological footprint (e.g., commercial crops, overharvesting of wildlife). Fourth, although people will be allocated ample land, the inevitable encroachment on adjacent forestland is best kept at a minimum, thus forest clearing laws should be created and enforced.

Lastly, little is known about the ecology of jaguar and prey populations in Panamá (e.g., Figel et al. 2009; Jiménez et al. 2010; Meyer et al. 2015, 2016; Fort 2016). Conservation efforts have mostly been directed at human-jaguar conflict with livestock (Moreno et al. 2015; Fort 2016;

Fort et al. 2018), while some research has revealed aspects of the jaguar’s ecology (Fort 2016).

Meyer et al. (2015) reported that terrestrial mammal communities in the forests of Central

Panamá were degraded, and that jaguar densities were low. Whereas, de la Torre et al. (2018) determined that jaguar populations in central Panamá were critically endangered. There remains uncertainty about the jaguar population and distribution or how the accelerating changes to the landscape affect their movements. Yet, scientific-based information on jaguar habitat use is required to make informed decisions about jaguar conservation amidst future development in the

144

Comarca. Oriol-Cotterill et al. (2015), for example, recommended using GPS-collar data to detect the spatio-temporal differences in habitat use and jaguar activity patterns in response to different levels of human-caused mortality risk. In addition, habitat suitability optimization modeling is another way to identify priority areas for jaguars within the landscape (K.C, unpublished data), aiming to provide suitable recommendations for the conservation of prime habitats and dispersal corridors (Sarkar et al. 2018).

Conclusion

Panamá is a biologically diverse country with high species richness. However, increasing anthropogenic pressures are threatening the forests, ecosystems, and ecosystem services. More recently, climate change has been added to the list of looming threats. The impacts of climate change have begun to affect the Indigenous Guna Yala communities living in San Blas. Heavy storms, flooding, and degraded marine resources coupled with increasing human populations have initiating a unique situation for these island communities. Having left their forest home seeking a new life on the islands over a century ago, the Guna are faced with the impending relocation back to the forest. The potential displacement of thousands of people from their island communities will likely have significant social, cultural, and economic consequences for the people.

Though the Guna are an autonomous community who manage their own resources, with little influence from the federal government, the implementation of a relocation process has been hindered by a lack of government support. Nonetheless, the resettlement of thousands of people will inevitably affect the forest and its biodiversity. The slow ongoing impacts of climate change on the forest, coupled with the proposed development of the landscape and associated

145 infrastructure, will act synergistically on species like the jaguar. As large, apex predators, jaguars are sensitive to changes within the landscape, since they have large habitat requirements that include ample prey to survive. The conservation of vast protected areas that provide connectivity and dispersal corridors are essential to the persistence of these big cats. Correspondingly, the protection of the jaguar maintains the trophic structure of the forest ecosystem and ecosystem services that are essential to human survival.

Proactively planning for the emerging and intrinsically connected issues of climate change, human displacement, and jaguar conservation, is a complex but essential management task.

Facilitating human-jaguar coexistence will be most effectively addressed by developing culturally acceptable and sustainable solutions combined with scientific research (Inskip &

Zimmermann 2009). The key to a successful relocation of people will need to embrace the idea of biocultural conservation, by merging science with indigenous culture. The Guna’s ecological knowledge can provide the tools for long-term sustainability and resource conservation in the

Comarca (Martin et al. 2010), but this must be coupled with biological data to integrate the needs of biodiversity (i.e., jaguars and prey). To increase an adaptive capacity for both ecological and human social systems in the face of climate change, an integration of worldviews and resource management frameworks will create a multifaceted approach to conservation planning and promote long-term success (Gavin et al. 2015). Acknowledging the Guna Yala community as key stakeholders in the development of a sustainable conservation approach is essential to ensure long-term jaguar conservation and minimize human-jaguar conflict in the Comarca.

Literature Cited

Adger, N. J. Barnet, K. Brown, N. Marshall, K. O’Brien. 2003. Cultural dimensions of climate change impacts and adaptation. Nature Climate Change 3:112-117.

146

Aldrich, M., C. Billington, M. Edwards, R. Laidlaw. 1997. Tropical Montane Cloud Forests: An Urgent Priority for Conservation. World Conservation Monitoring Centre 2:1–16.

ANAM. 2010. Atlas ambiental de la Republica de Panamá.

Apgar, J. M., 2010. Adaptive capacity for endogenous development of Kuna Yala, an indigenous biocultural system. Doctoral dissertation. Lincoln University.

Arenas, C. 2014. The Peninsula Principles in Action: Climate Change And Displacement in the Autonomous Region Of Gunayala, Panama. Mission Report.

Arenas, C. 2016. An Overview on the Relocation of Guna Indigenous Communities in Gunayala, Panama. Displacement Solution. Mission Report.

Arias-Alzate, A., J. F. González-Maya, J. Arroyo-Cabrales and E. Martínez-Meyer. 2016. Wild Felid Range Shift Due to Climatic Constraints in the Americas: a Bottleneck Explanation for Extinct Felids? Journal of Mammalian Evolution 4:427–438.

Amori, G. and S. Gippoliti. 2000. What do mammalogists want to save? Ten years of mammalian conservation biology. Biodiversity and Conservation 9:785–793.

Barnett, J. 2001. Adapting to climate change in Pacific island countries: the problem of uncertainty. World Development 29:977–993.

