<%BANNER%>

Are Forest Strips Cut out for the Job of Conserving Biodiversity? Evaluating the Case for Mammals with Hierarchical Baye...

Permanent Link: http://ufdc.ufl.edu/UFE0043890/00001

Material Information

Title: Are Forest Strips Cut out for the Job of Conserving Biodiversity? Evaluating the Case for Mammals with Hierarchical Bayesian Occupancy Models in the Chaco Forest
Physical Description: 1 online resource (57 p.)
Language: english
Creator: Nunez Regueiro, Mauricio M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: conservation -- mammals
Wildlife Ecology and Conservation -- Dissertations, Academic -- UF
Genre: Wildlife Ecology and Conservation thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Deforestation is a major cause of biodiversity loss, and the largest factor driving deforestation is expansion of agriculture. A key step toward successful conservation in agricultural areas is maximizing the biodiversity value of remaining forest. In subtropical and tropical regions, forest often is left in strips between agricultural fields under the assumption that biodiversity is sustained. In this thesis I examined how medium and large-sized mammals in the Argentine Chaco responded to forest strips retained along agricultural fields and how that response related to ecological traits of species. Presence/absence data from camera traps in continuous forest and forest strips showed that almost half of the total mammal assemblage that occurs in my study site was impacted by conversion of forest to strips. Five species found in forest were completely absent from strips. Hierarchical Bayesian occupancy models also demonstrated that occurrence of mammals decreased in forest strips with increasing distance from continuous forest. Species that occupied a variety of habitats and had broader diets tended to exhibit higher occupancy in strips than species that specialized on primary forest habitat or a more narrow range of foods. Almost 70% of the species cited for my 9 area either were not detected or occurred in less than 10% of my sampling units. This low occurrence of mammals in strips and in continuous forest raises mayor concerns related to the long term survival of mammals in the Chaco.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Mauricio M Nunez Regueiro.
Thesis: Thesis (M.S.)--University of Florida, 2011.
Local: Adviser: Branch, Lyn C.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043890:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043890/00001

Material Information

Title: Are Forest Strips Cut out for the Job of Conserving Biodiversity? Evaluating the Case for Mammals with Hierarchical Bayesian Occupancy Models in the Chaco Forest
Physical Description: 1 online resource (57 p.)
Language: english
Creator: Nunez Regueiro, Mauricio M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: conservation -- mammals
Wildlife Ecology and Conservation -- Dissertations, Academic -- UF
Genre: Wildlife Ecology and Conservation thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Deforestation is a major cause of biodiversity loss, and the largest factor driving deforestation is expansion of agriculture. A key step toward successful conservation in agricultural areas is maximizing the biodiversity value of remaining forest. In subtropical and tropical regions, forest often is left in strips between agricultural fields under the assumption that biodiversity is sustained. In this thesis I examined how medium and large-sized mammals in the Argentine Chaco responded to forest strips retained along agricultural fields and how that response related to ecological traits of species. Presence/absence data from camera traps in continuous forest and forest strips showed that almost half of the total mammal assemblage that occurs in my study site was impacted by conversion of forest to strips. Five species found in forest were completely absent from strips. Hierarchical Bayesian occupancy models also demonstrated that occurrence of mammals decreased in forest strips with increasing distance from continuous forest. Species that occupied a variety of habitats and had broader diets tended to exhibit higher occupancy in strips than species that specialized on primary forest habitat or a more narrow range of foods. Almost 70% of the species cited for my 9 area either were not detected or occurred in less than 10% of my sampling units. This low occurrence of mammals in strips and in continuous forest raises mayor concerns related to the long term survival of mammals in the Chaco.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Mauricio M Nunez Regueiro.
Thesis: Thesis (M.S.)--University of Florida, 2011.
Local: Adviser: Branch, Lyn C.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043890:00001


This item has the following downloads:


Full Text

PAGE 1

1 ARE FOREST STRIPS CU T OUT FOR THE J OB OF CONSERVING BIO DIVERSITY? EVALUATING THE CASE FOR MAMMALS WITH HIERARC HICAL BAYESIAN OCCUPANCY MODELS IN THE CHACO FOREST By MAURICIO M. NEZ REGUEIRO A THESIS PRESENTED TO THE GRAD UATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Mauricio M Nez Regueiro

PAGE 3

3 To the mammals of the Chaco forest

PAGE 4

4 TABLE OF CONTENTS pag e LIST OF TABLES ................................ ................................ ................................ ........... 6 LIST OF FIGURES ................................ ................................ ................................ ........ 7 ABSTRACT ................................ ................................ ................................ .................... 8 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ... 10 2 METHODS ................................ ................................ ................................ ............ 14 Research Design ................................ ................................ ................................ ... 14 Study Site ................................ ................................ ................................ .............. 14 Site Location and Sampling Sites ................................ ................................ .......... 16 Sampling Protocol for Camera Traps ................................ ................................ ..... 16 Assessment of Environmental Covariates ................................ ............................. 17 Species Traits for Vulnerability Analyses ................................ ............................... 18 Hierar chical Bayesian Occupancy Models ................................ ............................. 19 Relationship Between Vulnerability and Species Traits ................................ ......... 21 3 RESULTS ................................ ................................ ................................ .............. 23 Vegetation Structure ................................ ................................ .............................. 23 Species Detection ................................ ................................ ................................ .. 23 Relationship Between Species Occurrence a nd Location Along Transects ........... 24 Relationship Between Site of Occurrence and Species Traits ............................... 24 4 DISCUSSION ................................ ................................ ................................ ........ 26 Patterns of Occurrence and Distance Along Transects ................................ .......... 26 Patterns of Occurrence and Species Traits ................................ ............................ 28 General Patterns of Occurrence ................................ ................................ ............ 30 5 CONSERVATION IMPLICATIONS ................................ ................................ ........ 33 6 CONCLUSIONS ................................ ................................ ................................ .... 34 APPENDIX A CORRELATION MATRIX ................................ ................................ ...................... 45 B HIERARCHICAL OCCUPANCY BAYESIAN MODELS ................................ .......... 46

PAGE 5

5 C PARAMETER DISTAN CE AND SPECIES TRAITS ................................ ............... 51 BIBLIOGRAPHY ................................ ................................ ................................ .......... 52 BIOGRAPHICAL SKETCH ................................ ................................ ........................... 57

PAGE 6

6 LIST OF TABLES Table page 3 1 Number of detections, life history and ecological traits of mammals in the Chaco forest ................................ ................................ ................................ ..... 35 3 2 Results from hierarchic al Bayesian models.. ................................ ..................... 37

PAGE 7

7 LIST OF FIGURES Figure page 2 1 Diagram of sampling design.. ................................ ................................ ............ 38 2 2 Map of study area in the province of Salta, Argentina. ................................ ....... 39 3 1 Understory density by distance categories along the transect. .......................... 40 3 2 Canop y cover by distance categories along the transect. ................................ .. 41 3 3 Results of hierarchical occupancy models for modeled species. ...................... 42 3 4 Results of classification tree. ................................ ................................ ............. 43 3 5 with, Dietary, Habitat breadth, and Litter Size ................................ ................................ .................... 44

PAGE 8

8 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the R equirements for the Degree of Master of Science ARE FOREST STRIPS CU T OUT FOR THE J OB OF CONSERVING BIO DIVERSITY? EVALUATING THE CASE FOR MAMMALS WITH HIERARC HICAL BAYESIAN OCCUPANCY MODELS IN THE CHACO FOREST By Mauricio M Nez Regueiro December 2011 Chair: Lyn C Branch Major: Wildlife Ecology and Conservation Deforestation is a major cause of biodiversity loss and the largest factor driving deforestation is expansion of agriculture. A key step toward successful conservation in agricultural areas is maximizing the biodiversity value of remaining forest. In subtropical and tropical regions, forest often is left in strips between agricultural fields under the assumption that biodiversity is sustained. In this thesis I examined how medium and large sized mammals in the Argentine Chaco responded to forest strips retained along agricultural fields and how that respo nse related to ecological traits of species. Presence/absence data from camera traps in continuous forest and forest strips showed that almost half of the total mammal assemblage that occurs in my study site was impacted by conversion of forest to strips. Five species found in forest were completely absent from strips. Hierarchical Bayesian occupancy models also demonstrated that occurrence of mammals decreased in forest strips with increasing distance from continuous forest. Species that occupied a variety of habitats and had broader diets tended to exhibit higher occupancy in strips than species that specialized on primary forest habitat or a more narrow range of foods. Almost 70% of the species cited for my

PAGE 9

9 area either were not detected or occurred in les s than 10% of my sampling units. This low occurrence of mammals in strips and in continuous forest raises mayor concerns related to the long term survival of mammals in the Chaco.

