Citation
Predicting the Effect of Climate Change on Endemic Reptile Species in Mexico, an Individual and Community Level Approach

Material Information

Title:
Predicting the Effect of Climate Change on Endemic Reptile Species in Mexico, an Individual and Community Level Approach
Creator:
Bedoya Duran, Maria Juliana
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (111 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Interdisciplinary Ecology
Committee Chair:
CAMERON DEVITT,SUSAN ELIZABETH
Committee Co-Chair:
BRANCH,LYN CLARKE
Graduation Date:
8/9/2014

Subjects

Subjects / Keywords:
Biodiversity conservation ( jstor )
Climate change ( jstor )
Climate models ( jstor )
Ecology ( jstor )
Environmental conservation ( jstor )
Modeling ( jstor )
Protected areas ( jstor )
Reptiles ( jstor )
Species ( jstor )
Wildlife conservation ( jstor )
Interdisciplinary Ecology -- Dissertations, Academic -- UF
climate-change -- mexico -- reptiles
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Interdisciplinary Ecology thesis, M.S.

Notes

Abstract:
Over the past decade, numerous studies have predicted about the important impact of climate change on biodiversity. Nevertheless, a lack of information on the spatial distribution of many taxonomic groups make difficult to predict how these species will be affected. Here, I used Maxent, Random Forest, Generalized Boosting model and the Generalized Linear model as individual models and a generalized dissimilarity model as community-level model to analyze current and future diversity of endemic reptile species in Mexico. Future predictions were built using three time periods (2020, 2050, and 2080), two emission scenarios and two general climatic models. We found that richness is particularly high in the central-southern Pacific coast region, followed by the Yucatan peninsula. Under future climate, the Sierra Madre Occidental and few regions in the south will be particularly affected. When using the GDM, we found that current community composition did not follow a clear pattern like richness, instead it was predicted multiple areas with unique community composition, nevertheless communities in the tropical realm are predicted to be more similar than temperate populations. Here we provided richness and beta diversity information that can be used together to provide a framework for optimal design of protected areas. From a conservation perspective, the future effect of climate change in richness and turnover reveal further challenges for the current protected areas system in Mexico and propose a revision based not only in areas with high richness but also in areas where the predicted change will be the greatest in terms of species composition. ( en )
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.
Thesis:
Thesis (M.S.)--University of Florida, 2014.
Local:
Adviser: CAMERON DEVITT,SUSAN ELIZABETH.
Local:
Co-adviser: BRANCH,LYN CLARKE.
Statement of Responsibility:
by Maria Juliana Bedoya Duran.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Resource Identifier:
968131467 ( OCLC )
Classification:
LD1780 2014 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

PREDICTING THE EFFECT OF CLIMATE CHANGE ON ENDEMIC REPTILE SPECIES IN MEXICO, AN INDIVIDUAL AND COMMUNITY LEVEL APPROACH By MARIA JULIANA BEDOYA DURAN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORID A IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2014

PAGE 2

© 2014 Maria Juliana Bedoya Durá n

PAGE 3

To my family and people who support ed me in this process

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my advisor Dr. Susan Cameron for the opportunity to work in this project and the continuous support in all the writing process. Also, I w ould like to thank to my committee member professor Dr. Lyn Branch for her motiv ation and enthusiasm in my process of knowledge, mainly in the conservation biology field and Dr. Bette Loiselle for her helpful comments in th e final stage of this process . I thank Fulbright Colombia , Colciencias and to provide the financial support. I sp ecially thank to Universidad Autónoma de Mexico, especially Dr. Oscar Flores Villela for the provision of the data used in this research. I thank my labmates for their comments in the different stages of the project. Finally , I w ould like to thank my famil y and Oscar , the support and encouragement in each step was vital to overcome my objectives and finish this thesi s with success.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ . 4 LIST OF TABLES ................................ ................................ ................................ ........... 7 LIST OF FIGURES ................................ ................................ ................................ ........ 8 LIST OF ABBREVIATIONS ................................ ................................ .......................... 10 ABSTRACT ................................ ................................ ................................ .................. 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ... 13 2 PREDICTING CURRENT AND FUTURE PATTERNS OF RICHNESS OF ENDEMIC REPTILES IN MEXICO ................................ ................................ ........ 17 Methods ................................ ................................ ................................ ................ 19 Species Data ................................ ................................ ................................ ... 19 Environmental Data ................................ ................................ ......................... 19 Data Analysis ................................ ................................ ................................ .. 21 Species distribution models ................................ ................................ ...... 21 Consensus maps ................................ ................................ ...................... 22 Testing differences in models performance ................................ ............... 23 Current and future richness patterns ................................ ......................... 23 Representation of species in protected areas ................................ ........... 24 Results ................................ ................................ ................................ .................. 24 Variation among Individual Models and Comparisons ................................ ..... 24 Consensus Models ................................ ................................ .......................... 25 Species Richness (Alpha Diversity) ................................ ................................ . 26 Change in Protected Areas ................................ ................................ ............. 27 Analyzing the Change in Richness by National Parks ................................ ..... 28 Discussion ................................ ................................ ................................ ............. 28 Model Performance and Selection ................................ ................................ .. 28 Current Richness Patterns ................................ ................................ .............. 30 Climatic Models and Future Projection ................................ ............................ 31 Limitations an d Future Work ................................ ................................ ............ 33 3 COMMUNITY LEVEL PREDICTION OF CURRENT AND FUTURE COMPOSITION OF ENDEMIC REPTILES IN MEXICO ................................ ........ 53 Methods ................................ ................................ ................................ ................ 55 Species and Environmental Data ................................ ................................ .... 55 Predicting Beta Diversity Patterns ................................ ................................ ... 56

PAGE 6

6 Climate Change Assessment Using GDM ................................ ....................... 58 Results ................................ ................................ ................................ .................. 60 Discussion ................................ ................................ ................................ ............. 63 Climate Change ................................ ................................ .............................. 64 Protected Areas ................................ ................................ .............................. 65 Recommendation and Future Application ................................ ........................ 66 4 CONCLUSIONS ................................ ................................ ................................ .... 78 APPENDIX A ACCURACY OF EACH INDIVIDUAL MODEL BY SPECIE ................................ .... 80 B BIOGEOGRAPHIC REGIONS IN MEXICO ................................ ........................... 87 C FUTURE RICHNESS MAPS USING THE Hadcm3 GCM ................................ ...... 88 D PREDICTED CHANGE IN SPECIES RICHNESS BY NATIONAL PARK ............... 92 LIST OF REFERENCES ................................ ................................ ............................ 101 BIOGRAPHICAL SKETCH ................................ ................................ ......................... 111

PAGE 7

7 LIST OF TABLES Table page 2 1 Representation of endemic reptiles used in the analysis ................................ ... 36 2 2 Number of species with good (>0.70) and poor (<0.69) performance according to the 3 goodness of fit measurements used: Area under the curve (KAPPA) and True Sills Statistic (TSS). ........................ 36 2 3 Results of the Friedman test for differences between the performances measurem ents of the modeling techniques. ................................ ...................... 37 3 1 Principal Coordinates Analysis using the predicted dissimilarity values to build current beta diversity map. ................................ ................................ ........ 69

PAGE 8

8 LIST O F FIGURES Figure page 2 1 Frequency of records by species. Total data set: 183 species and 19.916 records. ................................ ................................ ................................ ............. 38 2 2 Box plot of predicted performances measures (AUC, KAPPA and TSS) by four modelling techniques ................................ ................................ .................. 39 2 3 Boxplot including all groups together and consensus techniques for the models tested with TSS scores, gray line indicates performance of 0.7. ............ 40 2 4 Map of current Richness patterns of Mexican endemic reptiles based on the Mean ensemble model. ................................ ................................ ..................... 41 2 5 Richness patterns of endemic reptiles using the general circulation model CGCM2, the emission scenario A2A and the three periods: 2020, 2050 and 2080. ................................ ................................ ................................ ................. 42 2 6 Richness patterns of endemic reptiles using the general circulation model CGCM2, the emission scenario B2A and the three periods: 2020, 2050 and 2080. ................................ ................................ ................................ ................. 43 2 7 Predicted change in species r ichness according to the CGCM2 model under the A2A scenario ................................ ................................ ............................... 44 2 8 Predicted change in species richness according to the CGCM2 model under the B2A scenario ................................ ................................ ............................... 45 2 9 Map of the protected areas included in the analysis. IUCN categories I and II (National parks and Scientific reserves). ................................ ........................... 46 2 10 Analysis of change in specie s Richness by IUCN categry for the CGCM2 GCM and A2a scenario. ................................ ................................ .................... 47 2 11 Analysis of change in species Richness by IUCN categry for the CGCM2 GCM and B2a scenario. ................................ ................................ .................... 48 2 12 Composite map using as an example the predicted change in richness for the last period, climatic scenario A2a and CGCM2 climatic model and the Protected areas system (IUCN Categories). ................................ ...................... 49 2 13 Change in Mean species richness by National Parks, we catologue those like smaller parks (Area: 10Km or less). . ................................ ................................ . 50

PAGE 9

9 2 14 Change in Mean species richne ss by National Parks, we catologue those like big parks (Area: greater that 10Km). . ................................ ................................ 51 2 15 Distribution of the dry forest and protected areas in Mexico. ............................. 52 3 1 Location of the 1543 sites included in the analysis. ................................ ........... 70 3 2 Observed response based on the overall beta diversity. The curved line in black represents the non linear link function where environmental and spatial distances relate the dissimilarities and the dots represents a site pair. .............. 71 3 3 Relative contribution of environmental variables and geogr aphic distance (Sum of coefficients of the I splines) for current species composition. The most important predictors are showing first.. ................................ ..................... 72 3 4 Fitted functions of observed overall beta divers ity. The maximum height reached by each function provides the amount of beta diversity along the gradient (holding all other variables constant). ................................ .................. 74 3 5 Beta diversity map using the ordination axes obtained from the predicted dissimilarity matrix. Comp 1 represents the first axe, Comp 2 represents the second axe and Comp 3 represents the third axe. ................................ ............ 74 3 6 Compositional turnover for ecasted among the three time periods evaluated 2020, 2050 and 2080. ................................ ................................ ....................... 75 3 7 Composite map of protected areas in Mexico using projected change in species turnover for the 2020 period.. ................................ ............................... 76 3 8 Composite map of protected areas in Mexico, classified following the IUCN, using projected change in species composition according to the last period evaluate 2080. ................................ ................................ ................................ ... 77

PAGE 10

10 LIST OF ABBREVIATIONS GBM Generalized boosting model GCMs General circulation models GDM Generalized dissimilarity model GLM Generalized linear model IPCC Intergovernmental panel on climate change MAXENT The maximum entropy model RF Random forest SDMs Species distribution models

PAGE 11

11 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science PREDICTING THE EFFECT OF CLIMATE CHANGE ON ENDEMIC REPTILE SPECIES IN MEXICO, AN INDIVIDUAL AND COMMUNITY LEVEL APPROACH By Maria Juliana Bedoya Durá n August 2014 Chair: Susan E. Cameron Devitt Major: Interdisciplinary Ecology Over the past decade, numerous studies have pr edicted about the important impact of climate change on biodiversity. Nevertheless, a lack of information on the spatial distribution of many taxonomic groups make difficult to predict how these species will be affected. Here, I use d Maxent, Random Forest, Generalized Boosting model and the Generalized Linear model as individual models and a generalized diss i milarity model as community level model to analyze current and future diversity of endemic reptile species in Mexico. Future predictions were built usi ng three time periods (2020, 2050, and 2080 ), two emission scenarios and two general c irculation models. We found that richness is particularly high in the central southern Pacific coast region, followed by the Yucatan peninsula. Under future climate, the Sierra Madre Occidental and few regions in the south will be particularly affected. When using the richness, instead it was predicted multiple areas with unique community composition, nevertheless communities in the tropical realm are predicted to be more similar than temperate populations.

PAGE 12

12 Here we provided richness and beta diversity information that can be used together to provide a framework for optimal design of protect ed areas. From a conservation perspective, the future effect of climate change in richness and turnover reveal further challenges for the current protected areas system in Mexico and propose a revision based not only in areas with high richness but also in areas where the predicted change will be the greatest in terms of species composition.

PAGE 13

13 CHAPTER 1 INTRODUCTION The increase in global temperature is predicted to be more severe in the near future and is projected to dramatically alter environmental pa tterns leading to a substantial loss of bio diversity (Mittermeier & Myers, 1998; Peterson et al. , 2002; Thomas et al. , 2004; Thuiller et al. , 2004; IPCC, 2012) . Currently, changes in climatic conditions combined with human per turbation is causing distributional shifts, community homogenization and in general altering communities at different level s (Walther e t al. 2002; Parmesan & Yohe 2003, McKinney & Lo ckwood 1999) . However, accurate analysis of the impact of climate change on different components of diversity are still sparse (Magurran et al. , 2010; Davey et al. , 2013) and the understanding of how those patterns are affected across multiple species is of special concern for conservation prioritization (Moilanen et al. , 200 9) . The large number of species under different threats or in danger of extinction and with high conservation priority around the world is alarming and is generating special attention (Mittermeier & Mittermeier, 1992) . Tropica l forest species are adapted to a relatively stable temperature range (Hoffmann et al. , 2003; Tewksbury et al. , 2008) , making them especially sensitive to increases in temperature (Deutsch et al. 2008; Feeley & Silman 2010; Tewksbury et al. 2008) . Therefore, it is predicted that areas with high levels of endemism, such as tropical forests, will be especially vulnerable to biodiversity loss caused by changes in environmen tal conditions (Mittermeier & Myers, 1998; Malcolm et al. , 2006; Jetz et al. , 2007) . Mexico occupy the first places in terms of b iological diversity around the world (Mittermeier & Goettsch, 1992; Ramamoorthy et al. , 1993) . Particularly, Mexico has the

PAGE 14

14 second highest number of reptile in the world with ap proximately 804 species, of which nearly 50% are endemic (Flores Villela & Canseco Márquez, 2004) . However, half of the species of reptiles are under some level of threat and just 46% of the endemic reptiles are protected in formal areas (Barrera Santos et al. , 2004) . Consequently, Mexican endemic reptiles are particularly vulnerable to changes in environmental conditions and habitat perturbation, which could cause an important loss of global reptile diversity. Given the multiple thre ats for species persistence, protected area networks become essential for biodiversity in the long term (Rodrigues & Gaston, 2002; Carroll et a l. , 2010) represent important biodiversity gradients. Predicting change in species composition and species richness through time is critical to fully understanding how climate change will affect biodiversity (Thuiller et al. , 2005) particularly in regions where species are numerous, highly threatened by habitat transformation or where there are large concentrations of species with small ranges (Raxworthy et al. , 2008) . SDMs describe the relationship between current species distribution s and climate and have become an important tool for conservation, ecology and biogeography (Araújo & Peterson, 2012) as well as becoming a useful tool to forecast the effect of climate change on species distribution (Araújo et al. , 2006; Thuiller et al. , 2006) . In this research w e used f our individual modeling tec hniques to predict current and future richness of endemic reptiles. We used t hree machine learning techniques; a G eneralized B oosting Regression M odel (GBM), Random Forest and a Max imum entropy model (Maxent) and a generalized linear model .

PAGE 15

15 Maxent is a te chnique that has been widely used in climate change predictions (Phillips et al. , 2006) which use presence only data with environmental variables to estim ate the species distribution, performing specially accurately the distribution of species with few point localities and in general has demonstrated high predictive performances among other single species distribution modelling techniques (Hernandez et al. , 2006; Thomassen et al. , 2010 ) . The GBM is characterized by its efficiency in fitting the data and because it combines different statistical techniques. he boosting (Friedman et al. , 2000) . T his technique , based the accuracy by iteratively estimating classifiers using a decision tree algorithm (Heikkinen et al. , 2006) . The Random Forest approach produc es thousands of random trees to finally select the most parsimonious tree that has a predictive value (Heikkinen et al. , 2006) . Finally we test ed a regression method, the generalized linear model, one of the most used techniques to predict species distribution that constitute a flexible regression model that allow other distribution for the response variable like Gaussian, Poisson, Binomial and that are able to incorporate additional information to predict (Guisan & Zimmermann, 2000) . Due to richness alone has a limited value for conservation, using species co mposition provide a better knowledge to propose conservation assessments in specific regions (Ferrier et al. , 2007) . The community modeling technique used to predict pattern s of beta diversity of endemic reptiles was the generalized dissimilarity model. This technique uses information from environmental variables and the geographic distance to predict species composition using a matrix regression technique .

PAGE 16

16 The compositional dissimilarity can be measured using any of the multiple indices proposed in the literature (Ferrier et al. , 2007) . Finally , we provide d specific information about the acc uracy of the different modeling techniques to select the best model in order to estimate change in species richness and composition , and to predict how climate change is going to have an effect in current diversity of Mexica n reptiles . Currently, t he Mexic o government is trying to develop strategies that cope for climate change and to allow maximum protection of biodiversity. This study provides a perspective about how different areas are going to be affected by climate change and offers important inputs fo r reserve design and conservation prioritization in Mexico.

PAGE 17

17 CHAPTER 2 PREDICTING CURRENT AND FUTURE PATTERNS OF RICHNESS OF ENDEMIC REPTILES IN MEXICO Evaluating distribution patterns of species is extremely important for conservation, particularly in ar eas where biological diversity is significant and highly threatened (Margules & Pressey, 2000) .However, evaluating species distributions across large areas, such as at the country level, is difficult due to high environmental heterogeneity coupled with insufficient sampling (Hortal et al. , 2007) which is the case of the most diverse countries (Ceballos et al. , 1998; Myers et al. , 2000) . The quantification and analysis of the relation between environment and species d istribution has become important for understanding patterns of diversity and extrapolating information to areas that are poorly sampled or not sampled at all which provides valuable information for conservation (Guisan & Zimmermann, 2000; Ferrier, 2002) . Although many factors influence species distributions, it is well known that climate strongly influences physical and biological systems . Current fluctuations of temper ature and dramatic flooding or dry events are occurring with more frequency around the world (IPCC, 2007) . Simultaneously, habitat fragmentation and deterioration threatened the permanence of specie s in their natural habitat (Warren et al. , 2001; Travis, 2003; Foufopoulos et al. , 2011) . S pecies distribution models (SDMs) combine records of species presence and/or absence with environme ntal information to predict their distribution and is the most used approach to modeling and extrapolating relationships between species and environmental variables (Ferrier et al. , 2002) . These SDMs are commonly used to identify important areas for conservation due to perturbation and climate change (Myers et al. , 2000; Dennis et al. , 2002) . P redict ing the

PAGE 18

18 impacts of climate change on species provides knowledge about wh ich areas are more at risk and help s managers to take actions now in order to be prepared for the future. Reptiles are an informative group of vertebrates to examine climate change issues due most of their physiological process are dependents on temperatur e alone. Characteristics like ectothermy, biogeographical histories, geographic distributions, dispersal capabilities, diversity and reproductive modes like oviparity make reptiles an interesting group for assessing those impacts (Lillywhite, 2013) . Additionally, across multi ple groups, variation in species richness is well explained by temperature alone (Qian, 2010) . Mexico is characterized as a biodiversi ty hotspot because is one of the (Mitter meier & Goettsch, 1992) . Mexico has around 804 reptile species, of which nearly 50% are endemic (Flores Villela & Canseco Márquez, 2004) . Mexico has experienced high rates of deforestation rates and resulting land use change which threatens its biological diversity (Fuller et al. , 2007) . Because of the geographic location and biogeographic history Mexico holds a great diversity terms of ecoregions and it has been estimated that around 90% of the humid forest has been transformed (Mas et al. , 2004) and only approximately a quarter of the protected areas contain intact seasonally dry tropical forest, one of the most diverse areas in terms of reptiles (Tre jo & Dirzo, 2000) . These concerns warrant special attention to determine important areas for conservation to guarantee conservation and protection for the long term in the future. This study quantifies the potential effects of climate change on the distri bution and richness of endemic reptiles in Mexico, identifies important areas with significant reptilian diversity and establishes a baseline for future monitoring and research. This

PAGE 19

19 paper highlights areas where reptiles will be impacted the most by climat e change and provides recommendations of potential areas for conservation in places that are currently unprotected. Finally, we will provide information about the performance of different modeling approaches and we will give insights on their congruence an d utility for climate change predictions. Methods Species Data Our main focus was the endemic species of reptiles occurring in all Mexico. Although important for conservation, many species had too few records to generate an accurate species distribution model (Ramirez Villegas et al 2012) so w e selected species with ten or more records for the analysis (Table 2 1). Consequently, the final database was formed by 183 of the 368 endemic reptile species that inhabit Mexico (Ochoa & Flores Villela 2006, Conabi o Conanp TNC Pronatura FCF, U AN L. 2007) with 19.916 total unique presence records with a higher proportion of species between 10 to 40 records (Figure 2 1). The dataset was previously verified in the map and in the field by experts. Those data were provid ed by the Museo de Zoología de la Facultad de Ciencias (MZFC) at the Universidad Nacional Autónoma de México. Environmental D ata We used fine scale climate layers of 2 by 2 km (4km2) resolution for current climatic scenarios (averaged from 1950 2000 years) from the World climatic database (worldclim.org; Hijmans et al. 2005) . The database contains 19 bioclimatic variables, but many are highly correlated. We performed a pairwise correlation analysis in order to select a subset of the most uncorrelated variables. Based on the correlation test, we selected seven bioclimatic variables (1) annu al mean temperature, (2) isothermality

PAGE 20

20 (mean diurnal range/temperature annual range), (3) maximum temperature of warmest month, (4) annual temperature range, (5) precipitation of wettest month, (6) precipitation of driest month, and (7) precipitation seas onality (the seasonality is the coefficient of variation of the monthly means). Future climate projections were derived from two different global circulation models (GCMs): the Coupled Global Climate Model (CGCM2) from the Canadian Centre for Climate Model ling and Analysis and the Hadley Centre Coupled Model version 3 (HadCM3). Both models have been widely used to model future species distribution (Thuiller et al. , 2006) . For each model, we used two emissions scenarios (SRES) A2a and B2a described in the 2001 Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report. Those scenarios are estab lished based on simulated climate system responses to increasing levels of greenhouse gases based on different hypotheses about projected population size, socioeconomic trends and technological advances. The a2a scenario projects a world with an increasing global This scenario predicts an increase in over the next century and suggest that climate change will be more severe than previously expected (IPCC, 2007) . The B2a scenario is also intermediate, but is based on adoption of local sustainability solutions across environmental, economic and social sectors. In this s cenario the global population is projected to increase but at a lower rate than the A2 scenario. This scenario predicts an increase in global mean temperature (IPCC, 2007) . We developed predictions for the following intervals 2020s (2010 2039), 2050s (2040 2069) and 2080s ( 2070 2099).