Barnett, J. and N. Adger. 2003. Climate dangers and atoll countries. Climatic Change 61:321– 337.

Bellard, C., C. Bertelsmeier, P. Leadley, W. Thuiller, C. Courchamp. 2012. Impacts of climate change on the future of biodiversity. Ecology Letters 15:365–377.

Berger, J., P. B. Stagey, L. Bellis and M. P. Johnson. 2001. A mammalian predator-prey imbalance: grizzly bear and wolf extinction affect avian neotropical migrants. Ecological Applications 11:947–960.

Bonan, G. 2008. Forests and Climate Change: Forcings, Feedbacks, and the Climate Benefits of Forests. Science 320:1444–1449.

Bradshaw, W. E. and C. M. Holzapfel. 2006. Climate change: evolutionary response to rapid climate change. Science 312:1477–1478.

Breslin, P. and M. Chapin. 1984. Conservation, Kuna Style. Grassroots Development 8:26–35. Bruijnzeel, L. A.. 2001. Hydrology of tropical montane cloud forests: a reassessment. Land use and water resources research 1:1–18.

147

Brummitt, N. and E. N. Lughadha. 2003. Biodiversity: where's hot and where's not. Conservation Biology 17:1442–1448.

Campbell, M., M. Alamgir and W. Laurance. 2017. Roads to ruin: Can we build roads that benefit people while not destroying nature?. Australasian Science 38:40.

Carter, N. H. and J. D. Linnell. 2016. Co-adaptation is key to coexisting with large carnivores. Trends in Ecology & Evolution 31:575–578.

Castillo, G. 2009. The Forests Dialogue. Introduction and Working Paper. Panama, Panama.

CATHALAC (Centro del Agua del Tropico Humedo para America Latina y El Caribe). 2008. Informe proyecto de asistencia técnica para la actualización del mapa de vegetación, uso y cobertura boscosa de Panamá. Informe final de análisis y actualización de la cobertura boscosa de Panamá 2008 ANAM. Panama.

Caviglia-Harris, J. and D. Harris. 2011. The impact of settlement design on tropical deforestation rates and resulting land cover patterns. Agricultural and Resource Economics Review 40:451–470.

CGK. 2008. Ubicacion geografica. Retrieved 30 December 2008. Available at http://www.congresogeneralkuna.org/ubicacion_geografica.htm.

Chape, S., J. Harrison, M. Spalding and I. Lysenko. 2005. Measuring the extent and effectiveness of protected areas as an indicator for meeting global biodiversity. Philosophical Transactions of the Royal Society of London B: Biological Sciences 360:443–455.

Chapin, M. 2000. Defending Kuna Yala: PEMASKY, the study project for the management of the wildlands of Kuna Yala, Panama. A case study for shifting the power: Decentralization and biodiversity conservation (No. 333.95097287 C463). Biodiversity Support Program, Washington, DC (EUA).

Cione, A. L., E. P. Tonni and L. H. Soibelzon. 2009. Did humans cause large mammal late Pleistocene-Holocene extinction in South America in a context of shrinking open areas?. Pages 125–144 in American Megafaunal Extinctions at the End of the Pleistocene. G. Haynes, editor. Springer Publishers, Vertebrate Paleobiology and Paleontology Series.

Cione, A. L., G. M. Gasparini, E. Soibelzon, L. H. Soibelzon and E. P. Tonni. 2015. The Great American Biotic Interchange: A South American Perspective. Springer, Dordrecht.

Crooks, K. R. and M. E. Soulé. 1999. Mesopredator release and avifaunal extinctions in a fragmented system. Nature 400:563–566. Crooks, K. R., C. L. Burdett, D. M. Theobald, C. Rondinini and L. Boitani. 2011. Global patterns of fragmentation and connectivity of mammalian carnivore habitat. Philosophical Transactions of the Royal Society B: Biological Sciences 366:2642–51.

148

Davis, E. 2010. Diagnostico de la Población Indígena de Panamá con bases en los Censos de Población y Vivienda. page 186. Instituto Nacional de Estadistica y Censo (INEC), Panama. de la Torre, J. A., J. F. González-Maya, H. Zarza, G. Ceballos and R. A. Medellín. 2018. The jaguar's spots are darker than they appear: assessing the global conservation status of the jaguar Panthera onca. Oryx 52:300–315.

Di Minin, E., R. Slotow, L. T. B. Hunter, F. Montesino Pouzols, T. Toivonen, P. H. Verburg, N. Leader-Williams, L. Petracca, A. Moilanen. 2016. Global priorities for national carnivore conservation under land use change. Scientific Reports 6:23814.

Elliott, H. and L. Tanguay. 2006. Traditional Kuna Agriculture in the Face of Occidental Influence: Impacts and Responses in Ukupsen. McGill University, Panama Field Study Report.

Ellis, E. C. 2011. Anthropogenic transformation of the terrestrial biosphere. Philosophical Transactions of the Royal Society A 369:1010–1035.

Espinosa, S., G. Celis and L. C. Branch. 2018. When roads appear jaguars decline: Increased access to an Amazonian wilderness area reduces potential for jaguar conservation. PLoS One 13:1–18.

Estes, J.A., C. H. Peterson and R. Steneck. 2010. Some effects of apex predators in higher- latitude coastal oceans. Pages 37–53 in Trophic Cascades: Predators, Prey, and the Changing Dynamics of Nature. J. Terborgh, J. A. Estes, editors. Island Press, Washington, DC, 2010.