PAGE 10

10 CHAPTER 1 INTRODUCTION Understanding and predicting distribution of animals in human modified landscapes is fundamental to much of ecological science, and lies at the heart of conservation and management of many species. This issue is particularly important in areas comprised of a mosaic of forest and human land uses. Globally, f orest area has decreased by some 13 million ha each year in the last decade ( FAO 2010 ) Deforestation is a major cause of loss of biological diversity and the largest factor driving deforestation is expansion of agriculture to supply an increasing demand for food by a wealthier and larger global popula tion ( Sachs et al. 2009 ; Tilman et al. 2001 ) Clarifying how wildlife species respond to different configurations of forest in agricultural landscapes will be essential for development of effective conservation policy to sustain wildlife in agriculturally dominated landscapes. As a result of massive loss and fragmentation of habitat, remnant forest strips are a common feature of agricultural landscapes throughout the wo rld ( Boughey et al. 2 011 ; Hawes et al. 2008; Seaman and Schulze 2010). In agricultural la ndscapes these strips of forest ; Boughey et al. 2011 ; L aurance and Laurance 1999 ) often extend from continuous primary forest into the patches ( Diamond 1975 ) Historically, strips of forest have been used to separate agricultural blocks to help prevent soil and wind erosion and spread of fire ( Zanuncio et al. 1998 ) Some authors suggest that forest strips also can represent a valuable c onservation asset to some small vertebrate s and insects like dung beetles, birds and small mammals by providing s uitable habitat ( Barlow et al. 2010a ; Boughey et al. 2011 ;

PAGE 11

11 de Lima and Gascon 1999 ; Lees and Peres 2008 ; Seaman and Schulze 2010 ) However, u nderstanding of the conservation value of forest strips in product ion landscap es is still poor (Hawes et al. 2008) especially for wide ranging animals and in sub tropical ecosystems If current deforestation rates continue, forest strips will be among the most common features of agricultural landscapes in tropical and subtropical systems and significant amounts of remaining forest will occur in strips In South America, this conversion process is particularly acute in the dry subtropical Chaco, which is the second largest forest area after the Amazonia. Argentina conta ins 60% of the Chaco forest This region has a deforestation rate that is 12 times larger than the world mean and 5 times larger than the continenta l mean deforestation rate ( Seghezzo et al. 2011 ) In northern Argentina current environment al norms require that for e very 100 ha of deforested land ~2 3 37 ha of forest must be left in strips surrounding agricultural plots ( Adamoli et al. 2011 ) If the remaining Chaco forest is converted to agricultural land in Salta province (NW Argentina) alone 1.6 2.6 million ha will be in forest stri ps, which is equivalent to 8 13% of the total national protected area of Argentina. This situation is not an isolated case. Strips of forest are widely accepted by large landowners and mandated by law throughout many tropical and subtropical areas of South America ( Hawes et al. 2008, M Nuez Regueiro Pers Obs ; Seghezzo et al. 2011 ) Although a large amount of evidence points to the negative effects of habitat reduction and fragmentation on species diversity ( Fahrig 2003 ) not all species are equally sensitive ( Thornton et al. 2011b ) Vulnerability of species to conversion of forest into strips of forests remnants may be related to the way in which species exploit habitat

PAGE 12

12 and other resources. Loss of habitat area and partic ularly interior forest habitat, may reduce the availability of optimum conditions and resources for some species lowering the quality of their habitat ( Fahrig 200 3 ) Such change s may strongly affect forest specialists ( i.e., species with smaller dietary and habitat breadth; Cagnolo et al. 2009 ) Species with large body mass also may be impacted by fragmentation because of their large space requirements ( Carbone et al. 2005 ; Cro oks 2002 ) Also, e xposure of hunted species often increases in fragmented landscapes which could, in turn, exacerbate the effects of hunting ( Peres 2001 ) Hunting is a significant threat throughout the tropics and also for subtropical systems (Altrichter et al. 2006). Species with lower reproductive rates have been shown to be the most vulnerable to hunting ( Altrichter 2005 ) Incorporating species traits into analyses of species occurrence ultimately will help understand how species react to conversion of forest to forest strips. Here, I examine how medium and large sized mam mals (body weight > 1 kg hereafter referred to as mammals ) respond to forest strips retained along agricultural fields in the Argentine Chaco and how that response relates to ecological traits of species. Particularly, I expect ed the community compositio n of mammals to be different in strips of forest and in continuous forest and that the most vulnerable species (i.e., species that are more likely to be absent from forest strips ) would be those with large body mass, narrower dietary and habitat breadth a nd species that are severely hunted. This work will contribute to conservation by providing insights into the response of medium to large mammals to conversion of forest to strips. Previous work has primarily focused on smaller vertebrates, insects and pla nts ( Barlow et al. 2010b ; Boughey et al. 2011 ; de Lima an d Gascon 1999 ; Hawes et al. 2008 ; Laurance and Laurance 1999 ;

PAGE 13

13 Lees and Peres 2008 ; Wehling and Diekmann 2009 ) Understanding this response is important both for evaluating impacts of landscape change on mammals and because medium to large mammals play key ecological role s in forest ( Noss et al. 1996 ; Rozylowicz et al. 2011 ) Also, my work provide s the first critical evaluation of biodiversity implica tions of current environmental norms in the endangered Chaco eco region that result in forest being confined to thin strips around agricultural fields.

PAGE 14

14 CHAPTER 2 METHODS Research Design To evaluate the response of mammals to conversion of forest to strips and to assess how that response changed with increasing distance from forest interior, I designed a comparative study in which I compared occurrence of mammals in forest and in strips along a transect (Figure 2 1 ) This comparison was carried out by creat ing mathematical models that accounted for variation in detection probability (as a function of vegetation structure) and modeled occurrence (collected from presence / absence data from camera traps) as a function of distance along the transect These hiera rchical Bayesian occupancy models were constructed for all species that were present in more than 10 different camera traps. To understand which life history traits were most closely associated with vulnerability of species to conversion of forest to strip s, I analyzed the relationship between occurrence of mammals and species traits (i.e., trophic level, body mass, dietary and habitat breadth, litter size, age to first reproduction and hunting pressure). This relationship was analyzed using two method s First, I conducted a classification tree analysis to group the entire assemblage of species according to their site of occurrence (i.e., strips + forest, forest only, or not detected) and their ecological traits. Second, I graphically represented the relat ionship between vulnerability of species ( obtained from occupancy models) and their ecological traits Study Site This study was conducted in Chaco forest of Salta Province, NW Argentina, a ighest deforestation

PAGE 15

15 rates (Figure 2 2) In this region annual deforestation rates have more than tripled in the last 10 years with expansion of the agricultural frontier as a result of increased international demand for soybeans ( Gasparri and Grau 2009 ) The new National Forestry Law of Argentina, which requires classification of forested areas for different la nd uses, potentially could result in deforestation of an additional 1.6 million ha in Salta in the next 6 9 years and leaves open the option of subsequent deforestation of additional 5.4 million ha. If deforestation continues, the entire Chaco forest of Sa lta will be reduced to network of remnant forest strips embedded in an agricultural landscape. The extent of my study area is ~9,000 km 2 and the geographical center is located at 24 0 0 dominated by soy and pasture in which strips of forest occur between all agricultural plots throughout the entire landscape. Large blocks of continuous forest surround this central agricultural area ( Figure 2 2 ). The native vegetation is representative of dry Chaco forests, dominated by quebrachos ( Schinopsisi lorentzeii and Aspidosperma quebrachoblanco ) and accompanied by Bulnesia sarmientoi Prosopis alba, Prosopis nigra, Ziziphus mistol Anadenanthera macrocarpa, Phyllostyllum rhamnoides, and Callicophy llum multiflorum ( Grau et al. 2008 ) Published accounts of terrestrial mammals for thi s area document 29 species >1 kg ( Canevari and Vaccaro 2007 ; Mares et al. 1989 ; Wallac e et al. 2010 ) I did not include in this list species associated with bodies of water (e.g., capybara, Hydrochoerus hydrochaeris ) or primarily arboreal species (e.g., m argay Leopardus wiedii ) because I did not adequately sample these species.