PAGE 21

21 These models were downloaded from the global climate model data portal (http://www.ccafs climate.org/). Data Analysis S pecies distribution m odels To model the current distribution of the endemic reptiles we use four widely used SDM modeling tec hniques performed with the biomod2 package (Thuiller et al., 2013) in the R environment (R Core Team, 2013) .We used three machine learning methods: (1) The maximum entropy model Maxent (Phillips et al. , 2006) ; a widely used method that gives accurate predictions for species with few records, requires presence only data and avoids over fitting (Phillips et al. , 2006) ; (2) A Generalized Boosting Model (GBM), another machine learning method which combines a classification and regression tree algorithm with a boosting algor (Ridgeway, 1999) , and (3) the random forest (R F) algorithm (Breiman, 2001) which uses decisions trees. Additionally, we selected one frequently used regression method (4) Generalized Linear Model GLM (McCullagh & Nelder, 1989) . Due to we had presence only data, we generated a random set of 10,000 background points (pseudo absences) to avoid over fitting and to get a good discrimination between presence and absences of species (VanDerWal et al. , 2009) . To evaluate the quality of predictions, we calibrated the models by using a random sample of 70% of the data and we used the remaining 30% to evaluate the predictions (Fielding & Bell, 1997) . The data splitting approach was replicated 3 times from which we calculated the means of each index of t he cross validation. To evaluate the model performance we use three goodness of fit measures: a threshold independent method (1) The area under the receiver operating characteristic curve (AUC; Hanley & McNeil

PAGE 22

22 1982) (KAPPA; Monserud & Leemans, 1992) and (3) the true skill statistic (TSS) which is defined as sen sitivity + specificity 1 (sensitivity: correctly classified presences ;specificity: correctly classified absences). The accuracy of predictive maps produced by the models was classified following ranges of agreement. There are a number of rules of thumb available to help interpreting measures of model performance between observed and projected events (Araújo et al. , 2005a) . in this case we followed the range of agreement by Landis & Koch, 1977 for the Kappa statistics propose an excellent agreement when Kappa >0.75; good 0.40 > K < 0.75; and poor when kappa is <0.40 . T he AUC range to interpret values was proposed by Swets, 1988 as excellent AUC > 0.9, fair 0.7 < AUC < 0.9 and poor AUC < 0 .7 and finally for the TSS we followed Allouche et al., 2006 (good > 0.7 or poor <0.69). Due to we were interested in compare among all of them we used just two categories ; G ood when the index presented a value equal or greater th an 0.7 and poor when the performance were less than 0.7. Species with poor performance across them were excluded from the final richness map. The predictions of the best model were converted into binary presence/absence maps using a threshold that maximize TSS. We assume a full dispersal scenario, it means the species will be able to track their thermal tolerance and move to those places. Consensus m aps Currently, lots of statistical methods are available to predict species distributions, and due to the te sted variability of the results among the modelling techniques (Thuiller et al. , 2004; Pearson et al. , 2006) , in this research we also used an ensemble strategy, which simultaneously applied sever al methods to build the prediction for each species (Araújo et al. , 2005b; Araújo & New, 2007) . The ensemble methods to forecast species

PAGE 23

23 distribution has proven advantages over the use of just one indi vidual technique (Araújo & New, 2007 ; Marmion et al. , 2009) . Here, we used the TSS criteria to select models with a good fit (TSS>0.7) ( Liu et al., 2005; Allouche et a l., 2006) . We tested three ensemble forecast methods; unweight ed mean, weighted mean and median which have been widely employed because of their improvement of species distributions (Coetzee et al., 2009; Marmion et al., 2009, 2008). Testing differences in m o dels p erformance Comparing model performance for large dataset is difficult to interpret on a species by species basis (Segurado & Araújo, 2004) . Based on research that grouping (Segurado & Araújo, 2004) and due to we had species from ten to 200 records, we grouped the species according to the following number of records: Low number of records (10 to 20 records); M edium (21 to 60) and high (61 to 600) holding 98, 25 and 59 species respectively. T o test the differences in models performance among groups we did a post hoc Friedman test . This non parametric test is used to calculate differences under three methods by each group in order to find out which pairs of groups are different from each other (Hollander & Wolfe, 1973) . Current and future richness p atterns Richness is the total count of species in an area. We calculated a richness map of endemic reptiles by summing the binary maps from the consensus models across all species. Higher values in the maps indicate that more endemic reptiles are predicted to occu r. Additionally, we calculated the change in richness by subtracting the consensus prediction maps from Present to 2020; 2020 to 2050 and 2050 to 2080. We generated 4 predictions for each time period for the two GCMs and two emission scenarios.

PAGE 24

24 Representat ion of species in protected a reas Protected area information was obtained from the Commission for Environmental Cooperation web site ( http://databasin.org/ ). This database includes information from the National Commiss ion for Knowledge and Use of Biodiversity (CONABIO) and contains a GIS database for each protected area, including federal, national, provincial, or territorial authority management in Mexico. Within the database, areas are classified by categories from I to IV and Unknowns primary use, following the International Union for the Conservation of Nature (IUCN; Dudley, 2008) . Our main interest was to understand how current areas under protection will be affected under climate change. To estimate the gain or loss of species by protected area, we calculated the mean change in richness per protected area for each time period. We considered areas classified as type I an d II because those areas are under the most controlled regulations to ensure their protection of their conservation values (Dudley, 2008) and correspond to scientific reserves and wilderness areas (I) and national parks (II). There were 48 (out of 66) National Parks (11088.05 km 2 out of 13985.17 KM 2 ) in IUCN categories I and II that we included in our analysis (Figure 2 9).We calculated the change in species richness by national park and by time per iod using the maps of predicted change (Present to 2020, 2020 to 2050, and 2050 to 2080). We calculated maximum, minimum, mean and range of species richness by park and by IUCN classifications. Results Variation a m ong Individual M odels and Comparisons In o rder to calculate the predictive performance of the modelling techniques, we considered the AUC, TSS and Kappa of each model. According to the performance by

PAGE 25

25 model, the AUC and TSS presented higher values (>0.7), but when using KAPPA, the predictive perfor mance was poor for almost all of them, including Maxent that was the model with the greatest number of species presenting good performances. Based on t he number of records t he Friedman tests show ed that significant d ifferences exist among the modeling tec hniques according to the three goodness of fit metrics for the three groups considered (p<0.000) (Table 2 3 ). Across methods, Maxent was the model with the highest predicted performance with an overall mean of AUC, TSS and KAPPA of 0. 9338 (±0.0510); 0. 833 4(±0.1125) and 0.3108(±0.1767) respectively. Random forest was the model with the poorest overall predictive performance with a mean of 0.8332 (±0.0930), 0.6415(±0.1791), and 0.3544(± 0.1601) for the AUC, TSS and Kappa indexes. However, Random forest had t he highest mean performance for the Kappa index followed by the Maxent model . For all the comparisons, the GLM and GBM model presented similar predictive performance. For all the groups, according to the AUC and TSS, the models presented an average above 0 .7, indicating a high predictive performance of each specie s distribution. Consensus M odels Although AUC and TSS were the predictive accuracy measure with higher values we selected the TSS as a measure of cut off due to the tested abilities like independen ce of prevalence and better performance than AUC using presence only records (Allouche et al. , 2006) . Accord ing to the Figure 2 3, the consensus models in general, presented a higher predictive performance than the models individually. Maxent was the model with the best predictive performance (0.8334±0.112) but, using the ens emble model, the predictions improved significantly, been similar the Mean and

PAGE 26

26 the weighted mean 0.9063 (±0.103) and 0.9055 (±0.103) respectively. We selected the Mean consensus method to predict the final species distribution. Species Richness (Alpha D ive rsity) According to the best consensus model (the Mean ensemble model), the maximum richness per grid cell (resolution of 4Km 2 ) was 36 out of th e 183 species evaluated (Figure 2 4). Based on the 183 species evaluated for Mexico, we projected an endemic di versity peak at 36 species per 4 km 2 , with the highest proportion in the southwest of Mexico, with a marked pattern in the Pacific coastal zone. This pattern clearly reflects the difference between the two biogeographical areas presented in Mexico, the Neo tropical region with a marked richness in the volcanic axis, the Sierra Madre del sur and the Depresión de Balsas, at the other hand the Nearctic area with fewer species, particularly the region of the Chihuahuan desert and the northwestern coastal plain. According to the Terrestrial ecoregions of Mexico the areas with the greatest reptile richness are the dry forests of the Planicie Costera and the Lomeríos del Pacífico S ur (see biogeographical areas in Appendix B 1 ) According to the future projections by the two different general circulation models and scenarios, the change in species richness varies in terms of the magnitude of change but not the general pattern. As expected, the change predicted by 2080 is the largest and the 2020 and 2050 periods are mo re similar. The prediction of species richness by the CGCM2 model, for the two scenarios showed that the areas with the greatest richness are distributed mainly in the pacific coast which correspond to the dry forest in the lowlands of the Sierra Madre del Sur particularly, and Sierra Madre Occidental (Figure 2 5). For both climatic scenarios (A2A

PAGE 27

27 and B2A) the dynamic in the central and desert area of Mexico is particularly interesting. This area is predicted to increase in richness with areas containing b etween 0 to 5 species predicted to increase. The dispersion of this pattern is from central Mexico towards the north and south particularly in the most populated areas like DF, Puebla and Morelos. In general, across scenarios, the general trend is that div ersity shifts towards the southern coast and the Gulf of Mexico. Even under a significant increase of change in environmental conditions, these areas will continue to be high in diversity. The differences in richness between 2020 and the present for both s cenarios show a decline in species number in the central part toward the Sierra Madre Occidental and in the lowlands of the Sierra Madre Oriental (Figure 2 7). For the change between 2050 and 2020 the loss of species is higher for the A2A scenario than the B2A. A marked pattern of species decrease is presented in the northern part of the Llanura Costera and Sierra Madre Occidental. Finally, the change in richness for the 2080 and 2050 period, shows a strong decrease in species richness in the Sierra Madre O riental; while slight increases in species number in the lowlands of sierra Madre del Sur, part of the Yucatan peninsula and Baja California Peninsula is predicted for both scenarios and models (Figure 2 7 and Figure 2 8). Change in Protected A reas Within IUCN categories I and II, the analysis predicted that for both types of categories the loss of species increase with the time. The average decrease in species richness is worst for the A2a scenario as expected with an average loss of 12 species (Ia) and 11 species (II) from 2050 2080 (Figure 2 10). This pattern is true for both categories indicating that national parks and strict natural reserves are predicted to experience drastic change in terms of endemic reptile diversity. Additionally, the change

PAGE 28

28 acc ording to the more conservative scenario B2A is similar although the change in species richness between 2050 2020 is particularly larger between 2080 2050 (Figure 2 11). Analyzing the Change in R ichness by National Park s Here we considered 48 of 66 Nationa l Parks that exist in Mexico. We excluded 18 parks that were particularly small. According to the change predicted for the different time periods and under both scenarios, the major tendency is to lose some species as demonstrated in the Figure 2 12, where the important National parks and reserves areas are predicted to change drastically, for the 2080 period. In general, the mean richness change under the different periods is predicted to be negative in most of the parks (Figure 2 13 and Figure 2 14). This pattern is particularly marked for the larger national parks, but there is a net loss predicted in almost all the parks. Discussion Model Performance and S election Species distribution modeling has been used widely around the world, and is particularly us eful in poorly sampled areas or where habitat loss occurs, to provide important information for conservation assessment and management (Araújo & Guisan, 2006) . However, it is also known that their robustness and differences between algorithms should be considered when conservation prioritization and planning depend solely on modeling (Loiselle et al. , 2003) . He re, we develop predictions of current and future distribution of endemic reptiles in Mexico as well as compare model performance in order to produce accurate predictions.

PAGE 29

29 The measure of model performances is the way to compare among best fit of models. Di fferent models are sensitive to different biases such as giving differential weight to commission and omission errors, how they relate to the independence of observations and other issues (Guisan & Thuiller, 2005; Hortal et al. , 2007; Jiménez Valverde et al. , 2008) .Here, we assess three different measures of model performance; TSS, AUC and KAPPA. According to the predictive per formance of the models evaluated, AUC presented higher values than TSS and Kappa. The TSS and Kappa predictive performance in this analysis (Figure 2 2 ). According to Jiménez Valverde in 2012 , calculating AUC when using pseudo absence data violates the AUC theory. In our case, we used TSS, which combine sensitivity and specificity and has been tested independent of prevalence (Allouche et al. , 2006b) to select the best SDM approach. Our results showed that on average, the individual method with the best predictive performance was Maxent (Figure 2 2 ). In gene ral, Maxent has demonstrated accuracy when estimating potential range shifts due to climate change and has been shown in other studies to accurately predict species distributions with few records (Hijmans & Graham, 2006; Phillips et al. , 2006) . In our case we had just few species with a high number of records, but the majority was species with few records (Figure 2 1). The other mo dels used in this analysis; GLM and GBM presented good predictive performances (AUC and TSS), but the RF model got a high variation in their predictions among species and the mean was less than 0.75, the selected value as a good threshold to perform specie s distribution. Because of this difference in predictive accuracy and to reduce model based uncertainty we used a consensus approach

PAGE 30

30 (Thuiller, 2004; Thuiller et al. , 2005) . Consensus models are a model averaging approach that combine several models outputs to overcome variability in predictions, which is part icularly important when generating climate change predictions (Araújo et al. , 2005a; Marmion et al. , 2009) .The mean and the weighted mean consensus approaches both had high goodness of fit, however, we only present the mean consensus models here. Current Richness P atterns According to the F igure 2 4, the area predicted to be the most diverse in terms of species richness was the tropica l dry forest in the pacific low lands . This area has been identified in numerous previous studies as an area of the highest diversity and endemism (CONABIO et al. , 2007; García et al. , 2007; Koleff et al. , 2008 ; García 2006 ) . García et al ( 2007 ) found that the majority of species were concentrated in protected areas along this ecoregion. However, most of the tr opical dry forest lacks formal protection and protected areas within the region are predicted to lose species in response to climate change (Figure 2 15 ). Mexico contains the largest extent of the tropical dry forest biome with the 38% of the total existed area in the Americas. Nevertheless, this vegetation type is highly threatened due to high rates of deforestation. Today in Mexico o nly 28% of the original forest remains (García et al. , 2007; Portillo Quintero & Sánchez Azofeifa, 2010) . Coastal regions in Mexico have high human populatio n density and support around 24% of the entire population (SEMARNAT INE, 2009) . Currently, much of the dry forest and coastal areas lack protection , we recommend urgent protection for this region as well as further surveys and monitoring in order to understand the real threats for endemic reptiles and other taxa.

PAGE 31

31 According to a previous study of the biodiversity in the region (Ochoa Ochoa & Flores Villela, 2006) , the areas with predicted highest endemic reptiles diversity were Sierra Madre del Sur and depresión de Balsas, followed by a similar pattern along the middle southern Mexico following a stronger pattern between the temperate and subtropical biogeographic realms . These overall patterns were similar to our study; however, some of these important areas in terms of richness are predicted to change in response to climate change . This issue of dynamic shifts in biodiversity is a si gnificant challenge to conservation prioritization areas that are highest in diversity today may not be the areas with the highest diversity in the future. Although our analysis excluded endemic species with few er than 10 records , the general results w er e similar to previous analyses (e.g. Ochoa Ochoa & Flores Villela 2006 ) . This highlights the importance of examining different methodologies and biological groups when available , to improve our knowledge about biodiversity patterns in order to formulate co nservation programs at different levels. Climatic Models and Future P rojection The spatial pattern of richness changed mainly through time and in intensity by climatic scenario. In general results were very similar across climate models and scenarios, alth ough there were some differences between scenarios for the 2080 period. Our results suggest that richness is predicted to change the most in the central western Mexico, followed by a strong change in the lowlands of the pacific coast and some areas in the Yucatan peninsula. For 2080, the northern portion of the Sierra Madre Occidental and part of the Chihuahuan desert are predicted to experience significant changes in richness. According to the future projections, and using just

PAGE 32

32 climate Durango and Zacateca s will be the most affected states in terms of change in climatic conditions threatened current protected areas system. One of the key obstacles to allowing endemic reptiles to cope climate change will be the lack of protected areas and a lack of connecti vity among them (Hannah, 2008; Ochoa Ochoa et al. , 2012) . Determining current and future spatial patterns of species distributions is relevant when including existing protected areas or potential undisturbed ecosystems ( Ortega Huerta and Peterson, 2004). Currently, the protected area system is insufficient for the areas with highest predicted richness. Given our prediction of significant future changes in richness, some important management strateg ies include reducing the predicted impacts by decreasing ant h ropogenic stress on ecosystems, increasing habitat quality and habitat connectivity and the creation of new protected areas in those regions (Hannah, 2008; Sinclair et al. , 2010) . In this case, it will be critical to provide connectivity both within and between biogeographic regions in Mexico. Recent studies suggest causal links between reptile extinction and climate change (Sinervo et al. , 2010) . In contrast, som e studies predict some benefits from increasing temperatures to reptiles living in cooler environments, by extending activity times, increasing body size or general energetic benefits (Chamaille Jammes et al. , 2006; Kearney & Porter, 2006) . On the other hand, there is evidence that tropical species may be highly threatened because their physiological function will be compromised due to an increase in temperature (Tewksbury et al. , 2008; Sinervo et al. , 2010) . Because of its diversity of temperate, tropical and subtropical habitats, Mexico is worthy of special attention and may provide insights to und erstand the different impacts of climate change on species.

PAGE 33

33 Limitations and Future W ork This research used different SDM modeling techniques (Maxent, GLM, RF, GBM), two climatic scenarios and two general circulation models in order to select the best appro ach that describe the current and future distribution of species. Past works that describe current patterns of diversity and predicts climate change effect on species mainly use a single model such as GARP (Ortega Huer ta & Peterson, 2 004; García, 2006; García et al. , 2007) and Maxent (Pineda & Lobo, 2009; Ochoa Ochoa et al. , 2012) . Our re search and findings are novel because they are the first to consider multiple models as well as develop a consensus model using endemic species and in general using a country wide assessment of endemic reptiles in Mexico . We hope our results provide new in sights in terms of future protection for biologically important areas. While our study attempts to make predictions of how richness will change through time in response to climate change, we acknowledge that is difficult to accurately and precisely predict the impact of climate change for different reasons. First, using reptiles as a target group results challenging, due to their plasticity as well as their ability to modify their behavior in response to climate change (Lillywhite, 2013) . However, our analysis using current av ailable climatic conditions and current distributions to predict future distributions, will give to us an idea about impacts of climate change, particularly for species that lack data about physiology or behavioral response. Second, we are assuming a full for some of the species due to their mobility capabilities or geographic barriers that will difficult their movement in order to follow their thermal range. Future projections would be more accurate if it is poss ible to integrate biotic and abiotic factors that affects the dispersal capabilities of species and also the range sizes (Sheldon et al. , 2011) . Third,

PAGE 34

34 change and habitat alteration is influencing the most biodiversity (Parmesan & Yohe, 2003) . In general land use effect will be intensified in th e context of climate change due to species may move in or out of areas that additionally, will experience disparate levels of human land use change, thus increasing or decreasing the migration benefit (Feeley & Silman, 2010) . Inside Mexican territories one of the major c oncerns is the fast deforestation or land use transformation. According to that, endemic reptiles will have following their thermal tolerances. How this driver acts and how this effect vary among reptile species and regions is the next step toward conserving reptiles and in general biodiversity. Despite the shortcomings of species distribution models, we hope our consensus models for endemic reptiles will help managers and f uture researchers to use this information as a baseline to understand how this important group of vertebrates will respond to climate change over the next century. Including potential areas of greatest change as target areas to sample in detail and to conn ect will be the first steps for managers. Specially, including information from different groups will help to identify and prioritize those areas. We predict change using different time frames, but due to we did not include how those areas are currently im pacted by land use transformation, the results of the most threatened areas will be worst. Economic and social incentives are changing current land use practices and are essentials to conservation in different countries including Mexico. Intensify those pr actices in order to protect current

PAGE 35

35 unprotected areas will definitely have a positive double effect, reducing emissions by carbon sequestration and allowing movement for species to cope for climate change.

PAGE 36

36 Table 2 1 . Representation of endemic reptiles used in the analysis Families No of Species National Status* IUCN Status # Anguidae 20 13(A:5,Pr:7,P:1) 8(EN:5,VU:3) Bipedidae 2 2(Pr:2) Colubridae 43 25(Pr:17,A:8) 4(EN:2,VU:2) Dibamidae 1 1(A:1) Dipsadidae 2 Elapidae 4 4(Pr:4) Emydidae 2 1( A:1) 1(EN:1) Gekkonidae 7 6(Pr:5,A:1) 1(NT:1) Iguanidae 6 4(Pr:2,A:2) 1(VU:1) Kinosternidae 2 1(PR:1) Leptotyphlopidae 2 Natricidae 1 Phrynosomatidae 40 13(A:4,P:1,Pr:8) 4(EN:1,VU:2,CR:1 Polychridae 12 10(8:Pr, A:2) 4(EN:2, VU:2) Scincidae 7 6 (Pr:5, A:1) 1(VU:1) Teiidae 7 5(Pr:5) Viperidae 16 14(A:6,Pr:7,P:1) 1(VU:1) Xantusiidae 5 3(Pr:2,A:1) 1(VU:1) Xenosauridae 3 2(Pr:2) * National status: P: Danger to extinction, A: Threatened, Pr: Special protection # IUCN Category: EN: Endanger, VU: Vulnerable, NT: Near Threatened, CR: Critically endangered, DD: Data Deficient. Table 2 2. Number of species with good (>0.70) and poor (<0.69) performance according to the 3 goodness of fit measurements used: Area under the curve d True Sills Statistic (TSS). Performance Index GBM GLM MAXENT RF Good AUC 180 181 185 166 Poor 2 4 0 19 Good TSS 130 129 158 77 Poor 52 56 27 108 Good KAPPA 2 0 6 1 Poor 180 185 179 180

PAGE 37

37 Table 2 3 . Results of the Friedman test for difference s between the performances measurements of the modeling techniques. Model performance measure F(maxT) P value TSS High 13.1674 2.20E 16 TSS Medium 6.6822 1.82E 10 TSS Low 8.5787 2.20E 16 AUC High 13.6734 2.2E 16 AUC Medium 6.037 2.33E 08 AUC Low 9.98 13 2.20E 16 KAPPA High 10.7392 2.20E 16 KAPPA Medium 5.2581 5.70E 07 KAPPA Low 8.5714 2.20E 16

PAGE 38

38 Figure 2 1. Frequency of records by species. Total data set: 183 species and 19.916 records.

PAGE 39

39 Figure 2 2. Boxplot of predicted performances measures (AU C, KAPPA and TSS) by four modelling techniques

PAGE 40

40 Figure 2 3. Boxplot including all groups together and consensus techniques for the models tested with TSS scores, gray line indicate s performance of 0.7.

PAGE 41

41 Figure 2 4. Map of current Richness patterns of Mexican endemic reptiles based on the Mean ensemble model.

PAGE 42

42 Figure 2 5. Richness patterns of endemic reptiles using the GCM CGCM2, the emission scenario A2A and three time periods: 20 20, 2050 and 2080.

PAGE 43

43 Figure 2 6. Richness patterns of endemic reptiles using the GCM CGCM2, the emission scenario B2A and three time periods: 2020, 2050 and 2080.

PAGE 44

44 Figure 2 7. Predicted change in species ric hness according to the CGCM2 GCM under the A2A emission scenario

PAGE 45

45 Figure 2 8. Predicted change in species richness according to the CGCM2 GCM under the B2A emission scenario

PAGE 46

46 Figure 2 9. Map of the protected areas included in the analysis. IUCN categories I and II are classified as Nationa l parks an d Scientific reserves .

PAGE 47

47 Figure 2 10. Analysis of change in s pecies Richness by IUCN categries I and II which corresponds to National parks and areas with strictic protection for the GCM CGCM2 and the A2a emission scenario.

PAGE 48

48 Figure 2 11. Analysis of chang e in species Richness by the IUCN categ or ies I and II for the GCM CGCM2 and the B2a emission scenario.

PAGE 49

49 Figure 2 12. Composite map using as an example the predicted change in richness for the last period evaluated , emission scenario A2a and GCM CGCM2 and the Protected areas system (IUCN Categories). The areas presented in green are the National parks, Categories I and II according to the IUCN classification. National parks included: Gogorrón, el Potosí, Los Mármoles, El Chico, Nevado de Colima, Pico de Tan sitaro, Isurgencia Jose María Morelos, Cerro de Garnica, Mariposa Monarca, Santuario del Agua y Forestal Presa Villa Victoria, Cuenca de los ríos Valle del Bravo Malacatepe Tilostoc Temascaltepec, Nevado de Toluca, Desierto del Carmen o de Nixcongo, Gru tas de Cacahuamilpa, Tenancingo Malinalco Zumpahuacan, Santuario del Agua y Forestal Subcuenca Río San Lorenzo, El Tepozteco,Iztaccihuatl Popocatepetl, Sistema Tetzcotzingo, Malinche o Matlalcueyatl, Santuario de Agua y Foresta l Manantiales Cascada Diama ntes.

PAGE 50

50 Figure 2 13. Change in Mean species richness by National Parks, we catologue as small parks areas with 10Km or less. P represents the time periods use d : P1:2020, P2: 2050 and P3:2080. The National parks included were: T: El Tepozteco, SPSM: Sierra de San Pedro Martir, SAV: Sistema Arrecifal Veracruzano, PT: Pico de Tancitaro, PO: Pico de Orizaba, NT:Nevado de Toluca,MM: Mariposa Monarca, MA:Los Mármoles, M: Malinche o Matlalcueyatl, LCH: Laguna de Chacahua, IP: Iztaccihuatl, H: Huatulco, G: Gogorron , CS: Cañón del Sumidero, CRB: Cañón de Río Blanco, CP: Cofre de Perote, CM: Cumbres de Monterey, CCH: Cobio Chichinautzin, BL: Bahía de Loreto, B: Bosencheve, AX: Arrecifes de Xcalak, APM: Arrecife dePuerto Morelos and AC: Arrecifes de Cozumel.

PAGE 51

51 Figure 2 14. Change in Mean species richness by National Parks, we catologue as big parks zones greater tha 10Km. P represents the time period used: P1 : 2020, P2 : 2050 and P3 :2080. The Nationalparks included were: VE: El Veladero, TU: Tulum, SO: Sierra de Organos , PO: El Potosi, PAL: Palenque, NC: Nevado de Colima, LZ: Lagunas de Zempoala, LM: Lagunas de Montebello, J: El Jabalí, IMHC: Insurgencia Miguel Idalgo y Castilla, IJMM: Insurgencia Jose Maria Morelos, IC: Isla Contoy, GCA: Grutas de Cacahuamilpa, DZ: Dzib ilchantun, DL: Desierto de los Leones, DCN: Desierto del Carmen o de Nixcongo, CP: Cabo Pulmo, COMCN: Costa Occidental de Mujeres, Punta Cancún y Punta Nizuc, CM: Cumbres de Majalca, CH: El Chivo, CG: Cerro de Garnica, CB: Cascada de Bassaseachic, CA: Cumb res de Ajusco, C1987: Constitución de 1857 and BJ: Benito Juarez.