Estes, J.A. et al. 2011. Trophic downgrading of planet Earth. Science 333:301–306.

Food and Agriculture Organization (FAO). 2010. Global Forest Resources Assessment. Food and Agricultural Organization of the United Nations. Rome, Italy.

Figel, J. J., E. Durán, D. B. Bray and J. R. Priscilano-Vázquez. 2009. New jaguar records from montane forest at a priority site in southern Mexico. CATnews 50:14–15.

Fort, J. L.. 2016. Large Carnivore Occupancy and Human-wildlife Conflict in Panamá. Doctoral dissertation. Southern Illinois University, Carbondale.

Fuller, A., T. Dawson, B. Helmuth, R. S. Hetem, D. Mitchell, S. K. Maloney. 2010. Physiological Mechanisms in Coping with Climate Change. Physiological and Biochemical Zoology 83:713–720.

149

Gavin, M. C., J. McCarter, A. Mead, F. Berkes. 2015. Defining biocultural approaches to conservation. Trends in Ecological Evolution 30:140–145.

Guzmán, H. M., C. Guevara and A. Castillo. 2003. Natural disturbances and mining of Panamanian coral reefs by indigenous people. Conservation Biology 17:1396–1401.

Guzmán, H. M. 2003. Caribbean coral reefs of Panama: present status and future perspectives. Pages 241–274 in Cortes, J., editor. Latin American coral reefs. Amsterdam, Elsevier Science, B.V.

Hamilton, L. S., J. O. Juvik and F. N. Scatena, editors. 1995. Tropical Montane Cloud Forests. Ecological Studies 110, Springer-Verlag, New York.

Hansen, A. J. and R. DeFries. 2007. Ecological mechanisms linking protected areas to surrounding lands. Ecological Applications 17:974–988.

Haruna, A., A. Pfaff, S. Van den Ende and L. Joppa. 2014. Evolving protected-area impacts in Panama: impact shifts show that plans require anticipation. Environmental Research Letters 9:1–11.

Harvey, C. A., et al. 2009. Integrating agricultural landscapes with biodiversity conservation in the Mesoamerican hotspot: Opportunities and an action agenda. Conservation Biology 22:8–15.

Haskell, D. E., C. R. Webster and D. J. Flaspohler. 2013. Relationship between Carnivore Distribution and Landscape Features in the Northern Highlands Ecological Landscape of Wisconsin. The American Midland Naturalist 169:1–16.

Hebblewhite, M., C. A. White, C. G. Nietvelt, J. A. McKenzie, T. E. Hurd, J. M. Fryxell, S. E. Bayley and P. C. Paquet. 2005. Human activity mediates a trophic cascade caused by wolves. Ecology 86:2135–2144.

Heltberg, R., P. Bennett Siegel and S. L. Jorgensen. 2010. Social Policies for Adaptation to Climate Change. Pages 259–275 in Social dimensions of climate change: equity and vulnerability in a warming world. R. Mearns and A Norton, editors. The World Bank, Washington, DC.

Hernandez, A. 1984. Migracion de colonos en Darien. Pages 105–118 in Colonizacion y Destruccion de Bosques en Panama, S. H. Moreno and A. McKay, editors. Asociación Panameña de Antropología, Panama City, Panama.

Hoegh-Guldberg, O. 1999. Climate change, coral bleaching and the future of the world’s coral reefs. Marine Freshwater Resources 50:839–866.

150

Howe, J. 2002. The Kuna gathering: Contemporary village politics in Panama (2nd ed.). Tucson, AZ, Fenestra Books.

Inskip C, Zimmermann A. 2009. Human-felid conflict: A review of patterns and priorities worldwide. Oryx 43:18–34.

International Organization of Migration. Retrieved on December 29, 2018. Available at https://www.iom.int/.

IPBES. 2018. Summary for policymakers of the thematic assessment report on land degradation and restoration of the Intergovernmental Science Policy Platform on Biodiversity and Ecosystem Services. In R. Scholes, L. Montanarella, A. Brainich, N. Barger, B. ten Brink, M. Cantele, B. Erasmus, J. Fisher, T. Gardner, T. G. Holland, F. Kohler, J. S. Kotiaho, G. Von Maltitz, G. Nangendo, R. Pandit, J. Parrotta, M. D. Potts, S. Prince, M. Sankaran and L. Willemen, editors. IPBES secretariat, Bonn, Germany.

IPCC. 1997. The Regional Impact of Climate Change: An Assessment of Vulnerability. Retrieved January 17, 2008, from Intergovernmental Panel on Climate Change-Special Report. Available at http://www.ipcc.ch/ipccreports/special-reports.htm.

IPCC. 2007. Climate change 2007: the physical basis. Summary for policy-makers. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, Cambridge University Press.

IPCC. 2014: Summary for Policymakers. In: Climate Change 2014: Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Edenhofer, O., R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel and J.C. Minx (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Jackson, J. B. C., M. K. Donovan, K. L. Cramer and V. V. Lam, editors. 2014. Status and Trends of Caribbean Coral Reefs: 1970–2012. Global Coral Reef Monitoring Network. IUCN, Gland, Switzerland.