PAGE 16

16 Site Loc ation and Sampling Sites From May 2010 to January 2011 I sampled occurrence of mammals in 12 transects placed in strips of primary forest between agricultural fields and extending into adjacent continuous primary forest ( Figure 2 2) Using Google Earth I identified all s trips of forest in my study area that ranged from 50 m to 100 m in wid th and were at least 16 km long in the study area Most of these strips were created between 1995 and 2004. For my study, I only considered strip s at least 8 km from the adjacent strip to assure biological and statistical independen ce. This distance is larger than the home range diameter of all but the largest carnivores (puma, jaguar; Canevari and Vaccaro 2007 ; Wallace et al. 2010 ) Therefore, the probability of sampling the sa me animal in different strip was low. I only sample d strips not severely degraded by fire ( i.e., < 15% of the area of the forest strips affected by fire ). To avoid biased results from nearby roads, transects were no closer than 3 km to any mayor road or hi ghway. Sixteen forest strips met the requirements described above From that pool, 1 2 strips were selected randomly for sampling. Samples were collected along tansects that incorporate d 8 km of the forest strip and 8 km of the ad jacent contiguous forest ( F igure 2 2 ). Each 16 km transect was treated as a block in occupancy models and each camera within a transect was considered a sampling unit Sampling Protocol for Camera Traps Five camera traps (Bushnell TrophyCam, Bushnell Corporation, Overland Park, KS, USA) were placed in each strip of forest and 5 cameras in the continuous forest (i. e., 10 cameras in each transect; Figure 2 2). In cont inuous forest, the first camera was placed at the limit with the forest strip and then cameras were placed at interval s of 1.6 km. In strips of forest, a camera was placed every 1.6 km beginning 1.6 km from the

PAGE 17

17 edge of the forest. Cameras were placed in the nearest appropriate location to the designated point s These locations included small and large game trails, water h oles, den sites, and other areas containing substantial signs of animal use such as tracks, digging, or scraping I placed the came ra sensor approximately 10 20 cm off the ground so that smaller species could not avoid detection by walking under the sens or. Camera traps remained at each sampling point along the transect for 16 days. I broke up the 16 day period into 4 day sessions to create a series of repeat detection/non detection data (i.e., a detection history) for use in modeling detection probabilit ies for each species ( MacKenzie et al. 2002a ) Assessment of Environmental Covariates I included covariates in hierarchical occupancy models that represented distance categories for the 1.6 km spacing of cameras along each transect (i.e., categories 1 5 for continuous forest and 6 10 for strip s of for est; Figure 2 2) and also measured forest structure. Covariates for forest structure (understory density and canopy cover) were included in probability of detection models. Detection of an animal likely varies as a function of density of understory vegetat ion. Also camera sensitivity can be affected by solar radiation which varies with canopy cover (see Bushnell TrphiCam, 2009 manual for more details). To evaluate forest structure in strips of forest and continuous forest, I took standardized digital photog raphs centered at each camera trap site ( Halchak et al. 2011 ; Lees and Peres 2008 ) For understory density es timations, I took 4 photographs in each cardinal direction parallel to the ground at 15 cm from the ground. For canopy cover analyses I took a single photograph perpendicular to the ground. I mages then were analyzed using ImageJ free ware (Java based imag e processing program developed at the National Institute of Health), which counted the number of forest (dark)

PAGE 18

18 pixels across the image and calculated a mean and SD for each image ( Halchak et al. 2011 ) Data from the 4 photographs of the understory were averaged for each site. To analyze differences in canopy cover and understory density of sites located in continuous forest v ersu s. strips, I first calculated a mea n value for each vegetation covariate from samples within the forest or strip portion of the transect. Then, using these transect level mean values, I conducted a paired t test to compare vegetation structure in forest v ersu s. strips. Species Traits for Vu lnerability Analyses For the 29 mammal species cited for my study area I determined trophic level, body mass, dietary and habitat breadth, litter size, and age to first reproduction from field guides and recently published studies ( Canevari and Vaccaro 2007 ; Mares et al. 1989 ; Wallace et al. 2010 ) Trophic level was categorized as follows : 1 = primarily browser/grazer or frugivore, 2 = omnivore, 3 = primarily carnivore/myrmecophage. I calculated body mass as the average of male and female body mass. When only a range was given, I took the midpoint of the range. I estimated dietary and habi tat breadth as the number of habitat or dietary categories a species uses based on published information ( Canevari and Vaccaro 2007 ; Mares et al. 1989 ; Wallace et al. 2010 ) Dietary breadth was calculated as the number of categories of different prey types eaten by each species. Categories included in the analysis were: grass, browse, cru stacean/fish, insects/arthropods, hard mast, soft mast, small mammals, large mammals, reptiles/amphibians, birds, carrion, and domestic crops. Habitat breadth was calculated by counting the number of habitat categories used by a species, including only tho se habitats commonly encountered in Chaco: primary dry forest, secondary/regenerating forest, savanna/pasture or cropland, and near urban

PAGE 19

19 environments. Litter size was the average number of young produced per year. Age to first reproduction was recorded as the average number of months from birth to first reproduction. Other reproductive parameters such as number of litters per year were not available for many of the species in my study area. Prior to analysis correlation s among species traits were examin ed with to determine if these traits were supplying redundant information. Most species traits examined in t his study were not correlated ( Appendix A ). Exceptions were as follows: a) Species with small body size also t end ed to h ave an early first reproduction and exploit many habitat types (R 2 = 0.61 and 0.46, respectively ; Appendix A ); b) In general, species that had many young per litter also consumed many dietary categories (R 2 = 0.54). Vulnerability to hunting was assessed by interviewing 27 informants distributed evenly across m y study area. Informants were local, long term (> 6 years) resident hunters and agricultural workers who are regular visitors to the sample area and are thoroughly familiar with the medium to large bodi ed vertebrate fauna. To assure that interviewees knew the species, I asked them to identify species present in the sample area from a selection of color plates of mammals and birds, including 10 species known to be absent from the study region. Informants recognized that the 10 species did not occur in the region >90% of the time. Informants grouped species into the following categories: 1 = rarely/never hunted or ki lled; 2 = occasionally hunted or killed, but not a preferred game species or actively perse cuted species; 3 = often hunted or killed. Hierarchical Bayesian Occupancy Models Species occupancy was analyzed with hierarchical occupancy mixed models using Bayesian inference and the history of detections collected from camera trap sampling.

PAGE 20

20 These mode ls can incorporate detection probabilities to overcome sampling biases related to differences in species detection and can include sampling designs that incorporate random block effects. If imperfect detection is not taken into account, some species may ap pear to be vulnerable than other species merely because these species are harder to detect ( Dorazio and Royle 2005 ; MacKenzie 2006 ; Thornton et al. 2011a ) Underestimation of the r elation ship between species occurrence and habitat covariates in logistic regression type of models has been documented when i mperfect detection was not taken into account ( Ke ry 2010 ; Royle and Dorazio 2009 ) This modeling approach also was appropriate for my study because of the nested study design (i.e., cameras were located within transects). I nterpretation of modeled o ccupancy differs in my study from traditional patch occupancy studies ( MacKenzie et al. 2002b ) because my study likely does not comply with the closure assumption for occupancy models (i.e., many of the mammal species in my study were capable or moving in and out of sites during sampling; MacKenzie 2006 ; MacKenzie et al. 2002b ) However, for ease of presentation, I will continue to use traditional occupancy terminology in this thesis. I modeled occupancy for species that had a na ve estimate of occupancy of >10% (n= 9 ). The se occupancy models had two components: probability of detection ( p ) and probability of occupa ncy ( i ) I modeled p as a linear function of two environmental covariates canopy cover and understory density. The full model included both environmental covariates. The final model for each species only included environmental covariates that had a stron g association with probability of detection (i.e., I discarded

PAGE 21

21 covariates in which credible intervals overlapped zero.). I modeled i as a function of a fixed effect (distance along the transect where the species was detected) and a random effect (transect strength of association between distance categories along a transect and occupancy. Negative values indicate that a species was more likely to occupy a given site as distance from the strip into the forest interior increase d and less likely to occupy the strip as distance from continuous forest into the strip increase d (i.e., as occupancy increase d from d10 to d1 ( Figure 2 1 more negative) Positive values in dicate the opposite pattern (i.e., occupancy decreased from d10 to d1) Transect was included as a random effect because sampling units within a transect may not be biologically independent. I did not include canopy cover and understory density because the model behaved better (i.e., achieved convergence at fewer number of iterations) with few parameters I chose to include distance along the transect instead of canopy cover or understory density because I was interested in understanding the overall pattern of species response to strips per se. Both components of the hierarchical model were analyzed simultaneously using program Winbugs ( Gilks et al. 1994 ) which was called remotely from program R with the package R2Winbugs ( Sturtz et al. 2005 ) For details on modeling approaches for occupancy and detection probability and for code details, see Appendix B Relation ship Between Vulnerability and Species Traits To better understand links between spec ies traits of mammals and their response to conversion of forest to strips, I used two approaches. First I divided all species of mammals that were detected in the study area or should occur in the study area into three categories. Category 1 comprised al l species not detected in my study that should

PAGE 22

22 occur in the area based on the known distribution of these species ( Canevari and Vaccaro 2007 ; Mares et al. 1989 ; Wallace et al. 2010 ) Local forest residents also confirmed that these species have occurred in the area in the near past (i.e., less than 5 years). The second category correspond ed to species that were detected only in continuous forest. Last category included species that were either detected in strips of forest and in continuous forest or only in strips of forest (one species). I used classification tree analysis to group specie s in to homogeneous sets using species ecological traits as independent variables ( Andersen et al. 2000 ) I used CRT growing method for classification tree with a Gini impurity measure that maximized the homogeneity of child node with respect to the value of the target variable. Second, for species that were modeled for occupancy (n=9) I graphically represented the relationship not analyze this relation ship statistically (e.g., with a linear regression) because of 1 ) the asymmetric nature of the posterior distribution around the mean v alues of parameter ) some variables wer e categorical (e.g., habitat/dietary breadth and hunting ) Methods like linear regression assume equivalent distances between categories, which is not likely to occur in case of categorical variables (i.e., species with habitat breadth value of 2 are not t wice as broad, in terms of diet, than a species with a value of 1 ).