PAGE 52

52 Figure 2 15. Distribution of the dry forest and protected areas in Mexico.

PAGE 53

53 CHAPTER 2 COMMUNITY LEVEL PREDICTION OF CURRENT AND FUTURE COMPOSITION OF ENDEMIC REPTILES IN MEXICO SDMs have been widely used to address different ecological questions; with emphasis on exploring the potential distribution of species under a variety of future climatic scenarios (Guisan & Zimmermann, 2000) . Typically, species are modeled individually (Guisan & Thuiller, 2005) , but it has been suggested that community level models may more accurately assess the impacts of climate c hange on biodiversity (Ferrier & Guisan, 2006; Baselga & Araújo, 2010; Fitzpatrick et al. , 2011) .Community level models use the available data of all species to predict distributional patterns, and thus their output is more useful than richness alone (Ferrier et al. , 2004; Ferrier & Guisan, 2006) . In addition, community level models can have better model performance for sparse datasets (Elith et al. , 2006) such as those of endemic or threatened species (Bonthoux et al. , 2013) and may improve accuracy of prediction, particularly, for rare or poorly sampled species (Ferrier et al. , 2002; Baselga & Araújo, 2009, 2010) . Some studies suggest potential benefits of using community models to predict species response to climate change particularly for temperate species (Baselga & Araújo, 2010) , however the potential benefit for tropical endemic species remains unknown. Understanding current patterns of species dist ribution and how climate change will affect those patterns is an important and fundamental requirement for conservation prioritization (Thessler et al. , 2005; Moilanen et al. , 2009) . Additionally it is important to understand how species composition changes across areas to accurately assess how species ranges change in response to climate change. However, species composition between areas are rarely avai lable and most of the time the information is sparse (McKnight et al. , 2007) . The use of modelling to relate environmental variables with

PAGE 54

54 biological data to explain the biodiversity patterns in specific regions has constituted one of the most important approaches to understand the future of biodiversity (Guisan & Zimmermann, 2000; Ferrier et al. , 2002; Ferrier & Guisan, 2006; Y ates et al. , 2010) . SDMs are relatively well known techniques that are particularly strong when areas are well sampled, so species with thorough field sampling are particularly well predicted (Guisan & Thuiller, 2005) , but in the case of rare or poorly sampled species may be less successful and anoth er approach may be needed (Elith et al., 2006, Bonthoux et al., 2013). Additionally, although species richness (the number of species in an area) is commonly used to characterize specific regions, as well as to understand impacts under environmental chang e, this will not necessary guarantee the protection of important areas with lower richness (Margules & Pressey, 2000; Ferrier et al. , 2004) . Including additional information about patterns of diversity, such as compositional dissimilarity (the difference in species composition between areas) may improve understanding of climate change impacts and thus constitutes a valuable approach in order to preserve ecosystem i ntegrity (Pre ssey et al. , 1993; Thessler et al. , 2005; Ferrier et al. , 2007; Fitzpatrick et al. , 2011) . The Generalized Dissimilarity Model (GDM) is a recently developed statistical approach to analyze and predict community composition al dissimilarity as a function of geographic and environmental distance between sites (Ferrier et al. , 2002, 2007) . This technique , employs a curvilinear relationship between composition and environmental distance instead of the classic linear relationship that most previous community models assumed (Ferrier et al. , 2002) . The benefit of this approach is that areas exhibiting high beta diversity levels ( compositional dissimil arities close to 1 ) are well predicted (Ferrier

PAGE 55

55 et al. , 2002) . Furthermore , GDM is effective when analyzing large datasets and species with low prevalence (Ferr ier et al. , 2007) . T he resulting analysis can identify environmental surrogate s for conservation (Overton et al. , 2009) , which is particularly important in places where biodiversity is high but not well sampled. Mexico is a tropical sub tropical country characterized by highly diverse topography (more than 15 physiographic provinc es) and climate (more than 60 types) (Ochoa Ochoa & Flores Villela, 2006) , supporting a rich and unique biodiversit y across a wide variety of taxonomic groups (Mittermeier & Goettsch, 1992; Aid et al. , 1997) . Globally, Mexico h as the second highest number of reptiles and almost half are endemic (Ochoa Ochoa & Flores Villela, 2006) .There is considerable interest and habitat loss and land use change. Additionally, there is some evidence that reptiles, particularly lizards, may be especially threatened by c limate change (Huey et al. , 2009; Sinervo et al. , 2010; Gadsden et al. , 2012) . In this study, we investigate the use of generalized dissimilarity models to analyze compositional dissimilarity patterns of endemic reptiles, and to predict turnover (beta diversity) under future climatic scenarios in Mexico. We account for geographic and environmental variation and assess which areas are important to protect both currently as well as in the future. Methods Species and Environme ntal D ata Mexico has approximately 804 species of reptiles, of which nearly 50% are endemic (Flores Villela & Canseco Márquez, 2004) . In this study we used a database of endemic reptiles provided by the Museo de Zoología (MZFC) at the Universidad

PAGE 56

56 Nacional Autónoma de Mé xico. We used a 10km by 10 km grid to build a site by species matrix. We eliminated multiple records of the same species in each site and grid cells with just one record (Ferrier et al. , 2007; Fitzpatrick et al. , 2011; Belbin, 201 3) . The final dataset used to perform this analysis included 182 species distributed in 1543 sites. We performed a pairwise correlation analysis and selected an uncorrelated subset (r >0.75) of seven climate variables from 19 bioclimatic variables from t he WorldClim database www.worldclim.org; (Hijmans et al. , 2005) . The final seven variables were: (1) annual mean temperature; (2) Isothermality; (3) the maximum temperature of warmest month; (4) temperature annual range; (5) precipitation of wettest month; (6) precipitation of driest month; and (7) precipitation seasonality (the seasonality is the coefficient of var iation of the monthly means). These seven variables were used as environmental gradients in the GDM analysis. Predicting Beta Diversity P atterns We used GDM to analyze and predict beta diversity patterns of endemic reptiles in Mexico in relation to environ mental gradients and to evaluate the impact of climate change on this pattern. We calculated and developed a spatial prediction of the compositional turnover. Several indices of compositional dissimilarity using either presences or abundance between pairs of are available (Legendre & Legendre, 1998) . We used the Bray Curtis dissimilarity index to generate a pairwise matrix of dissimilarity between all pairs of sites. For each environmental variable, the GDM uses a maximum likelihood estimator and I splines to transform each of the climatic variables and provide the best supported relationship between environment, compositional dissimilarity and

PAGE 57

57 the separation in geographic space. The GDM use s a Generalize linear Model (GLM) to model the Bray Curtis dissimilarity using the formula: (3 1) p is an I spline function to transform the environmental variables to fit the observed dissimilarities to predict the composition where (3 2) Where A is the number of species common to both sites (i and j); B is the number of species present only at site I; and C is the numbe r of species present only at site j. The link function accommodates the curvilinear relationship between the compositional dissimilarity and the predictors (geographical separation and climatic variables). The I splines per variable indicates the amount of compositional dissimilarity explained by each variable alone and also serves as an indicator of the importance in determining the overall beta diversity patterns (Ferrier et al. , 2007; Overton et al. , 2009; Fitzpatrick et al. , 2013) . Once th ese functions were estimated by the model, we then used them to predict patterns of compositional turnover (beta diversity) between any two pairs of sites in the study area. To perform the GDM, we downloaded the package available at < https://sites.google.c om/site/gdmsoftware/ > (release 1.0) in R 3.0.2 (R Core Team, 2013) Our analysis followed the recommendations of Ferrier et al . ( 2007 ) . We used presence only data to produce a site by species matrix with a 10km x 10km re solution. The coordinates of centroids of each of the cells with species were included as the geographical distance variable. This matrix was used to calculate the inter site dissimilarity using the Bray Curtis as the response variable for the GDM. We elim inated

PAGE 58

58 sites with just one or two records because sites with low richness due to the incomplete sampling may erroneously appea r completely dissimilar to one another (Fitzpatrick et al. , 2011) and this will have an important impact on the beta diversity calculation. We recognize that because of our use of endemic reptiles, some areas with low density of records demonstrate their specificity rather than low sampling effort. We assessed the plots of the splines for each predictor to understand the magnitude and rate of species turnover explained by each variable (holding all other variables constant) . To visualize the spatial pattern of current reptile composition we c onverted the GDM model output into a matrix of predicted dissimilarity and performed a principal coordinates analysis (PCoA). To visualize the spatial pattern, the first three PCoA axes were imported into ArcMap, transformed to a raster grid and projected as color bands (red green blue) to build the dissimilarity map (Ferrier et al. , 2007) . In the resulting map, the differences in turnover pattern are differentiated by color s; areas with similar colors have similar species composition and areas with dissimilar colors are predicted to have different composition. Climate C hange Assessment U sing GDM To predict future compositional turnover we used the same set of seven climate variables derived from the Coupled Global Climate Model (CGCM2) for the A2a emissions scenario. The CGCM2 model has been widely used to model future species distributions (Thuiller et al. , 2006) . The A2a emission scenario predicts a temp erature (IPCC, 2007) . We developed predictions for the years 2020, 2050 and 2080. The climate data was obtained from the global climate model data portal ( http://www.ccafs climate.org ) and was downscaled and centered on the current Worldclim climate (Ramirez & Jarvis, 2008) .

PAGE 59

59 To measure the effect of climate change on endemic reptile composition we applied the new climate data to our existing parameterized GDM model from the previous step. Here, we calculate how community composition is predicted to change in time giv en the amount of environmental separation between the current climate and the climate in 2020, 2050 and 2080 for 5500 randomly selected grid cells. In this case, we calculated predicted compositional dissimilarity not in space but in time by calculating th e predicted values from the comparison between current environmental condition in Time1 and future environmental condition in Time 2. This approach allowed us to composition. The GDM model assumes that forecasted changes in composition will occur regardless of the distance species would have to travel (full dispersal scenario) (Fitzpatr ick et al. , 2011) . We then developed maps of the predicted change in compositional dissimilarity for each of the three time periods, using the predicted dissimilarity values in each cell. The protected areas information was obtained from the Commission fo r Environmental Cooperation web site ( http://databasin.org/ ). This database includes polygons for areas with different types of protection that are managed by the federal, national, provincial, or territorial authoriti es . Within the database, areas are classified by categories from I to IV, following the International Union for the Conservation of Nature ( IUCN) (Dudley, 2008) . In order to assess the potential im pact of climate change on composition inside protected areas, we overlaid the National Protected A reas S ystems map onto the GDM map of the predicted dissimilarity for the three time periods, 2020, 2050 and 2080.

PAGE 60

60 All statistical analyses including the GDM m odel parameterization was done using the R environment , version 3.0.2 (R Core Team, 2013) and the visualization was performed using ArcGIS 10.2 (ESRI, 2013) . Results Our community data consisted of 1543 10km by 10 km sites (Fig ure 3 1), with 182 species, representing almost 50% of the endemic reptiles present in Mexico. Across species, the mean number of occurrences per 10 km grid cell was 4 , with a maximum of 27 . There were 631 sites where only two species were found. The entire study area had 83 , 220 grid cells. The plot of obser ved compositional dissimilarity and the predicted ecological distance identify the fit between observed and predicted dissimilarity. This result reveals a nonlinear fit of dissimilarities to the environment and the spatial distance (Fig ure 3 2 ). The GDM ap proach explain ed 44.5% of the variance in observed overall beta diversity . Sixth of the eight explanatory variables used to fit the model were finally selected by the GDM. Both environmental and geographical distance played an important role in predicting composition (Fig ure 3 3 ). The most important variables contributing to the overall patterns of current beta diversity of endemic reptiles in Mexico were the geographic distance (2.26: 47%) with the highest contribution followed by the Isothermality (1.58; 33%). The remained variables; temperature annual range, precipitation of the wettest month and the annual mean temperature were selected by the GDM, but, their contributions were low (Fig ure 3 3 ). The precipitation of driest month and the precipitation sea sonality were discarded in the final model due to their insignificant contribution to the overall beta diversity observed.

PAGE 61

61 Fig ure 3 4 shows the environmental variables chosen by the GDM model, and their relative importance in determining reptile species c omposition. The shape of each function explains how the rate of compositional turnover varies along the gradient (Fig ure 3 4 A to Figure 3 4 E ). The major variation in terms of beta diversity is presented at higher Isothermality values (Fig ure 3 4 A ). For geo graphic distance (Fig ure 3 4 ), sites near each other can be differentiated in terms of beta diversity, quite rapidly initially, and with a continuous but less pronounced differentiation as the distance increases between points. In general, for both variabl es, the slope is very strong, increasing with higher values. For the temperature annual range (Fig ure 3 4 C ), the change in beta diversity is not stronger along the gradient but from colder to hotter environments the turnover is rapid at the beginning and a fter certain temperature value the rate is maintained. Finally, for the other two variables selected by the model (Annual mean temperature and precipitation of the wettest month), the rate of change in beta diversity is not influenced at all by this gradie nts (Fig ure 3 4 D to Figure 3 4 E ). We selected the first three axes of our Principle Coordinates Analysis, together summarized 79% of the cumulative variation (Table 3 1 ). The first axis explained 33% followed by the second axis which explained 29% and t he third one explaining the 16%. We used these three axes to represent beta diversity using a composite map (Fig ure 3 5 ). The classification of communities showed a mixed pattern of community composition (Table D 5 ). Nevertheless, it is possible to observe a differential pattern between the tropical and temperate realms and a strong differential pattern in the Yucatan peninsula. Based on the model built under the current environmental conditions, we then predicted how community composition will change throu gh time in response to climate

PAGE 62

62 change (Time1 vs. Time2) for three time periods (2020, 2050 and 2080). The tendency for the first two periods (2020 and 2050) was similar. Central and Southern Mexico is predicted to experience major changes in terms of compo sition, with a greater amount of change predicted for the 2050 period (Figure 3 6 ). According to this predicted change for 2020 and 2050, the most affected biogeographic regions are the Gulf of Mexico, the Sierra Madre Oriental, the Altiplano Sur (Zacateca no Potosino), and the Trans Mexican Volcanic Belt (TMVB). The Southern most state of Chiapas is predicted to experience the greatest change among periods, including the 2080 prediction, where no major changes were calculated. According to the predicted cha nge in species turnover, for 2020, a high percentage of natural protected areas are predicted to be affected the most. The central and southern region is predicted to experience major changes compared to other regions (Fig ure 3 7 ). The Yucatan peninsula, w hich is predicted to change the most among three periods, according to the protected areas system, is covered in great percentage by protected areas of different types, enhancing the importance of connectivity among them in order to guarantee movement by s pecies that needs to follow specific thermal ranges. Consistent with the last period evaluated, the greatest predicted change in composition is predicted to occur in the area of the Yucatan peninsula, Chiapas, Tabasco, Campeche and Quintana Roo. Although t he protected area system is considerable (Fig ure 3 8 ), compared with other areas in Mexico, according to the IUCN classification of those areas, most of them are managed resources protected areas and

PAGE 63

63 areas where the real main use is unknown, in that way fu rther information about the mainly use is needed to propose more strict protection in case it was needed. Discussion In general the GDM approach provided ecological insights into the important drivers in the community characteristics of turnover of endemic reptiles in Mexico. The model suggested that those patterns were highly influenced by the geographical distance; therefore, the biogeographic gradient presented along Mexico is playing an important role in community composition. In terms of the climate, a measure of temperature (Isothermality) is the most influential, and is high in relatively warm environments. This pattern is expected for reptiles due to characteristic features of this group like ecthotermy, reproductive strategies, local abundance, ener getics and some more, which make them highly dependent on temperature (Lillywhite, 2013) . The GDM results attempt to explain the drivers of turnover (Fitzpatrick et al. , 2011) ., in t his case, temperature and geographic distance were the primary factors driving the patterns of compositional change. According to the current patterns of beta diversity (Fig ure 3 5 ), Mexican reptiles demonstrates a non aggregated pattern of community comp osition, reflecting the environmental heterogeneity among the different landscape units, with a slightly pattern between the Temperate and tropical realm and among the different biogeographic zones. This distribution can be explaining because Mexico contai ns both, montane and desert habitats that are distributed throughout Mexican territory and can be quite different in terms of species composition. Elevational gradients and biogeographic regions such as the Sierra Madre Oriental, Sierra Madre Occidental, S ierra Madre del Sur and the TMVB play an important role in the turnover patterns due to the restriction

PAGE 64

64 of some species to specific or restricted altitudinal ranges. This differentiation across the ecological gradients leads to isolation and speciation (Smith, 1997; Thomassen et al . , 2011) and in this case, endemic reptiles showed a high beta diversity variation along the different areas. Additionally, previous studies about patterns of alpha, beta and gamma diversity reveal similar spatial patterns of beta diversity (García et al. , 2007) . Ours is the first study to explicitly test a wide range of environmental factors to explain those patterns in Mex ico. Climate Change According to previous research about the effect of climate change on vertebrates (Araújo & Pearson, 2005) specially on amphibians and reptiles (Chen et al. , 2011) , species turnover of repti les was better explained by the differences in environmental variables, especially temperature, suggesting that those species will be least capa ble to track environmental change and shift their distribution under climate change, making them one of the most vulnerable groups given their strong reliance on temperature . These results are similar to those of Sinervo et al. (2010) which used different methods, but used data mostly from Mexico and concluded approximately 20% of lizard species worldwide will go extinct by 2080. According to our analysis of future change in composition based on climate, it was possible to detect the change in composition f or endemic reptiles among periods . In Mexico, most of the southern populations are predicted to have higher turnover due to climate change and these populations will experience the greatest change in composition between 2020 and 2050 (Fig ure 3 6 ), with a c onsistent change among periods in the tropical realm where reptile richness is higher, see Chapter 1 (Ochoa Ochoa & Flores Villela, 2006; Koleff et al. , 2008) .

PAGE 65

65 A similar study, using endemic Mexican amphibians, predicted that the change in beta diversity was not concentrated in any one region (Ochoa Ochoa et al 2012). Major changes in amph ibian composition for 2020 were predicted in a small region in the Yucatan peninsula, a small portion in the southern part of the Sierra Madre Occidental and the Northeast of Mexico. For 2050 period, they predicted strong changes in the southern middle and Northwest Mexico, but in general was predicted change in almost all the country. In contrast, our approach predicted for 2020 and 2050, a strong change in most of the Yucatan peninsula and the Gulf of Mexico, covering most of the wet tropical area and a s mall part of the temperate and arid zone in the central Mexico. Additionally, according to another study of ant community patterns in North and Central America using the same methodological approach (Fitzpatrick et al. , 2011) , the impact in terms of species composition under climate change was predicted to be particularly high in the Yucatan Peninsula, and in general for the tropical south east area, supporting ou r results. Consequently, the differences in predictions in terms of future beta diversity of Mexican species, demonstrate, the importance of considering different taxonomic groups to prioritize areas for long term conservation. Protected Areas ional Protected Area Commission (CONANP) is currently guiding the Mexican government to strengthen the conservation actions in protected area system (Comisión Nacional de Áreas Naturales Protegidas, 2010) . To more effectively protec t biodiversity in the face of climate change, it is necessary both, to establish protected areas as well as provid e connectivity am ong them (Ochoa Ochoa et al. , 2009; Comisión Nacional de Áreas Naturales Protegidas, 2010) . Even th ough we predict ed change in species composition usi ng climate alone, we consider this is the first step in

PAGE 66

66 order to lay a baseline to think about the most susceptible areas in terms of composition. Because land use is an additional concern and will influence the real availability of area to move, taking in to account this is the next step. According to our results most of the predicted change in composition of endemic reptiles is aggregated in one major area, so connectivity results particularly important because this will allow species to follow their therm al tolerance via corridors between protected areas. According to the composite map using the turnover for 2080 and current PA system (Fig ure 3 7 ) , almost half of the current protected area system is predicted to experience significant turnover . For the las t period evaluated, even the change is not as widespread as previous periods, Southern Mexico, which is characterized because the huge diversity is predicted to change the most. This turnover provide a significant challenge for the protected areas system a nd offer important basic knowledge to be included as part of the mitigation and adaptation strategies that currently Mexican government through the strategy of climate change for protected areas, Estrategia de cambio climático para á reas protegidas (ECCAP) is implemented. Recommendation and Future A pplication In order to compare results obtained from the individual models in chapter 1, we used the same set of data which were previously filter ed with a minimum of ten records (necessary to develop robust spe cies distribution model) . However, since GDM is a community approach it would be interesting to include more rare species in the analysis to increase the representativeness of endemic reptiles in Mexico. Therefore, including species with few records so wil l complement sites with few records and we will be able to exclude sites where fewer than 3 species had been recorded to account for the influences of incomplete sampling.

PAGE 67

67 The environmental variables selected by the GDM explained 44% of the variance in com munity composition. Th e inclusion of additional environmental, ecological or land cover data available for Mexico may explain additional variation and will produce a more robust prediction of community composition . In any case, compositional dissimilarity plays a much more substantial role in determining representativeness for conservation prioritization than would richness variation alone (Ferrier et al. , 2002) . Community dissimilarity of endemic reptiles can provide useful information for pa tterns of other taxa; however we suggest the inclusion of different taxon groups as proposed by Fitzpatr ic ik et al. 2013, to increase the utility of this method towards conservation prioritization in Mexico. We recommend a multi taxa approach to formulate a national conservation strategy in the face of climate change. S electing groups at higher taxonomic levels to get better coverage has been used in different groups in Australia (Belbin, 2013) . Additionally , w e suggest a comparison of results when higher taxonomic levels are includ ed in order to determine how much additional power is gained, if any. Additionally , using additional types of dissimilarity measure s other than Bray Curtis, such as those that inco rporate species replacement and species richness differences in beta diversity will improve our understanding of turnover patterns (Carvalho et al. , 2012) . This may be particularly useful for endemic reptiles, which generally ha ve restricted and aggregated distributions. As we suggested using ric hness alone in the first chapter, combining this analysis with land use change in Mexico to look vulnerability as a result of climate change will complement our results and will provide a better idea for managers to invest resources in the most vulnerable areas due to both factors.

PAGE 68

68 Compositional turnover has a direct relationship with endemism, therefore stu dy ing those patterns for endemic reptiles, as we did it here, is an important starting (Ferrier et al. , 2004) . Our analysis and others like it provide more robust information to prioritize areas for urgent conserv ation not only under climate change but also under other environmental stressors (Margules & P ressey, 2000) . Despite the widespread use of single SDM, by using community models like GDM, we provide a complementary approach using environmental and species records to predict composition which is one of the most useful biodiversity measurements for c onservation prioritization, particularly when highly contrasting environments are present, and when endemicity is the major characteristic of the studying group. Finally, recognizing the vulnerability of different species and different taxonomic groups und er change in climatic conditions will contribute with the objectives of the Mexican government. We recognize that using climate alone will predict certain changes that will be exacerbated according to the current and future use of the land on those areas, but providing this information is important to understand the vulnerability under different time periods according to an increase in CO2 emissions and certain increment in temperature. Therefore this kind of approach deserves to be a widely and more often considered technique to complement modeling individual species and land use effects.

PAGE 69

69 Table 3 1 . Principal Coordinates Analysis using the predicted dissimilarity values to build current beta diversity map. Axis Eigenvalues Proportion of variance Cumulative Variance 1 109.536 0.335 0.335 2 94.639 0.290 0.625 3 53.596 0.164 0.789

PAGE 70

70 Figure 3 1. Location of the 1543 sites included in the analysis.

PAGE 71

71 Figure 3 2 . Observed response based on the overall beta diversity. The curved line in black represents th e non linear link function where environmental and spatial distances relate the dissimilarities and the dots represents a site pair.

PAGE 72

72 Figure 3 3 . Relative contribution of environmental variables and geographic al distance (Sum of coefficients of the I spli nes) according to the current species composition. The most important predictors are showing first. Environmental variables included in the analysis: Bio1: annual mean temperature; Bio3: Isothermality (Mean diurnal range divided by the annual temperature r ange); Bio 5: the maximum temperature of warmest month, Bio7: temperature annual range; Bio13: precipitation of wettest month and Geographic or geographical distance).