Jiménez, C. F., H. Quintana, V. Pacheco, D. Melton, J. Torrealva and G. Tello. 2010. Camera trap survey of medium and large mammals in a montane rainforest of northern Peru. Revista peruana de Biología 17:191–196.

Kronik, J. and D. Verner. 2010. The Role of Indigenous Knowledge in Crafting Adaptation and Mitigation Strategies for Climate Change in Latin America. Pages 145–169 in Social dimensions of climate change: equity and vulnerability in a warming world. Robin Mearns and Andrew Norton, editors. The World Bank, Washington, DC. Leigh, E. G., A. O’Dea and G. T. Vermeij. 2013. Historical biogeography of the Isthmus of Panama. Biological Reviews 89:148–172.

151

Locatelli, B. 2016. Ecosystem Services and Climate Change. Pages 481–490 in Routledge Handbook of Ecosystem Services. M. Potschin, R. Haines‐Young, R. Fish and R. K. Turner, editors. Routledge, London and New York.

Mace, G. M., I. Bateman, S. Albon, A. Balmford, C. Brown, A. Church, R. Haines-Young, J. N. Pretty, K. Turner, B. Vira and J. Winn. 2011. Conceptual framework and methodology. In UK National Ecosystem Assessment: Technical Report. United Nations Environment Programme World Conservation Monitoring Centre, Cambridge.

Manoiu, V. M., C. Alexandru-Ioan and S. Răzvan-Mădălin. 2015. The Geoecological Structure Typical for The Depression Basin Of The Băile Herculane Resort, Romania.

Martinez Mauri, M. 2007. De Tule Nega a Kuna Yala. Mediacion, territorio y ecologia en Panama. Paginas 1903–2004. Universidad Autonoma de Barcelona & Ecole de Heute Etudes en Sciences Sociales.

Martin, J. F., E. D. Roy, S. A. Diemont, B. G. Ferguson. 2010. Traditional Ecological Knowledge (TEK): Ideas, inspiration, and designs for ecological engineering. Ecological Engineering 36:839–849.

McCarter, J., et al. 2018. Biocultural approaches to developing well-being indicators in Solomon Islands. Ecology and Society 23:32.

Mearns, R. and A. Norton. 2010. Equity and Vulnerability in a Warming World: Introduction and Overview. Pages 1–44 in Social dimensions of climate change: equity and vulnerability in a warming world. R. Mearns and A. Norton, editors. The World Bank, Washington, DC.

Meyer, N. F., et al. 2015. An assessment of the terrestrial mammal communities in forests of Central Panama, using camera-trap surveys. Journal for Nature Conservation 26:28–35.

Meyer, N. F., R. Moreno, E. Sanches, J. Ortega, E. Brown, and P. A. Jansen. 2016. Do protected areas in Panama support intact assemblages of ungulates?. Therya 7:65–76.

Midler, E., U. Pascual, S. Simonit. 2014. Forest ecosystems in national economies and contribution of REDD+ in a green economy transformation: the case of Panama Land-use intensification in forest-agriculture frontier landscapes: effects on ecosystem services and poverty alleviation (ESPA-Frontiers). Estelle Midler BC3-Basque Centre for Climate Change Unai Pascual BC3-Basque Centre for Climate Change.

Mooney, H., A. Larigauderie, M. Cesario, T. Elmquist, O. Hoegh-Guldberg, S. Lavorel, G. M. Mace, M. Palmer, R. Scholes and T. Yahara. 2009. Biodiversity, climate change, and conservation services. Current Opinion in Environmental Sustainability 1:46–54. Moreno, R., N. Meyer, M. Olmos, R. Hoogesteijn, and A. L. Hoogesteijn. 2015. Causes of jaguar killing in Panama a long term survey using interviews. CATnews.

152

Murray, J. and E. Ouellet-Décoste. 2008. Cambios Climáticos en Ukupseni, Comarca Kuna Yala: Investigación, Proyección, y Educación Climate Change in Ukupseni, Comarca Kuna Yala. McGill University.

Mustard, J., R. DeFries, T. Fisher, E. F. Moran. 2004. Land use and land cover change pathways and impacts. Pages 411–430 in G. Gutman, J. Janetos, C. O. Justice, E.F. Moran, J. Mustard, R. Rindfuss, D.L. Skole, B.L. Turner, M.A. Cochrane, editors. Land change science: observing, monitoring, and understanding trajectories of change on the earth’s surface. Springer-Verlag, Dordrecht, The Netherlands.

Myers. C. 1969. The Ecological Geography of Cloud Forest in Panama. American Museum Novitates, no 2396. The American Museum of Natural History, New York, NY.

Nelson, A. and K. Chomitz. 2011. Effectiveness of strict vs. multiple use protected areas in reducing tropical forest fires: a global analysis using matching methods. Plos One 6:e22722.

Oestreicher, J. S., et al. 2009. Avoiding deforestation in Panamanian protected areas: an analysis of protection effectiveness and implications for reducing emissions from deforestation and forest degradation. Global Environmental Change 19:279–91.

Olds, A. D., R. M. Connolly, K. A. Pitt, P. S. Maxwell. 2012. Habitat connectivity improves reserve performance. Conservation Letters 5:56–63.

Oliveira, R. S., C. B. Eller, P. R. L. Bittencourt, and M. Mulligan. 2014. The hydroclimatic and ecophysiological basis of cloud forest distributions under current and projected climates. Annals of Botany 113:909–920.