PAGE 23

23 CHAPTER 3 RESULTS Vegetation Structure Forest structure in continuous forest differed from fores t strips only for understory density (u nderstory density, F= 4 07 df= 1 0, p = 0.0 7; c anopy co ver, F = 2 4 df=10, p=0. 15 ). On average understory in continuous forest was denser than in forest strips (mean + SD, continuous forest, 236 + 3.1; strip, 227 + 8.9; units are expressed in number of forest pixels in the image. S ee methods for more details ). Also, understory density was more heterogeneous with increasing distance away from the forest edge into the strip of forest ( Figure 3 1 ). Canopy cover was highly variable among sites in both continuous forest and strips, and exhibited no obvious trends c omparable to those of understory density (mean + SD, continuous forest, 174 + 18.6; strip, 201 + 19.4; Figure 3 2 ). Species Detection I detected 23 species of mammals in my study site (Table 3 1). Seventeen species were detected in continuous forest and s trips. One species was found exclusively in forest strips ( n ine banded a rmadillo ). Five species (ocelot, tapir, r ed b rocket d eer tayra, chacoan peccary) were found only in continuous forest Six species that should occur in my study area (based on publish ed literature and interviews) were not detected (plains vizcacha, jaguar, giant armadillo, jaguarondi, tapeti, white lipped peccary). Most species were detected in less than 10 sampling units (Table 3 1). Accounting for detectability in the analysis resu lted in considerable increase in occupancy estimates, especially for species that were difficult to detect (Table 3 2). On average, estimates of occupancy corrected for detectability increased by 27% from the

PAGE 24

24 nave estimate of occupancy. For species that w ere easy to detect when present (i.e., many detection records at a given site ), such as the chacoan cavy, modeled occupancy estimates were similar to nave estimates. Modeled occupancy estimates for the species recorded in at least 10 samples units ranged from 0.14 (chacoan cavy) to 0.56 (giant anteater). Relationship Between Species Occurrence and Location Along Transects Species also var ied in their response to d istance from forest edge (T able 3 2 ). istance estimates for 5 species were negati ve, indicating that the species was more likely to occupy a given site as distance of the sample location fr om the forest interior decreased (i.e., from d1 0 to d1; Table 3 2 Figures 3 3 A and 3 3 B ) In some cases species occurrence decreased gradually alo ng the transect from forest into str i ps (e.g., skunk, gray brocket deer; Figure 3 3 B ), and in other cases a more abrupt decrease occurred in occurrence near the forest boundary (e.g., giant anteater Figure 3 3 A ). C redible intervals for the distan ce parameter were wide and overlapped zero, exce pt in the case of the giant ant eater and the collared peccary. These two species were most clearly associated with forest interior. Four species exhibited positive parameter estimates for distance, indicating that occurrence increased with distance from the forest interior into the strip ( i.e., from d1 to d10, Table 3 2, Fig ures 3 3 C and 3 3 D ). However c onfidence limits for the distance parameter also overlapped zero for these species, except in the case of th e pampas fox. Occurrence of this species strongly increased wi th distance from fores t ( Figure 3 3 C ) Relationship Between Site of Occurrence and Species Traits Habitat breadth was the most important variable for classifying all species recorded for my stu dy area in the classification tree by the environments in which these

PAGE 25

25 species were detected (i.e., not detected in study area but cited in the literature ) detected in forest only, o r detected in strips and forest; Figure 3 4 ). About half of the species th at were detected only in forest or were not detected in my study area even though they were cited for the area (n=6) were primary forest specialists and about half (n=5) were generalist species (i.e., they used other habitats as well as primary forest). In contrast, 94% of the species that occurred in strips as well as forest were categorized as generalist species. Other species traits (i.e., trophic level, body mass, dietary breadth, litter size, first reproduction, and hunting) were not used by the classi fication tree to group species. Although I could not statistically test for relationship s between species traits and occurrence of these mammals (n=9 modeled species) along transects some trends were apparent f rom graphical representation Species with narrow dietary and habitat breadth tended to be more likely to o ccupy forest interior sites (Figures 3 5 A and 3 5 B). Also species that had small litter sizes were more likely to occupy fore s t interior sites (Fig u re 3 5 C ) but this trait was correlat ed with diet breadth ( Appendix A ) Other ecological traits (i.e., age to first reproduction, body size, hunting pressure, and trophic level) did not show a linear relation ship with the distance parameter ( Appendix C ).

PAGE 26

26 CHAPTER 4 DISCUSSION Patterns of O ccurrence and Distance Along Transects My study indicate that at least 44% of the medium to large mammals currently present in my study area responded negatively to conversion of forest to strip s F ive species found in f orest were completely absent from strips, and five other species tended to decrease in occupancy along the strips as distance from the forest increased and exhibited highest occupancy in interior forest In contrast, three species appeared to benefit from strips. and the crab eating fox increased in strips as distance from the forest increased though this patter n was signifi cant only for the pampa s fox. In other parts of their geographical rang e, a ll three of these species occupy a variety of open habitats (e.g., savannas, grasslands, and scrublands ) as well as forest ( Wallace et al. 2010 ) The nine banded armadillo was the only species that I detected in strips, but not in forest. This species is widely distributed throughout tropical forest, as well as savannas ( Wallace et al. 2010 ) but only was detecte d in three cameras in my study and thus may not have been sampled adequately. Result s of my study both support and contrast with s tudies of species use of forest strips in other regions In studies of birds and medium to large mammals in riparian forest in tropical regions species richness has been found to be either similar or lower in forest strips compared to more continuous forest ( Hawes et al. 2008 ; Lees and Peres 2008 ; Seaman and Schulze 2010 ) and in upland (terra firme) forest, fewer bird species occupied strips than continuous forest ( Hawes et al. 2008 ) Forest strips in both habitats

PAGE 27

27 su pported fewer forest specialist birds than habitat generalists ( Hawes et al. 2008 ; Seaman and Schulze 2010 ) In contrast to my study, a study of use of forest strips by medium to large mammals in the Amazon reports that these mammals use strips more than continuous forest ( Barlow et al. 2010a ) Some of the mammals in the Amazon study were the same species found in my study area (e.g., collared pec cary, grey brocket deer, giant armadillo, and others). Differences between this study and mine in patterns of occurrence may relate to the landscape matrix, which was comprised of Eucalyptus plantations in the Amazonian study and pasture or soybean in the Chaco. Medium to large mammals apparently forage in the understory of the Eucalyptus matrix ( Barlow et al. 2010a ) S oybean s and pasture may provide less suitable habitat and may be less permeable to movement of forest species in my area. Barlow et al (2010) do not provide information on patterns of occurrence of individual species or forest specialists versus generalists so I cannot compare the results of our stud ies in detail. Declines in richness and abundance of species in forest strips has been documented for plants, dung beetles, birds and arboreal mammals as distance from continuous forest increases ( Barlow et al. 2010a ; Hawes et al. 2008 ; Laurance and Laurance 1999 ; Lees and Peres 2008 ; Wehling and Diekmann 2009 ) In demonstrating a loss of forest interior species with increasing distance from continuous forest, my results further strengthen s these patterns an d add s support to the notion that this may be wide spread across taxa. Structural changes in the vegetation may explain at least part of the patterns I documented. In my study understory became more open with distance from forest along the strips, and this may have provided more appropriate