PAGE 73

73 A B C D E

PAGE 74

74 Figure 3 4 . Fitted functions of observed overall beta diversity. The maximum height reached by each function provides the amount of beta diversity along the gradient (holding all other variables constant). The x axes are presented according to each variable unit and the y axes are presented in the units of the additive community d issimilarity (exponential link function calculated by the GDM). Figure 3 5 . Beta diversity map using the ordination axes obtained from the predicted dissimilarity matrix. Comp 1 represents the first axe, Comp 2 represents the second axe and Comp 3 repr esents the third axe.

PAGE 75

75 Figure 3 6 . Compositional turnover forecasted among the three time periods evaluated 2020, 2050 and 2080.

PAGE 76

76 Figure 3 7 . Composite map of protected areas in Mexico using projected change in species turnover for the 2020 period. T he current protected areas in Mexico were delimited with solid black lines. Areas in red are predicted to change the most in terms of turnover.

PAGE 77

77 Figure 3 8 . Composite map of protected areas in Mexico, classified following the IUCN, using projected cha nge in species composition according to the last period evaluate 2080.

PAGE 78

78 CHAPTER 4 CONCLUSIONS In this research we revealed the importance to take into account species richness and species composition to fully apprehen d the potential effects of cli mate ch ange on endemic reptiles due to both provide different kind of information and together can be used to provide a framework for optimal design of protected areas. Single distribution models provided individual responses in case specific species concerns are needed or in case richness alone justifies protection of certain areas . Additionally, beta diversity will help to characterize certain areas to protect specific ecosystems or regions because no similar representation is presented in other places. From a c onservation perspective, the future effect of climate change in richness and turnover reveal further challenges for the current protected areas system in Mexico and propose a revision based not only in areas with high richness but also in areas where the p redicted change is going to be the most in terms of composition. Finally, we recognize that reptiles is just part of a big picture due to the huge diversity present in Mexico, but, because of their characteristic feature like ectothermy, their declining an d extinction previously documented (Sinervo et al 2010) require especial concern. Consequently, understanding their response will enhance our knowledge about future potential response of them and other species, mainly because latitudinal response sometimes is not the obvious process that species use to handle climate change. We recognize the importance to include land use transformation in order to accurately predicts the real future of this important group, but this first step is definitely important in or der to detect potential areas of research. Additionally, given the high levels of habitat fragmentation already experienced in Mexico disentangle

PAGE 79

79 community composition and richness under change is a key step to support conservation and management of differ ent areas.

PAGE 80

80 APPENDIX A ACCURACY OF EACH INDIVIDUAL MODEL BY SPECIE Table A 1 . each model technique. G eneralised Boosting Model (GBM), Generalised Line ar Model (GLM), Maximum Entropy (MXT) and Breiman and Cutler's random forest for classification and regression (RF). Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXENT RF Abronia deppei 0.902 0.833 0.965 0.831 0.817 0.741 0.907 0.664 0.446 0.105 0.514 0.602 Abronia graminea 0.895 0.826 0.948 0.899 0.812 0.682 0.836 0.754 0.064 0.034 0.175 0.309 Abronia lythrochila 0.991 0.961 0.99 0.849 0.977 0.934 0.971 0.7 0.306 0.14 0.317 0.347 Abronia martindelcampoi 0.908 0.912 0.969 0.80 3 0.829 0.846 0.912 0.604 0.429 0.129 0.334 0.371 Abronia mixteca 0.926 0.931 0.975 0.69 0.85 0.879 0.946 0.386 0.371 0.147 0.268 0.331 Abronia oaxacae 0.959 0.948 0.951 0.77 0.904 0.898 0.917 0.541 0.069 0.088 0.19 0.221 Abronia smithi 0.902 0.992 0.99 4 0.858 0.813 0.987 0.99 0.714 0.213 0.288 0.356 0.426 Abronia taeniata 0.904 0.807 0.937 0.802 0.766 0.646 0.796 0.58 0.23 0.048 0.294 0.29 Adelphicos latifasciatum 0.859 0.854 0.894 0.697 0.734 0.749 0.734 0.403 0.254 0.046 0.108 0.165 Adelphicos nigr ilatum 0.962 0.962 0.983 0.968 0.908 0.929 0.936 0.934 0.659 0.509 0.738 0.735 Agkistrodon taylori 0.91 0 0.941 0.984 0.831 0.825 0.895 0.973 0.665 0.2 0.095 0.181 0.466 Anelytropsis papillosus 0.864 0.808 0.951 0.775 0.739 0.615 0.876 0.536 0.103 0.021 0 .132 0.092 Anolis anisolepis 0.84 0.891 0.895 0.725 0.693 0.787 0.782 0.453 0.342 0.152 0.339 0.303 Anolis barkeri 0.976 0.955 0.988 0.934 0.946 0.907 0.972 0.858 0.314 0.407 0.352 0.45 Anolis dunni 0.943 0.915 0.932 0.746 0.897 0.869 0.881 0.499 0.236 0.051 0.05 0.413 Anolis hobartsmithi 0.959 0.973 0.977 0.975 0.874 0.895 0.922 0.91 0.5 0.522 0.634 0.665 Anolis liogaster 0.971 0.969 0.991 0.897 0.915 0.938 0.964 0.785 0.495 0.538 0.506 0.379 Anolis matudai 0.902 0.859 0.978 0.804 0.813 0.719 0.956 0 .609 0.266 0.371 0.401 0.543 Anolis microlepidotus 0.642 0.684 0.736 0.663 0.414 0.524 0.645 0.331 0.031 0.013 0.017 0.209 Anolis naufragus 0.979 0.938 0.983 0.88 0.895 0.88 0.948 0.753 0.286 0.389 0.381 0.364 Anolis nebuloides 0.872 0.862 0.868 0.795 0 .714 0.732 0.712 0.538 0.054 0.061 0.113 0.252 Anolis pygmaeus 0.93 0.974 0.991 0.851 0.866 0.951 0.969 0.702 0.535 0.381 0.51 0.532

PAGE 81

81 Table A 1. Continued Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXENT RF Anolis subocular is 0.879 0.862 0.899 0.799 0.733 0.721 0.739 0.581 0.301 0.269 0.358 0.321 Aspidoscelis calidipes 0.918 0.946 0.938 0.889 0.842 0.89 0.847 0.764 0.389 0.384 0.453 0.448 Aspidoscelis communis 0.84 0.861 0.878 0.803 0.604 0.653 0.664 0.461 0.208 0.195 0.27 7 0.26 Aspidoscelis costata 0.863 0.873 0.894 0.903 0.635 0.645 0.681 0.646 0.317 0.314 0.379 0.389 Aspidoscelis guttata 0.901 0.921 0.927 0.937 0.705 0.72 0.743 0.757 0.396 0.442 0.496 0.556 Aspidoscelis lineattissima 0.934 0.956 0.961 0.927 0.801 0.81 5 0.845 0.809 0.307 0.41 0.431 0.416 Aspidoscelis opatae 0.912 0.87 0.98 0.791 0.83 0.745 0.939 0.583 0.675 0.387 0.714 0.577 Aspidoscelis parvisocia 0.912 0.813 0.957 0.854 0.731 0.625 0.835 0.696 0.38 0.142 0.427 0.409 Atropoides nummifer 0.937 0.92 0 .97 0.842 0.876 0.844 0.93 0.682 0.243 0.192 0.206 0.299 Barisia ciliaris 0.887 0.926 0.918 0.875 0.626 0.743 0.689 0.703 0.273 0.278 0.384 0.435 Barisia imbricata 0.883 0.887 0.893 0.89 0.654 0.661 0.681 0.591 0.285 0.312 0.321 0.347 Barisia jonesi 0.9 42 0.906 1 0.944 0.88 0.838 1 0.888 0.768 0.648 0.887 0.804 Barisia levicollis 0.896 0.89 0.989 0.808 0.784 0.782 0.972 0.616 0.301 0.366 0.381 0.393 Barisia planifrons 0.944 0.917 0.963 0.887 0.843 0.796 0.861 0.75 0.18 0.122 0.366 0.276 Barisia rudico llis 0.916 0.841 0.954 0.817 0.847 0.723 0.919 0.627 0.056 0.018 0.164 0.165 Bipes biporus 0.933 0.906 0.987 0.856 0.871 0.818 0.978 0.711 0.22 0.171 0.232 0.252 Bipes canaliculatus 0.866 0.898 0.936 0.86 0.749 0.804 0.851 0.72 0.435 0.045 0.104 0.599 C errophidion tzotzilorum 0.981 0.976 0.979 0.964 0.961 0.955 0.963 0.928 0.504 0.349 0.553 0.48 Chersodromus liebmani 0.828 0.842 0.969 0.74 0.647 0.689 0.846 0.482 0.238 0.094 0.311 0.251 Conophis vittatus 0.84 0.828 0.849 0.795 0.597 0.582 0.634 0.53 0. 159 0.177 0.189 0.272 Conopsis biserialis 0.864 0.886 0.888 0.848 0.64 0.673 0.658 0.576 0.222 0.227 0.278 0.345 Conopsis megalodon 0.805 0.78 0.847 0.688 0.599 0.535 0.675 0.36 0.199 0.029 0.182 0.233 Conopsis nasus 0.847 0.892 0.897 0.885 0.569 0.673 0.678 0.616 0.234 0.275 0.295 0.347 Crotalus aquilus 0.915 0.743 0.945 0.702 0.755 0.492 0.849 0.392 0.204 0.026 0.222 0.257 Crotalus basiliscus 0.822 0.819 0.861 0.733 0.572 0.557 0.63 0.443 0.121 0.066 0.14 0.196 Crotalus enyo 0.954 0.979 0.98 0.8 0.9 15 0.964 0.965 0.602 0.128 0.145 0.187 0.235

PAGE 82

82 Table A 1. Continued Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXENT RF Crotalus intermedius 0.856 0.841 0.906 0.688 0.681 0.638 0.752 0.364 0.104 0.061 0.105 0.185 Crotalus po lystictus 0.823 0.804 0.905 0.73 0.604 0.614 0.799 0.458 0.056 0.016 0.066 0.083 Crotalus ravus 0.877 0.884 0.912 0.861 0.725 0.734 0.748 0.675 0.116 0.095 0.192 0.228 Crotalus totonacus 0.921 0.935 0.964 0.904 0.823 0.843 0.878 0.768 0.191 0.26 0.294 0. 274 Crotalus transversus 0.925 0.963 0.976 0.718 0.856 0.931 0.952 0.443 0.243 0.071 0.184 0.466 Crotalus triseriatus 0.852 0.868 0.88 0.825 0.589 0.624 0.656 0.536 0.215 0.228 0.246 0.237 Ctenosaura acanthura 0.865 0.866 0.882 0.824 0.681 0.678 0.704 0 .589 0.194 0.241 0.267 0.219 Ctenosaura clarki 0.916 0.913 0.904 0.796 0.704 0.757 0.733 0.58 0.451 0.351 0.424 0.531 Ctenosaura hemilopha 0.89 0.977 0.995 0.962 0.685 0.954 0.99 0.921 0.37 0.369 0.44 0.404 Ctenosaura macrolopha 0.906 0.888 0.944 0.821 0.769 0.762 0.825 0.619 0.29 0.27 0.34 0.329 Ctenosaura oaxacana 0.934 0.935 0.965 0.861 0.856 0.852 0.918 0.705 0.329 0.219 0.296 0.276 Ctenosaura pectinata 0.854 0.866 0.872 0.867 0.631 0.646 0.646 0.605 0.196 0.209 0.225 0.253 Diploglossus enneagramm us 0.89 0.937 0.945 0.904 0.795 0.865 0.786 0.791 0.256 0.176 0.278 0.351 Elgaria paucicarinata 0.986 0.962 0.986 0.773 0.974 0.929 0.927 0.549 0.394 0.228 0.505 0.418 Ficimia olivacea 0.858 0.815 0.907 0.8 0.698 0.646 0.777 0.575 0.074 0.057 0.172 0.348 Geagras redimitus 0.893 0.881 0.937 0.785 0.799 0.789 0.885 0.57 0.296 0.098 0.263 0.252 Geophis dubius 0.662 0.685 0.783 0.536 0.413 0.443 0.676 0.083 0.039 0.008 0.01 0.111 Geophis latifrontalis 0.841 0.748 0.845 0.754 0.617 0.574 0.65 0.471 0.187 0. 019 0.254 0.303 Geophis semidoliatus 0.931 0.935 0.941 0.897 0.828 0.842 0.855 0.736 0.429 0.445 0.534 0.53 Gerrhonotus ophiurus 0.902 0.902 0.911 0.868 0.746 0.79 0.781 0.692 0.145 0.194 0.284 0.41 Gopherus flavomarginatus 0.867 0.986 0.98 0.831 0.741 0.984 0.968 0.664 0.218 0.169 0.091 0.525 Imantodes tenuissimus 0.913 0.857 0.997 0.942 0.827 0.717 0.995 0.885 0.361 0.309 0.448 0.456 Kinosternon creaseri 0.969 0.892 0.974 0.93 0.93 0.802 0.927 0.858 0.308 0.299 0.448 0.522 Kinosternon herrerai 0.905 0.932 0.922 0.901 0.792 0.836 0.855 0.774 0.18 0.121 0.229 0.296 Lampropeltis mexicana 0.855 0.806 0.837 0.572 0.715 0.582 0.642 0.157 0.078 0.021 0.112 0.022 Lepidophyma gaigeae 0.912 0.956 0.951 0.911 0.765 0.838 0.824 0.771 0.301 0.387 0.464 0.49 Le pidophyma occulor 0.875 0.796 0.892 0.775 0.767 0.633 0.801 0.55 0.139 0.077 0.147 0.413

PAGE 83

83 Table A 1. Continued Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXENT RF Lepidophyma sylvaticum 0.937 0.946 0.973 0.824 0.823 0.871 0. 888 0.632 0.231 0.328 0.403 0.453 Lepidophyma tuxtlae 0.965 0.97 0.992 0.939 0.919 0.938 0.931 0.853 0.607 0.567 0.678 0.66 Leptodeira maculata 0.854 0.867 0.872 0.832 0.649 0.681 0.658 0.56 0.133 0.148 0.174 0.184 Leptodeira punctata 0.923 0.901 0.994 0.896 0.853 0.829 0.981 0.784 0.431 0.288 0.503 0.419 Leptodeira splendida 0.839 0.644 0.836 0.742 0.626 0.373 0.617 0.438 0.138 0.015 0.091 0.295 Leptotyphlops maximus 0.886 0.795 0.932 0.871 0.711 0.588 0.836 0.709 0.061 0.041 0.137 0.232 Leptotyphlop s myopicus 0.932 0.952 0.963 0.867 0.791 0.847 0.863 0.707 0.286 0.277 0.341 0.441 Manolepis putnami 0.905 0.901 0.916 0.828 0.735 0.743 0.738 0.628 0.183 0.155 0.221 0.248 Mesaspis gadovii 0.906 0.928 0.915 0.877 0.717 0.819 0.765 0.723 0.52 0.322 0.56 0.499 Mesaspis juarezi 0.765 0.862 0.863 0.789 0.554 0.757 0.791 0.58 0.112 0.045 0.015 0.464 Mesaspis viridiflava 0.879 0.94 0.95 0.872 0.701 0.83 0.833 0.723 0.161 0.248 0.38 0.374 Micrurus distans 0.862 0.85 0.859 0.75 0.727 0.729 0.709 0.503 0.212 0 .105 0.277 0.300 Micrurus ephippifer NA 0.851 0.93 0.791 NA 0.7 0.833 0.563 NA 0.083 0.254 0.197 Micrurus laticollaris 0.856 0.784 0.898 0.72 0.688 0.611 0.797 0.442 0.175 0.016 0.184 0.178 Micrurus limbatus 0.998 0.913 0.998 0.915 0.998 0.829 0.998 0.8 32 0.633 0.325 0.577 0.489 Ophryacus melanurum 0.929 0.957 0.982 0.883 0.857 0.923 0.97 0.764 0.266 0.156 0.222 0.267 Ophryacus undulatus 0.907 0.815 0.887 0.692 0.779 0.652 0.72 0.384 0.09 0.107 0.123 0.141 Petrosaurus thalassinus 0.881 0.985 0.995 0.9 12 0.77 0.971 0.991 0.824 0.422 0.172 0.388 0.261 Phrynosoma braconnieri 0.957 0.914 0.96 0.864 0.903 0.825 0.87 0.727 0.222 0.128 0.321 0.124 Phrynosoma coronatum 0.965 0.97 0.977 0.931 0.904 0.934 0.937 0.841 0.346 0.449 0.398 0.442 Phrynosoma taurus 0.893 0.833 0.913 0.807 0.776 0.658 0.845 0.598 0.143 0.069 0.193 0.439 Phyllodactylus davisi 0.947 0.944 0.977 0.887 0.908 0.906 0.952 0.766 0.092 0.09 0.175 0.278 Phyllodactylus duellmani 0.846 0.94 0.939 0.874 0.747 0.886 0.907 0.748 0.463 0.147 0.343 0.633 Phyllodactylus homolepidurus 0.938 0.932 0.968 0.947 0.876 0.865 0.942 0.885 0.322 0.382 0.558 0.556 Phyllodactylus lanei 0.906 0.909 0.917 0.856 0.73 0.746 0.72 0.633 0.217 0.182 0.234 0.256 Phyllodactylus muralis 0.927 0.933 0.957 0.873 0.756 0 .833 0.82 0.729 0.42 0.341 0.457 0.377 Phyllodactylus unctus 0.987 0.994 0.995 0.996 0.973 0.989 0.99 0.984 0.372 0.499 0.555 0.553

PAGE 84

84 Table A 1. Continued Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXENT RF Phyllodactylus xa nti 0.937 0.986 0.972 0.977 0.875 0.969 0.945 0.937 0.486 0.472 0.532 0.564 Plestiodon copei 0.93 0.94 0.95 0.899 0.825 0.854 0.84 0.757 0.274 0.293 0.355 0.34 Plestiodon dugesi 0.858 0.875 0.931 0.779 0.703 0.695 0.82 0.548 0.199 0.091 0.18 0.178 Plest iodon ochoterenae 0.873 0.904 0.901 0.9 0.617 0.688 0.674 0.695 0.368 0.339 0.394 0.389 Plestiodon parvulus 0.948 0.892 0.958 0.848 0.859 0.752 0.829 0.69 0.23 0.044 0.338 0.445 Plestiodon lynxe 0.775 0.754 0.815 0.768 0.594 0.526 0.648 0.536 0.122 0.043 0.153 0.342 Pliocercus bicolor 0.885 0.888 0.909 0.862 0.78 0.748 0.8 0.712 0.168 0.104 0.262 0.479 Porthidium dunni 0.912 0.927 0.941 0.859 0.737 0.819 0.807 0.705 0.292 0.188 0.315 0.319 Porthidium yucatanicum 0.914 0.832 0.999 0.999 0.828 0.665 0.99 8 0.998 0.269 0.408 0.588 0.518 Pseudoleptodeira latifasciata 0.882 0.856 0.907 0.642 0.79 0.74 0.786 0.288 0.119 0.069 0.086 0.024 Rhadinaea forbesi 0.818 0.978 0.977 0.788 0.656 0.966 0.966 0.578 0.154 0.083 0.144 0.258 Rhadinaea fulvivittis 0.817 0.8 28 0.871 0.779 0.615 0.62 0.700 0.513 0.145 0.116 0.141 0.212 Rhadinaea gaigeae 0.916 0.827 0.938 0.801 0.744 0.653 0.819 0.568 0.321 0.067 0.374 0.444 Rhadinaea hesperia 0.857 0.808 0.88 0.586 0.676 0.625 0.712 0.191 0.127 0.026 0.097 0.117 Rhadinaea l aureata 0.876 0.883 0.921 0.796 0.709 0.724 0.812 0.551 0.278 0.078 0.222 0.212 Rhadinaea omiltemana 0.851 0.869 0.938 0.832 0.733 0.822 0.880 0.664 0.399 0.011 0.192 0.457 Rhadinaea taeniata 0.834 0.735 0.83 0.584 0.62 0.456 0.655 0.178 0.101 0.033 0.13 5 0.077 Salvadora intermedia 0.807 0.687 0.844 0.737 0.617 0.488 0.721 0.462 0.221 0.013 0.144 0.103 Salvadora lemniscata 0.917 0.944 0.961 0.87 0.817 0.857 0.862 0.721 0.17 0.265 0.343 0.307 Salvadora mexicana 0.870 0.891 0.902 0.810 0.704 0.714 0.722 0.591 0.102 0.15 0.156 0.148 Sceloporus adleri 0.971 0.944 0.988 0.874 0.922 0.892 0.944 0.746 0.521 0.22 0.615 0.553 Sceloporus aeneus 0.867 0.879 0.889 0.87 0.615 0.652 0.656 0.598 0.306 0.311 0.336 0.366 Sceloporus anahuacus 0.938 0.951 0.968 0.89 0. 859 0.897 0.894 0.766 0.289 0.179 0.253 0.311 Sceloporus asper 0.832 0.810 0.810 0.745 0.698 0.706 0.751 0.493 0.112 0.017 0.026 0.148 Sceloporus bicanthalis 0.919 0.936 0.936 0.897 0.780 0.823 0.798 0.729 0.234 0.305 0.327 0.375 Sceloporus bulleri 0.82 0.826 0.835 0.771 0.619 0.597 0.587 0.511 0.291 0.083 0.253 0.173 Sceloporus cautus 0.924 0.891 0.913 0.82 0.782 0.756 0.797 0.611 0.171 0.183 0.321 0.349

PAGE 85

85 Table A 1. Continued Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXE NT RF Sceloporus couchi 0.912 0.946 0.986 0.902 0.785 0.887 0.918 0.793 0.501 0.475 0.688 0.735 Sceloporus cozumelae 0.942 0.888 0.999 0.889 0.885 0.776 0.998 0.778 0.545 0.572 0.751 0.805 Sceloporus dugesi 0.877 0.904 0.909 0.887 0.67 0.737 0.751 0.713 0.272 0.254 0.321 0.351 Sceloporus edwardtaylori 0.926 0.964 0.973 0.873 0.825 0.903 0.888 0.724 0.362 0.380 0.433 0.337 Sceloporus exsul 0.868 0.967 0.972 0.832 0.743 0.964 0.968 0.665 0.366 0.199 0.058 0.555 Sceloporus gadoviae 0.858 0.917 0.928 0.87 7 0.653 0.773 0.775 0.695 0.19 0.219 0.295 0.37 Sceloporus heterolepis 0.909 0.865 0.92 0.802 0.73 0.654 0.727 0.558 0.149 0.137 0.306 0.33 Sceloporus insignis 0.713 0.738 0.846 0.748 0.496 0.516 0.733 0.497 0.227 0.055 0.094 0.305 Sceloporus internasal is 0.946 0.968 0.973 0.976 0.891 0.945 0.925 0.913 0.191 0.204 0.230 0.378 Sceloporus jalapae 0.904 0.900 0.916 0.900 0.709 0.672 0.737 0.672 0.341 0.299 0.381 0.397 Sceloporus licki 0.959 0.891 0.981 0.867 0.882 0.788 0.950 0.731 0.142 0.089 0.231 0.256 Sceloporus maculosus 0.97 0.937 0.995 0.915 0.932 0.886 0.984 0.828 0.501 0.108 0.541 0.548 Sceloporus megalepidurus 0.884 0.879 0.898 0.882 0.675 0.683 0.728 0.708 0.39 0.226 0.436 0.433 Sceloporus minor 0.921 0.948 0.952 0.951 0.755 0.845 0.806 0.812 0.356 0.355 0.399 0.465 Sceloporus nelsoni 0.920 0.951 0.953 0.896 0.762 0.879 0.873 0.749 0.177 0.158 0.183 0.317 Sceloporus ochoterenae 0.915 0.847 0.94 0.842 0.769 0.720 0.818 0.670 0.143 0.068 0.168 0.262 Sceloporus ornatus 0.971 0.942 0.989 0.963 0.91 0.885 0.951 0.913 0.41 0.187 0.476 0.345 Sceloporus palaciosi 0.923 0.916 0.939 0.826 0.774 0.805 0.793 0.625 0.237 0.283 0.244 0.266 Sceloporus parvus 0.902 0.938 0.946 0.906 0.756 0.816 0.814 0.722 0.193 0.326 0.336 0.359 Sceloporus pyrocephalus 0.899 0.933 0.939 0.919 0.73 0.785 0.779 0.776 0.27 0.318 0.385 0.372 Sceloporus salvini 0.919 0.917 0.957 0.841 0.797 0.777 0.829 0.678 0.237 0.181 0.237 0.243 Sceloporus samcolemani 0.951 0.924 0.988 0.861 0.904 0.849 0.949 0.719 0.342 0.29 0.412 0.33 Sceloporus slevini 0.981 0.992 0.996 0.945 0.951 0.975 0.987 0.882 0.529 0.486 0.594 0.597 Sceloporus smithi 0.931 0.955 0.959 0.751 0.845 0.894 0.888 0.470 0.182 0.201 0.238 0.2 Sceloporus sugillatus 0.846 0.869 0.922 0.753 0.642 0.728 0.731 0.503 0.24 1 0.175 0.238 0.267 Sceloporus utiformis 0.904 0.916 0.933 0.903 0.719 0.771 0.795 0.744 0.224 0.238 0.278 0.363 Sceloporus zosteromus 0.979 0.981 0.987 0.963 0.96 0.96 0.975 0.918 0.334 0.351 0.396 0.366