Oriol-Cotterill, A., M. Valeix, L. G. Frank, C. Riginos and D. W. Macdonald. 2015. Landscapes of Coexistence for terrestrial carnivores: the ecological consequences of being downgraded from ultimate to penultimate predator by humans. Oikos 124:1263–1273.

Pandolfi, J. M., et al. 2003. Global trajectories of the long-term decline of coral reef ecosystems. Science 301:955–958.

Parker, T., J. Carrion and R. Samudio. 2004. Biodiversity and Tropical Forestry Assessment of The Usaid/Panama Program: Environment, Biodiversity, Water, and Tropical Forest Conservation, Protection, and Management in Panama: Assessment and Recommendations. Chemonics International Inc.

Pedrono, M., B. Locatelli, D. Ezzine-de-Blas, D. Pesche, S. Morand, A. Binot. 2016. Impact of Climate Change on Ecosystem Services. Pages 252–261 in E. Torquebiau, editor. Climate Change and Agriculture Worldwide, Springer, Dordrecht.

PEMASKY/AEK. 1986. Kuna Yala Wildlands Project (PEMASKY) Request for Technical Cooperation to the World Wildlife Fund -U.S. for the Period of 1986 (1 year). February. 15- page document.

153

Perz, S. G. 2014. Sustainable development: the promise and perils of roads. Nature 513:178–179.

Petracca, L.S., S. Hernández-Potosme, L. Obando-Sampson, R. Salom-Pérez, H. Quigley, H. S. Robinson. 2014. Agricultural encroachment and lack of enforcement threaten connectivity of range-wide jaguar (Panthera onca) corridor. Journal of Nature Conservancy 22:436–444.

Pitman, R. T., et al. 2017. Cats, connectivity and conservation: incorporating datasets and integrating scales for wildlife management. Journal of Applied Ecology 54:1687–1698.

Potvin, C. and J. Mateo-Vega. 2013. Panama: curb indigenous fears of REDD+. Nature 500:400.

PRONAT. 2008. Programa Nacional de Administración Tierras. Panama.

Prugh, L. R., C. J. Stoner, C. W. Epps, W. T. Bean, W. J. Ripple, A. S. Laliberte and J. S. Brashares. 2009. The rise of the mesopredator. BioScience 59:779–790.

Purvis, A., J. L. Gittleman, G. Cowlishaw and G. M. Mace. 2000. Predicting extinction risk in declining species. Proceedings of the Royal Society of London. Series B: Biological Sciences 267:1947–1952.

Raleigh, C. and L. Jordan. 2010. Climate Change and Mitigation: Emerging Patterns in the Developing World. Pages 103–131 in Social dimensions of climate change: equity and vulnerability in a warming world. R. Mearns and A. Norton, editors. The World Bank, Washington, DC.

Ray, J. C., L. Hunter, and J. Zigouris. 2005. Setting Conservation and Research Priorities for Large African Carnivores. Wildlife Conservation Society, New York, USA.

Ripple, W. J., et al. 2014. Status and ecological effects of the world’s largest carnivores. Science 343:151–162.

Rodriguez, S. G., C. Méndez, E. Soibelzon, L. H. Soibelzon, S. Contreras, J. Friedrichs, C. Luna and A. E. Zurita. 2018. Panthera onca (Carnivora, Felidae) in the late Pleistocene-early Holocene of northern Argentina. Neues Jahrbuch für Geologie und Paläontologie- Abhandlungen 289:177–187.

Sarkar, M. S., A. Pandey, G. Singh, S. Lingwal, R. John, A. Hussain, G. S. Rawat and R. S. Rawal. 2018. Multiscale statistical approach to assess habitat suitability and connectivity of common leopard (Panthera pardus) in Kailash Sacred Landscape, India. Spatial statistics 28:304–318.

Schneider, A., et al. 2015. Recurrent evolution of melanism in South American felids. PLoS Genetics 11:e1004892.

154

Spracklen, B. D., M. Kalamandeen, D. Galbraith, E. Gloor, D. V. Spracklen. 2015. A Global Analysis of Deforestation in Moist Tropical Forest Protected Areas. PLoS One 10: e0143886.

Stadtmuller, T. 1987. Cloud Forests in the Humid Tropics. A bibliographic review. United Nations University, Tokyo, and CATIE, Turrialba, Costa Rica.

Stehli, F. G. and S. D. Webb, editors. 1985. The great American biotic interchange. Topics in geobiology. Plenum Press, New York.

Sterling, E., et al. 2017. Culturally grounded indicators of resilience in social-ecological systems. Environment and Society 8:63–95.

St-Laurent, G. P. 2013. Diversity of Perceptions on REDD+ Implementation at the Agriculture Frontier in Panama. International Journal of Forestry Research.

Sunquist, M. and F. Sunquist. 2002. Wild Cats of the World. University of Chicago Press.

The IUCN Red List of Species. 2015, 2016, 2017: Available at: http://www.iucnredlist.org; (downloaded March 2019).

UNDP (United Nations Development Programme). 2012. Third Consolidated Annual Progress Report on Activities Implemented under the UN-REDD Programme Fund Report of the Administrative Agent of the UN-REDD Programme Fund for the Period 1 January-31 December 2012.

Vergara-Asenjo, G., C. Potvin. 2014. Forest protection and tenure status: the key role of indigenous peoples and protected areas in Panama. Global Environmental Change 28:205– 215.