PAGE 28

28 conditions for species that use non forested habitats and poorer habitat for forest specialists Patterns of Occurrence and Species Traits Species response to conversion of forest to strips should relate to how species use space and exploit resources as well as their vulnerability to direct human impacts such as hunting ( Cagnolo et al. 2009 ; Fahrig 2003 ; Peres 2001 ) Studies of forest fragmentation have demonstrated that forest specialist species often are most affected by fragmentation ( Cagnolo et al. 2009 ) Species that use large areas also often disappear from fragments because their area requirements are not met ( Shahabuddin and Ponte 2005 ) Moreover, conversion of forest to fragments or strips can increase hunting pressure by increasing access ( Altrichter 2006 ; Peres 2001 ) In my study area species that occupied a variety of habitats and had broader diets tended to exhibit higher occupancy in strips than species that specialized on primary forest habitat or a more narrow range of foods. Similarly, other studies of mammals have found that species with smaller dietary and habitat breadth are most affected by fragmentation ( Cagnolo et al. 2009 ) Also responses of arboreal mammals to linear corridors are determined by traits such as diet, denning requirements, and degree of arboreality ( Laurance and Laurance 1999 ) Assuming the quality of habitat and dietary elements remains constant from forest to strips, as area in strips decreases (compared to forest), the quantity of habitat and dietary elements required by forest species could decrease. The patterns that I detected in vegetation structure also may indicate that abundance or quality of resources decrease with distance from the forest into the strips as a function of changes in habitat quality rat her than solely area.

PAGE 29

29 However, habitat and dietary breadth clearly do not explain all the patterns of mammal occurrence observed in my study. Almost half the species that were absent from the study area or were found only in forest (n=5) were not habitat specialists and these species varied widely in diet breadth. I did not analyze the relationship between occurrence and area requirements of species because home range data were missing for some mammals, instead, I examined body size, which generally correl at es with area requirements In my study area, body size was not a strong predictor of use of forest strips by mammals or of species that were absent from the study area. For example, small armadillos such as the t h ree banded armadillos were more common in forest than in strips, and the larger nine banded armadillo was only recorded in strips. Both large (e.g., jaguar and white lipped peccary) and small species (e.g., plains vizcacha and tapeti rabbit) were absent from the study area. All of these species, as well as many others such as the collared and Chacoan peccaries and tapirs are heavily hunted in Chaco ( Altrichter 2005 ; 2006, Nez Regueiro Pers. Obs. ; Altrichter and Almeida 2002 ) For this reason I expected to find a strong relation ship between occurrence of species and hunting pressure and th e lack of effect of hunting in species occurrence was a surprise One possible explanation for lack of evidence of impacts of hunting on mammal occurrence is that my estimate of hunting pressure (how frequently a species was killed) was imprecise (e.g. 16 of the 29 species are listed as heavily hunted, though hunting pressure certainly varies among these species). Also this measure confounds animal abundance and preference of hunters For example, species may be killed infrequently because their population s already have declined from hunting, not because hunters would not kill the species if they had the opportunity. Mammal species

PAGE 30

30 that have a faster reproductive rate often may be better able to cope with decrease in population size better than species with slower reproductive rates In areas with high hunting pressure, like my study site, reproductive traits might give an idea of how resilient a species is to hunting. The fact that my data showed that forest interior species also tended to have a lower repr oductive rate (as seen by the low number of young in a litter) could support the idea that species less resilient to hunting are more vulnerable to conversion of forest into strips. However, the number of young also was correlated with diet breadth, so the se factors are confounded. Separating the contributions of naturally low densities and hunting pressure on observed occurrence of mammals is difficult, but more attention needs to be paid to evaluating effects of hunting in Chaco General P atter n s of Occu rrence Almost 70% of species cited for my area were either not detected (n =6) or occurred in less than 10 sampling units (n=14) after 1920 camera/nights. All species that IUCN (2011) categorizes as threatened in my study area fall within this group (e.g., Chacoan peccary, endangered; giant armadillo, vulnerable; and tapir, vulnerable), with the exception of the giant ant eater (vulnerable), which occupied >10% of the sites but was significantly more likely to occupy continuous forest sites. This low rate of occurrence may occur because medium to large mammals naturally occur at very low densities in my study area and thus were not detected (i.e., sampling effect), or because this fauna has suffered from human impacts even in continuous forest. For some large bodied species that naturally occur at low densities throughout their range (e.g., jaguar and giant armadillo), lack of detection of these species may be related to sampling effect. However the puma, which is a large bodied wide ranging species that also

PAGE 31

31 g enerally occurs at low densities throughout its range (most of north and South America), was detected in 6 sampling units. For many mammals that the Chaco shares with tropical forest in Central and South America, Chaco forest (particularly in Argentina and Paraguay) corresponds to the southernmost part of their distributions (e.g., jaguar, ocelot, peccaries, tapir; IUCN 2011). At the limits of species distributions, optimal environment conditions and resources may occur with less frequency than in the cente r of their distribution ( Guisan and Thuiller 2005 ; Stearns 1976 ) resulting in lower population densities and thus limited detection in my sampling. However studies in protected area of the Bolivian Chaco demonstrate large mammals densities equal to or larger than many tropical forests to the north ( Maffei et al. 2004 ) Thus if low densities of mammals are related to the geographic boundaries of distributions, this phenomena is limited to the southernmost Chaco and not the entire Chaco region. Although t he low rate of occurrence th at I documented for many mammal species may be partially explained by samplin g effect, this low occurrence also likely relates to human impacts on the Argentine Chaco. Throughout the Argentine Chaco, forest has been highly modified and degraded as a consequence of grazing and logging ( Torrella et al. 2011 ) The few studies of medium to large mammals that have been conducted in this region primarily focus on harvest of peccarie s and conclude that hunting of C hacoan peccaries and white li pped peccaries is not sustainable ( Altrichter 2005 ; Altrichter and Almeida 2002 ) Likewise serious decl ines have occurred in jaguar populations ( Rabinowitz and Zeller 2010 ) These studies i n conjunction with my data point to the need to understand the complex factors that impact the Chaco fauna. For example, the effects of hunting likely are vastly underestimated in my study. I could not analyze the relationship

PAGE 32

32 between species occurrence a nd hunting, or other ecological traits, for a large set of the species that should occur in my area because the species were absent or the number of records in camera traps was too low. Assessment of the fauna of the Chaco is important because of the rich biodiversity and the enormous threat imposed by rapid loss of forest. The diversity and level of endemism of the medium and large terre strial mammals of the Chaco is at least comparable to other forests recognized as highly important for conservation such as the Amazon ( Mares 1992 ; Ojeda et al. 2003 ; Redford et al. 1990 ) For example, particular attention needs to be paid to Chacoan peccary, known only from fossil records (and was found still to exist in 1974), which is an endemic and endangered species detected in my study site only in continuous forest. Othe r species such as the Chacoan n aked tailed armadillo ( Cabassous chacoensis ), that occurs east of my sampling area are also endemic to this region ( Redford et al. 1990 ) Furthermore, deforestation rates have increased in recent years and are expected to continue (Gasparri 2009, Seghezzo 2 011, Torrella et al. 2011), threatening the long term existence of the Chaco and its biodiversity. The magnitude of the problem wh ich has only continued to grow, was captured more than 20 years ago by Redford et al. (1990), who referred to the Argentin e C

PAGE 33

33 CHAPTER 5 CONSERVATION IMPLICA TIONS According to my findings, i f forest continues to be converted to strips in agricultural landscapes forest specialist will disappear from the Chaco forest. Furthermore, the decrease in occurrence of numerous mammals with increasing distance from the forest interior highlights the importance of keeping primary forest strips short to help maintain biodiversity and functional co nnectivity for forest species in agricultural landscapes. Moreover, the observed response of mammals (i.e.,>44% of species impacted by conversion of forest to strips) could be an underestimate of the long term effect of forest transformation. Strips in my study area were relatively recent (i.e., < 10 year s old) and extinction debts may occur ( Tilman et al. 1994 ) Strips of forest will beco me much more common if deforestation continues as expected and therefore the total impact of strips on biodiversity likely will increase with time because of the predominance of this landscape configuration I do not advocate however that strips of forest should be eliminated. F orest strips retain (at least in the short term) some mammals and are preferable to complete deforestation. However, a lternative configurations for retaining fore st remnants need to be examined urgently in the mammalian fauna of the Chaco is to retained (e.g., conserving same area but in large interconnected blocks )

PAGE 34

34 CHAPTER 6 CONCLUSIONS Conversion of continuous forest to forest strip clearly has significant negative impacts on medium to large mammals. Occurrence of mammals in my study area decreased in forest strips with increasing distance from continuous forest, as reported for a variety of other taxa. In my study area species that occupied a multiple habitats and had broader diets tended to exhibit higher occupancy in strips th an species that specialized on primary forest habitat or a more narrow range of foods. However, these two factors do not explain all the patterns that I observed. Almost 70% of the species cited for my area either were not detected or occurred in less tha n 10% of my sampling units, and thus I could not evaluate the relationship between occurrence of these species and species traits. The low occurrence of these species could be related, in part, to naturally low abundance of these species because Argentine Chaco represents the southern most range limit of a variety of these species. However, other factors such as habitat degradation and hunting also likely influence occurrence of mammals. Finally low occurrence of mammals in strips and in continuous forest raises major concerns related to the long term survival of mammals in the Chaco, and points to the urgent need for conservation measures to preserve this rich fauna.