PAGE 86

86 Table A 1. Continued Species AUC TSS KAPPA GBM GLM MAXENT RF GBM GLM MAXENT RF GBM GLM MAXENT RF Scincella gemmingeri 0.904 0.908 0.917 0.900 0.707 0.729 0.741 0.717 0.315 0.293 0.372 0.419 Scincella silvicola 0.934 0.919 0.953 0.816 0.776 0.796 0.88 0.621 0.337 0.318 0.427 0.380 Sonora michoa canensis 0.818 0.780 0.863 0.663 0.650 0.598 0.736 0.336 0.039 0.036 0.034 0.037 Storeria hidalgoensis 0.910 0.843 0.952 0.641 0.803 0.695 0.874 0.283 0.104 0.062 0.218 0.128 Symphimus leucostomus 0.836 0.880 0.891 0.599 0.690 0.789 0.790 0.216 0.118 0.0 33 0.05 0.087 Symphimus mayae 0.913 0.956 0.998 0.958 0.829 0.912 0.997 0.916 0.289 0.326 0.452 0.638 Sympholis lippiens NA 0.877 0.91 0.656 NA 0.777 0.83 0.317 NA 0.034 0.05 0.045 Tantilla calamarina 0.857 0.814 0.864 0.699 0.652 0.613 0.682 0.373 0.12 3 0.040 0.111 0.155 Tantilla deppei 0.827 0.747 0.848 0.518 0.67 0.572 0.728 0.052 0.058 0.014 0.027 0.006 Tantilla striata 0.888 0.925 0.958 0.824 0.787 0.854 0.915 0.649 0.262 0.058 0.089 0.187 Terrapene coahuila 0.862 0.863 0.883 0.831 0.746 0.741 0. 816 0.664 0.303 0.145 0.493 0.42 Terrapene yucatana 0.984 0.960 0.99 0.965 0.959 0.921 0.971 0.918 0.256 0.314 0.401 0.446 Thamnophis chrysocephalus 0.831 0.783 0.863 0.677 0.589 0.541 0.646 0.344 0.112 0.078 0.159 0.200 Thamnophis godmani 0.84 0.771 0. 854 0.736 0.681 0.514 0.693 0.464 0.109 0.03 0.079 0.174 Thamnophis melanogaster 0.835 0.831 0.85 0.844 0.64 0.667 0.658 0.545 0.139 0.126 0.174 0.258 Thamnophis mendax 0.995 0.886 0.994 0.826 0.991 0.84 0.984 0.649 0.395 0.064 0.490 0.275 Thamnophis sc alaris 0.868 0.877 0.891 0.844 0.66 0.656 0.663 0.598 0.162 0.167 0.201 0.264 Thamnophis scaliger 0.899 0.928 0.933 0.776 0.802 0.848 0.811 0.523 0.091 0.129 0.142 0.267 Thamnophis sumichrasti NA 0.894 0.937 0.773 NA 0.757 0.816 0.52 NA 0.063 0.143 0.242 Thamnophis validus 0.854 0.863 0.883 0.684 0.694 0.708 0.783 0.365 0.093 0.152 0.168 0.13 Tropidodipsas fasciata 0.843 0.797 0.878 0.725 0.618 0.612 0.706 0.432 0.106 0.03 0.181 0.269 Uma exsul 0.999 0.979 1.000 1.000 0.996 0.957 1.000 0.999 0.764 0.79 5 0.915 0.912 Urosaurus gadovi 0.954 0.963 0.972 0.931 0.798 0.892 0.894 0.806 0.454 0.417 0.454 0.418 Urosaurus nigricaudus 0.981 0.977 0.987 0.981 0.954 0.95 0.963 0.940 0.491 0.481 0.560 0.560 Xantusia extorris 0.955 0.954 0.999 0.937 0.908 0.909 0.9 97 0.872 0.625 0.416 0.78 0.633 Xenosaurus newmanorum 0.898 0.923 0.996 0.832 0.815 0.863 0.993 0.665 0.235 0.218 0.565 0.548 Xenosaurus platyceps 0.924 0.922 0.985 0.892 0.833 0.872 0.97 0.772 0.189 0.122 0.299 0.401 Xenosaurus rectocollaris 0.824 0.94 0.963 0.833 0.658 0.893 0.933 0.666 0.209 0.166 0.100 0.597

PAGE 87

87 APPENDIX B BIOGEOGRAPHIC REGIONS IN MEXICO Figure B 1. Biogeographic regions in Mexico. Map obtained from Comisión Nacional para el Conocimiento y Uso de la Biodiversidad (CONABIO), (1997). ' Provincias biogeográficas de México'. Escala 1:4 000 000. Comisión Nacional para el Conocimiento y Uso de la Biodiversidad, México, D. F.

PAGE 88

88 APPENDIX C FUTURE RICHNESS MAPS USING THE Hadcm3 GCM Figure C 1 . Richness of endemic reptiles using the GCM Hadcm3 a nd the A2A emission scenario.

PAGE 89

89 Figure C 2 . Richness of endemic reptiles using the GCM Hadcm3 and the B2A emission scenario

PAGE 90

90 Figure C 3 . Predicted change in species richness according to the GCM Hadcm3 model under the A 2A scenario .

PAGE 91

91 Figure C 4 . Predict ed change in species richness according to the GCM Hadcm3 under the B2A scenario .

PAGE 92

92 APPENDIX D PREDICTED CHANGE IN SPECIES RICHNESS BY NATIONAL PARK National Parks used to predict the change in species richness according to the two GCM evaluated ( CGCM2 and Hadcm3 ) and the two emission scenarios (A2a and B2A) Table D 1 . Change in species richness using the general climatic model CGCM2 and the emission scenario A2A. NATIONAL PARK NAME Area (ha) 2020 CURRENT 2050 2020 2080 2050 Min Max Mean STD MIN MAX MEA N STD MIN MAX MEAN STD Arrecife de Puerto Morelos 9090 4 8 6.33 1.7 5 3 3.67 0.94 0 0 0 0 Arrecifes de Cozumel 12094 5 5 5.00 0 6 6 6 0 1 1 1 0 Arrecifes de Xcalak 17949 6 1 2.67 1.17 1 1 0.57 0.58 0 2 1.81 0.499 Bahía de Loreto 205683 0 7 3 .26 1.82 4 1 1.48 1.29 4 10 6.889 1.449 Benito Juárez 3272 3 3 3.00 0 3 3 3 0 3 3 3 0 Bosencheve 14600 2 1 1.50 0.5 0 2 1 1 6 6 6 0 Cañón de Rio Blanco 48800 8 2 2.47 1.95 6 4 0.322 1.78 9 2 2.743 2.475 Cañón del Sumidero 21840 7 2 0.69 1.52 4 5 0.438 1.9 6 4 0.125 2.197 Cabo Pulmo 7099 5 8 6.00 1.41 4 1 2.33 1.25 7 9 8.333 0.943 Cascada de Bassaseachic 5911 1 4 3.06 0.91 6 1 3.11 1.97 1 3 0.944 1.268 Cerro de Garnica 978 3 2 2.67 0.47 0 2 0.667 0.94 6 3 4.667 1.247 Cobio Chichinautzin 37195 1 1 1.00 0 1 1 1 0 5 5 5 0 Cofre de Perote 11550 5 1 1.42 1.33 4 2 1.26 1.31 6 0 2.789 1.281 Constitución de 1857 4950 1 0 0.94 0.24 1 3 1.25 0.66 0 2 1.375 0.781 Costa Occ. de I Mujeres, Pta. Cancú n Y Pta . N izuc 8621 6 6 6.00 0 3 3 3 0 2 2 2 0 Cumbres de Majalca 4801 5 0 2.06 1.3 2 1 0.47 0.78 5 1 3.882 1.131 Cumbres de Mo nterrey 177395 8 6 0.73 2.83 6 7 1.213 2.75 8 4 2.314 2.676 Cumbres del Ajusco 501 1 1 1.00 0 0 0 0 0 6 6 6 0 Desi erto de Los Leones 1524 3 2 0.17 1.77 2 1 0.67 0.94 6 1 3.667 2.134 Desierto del Carmen O de Nixcongo 475 4 4 4.00 0 0 2 1 1 5 3 4 1 Dzibilchantun 539 2 2 2.00 0 1 3 2 1 3 1 2 1

PAGE 93

93 Table D 1 . Continued NATIONAL PARK NAME Area (ha) 2020 CU RRENT 2050 2020 2080 2050 Min Max Mean STD MIN MAX MEAN STD MIN MAX MEAN STD El Chico 2729 4 1 2.20 0.98 1 2 1.6 0.49 5 1 3.3 1.552 El Jabali 5065 2 5 1.36 1.8 6 2 2.71 2.28 4 4 0.143 1.922 El Potosí 2171 4 2 2.67 0.67 1 4 2.556 1.17 7 2 4 1.633 El Tepozteco 23259 8 4 0.06 2.72 4 4 0.071 1.85 6 3 0.614 2.044 El Veladero 3617 6 0 3.63 1.93 6 3 3.75 1.09 1 6 2.875 1.763 Gogorrón 36965 4 2 0.63 1.13 2 3 0.342 1.26 8 0 3.915 2.106 Grutas de Cacahuamilpa 1624 0 1 0.75 0.4 3 2 1 1.25 0.43 2 4 3 0.707 Huatulco 11845 6 0 3.44 1.55 6 1 3.48 1.34 1 5 2.63 1.444 Insurg. Miguel Hidalgo Y Costilla 1920 2 2 0.50 1.38 2 1 0.17 0.9 6 3 3.833 1.213 Insurg. José Maria Morelos 7192 5 0 2.90 1.55 2 6 2.65 1.85 11 2 6.4 2.653 Isla Contoy 5125 6 7 6.50 0.5 3 3 3 0 0 0 0 0 Iztaccihuatl Popocatépetl 28980 3 2 0.15 1.08 2 6 2.09 1.91 8 0 4.011 1.917 Lagunas de Chacahua 14920 0 3 1.54 0.61 3 2 0.08 1.13 2 6 3.375 0.949 Lagunas de Montebello 6396 0 6 2.60 1. 8 8 3 5.55 1.28 4 1 1 1.483 Lagunas de Zempoala 4556 3 2 0.62 1.64 4 2 0.69 1.9 5 0 2.462 1.781 Los Mármoles 23514 5 4 0.30 1.89 7 5 1.91 2.63 6 6 0.703 2.561 Malinche ó Matlalcueyatl 45494 6 4 0.94 1.93 6 3 1.01 1.6 9 1 3.601 1.802 Mariposa Monarca 56258 7 0 4.00 2.06 2 6 2.269 1.85 10 3 6.462 1.715 Nevado De Colima 6525 1 2 0.11 1.02 8 2 3.89 1.37 3 4 0.737 1.859 Nevado De Toluca 53988 5 3 1.01 1.57 2 5 0.789 1.35 10 1 5.24 1.349 Palenque 1780 1 5 1.67 1.49 0 3 1.333 1.11 4 3 0.167 2.672 Pico De Orizaba 19601 5 1 1.69 1.45 3 3 0.67 1.22 4 2 0.689 1.553 Pico De Tancitaro 23448 4 4 1.06 1.75 5 4 1.27 1.98 7 1 3.817 1.952 Sierra De Organos 1125 0 0 0.00 0 0 0 0 0 1 1 1 0 Sierra de San Pedro Mart ir 72909 2 5 0.65 1.24 3 3 0.61 1.08 5 2 1.5 1.409 Sistema Arrecifal Veracruzano 52284 3 0 1.80 1.47 1 3 2 0.63 1 0 0.8 0.4 Tulum 648 1 0 0.50 0.5 4 3 3.5 0.5 0 1 0.5 0.5

PAGE 94

94 Table D 2. Change in species richness using the general climatic m odel CGCM2 and the emission scenario B2A NATIONAL PARK NAME Area (ha) 2020 CURRENT 2050 2020 2080 2050 MIN MAX MEAN STD MIN MAX MEAN STD MIN MAX MEAN STD Arrecife de Puerto Morelos 9090 2 3 2.33 0.47 3 5 4.33 0.94 4 4 4.00 0.00 Arrecifes de Cozumel 12094 4 4 4.00 0.00 0 0 0.00 0.00 2 2 2.00 0.00 Arrecifes de Xcalak 17949 2 3 0.43 1.50 3 0 1.81 1.05 2 1 0.81 0.85 Bahía de Loreto 205683 0 7 2.70 1.76 1 4 1.19 1.09 3 1 1.30 1.27 Benito Juárez 3272 1 1 1.00 0.00 0 0 0.00 0.00 2 2 2.00 0 .00 Bosencheve 14600 5 3 4.00 1.00 0 2 1.00 1.00 2 0 1.00 1.00 Cañón de Rio Blanco 48800 8 1 2.26 1.83 8 3 1.11 2.01 4 5 0.76 1.87 Cañón del Sumidero 21840 1 8 3.67 1.90 11 0 5.08 2.21 2 3 0.64 1.12 Cabo Pulmo 7099 0 2 1.00 0.82 6 7 6.67 0.47 3 2 2.67 0.47 Cascada de Bassaseachic 5911 1 6 2.94 1.61 4 1 2.44 1.01 3 1 2.06 0.78 Cerro de Garnica 978 5 3 3.67 0.94 1 3 2.00 0.82 0 1 0.67 0.47 Cobio Chichinautzin 37195 2 2 2.00 0.00 1 1 1.00 0.00 1 1 1.00 0.00 Cofre de Pero te 11550 3 3 0.42 1.44 5 0 2.34 1.28 3 4 0.29 1.57 Constitución de 1857 4950 1 1 0.88 0.48 1 3 1.69 0.58 1 2 0.88 0.93 Costa Occ. de I Mujeres, Pta. Cancún Y Pta. Nizuc 8621 0 0 0.00 0.00 6 6 6.00 0.00 4 4 4.00 0.00 Cumbres de Majalca 4801 3 4 0.12 1.78 0 4 1.59 1.09 10 6 7.82 1.15 Cumbres de Monterrey 177395 6 5 0.35 1.82 6 8 0.43 2.23 6 5 0.82 2.16 Cumbres del Ajusco 501 0 0 0.00 0.00 1 1 1.00 0.00 2 2 2.00 0.00 Desierto de Los Leones 1524 2 1 1.00 1.15 3 2 0.17 1.67 1 2 0.50 0.96 Desierto del Carmen O de Nixcongo 475 6 5 5.50 0.50 3 6 4.50 1.50 2 1 0.50 1.50 Dzibilchantun 539 2 1 1.50 0.50 0 1 0.50 0.50 2 3 2.50 0.50 El Chico 2729 2 0 0.50 0.67 1 0 0.20 0.40 3 2 0.50 1.57 El Jabali 5065 4 1 0.79 1.57 4 3 0.00 1.96 0 6 3.00 1.60 El Potosí 2171 2 1 0.56 1.34 2 1 0.44 1.17 3 2 2.44 0.50 El Tepozteco 23259 7 4 0.43 2.46 5 2 0.93 1.65 1 6 2.47 1.40 El Veladero 3617 6 1 3.13 2.26 5 0 2.75 1.64 3 1 0.50 1.12

PAGE 95

95 Table D 2. Continued NATIONA L PARK NAME Area (ha) 2020 CURRENT 2050 2020 2080 2050 MIN MAX MEAN STD MIN MAX MEAN STD MIN MAX MEAN STD Gogorrón 36965 6 0 2.25 1.23 3 2 0.22 1.43 3 3 0.24 1.26 Grutas de Cacahuamilpa 1624 0 1 0.75 0.43 2 1 0.50 1.12 1 3 1.75 0.83 Huatulco 11 845 7 1 3.41 1.45 0 5 2.00 1.33 5 1 1.33 1.54 Insurg. Miguel Hidalgo Y Costilla 1920 1 1 0.17 0.90 3 0 2.00 1.15 0 3 1.33 0.94 Insurg. José Maria Morelos 7192 5 0 1.90 1.51 3 3 0.10 1.41 3 4 0.95 1.75 Isla Contoy 5125 4 5 4.50 0.50 3 4 3.50 0.50 4 3 3.50 0.50 Iztaccihuatl Popocatépetl 28980 2 2 0.28 1.02 3 2 0.06 0.99 4 3 0.08 1.55 Lagunas de Chacahua 14920 3 3 0.88 1.17 3 2 2.06 1.01 3 8 5.79 0.96 Lagunas de Montebello 6396 4 8 5.85 1.19 14 8 11.70 2.03 4 10 7.15 1.62 Lagu nas de Zempoala 4556 3 1 1.00 1.18 3 1 0.92 1.14 2 1 0.23 0.89 Los Mármoles 23514 3 2 0.26 1.07 4 3 0.09 1.34 3 2 0.36 1.35 Malinche ó Matlalcueyatl 45494 5 3 0.89 1.60 6 3 1.34 1.73 2 7 1.65 2.07 Mariposa Monarca 56258 6 1 3.15 1.32 3 4 0.27 1.58 5 4 0.31 1.81 Nevado De Colima 6525 5 0 2.53 1.14 3 3 0.74 1.45 0 4 2.05 1.28 Nevado De Toluca 53988 4 1 1.06 1.18 3 3 0.24 1.41 4 2 0.14 1.32 Palenque 1780 0 6 4.00 1.91 6 4 4.67 0.75 0 2 0.83 0.69 Pico De Orizaba 19601 4 4 0.28 1.66 5 2 1.62 1.36 1 5 1.49 1.34 Pico De Tancitaro 23448 5 3 2.48 1.69 3 3 0.14 1.48 2 6 1.66 1.72 Sierra De Organos 1125 1 1 1.00 0.00 0 1 0.60 0.49 1 0 0.60 0.49 Sierra de San Pedro Martir 72909 2 5 0.89 1.57 4 4 1.20 1.20 6 1 1.44 1.39 Sistema Arrecifal Veracruzano 52284 3 1 1.00 1.41 1 1 0.60 0.80 1 0 0.60 0.49 Tulum 648 2 1 1.50 0.50 4 4 4.00 0.00 5 5 5.00 0.00

PAGE 96

96 Table D 3 . Change in species richness using the general climatic model Hadcm3 and the emission scen ario A2A . NATIONAL PARK NAME Area (ha) 2020 CURRENT 2050 2020 2080 2050 MIN MAX MEAN STD MIN MAX MEAN STD MIN MAX MEAN STD Arrecife de Puerto Morelos 9090 1 3 2.00 0.82 1 0 0.33 0.47 3 3 3.00 0.00 Arrecifes de Cozumel 12094 1 1 1.00 0.00 3 3 3.00 0.00 1 1 1.00 0.00 Arrecifes de Xcalak 17949 5 1 2.52 0.96 1 4 1.95 0.72 4 10 6.29 1.75 Bahía de Loreto 205683 2 6 2.00 1.94 3 4 1.04 1.73 1 11 4.33 2.34 Benito Juárez 3272 0 0 0.00 0.00 3 3 3.00 0.00 2 2 2.00 0.00 Bosencheve 14600 2 1 1.50 0.50 3 3 3.00 0.00 5 2 3.50 1.50 Cañón de Rio Blanco 48800 7 3 0.62 1.84 8 7 1.31 2.64 9 1 3.64 2.21 Cañón del Sumidero 21840 2 8 1.17 2.58 4 6 0.84 2.49 2 11 4.25 2.78 Cabo Pulmo 7099 6 8 7.00 0.82 1 0 0.33 0.47 2 4 3.00 0.82 Cascada de Bassaseachic 5911 4 3 1.78 1.81 2 1 0.50 0.83 3 7 4.33 1.15 Cerro de Garnica 978 4 3 3.67 0.47 6 3 1.33 3.68 8 0 4.33 3.30 Cobio Chichinautzin 37195 2 2 2.00 0.00 4 4 4.00 0.00 5 5 5.00 0.00 Cofre de Perote 11550 4 2 0.89 1.35 4 3 0.47 1.59 3 3 0.71 1.68 Constitución de 1857 4950 1 1 0.75 0.56 0 2 1.75 0.56 2 1 1.38 0.78 Costa Occ. de I Mujeres, Pta. Cancún Y Pta. Nizuc 8621 0 0 0.00 0.00 2 2 2.00 0.00 3 3 3.00 0.00 Cumbres de Majalca 4801 3 1 1.71 0.57 2 1 0.53 0.92 2 1 0.12 0.76 Cumbres de Monterrey 177395 5 5 0.26 1.62 6 5 0.29 1.85 6 6 0.59 2.86 Cumbres del Ajusco 501 1 1 1.00 0.00 1 1 1.00 0.00 4 4 4.00 0.00 Desierto de Los Leones 1524 1 2 0.50 1.26 1 4 1.17 1.67 6 3 4.50 0.96 Desierto del Carmen O de Nixcongo 475 2 2 2.00 0.00 7 4 5.50 1.50 2 1 0.50 1.50 Dzibilchantun 539 1 0 0.50 0.50 1 1 1.00 0.00 2 2 2.00 0.00 El Chico 2729 1 3 1.20 1.25 5 1 3.10 2.21 4 3 0.20 2.68 El Jabali 5065 7 0 3.36 2.26 1 3 1.50 1.24 1 4 1.57 1.64 El Potosí 2171 1 1 0.44 0.68 2 1 0.78 1.03 3 0 1.22 1.13 El Tepozteco 23259 6 4 0.39 2.09 9 6 1.07 2.57 10 8 0.01 4.54 El Veladero 3617 9 4 6.50 1.41 2 6 1.00 2.35 2 4 1.88 1.69

PAGE 97

97 Table D 3. Continued NATIONAL PARK NAME Area (ha) 2020 CU RRENT 2050 2020 2080 2050 MIN MAX MEAN STD MIN MAX MEAN STD MIN MAX MEAN STD Gogorrón 36965 1 4 1.35 1.25 6 1 3.22 1.31 3 6 0.15 2.03 Grutas de Cacahuamilpa 1624 1 1 0.00 0.71 1 4 3.00 1.22 0 6 2.50 2.60 Huatulco 11845 5 0 2.41 1.13 8 1 5.6 7 2.14 5 13 9.56 2.36 Insurg. Miguel Hidalgo Y Costilla 1920 1 3 0.17 1.46 0 4 2.50 1.26 6 3 4.83 0.90 Insurg. José Maria Morelos 7192 6 2 2.00 2.14 4 2 1.65 2.06 1 3 0.85 1.46 Isla Contoy 5125 2 4 3.00 1.00 1 2 1.50 0.50 2 2 2.00 0.00 Iztacci huatl Popocatépetl 28980 3 4 0.43 1.54 4 7 1.82 2.43 10 1 3.28 2.19 Lagunas de Chacahua 14920 3 2 0.35 1.07 1 6 2.58 1.81 4 8 6.02 1.09 Lagunas de Montebello 6396 4 1 2.90 1.04 1 4 2.35 0.96 1 4 1.95 1.12 Lagunas de Zempoala 4556 3 2 0.54 1. 50 1 5 2.77 1.25 10 2 4.69 2.01 Los Mármoles 23514 3 3 0.32 1.25 8 2 4.27 1.66 4 7 0.74 2.22 Malinche ó Matlalcueyatl 45494 8 5 1.00 2.71 5 1 1.73 1.52 7 2 4.10 1.60 Mariposa Monarca 56258 5 3 1.46 1.80 6 4 0.88 2.33 6 1 3.50 1.12 N evado De Colima 6525 7 2 5.00 1.56 5 1 2.11 1.37 3 4 0.95 2.14 Nevado De Toluca 53988 5 1 1.63 0.99 4 5 1.49 1.53 9 3 1.45 1.73 Palenque 1780 3 2 0.67 1.80 2 7 5.17 1.57 8 12 9.83 1.34 Pico De Orizaba 19601 5 1 0.93 1.40 4 3 0.34 2.07 7 2 1.72 1.89 Pico De Tancitaro 23448 9 1 4.69 1.87 5 4 0.17 2.33 7 2 2.34 1.83 Sierra De Organos 1125 0 1 0.80 0.40 2 1 1.40 0.49 2 1 1.20 0.40 Sierra de San Pedro Martir 72909 5 6 0.29 1.69 6 5 0.20 1.87 5 5 1.62 1.16 Sistema Arrecif al Veracruzano 52284 2 2 0.40 1.62 1 1 0.20 0.75 10 13 11.20 1.17 Tulum 648 3 3 3.00 0.00 1 2 1.50 0.50 3 5 4.00 1.00