Vergara-Asenjo, G. 2016. Land cover dynamics in Panama: Local lessons addressing global challenges in the context of REDD. Doctoral dissertation. McGill University.

Vergara-Asenjo, G, J. Mateo-Vega, A. Alvarado and C. Potvin. 2017. A participatory approach to elucidate the consequences of land invasions on REDD+ initiatives: A case study with Indigenous communities in Panama. PLoS One 12:e0189463.

Wali, A. 1993. “The transformation of a frontier: state and regional relationships in Panama, 1972–1990,” Human Organization 52:115–129.

Walther, G. R., E. Post, P. Convey, A. Menzel, C. Parmesan, T. J. C. Beebee, Fromentin, J. M., O. Hoegh-Guldberg and F. Bairlein. 2002. Ecological Responses to Recent Climate Change. Nature 416:389–95. Warner, K., C. Ehrhart, and A. de Sherbinin. 2009. In search of shelter: mapping de effects of climate change on human migration and displacement. CARE International.

155

Weir, J. T., E. Bermingham and D. Schluter. 2009. The great American biotic interchange in birds Proceedings of the National Academy of Sciences 106:21737–21742.

Woodburne, M. O., A. L. Cione, E. P. Tonni. 2006. Central American provincialism and the Great American Biotic Interchange, in Carranza-Castañeda Ó, and Lindsay EH eds. Advances in late Tertiary vertebrate paleontology in Mexico and the Great American Biotic Interchange: Universidad Nacional Autónoma de México, Instituto de Geología and Centro de Geociencias, Publicación Especial 4:73–101.

World Bank. 2008. Social Dimensions of Climate Change: Workshop Report 2008. Social Development Department, World Bank, Washington, DC. Available at: http://www.worldbank.org/sdcc.

World Bank. 2011. Forest Carbon Partnership Facility 2011 Annual Report. Carbon Finance Unit, World Bank, Washington, DC. Available at: http://www.forestcarbonpartnership.org/sites/forestcarbonpartnership.org/files/Documents/P DF/Oct2011/FCPF_Carbon_AR_FINAL_10_3.pdf.

World Bank. 2015. Indigenous Latin America in the Twenty-First Century. Washington, DC: World Bank. License: Creative Commons Attribution CC BY 3.0 IGO.

Zanin, M., F. Palomares and D. Brito. 2015. What we (don't) know about the effects of habitat loss and fragmentation on felids. Oryx 49:96–106.

Zanin, M., R. Sollmann, N. M. Tôrres, M. M. Furtado, A. T. A. Jacomo, L. Silveira and P. De Marco. 2015. Landscapes attributes and their consequences on jaguar Panthera onca and cattle depredation occurrence. European Journal of Wildlife Research 61:529–537.

156

Chapter V: Conclusion

The rapid conversion of the landscape coupled with a paucity of scientific investigation into jaguar (Panthera onca) and puma (Puma concolor) ecology in Panamá has prompted the need for this study. The present research findings highlight important considerations for the conservation of these sympatric predators by addressing key knowledge gaps in habitat selection in Panamá.

The annual jaguar habitat suitability model indicated that secondary forest was the most influential land cover type on jaguar habitat selection at broad scales (2nd order selection or home range selection). The significance of this finding is twofold: first, portions of secondary forest along the southern limits of the study area appeared to be key local determinants of jaguar distribution at the home-range level. Second, it reinforces the theory that broad scales are the most relevant limiting factors among the hierarchical scales of habitat selection (Rettie &

Messier 2000). Theory suggests that species should avoid factors with greater potential to reduce individual fitness at broad spatial and temporal scales. Thus, an individual will maximize fitness by avoiding the factors that are most limiting at each successive scale (i.e., fine) (Rettie &

Messier 2000).

Moreover, a fundamental component of jaguar dispersal limitation in my study area is the

Cordillera de San Blas. It not only marks the limits of undisturbed primary forest to the north, but also the human disturbance to the south. Yet, the jaguar’s fine scale tolerance for human settlement in the annual model is reflected by the low human population density that is scattered throughout the southern portion of the study area, and draws attention to the jaguar’s use of forest edge where intermittent human activity persists.

157

The jaguar seasonal models elucidated the strong influence of fine scales on jaguar habitat selection differential across seasons, most likely due to variances in prey productivity. This finding is corroborated by other studies that reported the seasonal distribution and abundance of prey are determinants of home range shifts in wide-ranging carnivores (Crawshaw & Quigley

1991; Davidson et al. 2012; Sollmann et al. 2012; Nuñez-Perez & Miller 2019).

The multi-scale seasonal models allowed me to identify the habitat features and associated scales that jaguars selected throughout the year. For example, during the wet season, jaguars showed an association with rivers and streams, which are affiliated with prey availability (Nuñez-Perez &

Miller 2019). During the dry season, jaguars used limited amounts of fallow land which may provide areas for dispersal, refuge and resource acquisition. This finding suggests that fallow lands may provide important ecological value for jaguars if left to develop into successional forest.

My seasonal jaguar models indicated that the long-term survival of jaguars in the narrowest section of the country will rely on the persistence of primary forest. As such, my analysis indicated that the most suitable habitat for the species included the Guna Yala territory. In addition, the cover type fallow should be considered in land management strategies, as it may be valuable habitat for planning more resilient landscapes for forest corridors or patches deemed important for jaguar movement.