PAGE 35

35 Table 3 1. Number of detections, life history and ecological traits of mammals in the C haco forest that either are present at my site or should occur based on published literature and interviews. Key to abbreviations: Site indicates the location where a species was detected (F, Continuous forest; S, Strip of forest; F+S, Continuous forest an d strips of forest; A, Absent fro m study area). Cf is the total number of cameras where detection occurred in continuous forest; Cs is the total number of cameras where detection oc curred in strips of forest; Tf is the total number of transects where a spe cies was detected in forest; T s is the total number of transects where a species was detected in strips ; Trophic is a species trophic level; Body is adult body mass (kg); Diet diet breath (larger values indicate more diet categories); Hab habitat breadt h; Litter litter size (number of young produced per litter); FR e p, age at first reproduction; Hunt, hunting pressure (larger values show larger hunting pressure). See methods for full description of species traits. Species Common name Site Cf Cs Tf Ts T rophic Body Diet Hab Litter FRep Hunt Mazama gouazoubira Gray Brocket Deer F+S 20 17 9 7 1 20.4 2 3 1.0 13.5 3 Lycalopex gymnocercus Pampa s Fox F+S 11 16 5 9 2 4.6 6 4 4.0 12.0 2 Leopardus geoffroyi Geoffroy's Cat F+S 7 11 5 7 3 3.4 3 3 2.0 21.0 2 Cone patus chinga Molina's Hog nosed Skunk F+S 10 6 4 3 2 2.3 6 3 3.5 10.5 1 Myrmecophaga tridactyla Giant Anteater F+S 12 2 7 2 3 30.9 1 2 1.0 39.0 2 Dolichotis salinicola Chacoan Cavy F+S 8 5 1 1 1 2.1 3 3 1.5 2.8 2 Cerdocyon thous Crab eating Fox F+S 5 7 5 6 2 5.2 6 3 4.0 11.0 2 Pecari tajacu Collared Peccary F+S 9 3 6 2 1 24.9 3 1 2.0 18.0 3 Tolypeutes matacus Southern Three banded Armadillo F+S 8 3 4 2 2 1.2 4 2 1.0 9.0 3 Chaetophractus villosus Large Hairy Armadillo F+S 4 2 3 2 2 2.3 3 3 1.6 9.0 3 Puma concolor Puma F+S 5 1 3 1 3 54.0 4 2 2.8 27.0 3

PAGE 36

36 Table 3 1 Continued Common name Site Cf Cs Tf Ts Trophic Body Diet Hab Litter FRep Hunt Dasypus novemcinctus Nine ba nded Armadillo S 0 3 0 2 2 4.0 4 2 4.0 2.0 3 Didelphis albiventris White eared Opossum F+S 1 2 1 1 2 1.0 6 4 6.9 10.0 2 Euphractus sexcinctus Six Banded Armadillo F+S 2 1 2 1 2 4.4 6 2 1.5 9.0 3 Galictis cuja Lesser Grison F+S 1 2 1 2 3 1.9 2 2 3.0 20. 0 1 Catagonus wagneri Chacoan Peccary F 3 0 3 0 1 35.6 2 1 2.7 30.0 3 Chaetophractus vellerosus Screaming Hairy Armadillo F+S 1 1 1 1 2 1.0 6 2 2.0 9.0 3 Dasyprocta punctata Central American Agouti F+S 1 1 1 1 1 3.0 3 2 1.3 16.2 1 Nasua nasua South Ame rican Coati F+S 1 1 1 1 2 4.6 6 2 3.7 24.0 1 Tapirus terrestris Tapir F 2 0 2 0 1 207.7 3 1 1.0 36.0 3 Eira barbara Tayra F 1 0 1 0 2 5.1 5 3 2.0 20.0 1 Leopardus pardalis Ocelot F 1 0 1 0 3 11.9 5 1 1.7 23.0 1 Mazama americana Red Brocket Deer F 1 0 1 0 1 33.1 3 2 1.2 15.0 3 Lagostomus maximus Plains Viz cacha A 0 0 1 4.4 2 3 1.9 12.4 3 Panthera onca Jaguar A 0 0 3 83.9 3 2 2.0 30.0 3 Priodontes maximus Giant Armadillo A 0 0 2 27.7 2 1 1.5 10.5 3 Puma yagouaroundi Jaguarundi A 0 0 3 64.0 3 1 2.5 18.0 2 Sylvilagus brasiliensis Tapeti A 0 0 1 1.0 4 3 1.2 13.0 3 Tayassu pecari White lipped Peccary A 0 0 1 27.7 3 1 1.0 18.0 3

PAGE 37

37 Table 3 2. Results from hierarchical Bayesian models. Paramete istanc presented with 95% Bayesian credible intervals. Negative values indicate that a species was more likely to occupy a given site as distance from the strip into the forest interior increased and less likely to occupy the strip as distance fr om continuous forest into the strip increased (i.e., as occupancy increased from d10 to d1 (Fig ure 2 1 te the opposite pattern (i.e., O ccupancy decreased from d10 to d1). Na ve Occupancy Estimates of Occupancy 95% Credible intervals 95% Credible intervals Giant Ant eater 0.15 0.56 0.39 0.76 5.73 14.58 0.52 Collared Peccary 0.13 0.38 0.18 0.60 2.91 7.50 0.15 Southern Three banded Armadillo 0.12 0.48 0.28 0. 70 2.14 9.60 3.07 Gray Brocket Deer 0.40 0.60 0.48 0.73 0.58 2.27 0.44 Molina's Hog nosed Skunk 0.17 0.41 0.27 0.61 0.42 4.61 3.19 Chacoan Cavy 0.14 0.14 0.14 0.15 1.02 0.89 3.54 Pampa s Fox 0.29 0.51 0.36 0.70 1.86 0.14 7.52 Geoffroy's Cat 0.20 0.54 0.27 0.76 2.35 0.14 3.50 Crab eating Fox 0.13 0.57 0.37 0.77 5.47 0.14 13.79

PAGE 38

38 F i g 2 Fig ure 2 1 Diagram of sampling design. X marks the specific locations of camera traps. Diagram also shows dist ance categories along a transect. Distance categories 1 5 correspond to continuous forest and distance categories 6 10 to strips of forest. Strips are 50 100 m wide. Diagram not to scale.

PAGE 39

39 Figure 2 2. Map of study area in the province of Salta Argentin a

PAGE 40

40 Fig ure 3 1 Understory density (number of dark pixels) by d istance categories(at 1.6 km increments) along the transect. Higher values of understory density represent more dense vegetation.

PAGE 41

41 Figure 3 2 Canopy cover (n umber of dark pixels) by dist ance categories along the transect. Higher values of canopy cover represent more dense vegetation.

PAGE 42

42 Fig ure 3 3 Results of hierarchical occupancy models for modeled species. Distance categories are presented in standard deviation units. Negative val ues of the distance parameter indicate that a species was more likely to occupy a given site as distance from the strip into the forest interior increased and less likely to occupy the strip as distance from continuous forest into the strip increased (i.e. A s occupancy increased from d10 to d1 (Fig ure 2 1 parameter became more negative). Positive values indicate the opposite pattern (i.e., occupancy decreased from d10 to d1). C D A B

PAGE 43

43 Fig ure 3 4 Results of c lassification tree The s plitting va riable in the classif ication tree is h abitat b readth Species with habitat breadth < = 1 correspond to forest specialists. Similarly, species with habitat breadth >1 correspond to generalist species. S pecies that were detected in forest strips and in contin u ous for est are represented by F+S (n= 18). Category A (n=6) corresponds to species that were not detected in the study area but have been reported in published literature for the area, and category F represent s species that were only found in continuous for est (n=5).