PAGE 98

98 Table D 4. Change in species richness using the general climatic model Hadcm3 and the emission scenario B2A . NATIONAL PARK NAM E Area (ha) 2020 CURRENT 2050 2020 2080 2050 MIN MAX MEAN STD MIN MAX MEAN STD MIN MAX MEAN STD Arrecife de Puerto Morelos 9090 1 4 2.67 1.25 0 2 1.00 0.82 9 7 8.00 0.82 Arrecifes de Cozumel 12094 1 1 1.00 0.00 3 3 3.00 0.00 9 9 9.00 0.00 Ar recifes de Xcalak 17949 4 2 2.95 0.72 0 1 0.57 0.49 5 3 4.57 0.66 Bahía de Loreto 205683 0 5 1.96 1.57 3 2 0.15 1.41 7 3 4.85 0.97 Benito Juárez 3272 0 0 0.00 0.00 0 0 0.00 0.00 2 2 2.00 0.00 Bosencheve 14600 1 0 0.50 0.50 2 2 2.00 0.00 9 8 8.50 0.50 Cañón de Rio Blanco 48800 10 4 2.20 2.40 2 7 2.70 1.84 14 2 7.68 2.19 Cañón del Sumidero 21840 5 5 0.39 1.88 2 9 2.23 2.30 16 1 5.17 3.76 Cabo Pulmo 7099 4 6 5.33 0.94 3 2 2.67 0.47 5 3 4.00 0.82 Cascada de Bassaseachi c 5911 3 3 0.17 2.22 2 2 0.39 1.11 8 4 7.06 1.13 Cerro de Garnica 978 3 1 2.00 0.82 1 2 1.00 1.41 9 7 8.33 0.94 Cobio Chichinautzin 37195 2 2 2.00 0.00 3 3 3.00 0.00 8 8 8.00 0.00 Cofre de Perote 11550 5 1 2.74 1.14 2 6 2.53 1.96 11 4 6.92 1.91 Constitución de 1857 4950 1 2 1.19 0.39 2 1 1.38 0.48 4 3 3.94 0.24 Costa Occ. de I Mujeres, Pta. Cancún Y Pta. Nizuc 8621 2 2 2.00 0.00 2 2 2.00 0.00 9 9 9.00 0.00 Cumbres de Majalca 4801 6 1 3.29 1.13 2 4 2.88 0.58 7 5 6.06 0.64 Cumbres de Monterrey 177395 5 5 0.58 1.70 4 5 0.34 1.46 10 2 3.62 1.76 Cumbres del Ajusco 501 1 1 1.00 0.00 1 1 1.00 0.00 9 9 9.00 0.00 Desierto de Los Leones 1524 1 1 0.33 0.75 0 5 3.00 1.63 13 7 9.67 1.89 Desierto del Carmen O de Nixcongo 475 4 2 3.00 1.00 3 0 1.50 1.50 6 6 6.00 0.00 Dzibilchantun 539 2 1 1.50 0.50 2 3 2.50 0.50 13 13 13.00 0.00 El Chico 2729 2 1 0.60 0.92 3 5 0.30 2.69 8 1 4.50 2.54 El Jabali 5065 6 2 1.50 2.56 2 3 0.14 1.36 8 4 5. 14 1.06 El Potosí 2171 1 1 0.11 0.74 0 3 1.11 0.87 6 4 4.89 0.57 El Tepozteco 23259 6 4 0.06 2.36 7 9 2.99 2.92 14 5 9.90 2.09

PAGE 99

99 Table D 4. Continued NATIONAL PARK NAME Area (ha) 2020 CURRENT 2050 2020 2080 2050 MIN MAX MEAN STD MIN MAX MEAN STD MIN MAX MEAN STD El Veladero 3617 11 2 7.50 2.65 2 7 3.75 1.64 15 10 12.63 1.41 Gogorrón 36965 2 2 0.21 0.97 4 2 1.09 1.11 6 1 3.77 1.19 Grutas de Cacahuamilpa 1624 1 0 0.25 0.43 5 5 5.00 0.00 7 6 6.25 0.43 Huatulco 11845 8 4 5.78 1.20 3 2 0.22 0.99 14 10 12.33 1.25 Insurg. Miguel Hidalgo Y Costilla 1920 2 1 0.67 1.11 1 6 4.50 1.71 11 8 9.83 1.07 Insurg. José Maria Morelos 7192 4 4 0.50 1.77 5 0 2.20 1.72 9 2 5.00 1.90 Isla Contoy 5125 2 3 2.50 0.50 2 3 2.50 0.50 9 8 8.50 0.50 Iztaccihuatl Popocatépetl 28980 3 6 1.92 2.08 3 6 2.16 1.68 18 8 14.01 1.78 Lagunas de Chacahua 14920 4 0 2.06 0.72 0 5 3.88 0.78 9 7 7.98 0.52 Lagunas de Montebello 6396 6 1 3.35 1.65 2 6 4.40 1.11 8 2 5.25 1.41 L agunas de Zempoala 4556 3 2 1.15 1.35 2 5 3.46 1.01 13 8 9.92 1.49 Los Mármoles 23514 4 3 1.58 1.49 6 1 1.84 1.41 7 2 2.49 1.95 Malinche ó Matlalcueyatl 45494 7 6 1.80 2.53 8 4 0.66 2.07 11 4 6.42 1.26 Mariposa Monarca 56258 5 1 1.65 1.17 3 4 0.88 2.01 11 5 7.65 1.52 Nevado De Colima 6525 6 1 3.53 1.23 5 1 2.79 2.28 8 3 5.00 1.30 Nevado De Toluca 53988 3 2 0.19 1.10 3 4 0.58 1.35 12 0 4.51 1.76 Palenque 1780 1 6 3.00 1.53 3 2 0.00 1.83 1 1 0.50 0.76 Pico De Or izaba 19601 7 2 2.31 1.88 4 4 2.48 1.63 8 2 5.72 1.42 Pico De Tancitaro 23448 5 2 2.31 1.77 6 3 1.87 2.33 13 5 9.96 1.67 Sierra De Organos 1125 0 0 0.00 0.00 1 1 1.00 0.00 2 2 2.00 0.00 Sierra de San Pedro Martir 72909 2 5 0.97 1.39 9 2 3.16 2.81 8 0 4.43 1.83 Sistema Arrecifal Veracruzano 52284 2 3 0.40 1.85 1 1 0.60 0.80 4 0 1.80 1.33 Tulum 648 4 4 4.00 0.00 0 0 0.00 0.00 8 7 7.50 0.50

PAGE 100

100 Table D 5 . Correlation among bioclimatic variables and the ordination axes o btained with the P rincipal coordinates analysis . In parenthesis the p values Variable Ordination Axe 1 Ordination Axe 2 Ordination Axe 3 Bio 1 0.001(0.99) 0.007(0.79) 0.015(0.56) Bio3 0.004(0.87) 0.009(0.74) 0.005(0.84) Bio5 0.005(0.84) 0.007(0.79) 0.017(0.52) Bio7 0.011(0.67) 0.001(0.96) 0.004(0.89) Bio13 0.010(0.70) 0.017(0.50) 0.004(0.86) Bio14 0.034(0.18) 0.023(0.37) 0.025(0.32) Bio15 0.034(0.18) 0.008(0.74) 0.028(0.27)

PAGE 101

101 LIST OF REFERENCES Aid CS, Carter MF, Peterson AT (1997) Protected areas of Western Mexico: status, management, and needs . Allouche O, Tsoar A, Kadmon R (2006) Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). J ournal of applied ecology , 46 , 1223 1232. Araújo MB, Guisan A (2006) Five (or so) challenges for species distribution modelling. Journal of Biogeography , 33 , 1677 1688. Araújo MB, New M (2007) Ensemble forecasting of species distributions. Trends in Ecolog y & Evolution , 22 , 42 47. Ecography , 28 , 693 695. Araújo MB, Peterson AT (2012) Uses and misuses of bioclimatic envelop modeling. Ecology , 93 , 1527 1539. Araújo MB, Pearson RG, Erhard M, Thuiller W (2005a) Validation of species climate impact models under climate change. Global Change Biology , 11 , 1504 1513. Araújo MB, Whittaker RJ, Ladle RJ, Erhard M (2005b) Reducing uncertainty in projections of extinction risk from climate change. Global Ecology and Biogeography , 14 , 529 538. Araújo MB, Thuiller W, Pearson RG (2006) Climate warming and the decline of amphibians and reptiles in Europe. Journal of Biogeography , 33 , 1712 1728. Barrera Santos G, Pacheco J, Ceballos G (2004) La c onservación de los reptiles y anfibios de méxico. Biodiversitas , 57 , 1 6. Baselga A, Araújo MB (2009) Individualistic vs community modelling of species distributions under climate change. Ecography , 32 , 55 65. Baselga A, Araújo MB (2010) Do community level models describe community variation effectively? Journal of Biogeography , 37 , 1842 1850. Belbin L (2013) Atlas of living Australia website. Bonthoux S, Baselga A, Balent G (2013) Assessing community level and single species models predictions of species d istributions and assemblage composition after 25 years of land cover change. PloS one , 8 , e54179.

PAGE 102

102 Breiman L (2001) Random Forests. Machine Learning , 45 , 5 32. Carroll C, Dunk JR, Moilanen A (2010) Optimizing resiliency of reserve networks to climate change : multispecies conservation planning in the Pacific Northwest, USA. Global Change Biology , 16 , 891 904. Carvalho JC, Cardoso P, Gomes P (2012) Determining the relative roles of species replacement and species richness differences in generating beta diversi ty patterns. Global Ecology and Biogeography , 21 , 760 771. Ceballos G, Rodríguez P, Medellín R (1998) Assesing conservation priorities in megadiverse Mexico: Mammalian diversity, endemicity,and endangerment. Ecological Applications , 8 , 8 17. Chamaille Jamm es S, Massot M, Aragon P, Clobert J (2006) Global warming and positive fitness response in mountain populations of common lizards Lacerta vivipara. Global Change Biology , 12 , 392 402. Chen S, Jiang G, Zhang J, Li Y, Qian H (2011) Species turnover of amphib ians and reptiles in eastern China: disentangling the relative effects of geographic distance and environmental difference. Ecological Research , 26 , 949 956. Coetzee BWT, Robertson MP, Erasmus BFN, van Rensburg BJ, Thuiller W (2009) Ensemble models predict Important Bird Areas in southern Africa will become less effective for conserving endemic birds under climate change. Global Ecology and Biogeography , 18 , 701 710. Comisión Nacional de Áreas Naturales Protegidas (2010) Estrategia de cambio climático para áreas protegidas , 2da edn. Secretaría de Medio Ambiente y Recursos Naturales, Mexico. C onabio , C onanp , T nc , P ronatura , F cf , U anl (2007) Análisis de vacíos y omisiones en conservación de la biodiversidad terrestre de México: espacios y especies (eds Koleff P, Andrés Lira Noriega, César Cantú, Tania Urquiza, Melanie Kolb). Comisión Nacional para el Conocimiento y Uso de la Biodiversidad, Comisión Nacional de Áreas Naturales Protegidas, The Nature Conservancy Programa México, Pronatura, A.C., Facultad de Cien cias Forestales, Universidad Autónoma de Nuevo León, México. Davey CM, Devictor V, Jonzén N, Lindström Å, Smith HG (2013) Impact of climate change on communities: revealing species contribution (ed Webb T). Journal of Animal Ecology , 82 , 551 561. Dennis RL H, Shreeve TG, Sparks TH, Lhonore JE (2002) A comparison of geographical and neighbourhood models for improving atlas databases. The case of the French butterfly atlas. Biological Conservation , 108 , 143 159.

PAGE 103

103 Deutsch CA, Tewksbury JJ, Huey RB, Sheldon KS, G halambor CK, Haak DC, Martin PR (2008) Impacts of climate warming on terrestrial ectotherms across latitude. Proceedings of the National Academy of Sciences of the United States of America , 105 , 6668 6672. Dudley N (2008) Guidelines for applying protected area management categories . IUCN, Gland, Switzerland. Elith J, Graham CH, Anderson RP et al. (2006) Novel methods improve prediction of Ecography , 29 , 129 151. ESRI (2013) ArcMap 10.1. Environmental Systems Rese arch Institute. Feeley KJ, Silman MR (2010) Land use and climate change effects on population size and extinction risk of Andean plants. Global Change Biology , 16 , 3215 3222. Ferrier S, Guisan A (2006) Spatial modelling of biodiversity at the community lev el. Journal of Applied Ecology , 43 , 393 404. Ferrier S, Drielsma M, Manion G, Watson G (2002) Extended statistical approaches to modelling spatial pattern in biodiversity in northeast New South Wales. II. Community level modelling. Biodiversity and Conserv ation , 11 , 2309 2338. Ferrier S, Powell GVN, Richardson KS et al. (2004) Mapping More of Terrestrial Biodiversity for Global Conservation Assessment. BioScience , 54 , 1101 1109. Ferrier S, Manion G, Elith J, Richardson K (2007) Using generalized dissimilari ty modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment. Diversity and Distributions , 13 , 252 264. Fielding AH, Bell JF (1997) A review of methods for the assessment of prediction errors in conservation presence/a bsence models. Environmental Conservation , 24 , 38 49. Fitzpatrick MC, Sanders NJ, Ferrier S, Longino JT, Weiser MD, Dunn R (2011) Forecasting the future of biodiversity: a test of single and multi species models for ants in North America. Ecography , 34 , 8 36 847. Fitzpatrick MC, Sanders NJ, Normand S, Svenning J, Ferrier S, Gove AD, Dunn RR (2013) Environmental and historical imprints on beta diversity: insights from variation in rates of species turnover along gradients. Proceedings of the Royal Society B , 280 , 1 8. Flores Villela O, Canseco Márquez L (2004) Nuevas Especies y Cambios taxonómicos para la herpetofauna de México. Acta Zoológica Méxicana , 20 , 115 144.

PAGE 104

104 Foufopoulos J, Kilpatrick a M, Ives AR (2011) Climate change and elevated extinction rates of reptiles from Mediterranean Islands. The American naturalist , 177 , 119 29. Friedman BJ, Hastie T, Tibshirani R (2000) Additive logistic regression: A statistical view of boosting. The Annals of Statistics , 28 , 337 407. Fuller T, Sanchezcordero V, Illoldira ngel P, Linaje M, Sarkar S (2007) The cost of postponing biodiversity conservation in Mexico. Biological Conservation , 134 , 593 600. Gadsden H, Ballesteros Barrera, C., Hinojosa de la Garza, O., Castañeda G, García De la Peña C, Lemos Espinal JA (2012) Eff ects of land cover transformation and climate change on the distribution of two endemic lizards, Crotaphytus antiquus and Sceloporus cyanostictus, of northern Mexico. Journal of Arid Environments , 83 , 1 9. García A (2006) Using ecological niche modelling t o identify diversity hotspots for the herpetofauna of Pacific lowlands and adjacent interior valleys of Mexico. Biological Conservation , 130 , 25 46. García A, Solano Rodríguez H, Flores Villela O (2007) Patterns of alpha, beta and gamma diversity of the he valleys. Animal Biodiversity and Conservation , 30.2 , 169 177. Guisan A, Thuiller W (2005) Predicting species distribution: offering more than simple habitat models. Ecology Letters , 8 , 993 1009. Guisan A , Zimmermann NE (2000) Predictive habitat distribution models in ecology. Ecological Modelling , 135 , 147 186. Hanley JA, McNeil BJ (1982) The meaning and use of the area under a receiver operating characterisic (ROC) curve. Radiology , 143 . Hannah L (2008) Protected areas and climate change. Annals of the New York Academy of Sciences , 1134 , 201 12. Heikkinen RK, Luoto M, Araujo MB, Virkkala R, Thuiller W, Sykes MT (2006) Methods and uncertainties in bioclimatic envelope modelling under climate change. Progre ss in Physical Geography , 30 , 751 777. Hernandez PA, Graham CH, Master LL, Albert DL (2006) The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography , 29 , 773 785.

PAGE 105

105 Hijmans RJ, Graham CH (2006) The ability of climate envelope models to predict the effect of climate change on species distributions. Global Change Biology , 12 , 2272 2281. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate s urfaces for global land areas. International Journal of Climatology , 25 , 1965 1978. Hoffmann AA, Hallas RJ, Dean JA, Schiffer M (2003) Low potential for climatic stress adaptation in a rainforest Drosophila species. Science , 301 , 100 102. Hollander M, Wolf e DA (1973) Nonparametric Statistical Methods . John Wiley&Sons, New York. Hortal J, Lobo JM, Jiménez Valverde A (2007) Limitations of biodiversity databases: case study on seed plant diversity in Tenerife, Canary Islands. Conservation biology , 21 , 853 63. Huey RB, Deutsch C a, Tewksbury JJ, Vitt LJ, Hertz PE, Alvarez Pérez HJ, Garland T (2009) Why tropical forest lizards are vulnerable to climate warming. Proceedings of the Royal Society B , 276 , 1939 48. IPCC (2007) Cambio climático 2007: Informe de síntesi s. Contribución de los Grupos de trabajo I, II y III al Cuarto Informe de evaluación del Grupo Intergubernamental de Expertos sobre el Cambio Climático (eds Pachauri RK, Reisinger A). IPCC, Ginebra, Suiza. IPCC (2012) Managing the Risks of Extreme Events a nd Disasters to Advance Climate Change Adaptation. A Special Report of workin groups I and II of the Intergovernmental Panel on climate Change Change (eds Field C., Barros V, Stocker TF, Qin D, Dokken DJ, Ebi, K L, Mastrandrea MD, Mach KL, Plattner GK, All en SK, Tignor M, Midgley PM). Cambridge University Press, Cambridge. Jetz W, Wilcove DS, Dobson AP, Jetz, W., Wilcove, D. S., & Dobson AP (2007) Projected impacts of climate and land use change on the global diversity of birds. PLoS biology , 5 , e157. Jimén ez Valverde A (2012) Insights into the area under the receiver operating characteristic curve (AUC) as a discrimination measure in species distribution modelling. Global Ecology and Biogeography , 21 , 498 507. Jiménez Valverde A, Lobo JM, Hortal J (2008) No t as good as they seem: the importance of concepts in species distribution modelling. Diversity and Distributions , 14 , 885 890.

PAGE 106

106 Kearney M, Porter WP (2006) Mapping the fundamental niche: physiology, climate, and the distribution of a nocturnal lizard. Ecol ogy , 85 , 3119 3131. Koleff P, Soberón J, Arita H et al. (2008) Patrones de diversidad espacial en grupos selectos de especies. In: Capital natural de México, volumen I: Conocimiento actual de la biodiversidad , Vol. I, pp. 323 364. CONABIO, México. Landis J R, Koch GG (1977) The measurement of observer agreement for categorical data. Biometrics , 33 , 159 174. Legendre P, Legendre L (1998) Numerical ecology , 2nd edn. Elsevier Ltd. Lillywhite HB (2013) Climatic change and reptiles. In: The balance of Nature and Human Impact (ed Rohde K), pp. 279 294. Cambridge University Press, Washington, DC. Liu C, Berry PM, Dawson TP, Pearson RG (2005) Selecting thresholds of occurrence in the prediction of species distributions. Ecography , 28 , 385 393. Loiselle BA, Howell CA, Graham CH, Goerck JM, Brooks T, Smith KG, Williams PH (2003) Avoiding Pitfalls of Using Species Distribution Models in Conservation Planning. Conservation Biology , 17 , 1591 1600. Magurran AE, Baillie SR, Buckland ST et al. (2010) Long term datasets in bio diversity research and monitoring: assessing change in ecological communities through time. Trends in ecology & evolution , 25 , 574 82. Malcolm JR, Liu C, Neilson RP, Hansen L, Hannah L (2006) Global Warming and Extinctions of Endemic Species from Biodivers ity Hotspots. Conservation Biology , 20 , 538 548. Margules CR, Pressey RL (2000) Systematic conservation planning. Nature , 405 , 243 53. Marmion M, Parviainen M, Luoto M, Heikkinen RK, Thuiller W (2009) Evaluation of consensus methods in predictive species d istribution modelling. Diversity and Distributions , 15 , 59 69. Mas J F, Velázquez A, Díaz Gallegos JR et al. (2004) Assessing land use/cover changes: a nationwide multidate spatial database for Mexico. International Journal of Applied Earth Observation and Geoinformation , 5 , 249 261. McCullagh P, Nelder JA (1989) Generalized Linear Models, Second Edition . Chapman and Hall.

PAGE 107

107 McKnight MW, White PS, McDonald RI, Lamoreux JF, Sechrest W, Ridgely RS, Stuart SN (2007) Putting beta diversity on the map: broad scale congruence and coincidence in the extremes. PLoS biology , 5 , e272. Mittermeier RA, Goettsch C (1992) La importancia de la diversidad biológica de México. México ante los Retos de la Biodiversidad. 63 73. Mittermeier RA, Mittermeier CG (1992) La Importanci a de la Diversidad Biologica de Mexico. In: Mexico ante los retos de la Biodiversidad (eds Sarukhan J, Dirzo R). Comisión Nacional para el Conocimiento y Uso de la Biodiversidad, Mexico, D. F. Mittermeier RA, Myers N (1998) Biodiversity hotspots and major tropical wilderness areas: approaches to setting conservation priorities. Conservation Biology , 12 , 516 520. Moilanen A, Wilson KA, Possingham HP (2009) Spatial Conservation Prioritization, Quantitative Methods and Computational Tools . Oxford University Pr ess, New York. Monserud RA, Leemans R (1992) Comparing global vegetation maps with the Kappa statistic. Ecological Modelling , 62 , 275 293. Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J (2000) Biodiversity hotspots for conservation prioriti es. Nature , 403 , 853 8. Ochoa Ochoa LM, Flores Villela O (2006) Áreas de diversidad y endemismo de la herpetofauna Mexicana . UNAM, CONABIO. Ochoa Ochoa LM, Urbina Cardona JN, Vázquez L B, Flores Villela O, Bezaury Creel J (2009) The effects of governmental protected areas and social initiatives for land protection on the conservation of Mexican amphibians. PloS one , 4 , e6878. Ochoa Ochoa LM, Rodríguez P, Mora F, Flores Villela O, Whittaker RJ (2012) Climate change and amphibian diversity patterns in Mexico. Biological Conservation , 150 , 94 102. Ortega Huerta MA, Peterson AT (2004) Modelling spatial patterns of biodiversity for conservation prioritization in North eastern Mexico. Diversity and Distributions , 10 , 39 54. Overton JM, Barker GM, Price R (2009) Es timating and conserving patterns of invertebrate diversity: a test case of New Zealand land snails. Diversity and Distributions , 15 , 731 741. Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature , 421 , 37 42.

PAGE 108

108 Pearson RG, Thuiller W, Araújo MB et al. (2006) Model based uncertainty in species range prediction. Journal of Biogeography , 33 , 1704 1711. Peterson AT, Ball LG, Cohoon KP (2002) Predicting distributions of Mexican birds using ecological nic he modelling methods. Ibis , 144 , E27 E32. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling , 190 , 231 259. Pineda E, Lobo JM (2009) Assessing the accuracy of species distribution models to predict amphibian species richness patterns. Journal of Animal Ecology , 78 , 182 190. Portillo Quintero CA, Sánchez Azofeifa GA (2010) Extent and conservation of tropical dry forests in the Americas. Biological Conservation , 143 , 144 155. Pressey RL, Humphries CJ, Margules CR, Vane Wright RI, Williams PH (1993) Beyond opportunism: Key principles for systematic reserve selection. Trends in ecology & evolution , 8 , 124 8. Qian H (2010) Environment richness relationships for mammals, birds, reptiles, and amphibians at global and regional scales. Ecological Research , 25 , 629 637. R Core Team (2013) R: A language and environment for statistical computing. Copyright (C). Ramamoorthy TP, Bye R, Lott A, Fa J (Eds. . (1993) Biological diversity of Mexico: or igin and distribution. Oxford University Press, New York. Ramirez J, Jarvis A (2008) High Resolution Statistically Downscaled Future Climate Surfaces. Ramirez Villegas J, Challinor A (2012) Assessing relevant climate data for agricultural applications. Agr icultural and Forest Meteorology , 161 , 26 45. Raxworthy CJ, Pearson RG, Rabibisoa N et al. (2008) Extinction vulnerability of tropical montane endemism from warming and upslope displacement: a preliminary appraisal for the highest massif in Madagascar. Glo bal Change Biology , 14 , 1703 1720. Ridgeway G (1999) 2 Boosting for classification. 172 181. Rodrigues ASL, Gaston KJ (2002) Maximising phylogenetic diversity in the selection of networks of conservation areas. Biological Conservation , 105 , 103 111.