In the multi-scale optimized puma models, primary forest and its attributes such as elevation and hydrologic features at fine scales (3rd order selection or patch selection, and 4th order selection or resource selection) were consistent predictors of puma occurrence. Additionally, agropecuary landscapes (i.e., agriculture and livestock) emerged as an important variable for pumas at broad

158 scale (i.e., home range). The pumas’ association with agropecuary land is likely attributed to the seasonal distribution and abundance of prey (Sunquist 1981; Fort 2016; Azevedo et al. 2018).

While the puma’s selection of primary forest and its environmental heterogeneity (i.e.,

hydrologic features, elevation, aspect) at fine scales function as the primary driver for the

occurrence of the species in natural habitats, the selection of open habitats (e.g., agropecuary

lands) at broad scales, demonstrates the behavioral plasticity and adaptability of the species to a

diverse range of landscapes (Scognamillo et al. 2003; De Angelo et al. 2011; Elbroch & Wittmer

2012). Further, as seasonal changes in habitat selection are likely determined by changes in the

distribution and abundance of prey, interspecific competition with the jaguar may cause pumas to

suffer displacement at food sources. Thus, pumas may seek suitable habitat for food or other

resources outside of the natural forest. However, given the expansion of land-use change, further

monitoring will provide insight into how pumas respond to the pervasive human encroachment.

My findings highlight the importance of modeling habitat selection using a multi-scale optimized

approach for developing jaguar and puma conservation and management strategies in the region.

The study shows that secondary forest is an important predictor variable in limiting the spatial

distribution of jaguars across scales. This finding suggests there exists a short dispersal range in a

north-south direction, as a thin strip of forest habitat is framed by the Caribbean Sea to the north

and the Cordillera San Blas to the south. In turn, this has implications for the survival of the

species, as it encounters a highly dynamic human-wildlife interface on either side. Moreover, the

influence of fine scales on jaguar habitat selection across seasons revealed the most likely

association between biotic and abiotic factors affecting patch selection (3rd order) of the species

living in a stochastic, or unpredictable environment. Unlike jaguars, pumas incorporated

anthropogenic lands at the interface between primary forest and open areas at broad scales,

159 suggesting a correlation with prey distribution and availability and/or interspecific competition.

In this context, my study shows that species-environment relationships are scale dependent and that studies of species distribution would benefit by using a multi-scale optimized sampling scheme, enabling concurrent inferences about habitat selection at different spatial scales.

While I acknowledge the limitations of my study design, such as the spatial distance between sampling units (camera stations) and the homogeneity of the study area, which precluded the inclusion of additional vegetation cover types as potential habitat for jaguars and pumas, an examination of the scales at which the variables in the models performed, in accordance with the literature, provides biological clues into the marginal and outlying habitat use of jaguars and pumas in the study area. Although the multi-scale modeling approach within habitat selection studies is still new (McGarigal et al. 2016), it has been effective for predicting large, wide- ranging carnivore species distributions across human-dominated landscapes (Torres et al. 2012;

Hearn et al. 2018; Macdonald et al. 2018).

This research contributes to the growing number of studies that demonstrate strong conceptual and inferential advantages of multi-scale approaches (McGarigal et al. 2016; Zeller et al. 2017;

Hearn et al. 2018; Macdonald et al. 2018). The results include a set of maps from which spatial information on jaguar and puma distribution can be extracted. Despite the shortcomings, the models have captured differences in habitat selection and variation in habitat use among seasons within the study area of these sympatric predators. It has also provided a framework for potential applicability throughout the country for predicting species distributions and identifying important habitat for conservation planning. In agreement with other studies, the application of habitat suitability models using a multi-scale optimized approach shows promise (Torres et al. 2012;

Jędrzejewski et al. 2017; Hearn et al. 2018; Macdonald et al. 2018).

160

Recommendations – My findings expand on the existing ecological knowledge of habitat selection by these sympatric predators. However, it is important to evaluate and validate the models by (i) selecting more sampling areas throughout the country, (ii) adding different land cover types that were not represent in my model (study area), but may have ecological value for the species in different geographic areas, (iii) implementing a program that combines camera- trap with GPS-telemetry methods, which track species movements with higher precision than by using camera-trap methods alone. In the meantime, extrapolation of my data across the country should be taken with caution. Following the above steps, resulting predictive maps will be of more value in determining essential areas of conservation, wildlife corridors, and areas of high human-wildlife conflict country-wide. Lastly, I believe that my novel approach may be used as a guideline for future wildlife monitoring and threatened species conservation when evaluating impacts of anthropogenic disturbance, management actions and climatic change. This is particularly applicable to eastern Panamá, where a narrowing section of one of the largest remaining contiguous strips of natural forest is under pressure from increasing human densities, fragmentation, and climate change, threatening an area of highly suitable puma and jaguar habitat.