PAGE 44

44 Fig ure 3 5 Relation ship between pa (see Table 2 for explanation of distance parameter) with, Dietary (A), Habitat (B) breadth, and Litte r Size (C) Key to abbreviation: Pt, Peca ry tajacu ; Mt, Myrmecophaga tridact y la ; Tm, Tolypeutes matacus; Ds, Dolichotis salinicola ; Mg, Mazama gouazoubira ; Cch, Coenopatus chinga ; Lge, Leopardus geoffroyi ; Ct, Cerdocyon thous ; Lgy, Lycalopex gymnocercus Error bars represent Bayesian credible in tervals. A B C Parameter estimate of effect

PAGE 45

45 APPENDIX A CORRELATION MATRIX breadth; Hab, habitat breadth; Litter, litter size; FRep, age to first reproduction; Hunt, hunting pressure. Trophic Body Diet Hab Litter FRep Hunt Trophic 1.00 0.05 0.13 0.02 0.24 0.27 0.34 Body 0.05 1.00 0.28 0.46 0.24 0.61 0.29 Diet 0.13 0.28 1.00 0.37 0.54 0.36 0.30 Hab 0. 02 0.46 0.37 1.00 0.44 0.42 0.20 Litter 0.24 0.24 0.54 0.44 1.00 0.21 0.31 FRep 0.27 0.61 0.36 0.42 0.21 1.00 0.10 Hunt 0.34 0.29 0.30 0.20 0.31 0.10 1.00

PAGE 46

46 APPENDIX B HIERARCHICAL OCCUPAN CY BAYESIAN MODELS To formalize the distinction between v ariation in detectability and variation in occurrence I introduced a distinction between observation (i.e., the observed occupancy state, y ) and the underlying state variable (i.e., the true occupancy state, z ) which is only observed imperfectly ( Royle and Dorazio 2009 ) I distinguish ed between detection prob ability, the probability of observing the species given that it is present, i.e., p = Pr( y i = 1| z i = 1) and occurrence probability, which is = Pr( z i = 1) In particular, I describe d the relationship between y and z in terms of p and by the compound Bernoulli model wh ich, for the observation model is : y i | z i ~ Bernoulli ( z i p ) and, for the process or state model, z i ~ Bernoulli ( i ) ; ( z i i is occupied if z i = 1, site is unoccupied if z i = 0). I modeled i as a function of both fixed and random effects: logit ( i ) = 0 + cov x i + t 0 is an intercept term, 0 is a vector of regression parameters associated with covariate s, x i is a vector of distance covariates at sampling unit i and t is a random effect of transect t between distance categories along a transect and occupancy. Negative values indicate that a species was more likely to occupy a given site as distance from the strip into the

PAGE 47

47 forest interior increased and less likely to occupy the strip as distance from continuous forest into the strip increased (i.e., A s occupancy increased from d10 to d1 (Fig ure 2 1 ), te the opposite pattern (i.e., O ccupancy decreased from d10 to d1 ).I effectively block ed each site by transect by including a random transect effect in the mode l, which can be a useful way to deal with potential spatial autocorrelation I included the additional observation model that accounts for imperfect detection. I modeled detection or non detection of a species as a Bernoulli random variable: y i ~ Bernoull i ( p ) I modeled p i as a linear function of covariates likely to influence detection probability as: logit ( p i ) = 0 + cov i Where 0 is an intercept term, cov is a vector of regression parameter associated with environmental i is a vector of detection covariates at site i. I included two environmental covariates: canopy cover and understory density Prior distribution For all fixed effects in Bayesian models, I specified vague normal prior distribution w ith a mean 0 and variance 1000 ( Royle and Dorazio 2009 ) I used a non informative Uniform (0, 1) prior distribution on the intercept ( Rota et al. 2011 ) I specified all random effects to have a mean zero and a non informative Uniform (0, 10) prior distribution for each standard deviation parameter.

PAGE 48

48 #Code for a hierarchical occupancy model dist.mean = mean(mazama$d stance) #variables are centered dist.sd = sd(mazama$dstance) under.mean = mean(mazama$und_dens) under.sd = sd(mazama$und_dens) canopy.mean = mean(mazama$cnp_cvr) canopy.sd = sd(mazama$cnp_cvr) trans < as.numeric(factor(mazama$trans)) data < list( n=length(mazama$Site), #I specify the length so it can reiterate j=mazama$count,#number of detections/camara y=mazama$total, distance=((mazama$dstance dist.mean)/dist.sd), canopy=((mazama$cnp_cvr canopy.mean)/canopy.sd), understory=((mazama$un d_dens under.mean)/under.sd), trans=trans, ntrans=max(trans) ) inits < function(){ # load inits for WinBugs list( bMean=runif(1), b1=rnorm(1), aMean=runif(1), a1=rnorm(1), a2=rnorm(1), z=as.integer(mazama$total>0 ), transect.effect=rnorm(max(trans)), sd.transect=runif(1,0,5) ) } params < list( 'b0', 'b1', 'a0', 'a1',

PAGE 49

49 'a2', 'bMean', 'occ.fs' ) # Define model cat(" model{ #Prior distributions for fixed effects (B&A) bMean ~ dunif(0,1) b0 < log(bMean) log(1 bMean) #log transform parameter b1 ~ dnorm(0,.001) #set priors for b aMean ~ dunif(0,1) a0 < log(aMean) log(1 aMean) a1 ~ dnorm(0,.001) a2 ~ dnorm(0,.001) #Prior distributions for random effect variance parameters sd.transect ~ dunif(0,10) tau.transect < pow(sd.transect, 2) #Prior distributions for random effects for(r in 1:ntrans){ transect.effect[r] ~ dnorm(b0, tau.transect)} for (i in 1:n) { #start initial loop over the n sites # T rue state model for the partially observed true state z[i] ~ dbern(psi[i]) # True occupancy z at site i psi[i] < 1/(1+exp( fu[i])) fu[i]< b1*distance[i] + transect.effect[trans[i]] #observation model mu.y[i] < z[i] p[i] p[i]< 1/(1+exp( cu[i])) cu[i] < a0 + a1*understory[i] + a2*canopy[i] y[i] ~ dbin(mu.y[i],j[i])

PAGE 50

50 }#for loop occ.fs < (sum(z[]))/n # Number of occupied sites } #model ",fill=TRUE, file="trainingmodel.txt") out = bugs(data, inits, params, model.file='trainingmodel.txt', debug=T, n.chains=3, n.iter=100000, n.burnin=20 000, n.thin=3, DIC=F, working.directory=getwd())

PAGE 51

51 APPEND IX C PARAMETER DISTANCE A ND SPECIES TR AITS Relation between parameter estimate of Distance with Age to first reproduction (A), Log10 of Body Size (B), Hunting Pressure (C), and Trophic Breadth (D). Interpretation tions: Pt, Pecary tajacu; Mt, Myrmecophaga tridactila; Tm, Tolypeutes matacus; Ds, Dolichotis salinicola; Mg, Mazama gouazoubira; Cch, Coenopatus chinga; Lge, Leopardus geoffroyi; Ct, Cerdocyon thous; Lgy, Lycalopex gymnocercus. P ar a m et er es ti m at e of ist a nc ef fe ct B C A D

PAGE 52

52 BIBLIOGRAPHY Adamoli, J.M., Ginzburg, R., Torrella, S., 2011. Escenarios productivos y ambientales del Chaco Argentino: 1977 2010, ed. `, p. 101. Universidad de Buenos Aires, Buenos Aires. Altrichter, M., 2005. The sustainability of subsistence hunting of pecca ries in the Argentine Chaco. Biological Conservation 126, 351 362. Altrichter, M., 2006. Wildlife in the life of local people of the semi arid Argentine Chaco. Biodiversity and Conservation 15, 2719 2736. Altrichter, M., Almeida, R., 2002. Exploitation o f white lipped peccaries Tayassu pecari (Artiodactyla : Tayassuidae) on the Osa Peninsula, Costa Rica. Oryx 36, 126 132. Andersen, M.C., Watts, J.M., Freilich, J.E., Yool, S.R., Wakefield, G.I., McCauley, J.F., Fahnestock, P.B., 2000. Regression tree mode ling of desert tortoise habitat in the central Mojave Desert. Ecological Applications 10, 890 900. Barlow, J., Louzada, J., Parry, L., Hernandez, M.I.M., Hawes, J., Peres, C.A., Vaz de Mello, F.Z., Gardner, T.A., 2010b. Improving the design and management of forest strips in human dominated tropical landscapes: a field test on Amazonian dung beetles. (vol 47, pg 779, 2010). Journal of Applied Ecology 47, 1158 1158. Boughey, K.L., Lake, I.R., Haysom, K.A., Dolman, P.M., 2011. Improving the biodiversity ben efits of hedgerows: How physical characteristics and the proximity of foraging habitat affect the use of linear features by bats. Biological Conservation 144, 1790 1798. Cagnolo, L., Valladares, G., Salvo, A., Cabido, M., Zak, M., 2009. Habitat Fragmentat ion and Species Loss across Three Interacting Trophic Levels: Effects of Life History and Food Web Traits. Conservation Biology 23, 1167 1175. Canevari, M., Vaccaro, O., 2007. Guia de mammiferos del sur de America del Sur, Buenos Aires. Carbone, C., Cowl ishaw, G., Isaac, N.J.B., Rowcliffe, J.M., 2005. How far do animals go? Determinants of day range in mammals. American Naturalist 165, 290 297. Crooks, K.R., 2002. Relative sensitivities of mammalian carnivores to habitat fragmentation. Conservation Biolo gy 16, 488 502. de Lima, M.G., Gascon, C., 1999. The conservation value of linear forest remnants in central Amazonia. Biological Conservation 91, 241 247.