PAGE 109

109 Segura do P, Araújo MB (2004) An evaluation of methods for modelling species distributions. Journal of Biogeography , 31 , 1555 1568. S emarnat Ine (2009) México Cuarta Comunicación Nacional ante la Convención Marco de las Naciones Unidas sobre el Cambio Climático , 1st edn (eds Koob R, Pont M del). Secretaría de Medio Ambiente y Recursos Naturales, Instituto Nacional de Ecología, Mexico, D. F. Sheldon KS, Yang S, Tewksbury JJ (2011) Climate change and community disassembly: impacts of warming on tropical and tempera te montane community structure. Ecology letters , 14 , 1191 200. Sinclair SJ, White MD, Newell GR (2010) How Useful Are Species Distribution Models Ecology and Society , 15 . Sinervo B, Méndez de la Cruz F, Mil es DB et al. (2010) Erosion of lizard diversity by climate change and altered thermal niches. Science , 328 , 894 9. Smith TB (1997) A Role for Ecotones in Generating Rainforest Biodiversity. Science , 276 , 1855 1857. Tewksbury JJ, Huey RB, Deutsch CA (2008) Putting the heat on tropical animals the scale of prediction. Ecology , 320 , 1296 1297. Thessler S, Ruokolainen K, Tuomisto H, Tomppo E (2005) Mapping gradual landscape scale floristic changes in Amazonian primary rain forests by combining ordination and re mote sensing. Global Ecology and Biogeography , 14 , 315 325. Thomas CD, Cameron A, Green RE et al. (2004) Extinction risk from climate change. Nature , 427 , 145 8. Thomassen H a., Buermann W, Milá B et al. (2010) Modeling environmentally associated morpholo gical and genetic variation in a rainforest bird, and its application to conservation prioritization. Evolutionary Applications , 3 , 1 16. Thomassen HA, Fuller T, Buermann W et al. (2011) Mapping evolutionary process: a multi taxa approach to conservation prioritization. Evolutionary Applications , 4 , 397 413. spec ies distribution modeling. 89. Thuiller W (2003) BIOMOD optimizing predictions of species distributions and projecting potential future shifts under global change. Global Change Biology , 9 , 1353 1362.

PAGE 110

110 Thuiller W (2004) Patterns and uncertainties of speci change. Global Change Biology , 10 , 2020 2027. Thuiller W, Brotons L, Araújo MB, Lavorel S (2004) Effects of restricting environmental range of data to project current and future species distributions. Ecography , 2 , 165 172. T huiller W, Lavorel S, Araújo MB (2005) Niche properties and geographical extent as predictors of species sensitivity to climate change. Global Ecology and Biogeography , 14 , 347 357. Thuiller W, Midgley GF, Hughes GO, Bomhard B, Drew G, Rutherford MC, Woodw ard FI (2006) Endemic species and ecosystem sensitivity to climate change in Namibia. Global Change Biology , 12 , 759 776. Travis JMJ (2003) Climate change and habitat destruction: a deadly anthropogenic cocktail. Proceedings of the Royal Society B , 270 , 46 7 73. Trejo I, Dirzo R (2000) Deforestation of seasonally dry tropical forest local analysis in Mexico. Biological Conservation , 94 , 133 142. VanDerWal J, Shoo LP, Williams SE (2009) New approaches to understanding late Quaternary climate fluctuations and refugial dynamics in Australian wet tropical rain forests. J ournal of Biogeography , 36 , 291 301. Walther G, Post E, Convey P et al. (2002) Ecological responses to recent climate change. Nature , 416 , 389 395. Warren MS, Hill JK, Thomas JA et al. (2001) Rapid responses of British butterflies to opposing forces of cli mate and habitat change. Nature , 414 , 65 69. Yates CJ, McNeill A, Elith J, Midgley GF (2010) Assessing the impacts of climate change and land transformation on Banksia in the South West Australian Floristic Region. Diversity and Distributions , 16 , 187 201.

PAGE 111

111 BIOGRAPHICAL SKETCH Maria Juliana Bedoya ear ned her Bachelor of Science in b iology from Universidad del Valle in Cali, Colombia in 2007. Juliana was supported by Fulbright and Colciencias in Colombia to begin her Master s in i nterdisciplinary e cology a t the University of Florida in August 2012 and graduated in August 2014.



PAGE 1

Bodysizeandactivitytimesmediatemammalian responsestoclimatechange CHRISTYM.MCCAINandSARAHR.B.KING DepartmentofEcology&EvolutionaryBiologyandCUMuseumofNaturalHistory,UniversityofColorado,265UCB,Boulder, CO80309,USA Abstract Modelpredictionsofextinctionrisksfromanthropogenicclimatechangearedire,butstilloverlysimplistic.Toreliablypredictat-riskspeciesweneedtoknowwhichspeciesarecurrentlyresponding,whicharenot,andwhattraits aremediatingtheresponses.Formammals,wehaveyettoidentifyoverarchingphysiological,behavioral,orbiogeographictraitsdeterminingspecies’responsestoclimatechange,buttheymustexist.Todate,73mammalspeciesin NorthAmericaandeightadditionalspeciesworldwidehavebeenassessedforresponsestoclimatechange,including localextirpations,rangecontractionsandshifts,decreasedabundance,phenologicalshifts,morphologicalorgenetic changes.Only52%ofthosespecieshaverespondedasexpected,7%respondedoppositetoexpectations,andthe remaining41%havenotresponded.Whichmammalsareandarenotrespondingtoclimatechangeismediatedpredominantlybybodysizeandactivitytimes(phylogeneticmultivariatelogisticregressions, P < 0.0001).Largemammalsrespondmore,forexample,anelkis27timesmorelikelytorespondtoclimatechangethanashrew.Obligate diurnalandnocturnalmammalsaremorethantwiceaslikelytorespondasmammalswithexibleactivitytimes ( P < 0.0001).Amongtheothertraitsexamined,specieswithhigherlatitudinalandelevationalrangesweremore likelytorespondtoclimatechangeinsomeanalyses,whereashibernation,heterothermy,burrowing,nesting,and studylocationdidnotinuenceresponses.Theseresultsindicatethatsomemammalspeciescanbehaviorallyescape climatechangewhereasotherscannot,analogoustopaleontology’sclimateshelteringhypothesis.Includingbodysize andactivityexibilitytraitsintofutureextinctionriskforecastsshouldsubstantiallyimprovetheirpredictiveutility forconservationandmanagement. Keywords: behavior,elevation,latitude,mammal,microclimate,physiology,thermalniches Received22September2013andaccepted4December2013 Introduction Inthepast50years,temperatureshaveincreased,precipitationregimeshavechanged,arcticicehasshrunk, andextremeweathereventshaveincreasedinfrequencyduetoanthropogenicmodicationofclimate (Trenberth etal. ,2007;USGCRP,2009;Duffy&Tebaldi, 2012).Variousplantsandanimalshaveresponded withinindividualpopulationsandacrosstheirgeographicandelevationalranges(Grabherr etal. ,1994; Walther etal. ,2002;Root etal. ,2003;Parmesan,2006; Lenoir etal. ,2008;Chen etal. ,2009;Myers etal. ,2009). Mostpredictivemodelingofmammalsstillassumes thatallspecieswillrespondtoclimatechange,will respondsimilarlyalbeitpossiblyatdifferentrates,and directionallyinstepwithtemperature(e.g.,Lawler etal. ,2009;McCain&Colwell,2011).Incontrast,ectothermvertebratemodelinghasincorporatedphysiologicaldifferencesamongspeciestomakediscriminating predictionsaboutspeciesresponsestoclimatechange (Buckley,2008;Arag on etal. ,2010;Huey etal. ,2012). Individualmammalstudies(e.g.,Moritz etal. ,2008; Myers etal. ,2009)demonstratethatsomemammals haverespondedtoclimatechangewhileothershave not,eventhoughallwerepredictedtorespondsimilarlyinclimateenvelopemodeling.Despitethe acknowledgementofthepotentialimportanceofspecies’traitsinmodifyingresponsestoclimatechange (e.g.,Parmesan,2006;Buckley,2008;Moritz etal. ,2008; Angert etal. ,2011;Huey etal. ,2012),effortstopinpoint suchtraitsforendothermssofarareunsuccessful(Angert etal. ,2011).Nonetheless,abetterunderstandingof whichspeciesarerespondingandifthereareconsistent biologicalreasonsforthoseresponsesisaconservation andmanagementimperative.Itisclearfromanalyses ofmammalianfossilrecordsthatnotallmammals respondedsimilarlyordirectionallywithpaleo-temperaturechange(e.g.,Lyons,2003;Barnosky etal. ,2004; Blois etal. ,2010),butmanyphysiologicalorbehavioral traitscannotbeanalyzedinthefossilrecordduetolack ofpertinentbiologicalinformationortaphonomic biases.Threesetsofgeneralfactorscouldinuence whetherparticularspeciesexhibitresponses — (i)species’traits,particularlythosethatinuenceinteractions Correspondence:ChristyM.McCain,tel.+13037351016;fax+1 3034924195,e-mail:christy.mccain@colorado.edu © 2014JohnWiley&SonsLtd 1760 GlobalChangeBiology(2014) 20 ,1760–1769,doi:10.1111/gcb.12499 Global Change Biology

PAGE 2

withclimate;(ii)locations,particularlyareasthatare experiencinggreaterclimatechange;and(iii)phylogeny,inthatparticularmammalcladesmayberespondingmorethanothersduetosharedevolutionary history.Documentationofmammalianresponsesto recentclimatechangehaveaccumulatedrapidlyin NorthAmerica(147populationsof73species;AppendixS1),andeightspecieshavebeenstudiedinother regions(e.g.,Adamik&Kr al,2008;Lundy etal. ,2010; Moyes etal. ,2011;AppendixS2).Thesedocumented climatechangeresponses,includinglocalpopulation extirpations,rangecontractions,rangeshifts,anddirectionalchangeinabundance,phenology,bodysize,and geneticdiversity(AppendixS1),offeranopportunityto testforspecies’traits,localitytrends,andcladeassociationswhichmayunderliethevariabilityinresponsesto currentclimatechange. Manyspecies’traitscouldinuenceclimatechange responsefrommetabolismtofoodwebinteractions, buttraitsthatdirectlyimpacttheclimateexperienced bytheindividualmaybemostinuential.Paleontologistshavetheorizedthatcertaincharacteristicsofmammals,includingsmallbodysize,nocturnalbehavior, andburrowing,mayhaveallowedthemto‘shelter’ fromtherapidclimatechangeassociatedwiththeK-T extinctioneventintheCenozoicincontrasttothemuch largerdinosaurswhichperished(Robertson etal. , 2004).Evidencelinkingmammalbodysizetodifferentialextinctionratesandlarge-scalerangeshiftsduring theK-TandPleistoceneextinctioneventslendssome supporttotheclimateshelteringtheory(e.g.,Lyons, 2003;Barnosky etal. ,2004;Gingerich,2006;Blois etal. , 2010).Inthecurrentanthropogenicextinctioncrisis, largebodysizehasbeencorrelatedwithhigherextinctionriskstatusontheIUCNRedList,whereasother mammalianspecies’traitslikeburrowingandhibernatingbehaviorshavebeencorrelatedwithlowerextinctionriskstatus(e.g.,Cardillo etal. ,2005;Liow etal. , 2009).Bodysizeofterrestrialmammalsspansordersof magnitudefrom2gto1000kg,andtheenvironments availabletodifferentsizedmammalsvaryconsiderably.Smallmammalscanlivewithinandunder vegetationandsoilwhichmediatestheexperienced temperatureandhumiditylevels,whereaslargemammalsarenecessarilyabovethevegetationandhave fewermicroclimateopportunities(e.g.,Cardillo etal. , 2005;Feldhamer etal. ,2007).Alternatively,largemammalsaremoremobileandmaybemoreabletotrack climatechangeandtodetectisolatedrefugesthatmay beinaccessibletolessmobile,smallmammals(Angert etal. ,2011;Schloss etal. ,2012).Whenamammalis activecanalsoinuencethetemperatureandhumidity rangesitencounters.Particularlyforspeciesobligately activeatcertaintimesoftheday,theymustexperience theabbreviatedrangeoftemperaturesandhumidityof thattimeinterval.Incontrast,speciesthatareexible intheiractivitytimes,forexamplenocturnalinthe summeranddiurnalinthewinter,canselecttherange oftemperaturesandhumidityinwhichtheyareactive (e.g.,Nowak,1991;Feldhamer etal. ,2007).Mammals thatburrowunderthesoiloractivelyconstructnests areabletoliveinamoderatedmicroclimatenotavailabletomammalswithoutthosetraits(e.g.,Kay,1977; Bulova,2002;Cardillo etal. ,2005;Liow etal. ,2009). Lastly,mammalscapableofheterothermy,eitherseasonal(hibernation)ordaily(torpor,estivation),can escaperepeatedextremetemperatureandprecipitation uctuationsthatothermammalscannot(e.g.,marmots vs.caribouinAlaska;e.g.,Feldhamer etal. ,2007;Liow etal. ,2009).Allofthesetraitsinuencetherangeof temperaturesandhumidityamammalexperiencesina particularlocationwheresomeorganismshaveawider arrayofmicroclimaticoptionsandothershavefewer. Thosespeciesthatcanbroadentheirmicroclimate choicesandtherangeoftemperaturesandhumidity experiencedthroughmodicationofhabitatchoices, activitytimes,andbehavioraltemperatureregulation maybelesssusceptibletoclimatechange,whilethose specieswithareducedrangeofoptionsmayhave greaterexposuretoclimatechangeandthereforebe moresensitiveandresponsive. Responsestoclimatechangearepredictedtobeconcentratedgeographically,particularlyathighlatitudes andelevations,andatrangeedges(Grabherr etal. , 1994;Walther etal. ,2002;Parmesan&Yohe,2003;Root etal. ,2003;Hickling etal. ,2006).Thus,variationin responserateamongmammalsmayberelatedtothe severityoftheclimatechangeinthestudylocation.For example,temperaturesareincreasingatafasterrateat highlatitudes(Trenberth etal. ,2007;USGCRP,2009), andmammalpopulationsathigherlatitudesmay respondmore(e.g.,Post&Forchhammer,2008;Gleason&Rode,2009).Speciesrestrictedtothehighest latitudesandhighestelevationsarealreadyneartheir climaticandgeographiclimits,andmayalsobe respondingatahigherrate(Grabherr etal. ,1994; Parmesan&Yohe,2003;Chen etal. ,2009).Similarly, variationinpopulationresponsestoclimatechangeare predictedacrossaspecies’range,aspopulationsatthe warmestrangeedgemaybeexpectedtohavelarger responsesthanpopulationsinthemiddleofaspecies range(e.g.,Beever etal. ,2011).Thus,mammalswithin biogeographicregionsoftheirrangethatareexperiencinglargerchangesinclimatemaybepredictedtoshow strongerresponsestoclimatechange. Finally,whichspeciesdoanddonotrespondto climatechangemaybearesultofphylogeneticrelatedness(Blomberg&Garland,2002;Blomberg etal. ,2003;© 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–1769MAMMALTRAITSDETERMINECLIMATECHANGERESPONSES 1761

PAGE 3

Ives&Garland,2010).Ifcertaingroupsofmammals areconsistentlyrespondingmorethanothersduetoa sharedtraitsleadingtoclimatechangesusceptibility, suchatrendcouldleadtomorepersuasivearguments aboutspeciesriskinareasoftheworldcurrently understudiedforclimatechangeresponse.InNorth America,the73speciesstudiedforresponsesbelongto 11ordersand16families,representingabouthalfofthe worldwidemammalorders.Theclade-basedtrendsin climatechangeresponsecanbeassessedthroughestimatesofphylogeneticsignalinpredictorvariablesand withcomparisonsbetweenphylogenetically-corrected anduncorrectedanalyses(Ives&Garland,2010). Giventhatsomemammalsareandsomearenot respondingtocurrentclimatechange,hereweassess whethervariabilityinspecies’traits,geographiclocation,andmammaliancladeafliationareconsistently relatedtomeasuredresponses.Weassessseveralmammalspecies’traitswhichmayinuencetheirinteraction withachangingclimate,includingbodysize,activity times,burrowingbehavior,andhibernation.Weassess severalbiogeographicfactors,includinglatitudeofgeographicrange,locationofthestudypopulationwithin thegeographicrange,andelevationalrange.Lastly,we assessthedegreeofresponseduetophylogeneticrelatedness.MaterialsandmethodsMammaldataTerrestrialmammalresponsestoanthropogenicclimate changewerelocatedthroughstandardizedliteraturesearches (ISIWebofKnowledge,GoogleScholar,andScirus)usingvariouscombinationsofkeywords(climatechange,globalwarming,mammals)andthosecombinationsalsowithlocations specied(USA,Canada).Intotalweexaminedover54000 citationsofwhichabout1050wererelativelyrelevant.From these,weonlyincludedstudiesthatempiricallyexamined andstatisticallytestedpre-andpost-anthropogenicclimate changeresurveydataorlong-termmonitoringdataacrossa timeperiodofanthropogenicclimaticchangeonnativespeciesbasedontheindividualauthor’sassessmentsofpertinent timeintervals.Wedidnotincludestudiesthatcouldnot excludeanthropogenichabitatchangeasthemaindriverof change,thosethatdidnothavestrongsimilaritiesindata collectionmethodologybetweenthepre-andpost-climate changeperiods,andthosethatonlyexaminedshiftsin responsetocurrentclimateandextendedtheirresultstopredictionsinthefuture(seeDataS1formoredetail).Almostall oftheincludedstudieswerefromNorthAmerica(NA:US, Canada,Greenland;FigureS1,AppendixS1;73species), whereasonlyafewstudiesmetthosecriteriaoutsideofNA (AppendixS2;8atspecieslevel).Tobeconsistentaboutthe degreeofclimatechangeexperiencedbythemammals,here weconcentratedontheresponsesofNAmammals,although theresultsdidnotchangewiththeinclusionofthe8non-NA species’responses.Astrongpublicationbiasisnotapparent intheNAdataasnearlyanequalnumberofpopulationsand speciesareincludedthatrespondedtoclimatechangeasthose thatdidnotrespondorrespondedoppositetopredictions (FigureS2a).Thisisnotthecaseinnon-NAdataasonlyone studypublishedanon-signicantresponse(Poroshin etal. , 2010;AppendixS2).Additionally,species’traitsinresponse datasetsarenearlyidentical(bodysize)orcloselysimilarin distribution(activitytimes,latitudinalmidpointsandmaxima)toallNAterrestrialmammals,thusareunbiasedinrelationshiptotheanalyzedpredictorvariables(FigureS2b e). Dataqualityandstatisticalinferencesstillvariedamong includedstudies;therefore,wealsoanalyzedabestdatasubset(AppendixS4)aswellastwoadditionalrestricteddatasets onextinctionriskresponsesandcontrastingcontractingand expandingelevationalranges(seebelow). Studiespredictedsevenclimatechangeresponses:local populationextirpations,rangecontractions,rangeshifts,and directionalchangeinabundance,phenology,bodysize,or geneticdiversity.Responseswere‘expected’ifthatwasthe predictedresponsespeciedintheparticularstudyoraccordingtoclimatechangetheory(e.g.,upwardelevationalor polewardrangeshift),‘unexpected’iftheyweretheopposite responsethanpredicted(e.g.,rangeexpansion),and‘no change’ifnoresponsewasdetected.Localpopulationextirpationsweredenedascompletelossofaspeciesfromaknown localityorregion.Rangecontractionsweredenedasanoveralldecreaseinareaofthelatitudinalorelevationalrangeofa species.Rangeshiftsweredenedasalatitudinalorelevationalshiftupwardordownwardofaspecies’rangebeyond historicallimitswithoutnecessarilychangingrangesize.At thelatitudinalorregionallevel,manystudiesdidnotassess theentiretyofaspecies’rangebutratherdetectedchangesat localsites(e.g.,Gleason&Rode,2009;Myers etal. ,2009), thereforetheseareconsideredrangeshifts.Sincerangecontractionandrangeshiftcanbothbeexpectedresponsesalong elevationalgradients(e.g.,Moritz etal. ,2008;Rowe etal. , 2010;Beever etal. ,2011)(andlatitudinalgradients,although nonearetestedsofarformammals),wedenedthe‘expected’ trendascontractioniftheupperrangelimitabuttedorwas closetothemountaintop,whereasthe‘expected’trendfor rangeshiftwasupslopeiftheupperlimitwasatasufciently lowelevationtoallowbothrangelimitstoshiftupwardwithoutachangeinrangesize.The‘unexpected’responsetocontractionwasrangeexpansion,withanincreaseinrangesize, andthe‘unexpected’responsetoanupsloperangeshiftwasa downsloperangeshiftwithoutanincreaseinrangesize. Populationabundancechangesweredetectedwith increasesordecreasesinlocalabundanceincomparisonto valuesmeasuredbeforedetectableclimatechange,andpredictionsdependedonthehypothesespresentedinthestudy. Mammalianphenologicalchangesassessedshiftsintimingof reproductiveevents,hibernation,andotherphysiological events,andinallcasesincludedherewereexpectedto advanceintheirtimingwithincreasingtemperatures.Mammalianmorphologicalchangesassociatedwithclimatechange weremostlyconcernedwithbodysizeshifts.Smallerbody sizeswereexpectedwithincreasingtemperatures(e.g.,Smith© 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–17691762 C.M.MCCAIN&S.R.B.KING

PAGE 4

etal. ,1998)orincreasesinbodysizewereexpectedwitha longergrowingseason(e.g.,Ozgul etal. ,2010),althoughsome studiesalsodiscussedothermorphologicalandgenetictrends withclimatechange. Weexaminedalltheseresponsesasawholeforalldatasets, including147populationresponsesfrom73species(AppendixS1andS3).Wealsoanalyzedabestdatasubset(80populationsof49species;AppendixS4),whichincludedonlythose studieswithgreaterthan5yearsofsampling,leastamountof potentialanthropogenichabitatchange,andwiththestrongest statisticalinference — meaningthestudyshowed(i)astrong, measurablerelationshipbetweenclimatechangeandthe mammalianresponsevariable;(ii)thesamplingeffortper methodofassessingmammalianresponsesacrossthemonitoringperiodorthehistoricalandcurrentperiodswasrelativelyconsistent;and(iii)theclimatechangesignalwas showntobestrongerstatisticallythanotherexplanatoryvariables.Becausenotalloftheexpectedresponsestoclimate changearepotentiallydetrimentaltothespecies,wealsoanalyzedasubsetofresponsesthatarepotentiallylinkedtoa higherextinctionrisk:localpopulationextirpations,contractingranges,anddecreasingpopulationsizes(66populationsof 35species;AppendixS3).Lastly,wecomparedmammalsthat hadcontractedtothosethathadexpandedtheirelevational ranges(36populationsof19species;AppendixS3). Species’traitsandbiogeographiccharacteristicswere assessedfromthestudiesthemselvesandfromtheliterature. BodysizeswerefromtheMOMdatabase(Smith etal. ,2003) except Tamiasminimus whichwasfromthePanTHERIA database(Jones etal. ,2009).Dailyactivitytimes(obligate diurnal,obligatenocturnal,andexible[crepuscular,combinationorexibilityofnocturnal,diurnal,crepuscular])were fromthePanTHERIAdatabase(Jones etal. ,2009),species accountsinthejournal MammalianSpecies ,andchecked againstadditionalliteraturesources(Hall,1981;Nowak, 1991;Armstrong etal. ,2011).Burrowing,nestingbehavior, hibernation,andtorporwerecompiledfromvariousliteraturesources(Hall,1981;Nowak,1991;Matocq&Murphy, 2007;Armstrong etal. ,2011)and MammalianSpecies accounts. Latitudinalrangemidpointsandmaxima,andstudylocationswithinNAgeographicrangeswerecalculatedfromthe PanTHERIAdatabase(Jones etal. ,2009)butmodiedtoonly includedistributionswithinNA(Hall,1981;Matocq&Murphy,2007;IUCN,2011).Forcontractingandexpanding ranges,elevationalrangemidpointwasaninaccuratereectionofoverallrangepositionduetothebroadelevational rangesofmostspecies.Therefore,arangewasnotedas(i) predominantlylowelevation( < midpointofmountain)orall elevationsvs.;(ii)predominantlyhighelevation( > midpoint ofmountain).Becauseactivitytimesmaynotbeaswellknownamongspeciesthataredifculttoseeandmonitor, thedeterminationofexiblespeciesmightbeunder-represented.Therefore,weaddeda‘potentiallyexible’category tothespeciesifaliteraturesourcesuggestedthatitmaybe activeatothertimesorifasisterspeciesorcloselyrelated specieswaswell-knowntobeexibleinitsactivitytimes (AppendixS3andS4).Wethenconductedcomparison analysestodetectifthisbroaderdatasetinuencedthe resultsforthetwodatasetswithsufcientnumbersofpotentiallyexiblespecies(alldata,bestdata).StatisticalanalysesAnalyseswereconductedatthelevelofspeciestoaccommodatephylogeneticcorrections(Blomberg&Garland,2002; Blomberg etal. ,2003;Ives&Garland,2010)andavoidmultiplepopulationbias.Compositeresponseswereusedforspecieswithmultiplestudypopulations.Ifoneormore populationshaddetectionsofexpectedresponsestoclimate change,thenapositiveresponsewasnotedforthatspecies evenifotherpopulationsdetectednochange.Thesecompositeresponseschangedbetweenthecompletedataset(AppendixS3)andthebestsubset(AppendixS4)forafewspecies withmultiplestudiedpopulationsofvaryingquality. Theclimatechangeresponseswereanalyzedasabinomial — eitheranexpectedresponse(1)ornoresponse(0:unexpectedornoresponse)perspecies.Theunexpectedresponses weretoofewtobeanalyzedseparately(eightspecies)and werealmostallelevationalexpansions(AppendixS1;Figure S1).Thus,thesewereanalyzedspecicallyinthecomparison ofcontractingandexpandingelevationalranges(Appendix S3).Onlytheelevationalshiftswerequantiableasmeters changed,butpreliminaryanalyseswereweak,mainlydueto smallsamplesizesandrestrictedshiftmagnitudes.More quantiedshiftstudiesareneededbeforerobusteffect-size meta-analysesarepossible.Asnotedintheappendices,for speciestraitandbiogeographicvariables,predictorswereanalyzedascontinuousquantitativevariables,two-statevariables orthree-statevariables.Foractivitytimes(obligatediurnal, obligatenocturnal,andexible),twodichotomieswere strongest(exiblevs.obligatediurnalandnocturnal,andobligatediurnalvs.exibleandnocturnal)sobothcontrastswere includedastwo-statevariablesinthemultivariatemodels. Inordertodetectthebest-tmodelsandthestrongest predictorvariables,werstconductedmultivariate,nominal logisticregressionsforallpossiblemodelsusingallcombinationsofpredictorvariables(Burnham&Andersen,2002). Second,wecalculatedAICcweightsforeachmodelandthe modelwiththelargestAICcweightwasusedasthebest-t model(Burnham&Andersen,2002).Third,todetectthe inuenceofphylogeneticrelatednessoneachpredictorvariableandwithinthebest-tmodels,weranbothmultivariate ordinaryandphylogeneticlogisticregressions(OLR,PLR) usingtheprogram PLogreg inMATLAB(Ives&Garland, 2010).Phylogenieswerefromthemammalsupertree (Bininda-Emonds etal. ,2007)prunedtothetaxaincludedin eachdataset. PLogreg alsoestimatedthephylogeneticsignal withcondenceintervalsforeachvariableormodel,which allowsdeterminationofresponsedifferencesamongmammalianclades.Phylogeneticsignalin PLogreg variesfromno signal( 4)tostrongsignal( > 0)(Ives&Garland,2010).Body sizewaslog-transformed,allpredictorvariableswerestandardizedtoameanofzeroandavarianceofone,andall bootstrappedcondenceintervalscalculatedwith2000simulationsfollowingIves&Garland(2010). PLogreg wasnot usedintheoriginalassessmentofthebest-tmodels,© 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–1769MAMMALTRAITSDETERMINECLIMATECHANGERESPONSES 1763