In eastern Panamá, climate change poses fundamental challenges to the near threatened jaguar and the Guna Yala Indigenous people. Although, in the nation, protected areas and the indigenous territories are the most effective tenure regimes for avoiding deforestation, as argued by Vergara-Asenjo & Potvin (2014), the Panamanian forest, its biodiversity, and ecosystem services are still threatened by a range of anthropogenic pressures. There are significant negative social and cultural consequences from the slow-onset impacts of sea level rise on the Guna Yala community, followed by an inexorable impact on the jaguar population resulting from the

161 pending relocation of the Guna people from their small island-states, to a narrow strip of mainland forest along the Caribbean coast of eastern Panamá. I advocate first, that any resettlement strategy should aim to secure a location where people’s livelihoods, traditions, and cultures are not significantly altered as pointed out by Gavin et al. (2015), and second, that scientific understanding of jaguar (and the entire carnivore guild) distribution and adaptive responses (micro-evolution and phenotypic plasticity) to altered landscapes and environmental changes is needed to prepare for managing this shift in land use due to climate change. The maintenance of effective densities of large carnivores secures the maintenance of the structure and function of the ecosystem in which they live (Ripple et al. 2014). Finally, as postulated by

Bellard et al. (2012), to provide the capacity of a species to adapt and persist, conservation planning should prioritize the protection of the heterogeneity of habitats and the genetic diversity within a species.

The jaguar population in eastern Panamá is part of the larger meta-population of the America’s and my findings must be interpreted in this context. Amid human–wildlife interface dynamics, particularly relating to the relocation of the Guna Yala Indigenous people, I anticipate that increased human density and fragmentation of a narrowing section of one of the largest remaining contiguous strips of natural forest will create a challenging matrix for jaguars to navigate. The development of a human settlement (i.e., housing, roads, crops) may cause deleterious effects to the forest and its biodiversity. Most notably, the genetic erosion of the

Panamanian jaguar population, owing to the increasingly limited space for its dispersal and growth, if careful land use management is not implemented.

162

Literature Cited

Azevedo FC, Lemos FG, Freitas‐Junior MC, Rocha DG, Azevedo FCC. 2018. Puma activity patterns and temporal overlap with prey in a human‐modified landscape at Southeastern Brazil. Journal of Zoology 305:246–255.

Bellard C, Bertelsmeier C, Leadley P, Thuiller W, Courchamp C. 2012. Impacts of climate change on the future of biodiversity. Ecology Letters 15:365–377.

Crawshaw PG, Quigley HB. 1991. Jaguar spacing, activity, and habitat use in a seasonally flooded environment in Brazil. Journal of the Zoological Society of London 223:357–370.

Davidson Z, Valeix M, Loveridge AJ, Hunt JE, Johnson PJ, Madzikanda H, Macdonald DW. 2012. Environmental determinants of habitat and kill site selection in a large carnivore: scale matters. Journal of Mammalogy 93:677–685.

Dobbins MT, Steinberg MK, Broadbent EN, Ryan SJ. 2018. Habitat use, activity patterns and human interactions with jaguars Panthera onca in southern Belize. Oryx 52:276-281.

Fort JL. 2016. Large Carnivore Occupancy and Human-Wildlife Conflict in Panama Large Carnivore Occupancy And Human-Wildlife Conflict In Panamá. Doctoral dissertation. Southern Illinois University Carbondale.

Gavin MC, McCarter J, Mead A, Berkes F. 2015. Defining biocultural approaches to conservation. Trends in Ecological Evolution 30:140–145.

Hearn AJ, Cushman SA, Ross J, Goossens B, Hunter LTB, Macdonald DW. 2018. Spatio- temporal ecology of sympatric felids on Borneo. Evidence for resource partitioning? PLoS One 13:e0200828.

Macdonald DW et al. 2018. Multi-scale habitat selection modeling identifies threats and conservation opportunities for the Sunda clouded leopard (Neofelis diardi). Biological Conservation 227:92–103.

McGarigal K, Wan HY, Zeller KA, Timm BC, Cushman SA. 2016. Multi-scale habitat selection modeling: a review and outlook. Landscape Ecology 31:1161–1175.

Nuñez-Perez R, Miller B. 2019. Movements and Home Range of Jaguars (Panthera onca) and Mountain Lions (Puma concolor) in a Tropical Dry Forest of Western Mexico. Pages 243– 262 in Movement Ecology of Neotropical Forest Mammals. Springer, Cham.

Rettie WJ, Messier F. 2000. Hierarchical habitat selection by woodland caribou: its relationship to limiting factors. Ecography 23:466–478.

Ripple WJ et al. 2014. Status and ecological effects of the world’s largest carnivores. Science 343:1241484.

163

Sollmann R, Furtado MM, Hofer H, Jácomo ATA, Tôrres NM, Silveira L. 2012. Using occupancy models to investigate space partitioning between two sympatric large predators, the jaguar and puma in central Brazil. Mammalian Biology 77:41–46.

Sunquist ME. 1981. The social organization of tigers (Panthera tigris) in Royal Chitawan National park, Nepal. Smithsonian Contributions to Zoology. Smithsonian Institution Press, Washington D.C.

Tôrres NM, De Marco P, Santos T, Silveira L, de Almeida Jácomo AT, Diniz‐Filho JA. 2012. Can species distribution modelling provide estimates of population densities? A case study with jaguars in the Neotropics. Diversity and Distributions 18:615–627.

Vergara-Asenjo G, Potvin C. 2014. Forest protection and tenure status: the key role of indigenous peoples and protected areas in Panama. Global Environmental Change 28:205– 215.

Zeller KA, Vickers TW, Ernest HB, Boyce WM. 2017. Multi-level, multi-scale resource selection functions and resistance surfaces for conservation planning: Pumas as a case study. PloS One 12:e0179570.