PAGE 53

53 Diamond, J.M., 1975. The island dilemma: Lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7, 129 146. Dorazio, R.M., Royle, J.A., 2005. Estimating Size and Composition of Biological Communities by Modeling the Occurrence of Species. Journal of the American Statistical Association 100, 389 398. Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics 34, 487 515. FAO, 2010. Global forest resources assessment ed. F. department. Gasparri, N.I., Grau, H.R., 2009. Deforestation and fragmentatio n of Chaco dry forest in NW Argentina (1972 2007). Forest Ecology and Management 258, 913 921. Gilks, W.R., Thomas, A., Spiegelhalter, D.J., 1994. A language and program for complex Bayesian modeling. Statistician 43, 169 177. Grau, H.R., Gasparri, N.I., Aide, T.M., 2008. Balancing food production and nature conservation in the Neotropical dry forests of northern Argentina. Global Change Biology 14, 985 997. Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than simple habita t models. Ecology Letters 8, 993 1009. Halchak, J.L., Seliskar, D.M., Gallagher, J.L., 2011. Root system architecture of kosteletzkya pentacarpos (Malvaceae) and belowground environmental influences on root and aerial growth dynamics. American Journal of Botany 98, 163 174. Hawes, J., Barlow, J., Gardner, T.A., Peres, C.A., 2008. The value of forest strips for understorey birds in an Amazonian plantation landscape. Biological Conservation 141, 2262 2278. Kery, M., 2010. Introduction to WinBUGS for ecolog ists, London. Laurance, S.G., Laurance, W.F., 1999. Tropical wildlife corridors: use of linear rainforest remnants by arboreal mammals. Biological Conservation 91, 231 239. Lees, A.C., Peres, C.A., 2008. Conservation value of remnant riparian forest corr idors of varying quality for Amazonian birds and mammals. Conservation Biology 22, 439 449. MacKenzie, D.I., 2006. Modeling the probability of resource use: The effect of, and dealing with, detecting a species imperfectly. Journal of Wildlife Management 7 0, 367 374.

PAGE 54

54 MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege, S., Ro yle, J.A., Landtimm, C.A., 2002 Estimating site occupancy rates when detection probabilities are less than one. Ecology 83, 2248 2255. Maffei, L., Cuellar, E., Noss, A., 2004. One t housand jaguars (Panthera onca) in Bolivia's Chaco? Camera trapping in the Kaa Iya National Park. Journal of Zoology 262, 295 304. Mares, M.A., 1992. Neotropical mammals and the myth of amazonian biodiversity. Science 255, 976 979. Mares, M.A., Ojeda, R. A., Barquez, R.M., 1989. Guide to mammals of Salta Province, Argentina. Publishing division of the University, Norman. Noss, R.F., Quigley, H.B., Hornocker, M.G., Merrill, T., Paquet, P.C., 1996. Conservation biology and carnivore conservation in the Rock y Mountains. Conservation Biology 10, 949 963. Ojeda, R.A., Stadler, J., Brandl, R., 2003. Diversity of mammals in the tropical temperate Neotropics: hotspots on a regional scale. Biodiversity and Conservation 12, 1431 1444. Peres, C.A., 2001. Synergisti c effects of subsistence hunting and habitat fragmentation on Amazonian forest vertebrates. Conservation Biology 15, 1490 1505. Rabinowitz, A., Zeller, K.A., 2010. A range wide model of landscape connectivity and conservation for the jaguar, Panthera onca Biological Conservation 143, 939 945. Redford, K.H., Taber, A., Simonetti, J.A., 1990. There is more to biodiversity than the tropical rain forests. Conservation Biology 4, 328 330. Rota, C.T., Fletcher, R.J., Evans, J.M., Hutto, R.L., 2011. Does accou nting for imperfect detection improve species distribution models? Ecography 34, 659 670. Royle, J.A., Dorazio, R.M., 2009. Hierarchical modeling and inference in ecology, London. Rozylowicz, L., Popescu, V.D., Patroescu, M., Chisamera, G., 2011. The pot ential of large carnivores as conservation surrogates in the Romanian Carpathians. Biodiversity and Conservation 20, 561 579.

PAGE 55

55 Sachs, J.D., Baillie, J.E.M., Sutherland, W.J., Armsworth, P.R., Ash, N., Beddington, J., Blackburn, T.M., Collen, B., Gardine r, B., Gaston, K.J., Godfray, H.C.J., Green, R.E., Harvey, P.H., House, B., Knapp, S., Kumpel, N.F., Macdonald, D.W., Mace, G.M., Mallet, J., Matthews, A., May, R.M., Petchey, O., Purvis, A., Roe, D., Safi, K., Turner, K., Walpole, M., Watson, R., Jones, K .E., 2009. Biodiversity Conservation and the Millennium Development Goals. Science 325, 1502 1503. Seaman, B.S., Schulze, C.H., 2010. The importance of gallery forests in the tropical lowlands of Costa Rica for understorey forest birds. Biological Conserv ation 143, 391 398. Seghezzo, L., Volante, J.N., Paruelo, J.M., Somma, D.J., Buliubasich, E.C., Rodrguez, H.E., Gagnon, S., Hufty, M., 2011. Native Forests and Agriculture in Salta (Argentina). The Journal of Environment & Development 20, 251 277. Shaha buddin, G., Ponte, C.A., 2005. Frugivorous butterfly species in tropical forest fragments: correlates of vulnerability to extinction. Biodiversity and Conservation 14, 1137 1152. Stearns, S.C., 1976. Life history tactics review of ideas. Quarterly Revie w of Biology 51, 3 47. Sturtz, S., Ligges, U., Gelman, A., 2005. R2WinBUGS: A package for running WinBUGS from R. Journal of Statistical Software 12, 1 16. Thornton, D., Branch, L., Sunquist, M., 2011a. Passive sampling effects and landscape location alter associations between species traits and response to fragmentation. Ecological Applications 21, 817 829. Thornton, D.H., Branch, L.C., Sunquist, M.E., 2011 b. The relative influence of habitat loss and fragmentation: Do tropical mammals meet the temperate paradigm? Ecological Applications 21, 2324 2333. Tilman, D., Fargione, J., Wolff, B., D'Antonio, C., Dobson, A., Howarth, R., Schindler, D., Schlesinger, W .H., Simberloff, D., Swackhamer, D., 2001. Forecasting agriculturally driven global environmental change. Science 292, 281 284. Tilman, D., May, R.M., Lehman, C.L., Nowak, M.A., 1994. Habitat destriction and the extintion bedt. Nature 371, 65 66. Torrell a, S.A., Oakley, L.J., Ginzburg, R.G., Adamoli, J.M., Galetto, L., 2011. Structure, composition and state of conservation of the woody plant community of the "tres quebrachos" forests in the Subhumid Central Chaco. Ecologia Austral 21, 179.

PAGE 56

56 Wallace, R., B ., Gomez, H., Zulia R, P., Rumiz, D., 2010. Distribucion, ecologia y conservacion de los mamiferos medianos y grandes de Bolivia. Santa Cruz de la Sierra. Wehling, S., Diekmann, M., 2009. Importance of hedgerows as habitat corridors for forest plants in agricultural landscapes. Biological Conservation 142, 2522 2530. Zanuncio, T.V., Zanuncio, J.C., Miranda, M.M.M., Medeiros, A.G.D., 1998. Effect of plantation age on diversity and population fluctuation of Lepidoptera collected in Eucalyptus plantations in Brazil. Forest Ecology and Management 108, 91 98.

PAGE 57

57 BIOGRAPHICAL SKETCH Mauricio finished his licenciatura in biological sciences in 2009 at the National University of Salta, Argentina. He then pursued a Master of Science degree in the Wildlife Ecology and Conservation Department at the University of Florida and completed the degree in 2011.