PAGE 5

becauseamaximumofonlythreepredictorsvariableswas possibleinasinglemodelwithoursamplesizestoproduce reliableresults(e.g.,modelconvergence,completedsimulations)and PLogreg cannotcalculateAICcweights(Ives& Garland,2010).Lastly,duetothesmallsamplesizesforcontractingandexpandingelevationalranges( n = 19),only3 variableswereneededforacompletetmodel( r2= 1.0), thusAICcweightswerecalculatedonlyformodelsofthree variablesorless.Formoredetailsonanalysesandmodels seetheDataS1.ResultsOnly52%ofmammalpopulationsrespondedas expectedtorecentclimatechange.Incontrast,7% respondedoppositetoexpectationsand41%hadno detectableresponse(FigureS1;AppendixS1).The majorityofmammalianpopulationswereassessedfor rangecontraction,rangeshifts,andabundancechanges (82%)withfewerstudiesexaminingotherresponses (localextirpations,phenology,morphologicalorgenetic changes:18%).Thebest-tmodelsforthevariousdatasetsallincludedtwospecies’traits,bodysizeandactivitytimes,whereaslatitudinalandelevationalrange wereonlyincludedinafewmodels(TablesS1andS2). Incontrast,burrowing,nesting,hibernation,heterothermy,andstudylocationwithingeographicrangewere notincludedinanyofthebest-tmodels,andwere rarelysignicantinsinglevariablemodels(TablesS1 andS2).Forthedatasetofallspeciesandresponses(73 species),thebest-tmodelincludedonlybodysizeand exibleactivitytimes( r2= 0.22; X2= 22.09, P < 0.0001). Thebest-tmodelsforthebestsubsetofspecies(49species)andtheextinctionriskdataset(35species)included bodysize,exibleactivitytimes,latitudinalrange ( r2= 0.50; X2= 32.60, P < 0.0001; r2= 0.71; X2= 29.54, P < 0.0001,respectively).Lastly,thebest-tmodelfor thespeciescontractingandexpandingtheirelevational ranges(19species)includedbodysize,diurnalactivity times,elevationalrange( r2= 1.0; X2= 23.70, P < 0.0001).Thestatisticalsignicanceofthebest-tmodels andstrongestpredictorvariableswereidenticalwith andwithoutphylogeneticcorrection(TablesS1andS2). Bodysizewasthestrongestsinglepredictorofclimatechangeresponseacrossalldatasets(Fig.1). Expectedresponsestoclimatechangeincreasedwith bodysizewithandwithoutphylogeneticcorrectionsin allbest-tmodels(TableS2;OLR P -values:0.005 – < 0.0001;PLR P -values:0.045 – < 0.0001).Forthemammalsstudied,bodysizesrangedfrom2.5gto388kg, andmammalsatabout100gshiftedtopredominantly respondingtoclimatechange.Thelargestmammals, e.g.,elkandpolarbear,were27timesmorelikelyto respondtoclimatechangethanwerethesmallest mammals,e.g.,shrewsandmice. Specieswithexibleactivitytimeshadsignicantly lowerresponseratestoclimatechangethanobligate diurnalornocturnalspecies(Fig.2a;multivariateOLR (a) (b) (c) (d) Fig.1 Responsestoanthropogenicclimatechangeincreasewith mammalianbodysizeshownusingordinary(redline)andphylogeneticlogisticregression(blackline).(a)Allspeciesandall responsestoclimatechange.(b)Bestdatasubsetandallresponses toclimatechange.(c)Onlythosespeciestestedforresponses potentiallylinkedtoextinctionrisk(localpopulationextirpation, rangecontraction,decreasingpopulationsize).For(a – c),climate changeresponses:expectedresponse = 1,andnoresponseor unexpectedresponse = 0.(d)Largerspeciesaremorelikelyto contracttheirelevationalranges(1),whereassmallermammals aremorelikelytoexpandtheirelevationalranges(0). P -values arefromsinglevariable,ordinarylogisticregressionmodels. Zerosintheguresindicatewherecertainbodysizerangeshave anaverageresponsetoclimatechangeofzero(i.e.noresponse). © 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–17691764 C.M.MCCAIN&S.R.B.KING

PAGE 6

P -values:0.012 – 0.001;multivariatePLR P -values:0.018 – 0.004).Diurnalspeciescontractedtheirelevational ranges,whereasexibleandnocturnalspecieswere morelikelytoexpandtheirranges(Fig.2b;multivariate P -values:OLR < 0.0001;PLR = 0.020).Analyses wererobusttoincludingmorespeciesaspotentially exible(singleandmultivariatemodelsstillsignicant; TablesS1andS2).Activitywasnotjustafunctionof bodysizeasthetwovariableswerenotsignicantly associated( r = 0.012, P = 0.349),andbothvariables wereincludedandsignicantinallfourbest-t,multivariatemodels. Thebiogeographicvariableswerenotasstronglyor consistentlyrelatedtomammalianresponsestoclimate changeaswerespecies’traits(TablesS1andS2).Only latitudinalrangemeasuredaseithermidpointormaximumhadsomeconsistentsupportamongdatasets (Fig.3).Climatechangeresponsesincreasedwith increasinglatitudeofspecies’rangesinmostsinglevariablemodels(exceptallresponses),andinbest-t modelsforthebestdataandextinctionriskdata(Table S2;best:OLR P < 0.0001and0.002;PLR P < 0.0001and 0.021,respectively).Forchangingelevationalranges (Fig.4),specieswithhigherelevationrangespredominantlycontractedtheirranges,whereasspecieswith lowerelevationrangesexpandedtheirranges(OLR P = 0.012,PLR P = 0.033). Estimatesofphylogeneticsignalusing PLogreg were lowamongalldatasetsandvariables:allbootstrapped condenceintervalsincluded 4,whichcorrespondsto nodetectablephylogeneticsignal(TablesS1andS2; Ives&Garland,2010).Nonetheless,allpointestimates ofphylogeneticsignalwerelargerthan 4anddenote asmalllevelofphylogeneticrelatednessinuencing thedistributionofclimatechangeresponses.Thus,variabilityinresponsesexistedwithinandacrossclades, andparticularmammaliancladeswerenotassociated stronglywithclimatechangeresponses.Regardless, boththeordinaryandphylogeneticlogisticregressions consistentlysupportedthesamevariablesasstrongly linkedtothedifferencesinclimatechangeresponses, albeitthephylogenetically-corrected P -valuesare generallyslightlylarger(TablesS1andS2). (a) (b) Fig.2 Responsestoanthropogenicclimatechangedecrease withexibleactivitytimesusingordinaryandphylogenetic logisticregressions.(a)Forallspecies(grey; n = 72, P = 0.001), thebestsubset(black; n = 49, P < 0.0001)andextinctionrisk responses(white; n = 35, P = 0.016).Climatechangeresponses: expectedresponses = 1,andnoresponseorunexpected response = 0.(b)Thespeciescontractingtheirelevational ranges(1)arealldiurnal,whereasspeciesthatexpandedtheir elevationalranges(0)wereeithernocturnalorexibleintheir activitytimes( n = 19, P < 0.0001). P -valuesarefromsinglevariable,ordinarylogisticregressionmodels. (a) (b) Fig.3 Responsestoanthropogenicclimatechangeincrease withthelatitudeofaspecies’rangeshownforlatitudinalmaximum(midpointsarenearlyidentical;notshown)usingordinary(redline)andphylogeneticlogisticregression(blackline). (a)Bestdataandallresponsestoclimatechange( n = 49, P < 0.0001);(b)onlythosespeciestestedforresponsespotentiallylinkedtoextinctionrisk(localpopulationextirpation, rangecontraction,decreasingpopulationsize):( n = 35, P = 0.005).Climatechangeresponses:expectedresponse = 1; noresponseorunexpectedresponse = 0. P -valuesarefromsinglevariable,ordinarylogisticregressionmodels. © 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–1769MAMMALTRAITSDETERMINECLIMATECHANGERESPONSES 1765

PAGE 7

DiscussionMammalianstudiesexaminingcurrentclimatechange responsesclearlydemonstratethatnotallmammalsare respondingorrespondingsimilarlytoclimatechange. Somemammalsarerespondingnegativelywithlocal populationextirpations,rangecontractions,and decreasingpopulationsizes,forexamplepolarbears, pika,andShadowchipmunks(Moritz etal. ,2008;Gleason&Rode,2009;Beever etal. ,2011).Othermammals arerespondingpositivelybyincreasingranges,populationsize,orgrowthrates,forexampleseveralshrews andYellow-belliedmarmots(Moritz etal. ,2008;Ozgul etal. ,2010;Rowe etal. ,2010).Andasubstantialportion ofmammalshavenotrespondedatall,nearlyhalfof thosestudiedsofar.Otherstudiesdocumentingmultiplespecies’responsestoclimatechangeinothertaxonomicgroupsincludinginsectsandplantshave detectedrespondersandnon-responders(Grabherr etal. ,1994;Lenoir etal. ,2008;Chen etal. ,2009;Angert etal. ,2011).Identicationofbiologicalreasonsunderlyingthisvariationinresponsetoclimatechangewillaid inourabilitytodeveloprobustandtargetedconservationplansandmakemoreinformedmanagement decisionsforthosespeciesmostatrisk. Variabilityinclimatechangeresponsemaybedueto species’traitsthatinuencehowtheyinteractwith theirclimaticenvironment(sensitivity)andhowpronouncedtheclimatechangeisinaparticularsite(exposure)(e.g.,Huey etal. ,2012).Forectothermvertebrates, researchershavesuggestedthatbehavioralthermoregulation,acclimation,andadaptationaswellasotherlife historycharacteristicscouldhavecascadinginuences onspecies’responses(e.g.,Buckley,2008;Huey etal. , 2012).Forendothermstherehavebeenfewersuggestionsofimportantspecies’traitsduetotheassumption thattheyarenotasdirectlytiedtoambientconditions asectotherms.Andpreviousattemptstodetectspecies’ traitslinkedtothemagnitudeofclimatechange responsesinbirdsandmammalshavenotbeen successful(Angert etal. ,2011).Paleontologistshave suggestedthatcertainmammaltraitslikebodysize, burrowing,andnocturnalityhelpedthemsurvive extremeclimaticevents(K-Textinction;Robertson etal. ,2004).Insupportoftheshelteringhypotheses,we robustlydetectedtwomammaliantraitsthatinuence anendotherm’sresponsebasedonvariabilityinaspecies’exposureandsensitivitytoprevailingclimatic conditions — bodysizeandactivitytimes. Bodysizeandactivitytimeswereconsistentlyand stronglyrelatedtothevariabilityinmammalian responsestoclimatechange.Expectedresponses increasedwithbodysizeforalltypesofresponsesand justforresponsesindicatingpotentialextinctionrisk (i.e.localextirpations,rangecontractions,andpopulationdeclines).Pleistocenemega-faunalextinction includedpredominantlylarge-bodiedmammals(Barnosky etal. ,2004).ThelargestPleistocenerangeshifts weredetectedamongthelargebodiedmammals, includingthelargerrodents(Lyons,2003).Incontrast, thesmallestbodiedmammalsareconsistentlyrepresentedintheNorthAmericanfossilrecordacrossthe past80Myr(Alroy,1998),andtheysurvivedtheCretaceous-Tertiaryboundary,whilelarger,non-aviandinosaursbecameextinct(Robertson etal. ,2004;Lloyd etal. ,2008).DuringthePleistocene,smallmammals didnotshowanincreasedextinctionrateorasmany largerangeshiftsaslargermammals(Lyons,2003; Barnosky etal. ,2004;Blois etal. ,2010).Thisbodysize trendmayindicateclimaticinteractionsnotpreviously consideredinmammalianresponsestocurrentclimate change.Smallermammalsmayexperienceclimate differentlythanlargermammalsduetodifferential availabilityofmicrohabitatsandthustheirmicroclimatesnearandunderthesoilandvegetationwhere bothtemperatureandhumidityismoderated(Kay, 1977;Bulova,2002;Feldhamer etal. ,2007;Porter& Kearney,2009;Scherrer&K   orner,2011;Suggitt etal. , 2011;Huey etal. ,2012;Scheffers etal. ,2013).Larger (a) (b) Fig.4 Mammalspeciesthathavecontractedtheirelevational rangewerepredominatelydetectedathighelevation(a), whereasspeciesthathaveexpandedtheirelevationalrange werepredominatelydetectedatlowelevationoracrossboth lowandhighelevations(b).Higherresponseratesareshownin redandlowerresponseratesinyellow;andarrowedlinesrepresenttheelevationalrangesofthegroupsincludedineach analysis.Thisrelationshipbetweencontractingandexpanding elevationalrangeswassignicantinsinglevariable,ordinary logisticregressionmodels( P = 0.031). © 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–17691766 C.M.MCCAIN&S.R.B.KING

PAGE 8

mammalsmayhavelessexibilityintheclimateand temperaturestheyencounterabovetheunderstoryvegetationduetotheirnecessaryexposuretoambientair temperature,humidityanddaily-seasonalvariability. Thismaysuggestthatclimategrainsizescaleswith bodysize(sensuRitchie,2010),andthatthelargergrain ofclimateencompassesmoreclimatechangethandoes thesmallergrainofclimateavailabletosmallermammals.Additionally,mammalbodysizehasrepercussionsonmanylifehistorycharacteristicsfrom metabolicandreproductiveratestopopulationsizes andlongevitythatmaynotbeasdirectlyrelatedto climate,buthavepredictablerelationshipswithrarity andgeneralextinctionrisk(e.g.,Cardillo etal. ,2005; Feldhamer etal. ,2007;Moritz etal. ,2008;Liow etal. , 2009).Largemammalsmaybemoremobileandthereforebetterabletoexpandtheirrangestotrackchanges andencounterrefuges(e.g.,Angert etal. ,2011;Schloss etal. ,2012).However,notonlymaythisbeconfounded byhabitatlossorfragmentation,butourresultthat largemammalsarecontractingtheirrangesand decreasinginabundancemorethansmallmammals indicatesastrongernegativeimpactofclimatechange thanasimpleincreasedmobilityresponse. Thevariabilityofactivitytimesamongmammals fromnocturnalordiurnaltobehavioralexibilityalso hadstrongrepercussionsontheheterogeneityin climatechangeresponses.Thespecieswithexibilityin activitytimes — abilitytoswitchbetweensomecombinationofnocturnal,diurnal,andcrepuscularactivity — weretheleastlikelytorespondorrespondnegatively toclimatechange.Infact,manyofthesespecieswere thosethatrespondedpositivelybyincreasingtheirelevationalrangesorabundances.Whereasthosethat wereobligatelyactiveataparticulartimeofday respondedatmuchhigherratesandwithmorenegativeresponses.Inparticular,onmountainsitappears thatthosespeciesobligatelydiurnalaremostatrisk, andthismaybeduetothecoldadaptedspeciesathigh elevationpossessingalowerphysiologicaltolerancefor thewarmestdailytemperatures(e.g.,pika:Beever etal. , 2011andreferencestherein).Thisactivitytraitresult, althoughsurprising,pinpointsanimportantmammaliantraitdifferencethathascascadingimportanceon theabilityofaspeciestobehaviorallyinuencetheir experiencedclimate.Flexiblespeciescanselectaparticularsetofclimaticconditionsthatmaybesignicantly differentfromtheambientconditionsfromwhicha nocturnalordiurnalspeciescannotescape.This species’traitmayhavebeensimilarlyimportantin paleo-mammals,butisnotatraiteasilyidentiedwith fossilevidence. Temperaturesaremoremoderateandlessvariable nearthesoilandundervegetativegroundcoverduring bothsummerandwinter,andarealsomorestablein undergroundenvironments(Kay,1977;Bulova,2002; Feldhamer etal. ,2007),thusburrowingandnesting mammalsmayhavelessexposuretochangingclimate. Butthesetraitswerenotstronglylinkedtocurrent climatechangeresponses.Thismaybebecausethisisa basaltraitwithinmostmammals.MostMesozoicmammalsarethoughttobeburrowersandnesters,andthis basaltraitdominatesinextantmammals(Hall,1981; Nowak,1991).Amongthemammalsstudiedfor climatechangeresponses,fewerthanvespeciesdid notburrowornest(e.g.,artiodactyls,somelargecarnivores),whiletherestburrowed,builtnests,orboth. Burrowingandnestingtraitsmaybelessimportantfor climateresponseswithinmammalsthanamongvertebrateclades.Hibernationanddailytorporallowsome mammalstosurviveextremetemperaturesandlow foodresourceconditionsbymetabolicmanipulations (Feldhamer etal. ,2007;Armstrong etal. ,2011).Despite thestrongassociationwithclimaticadaptation,hibernationandtorportraitswerenotlinkedtothevariabilityinresponsestocurrentclimatechange,suggesting thatspecieswithorwithoutheterothermycapabilities arenotparticularlymoreorlesssensitivetoclimate change,atleastinNorthAmerica. Responsestoclimatechangearepredictedtobeconcentratedgeographically(Grabherr etal. ,1994;Walther etal. ,2002;Parmesan&Yohe,2003;Root etal. ,2003; Hickling etal. ,2006),althoughbiogeographicfactors wereshowntobesecondarytospecies’traitsinthis study.Higherlatitudespeciesweremorelikelyto respondtoclimatechange,butonlyinthebestdata subsetandextinctionriskresponses.Highelevation specieswerestronglyassociatedwithrangecontraction formontanemammals.Despitethestrongertrendsfor bodysizeandactivity,highlatitudeandelevationstill appeartobehighlysusceptibleenvironments. Otherreasonsfornodetectableclimatechange responsemaybeduetoadditionalanthropogenic causes,forexamplesurveydataareincompleteor sparse,oranthropogenichabitatmodicationsarestronger(Walther etal. ,2002;Parmesan&Yohe,2003;Root etal. ,2003;Lenoir etal. ,2008;Rowe etal. ,2010).Additionally,timelags,non-climaticallydeterminedrange limitorabundancetrends,strongphenotypicplasticity, andlargertolerancelimitsthancurrentlyexpressedmay leadtonotdetectingaclimatechangeresponse(Walther etal. ,2002;Parmesan&Yohe,2003;Root etal. ,2003; Lenoir etal. ,2008;Rowe etal. ,2010).Inouranalyses,we controlfordataqualityandresponsestootheranthropogenichabitatchangesthroughacomparisonofall responsedatatoabestdatasubset.Wecannotcontrolor testfortimelagsorlargertolerancelimitsthancurrently exhibiteduntilmoredatabecomeavailable.© 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–1769MAMMALTRAITSDETERMINECLIMATECHANGERESPONSES 1767

PAGE 9

Insummary,ourresultssuggestthatlarge-bodied, obligatorydiurnalornocturnalmammalsarerapidly respondingtocurrentclimatechangeandmanyof theseresponsesindicatehigherextinctionrisks.Mostof thesemammalsarethecharismaticfaunaofNorth America:bighornsheep,pika,andpolarbears.Those leastlikelytorespondorthatareexpandingtheir rangesarethosesmall,exible,unseenmammals:the shrewsandmicelivinginthesoilandunderthevegetation.Microclimateavailabilityandbehavioralchoices onactivitytimes,evenfortwospeciesonthesameplot ofland,therefore,havecascadingimplicationsforthe climateexperienced,anddictatethenecessityof respondingbytrackingaclimatenicheorescaping fromchange(Porter&Kearney,2009;Kearney etal. , 2010;Scherrer&K   orner,2011;Suggitt etal. ,2011;Huey etal. ,2012;Scheffers etal. ,2013).Moreresearchdocumentingtheclimatechangeresponsesinmammals throughouttheworldisurgentlyneeded,andthese resultsgiveaframeworkforassessingmammaltraits andthepertinentclimatechangescalemostlikelyassociatedwithdifferentialresponsestoclimatechange. Sincemostpredictivemodelsofclimatechangerisk treatallmammalsasequallylikelytorespond(e.g., Moritz etal. ,2008;McCain&Colwell,2011),including bodysizeandactivitytimeswillimprovepredictions withevolutionaryandbiologicalrealismtoincrease utilityforconservationplanning.AcknowledgementsThisworkwassupportedbytheUSNationalScienceFoundation(McCain:DEB0949601).Wethankallresearcherswhose datawereusedinthisstudy,andRobertAnderson,Robert Colwell,DanielDoak,JaelynEberle,NormanSlade,andanonymousreviewersforfeedbackonmanuscriptdrafts.AuthorcontributionsC.M.M.formulatedtheideas,conductedtheanalyses, andwrotethemanuscript.S.R.B.K.conductedthe literaturesearch,helpedcollectthepapersanddata, andprovidedfeedbackonideas,analyses,andmanuscriptrevisions.ReferencesAdamikP,Kr alM(2008)Climate-andresource-drivenlong-termchangesindormice populationsnegativelyaffecthole-nestingsongbirds. JournalofZoology , 275 ,209 – 215. AlroyJ(1998)Cope’sruleandthedynamicsofbodymassevolutioninNorthAmericanfossilmammals. Science , 280 ,731 – 734. AngertAL,CrozierLG,RisslerLJ,GilmanSE,TewksburyJJ,ChuncoAJ(2011)Do species’traitspredictrecentshiftsatexpandingrangeedges? EcologyLetters , 14 , 677 – 689. Arag onP,Rodr guezMA,Olalla-T arragaMA,LoboJM(2010)Predictedimpactof climatechangeonthreatenedterrestrialvertebratesincentralSpainhighlights differencesbetweenendothermsandectotherms. AnimalConservation , 13 ,363 – 373. ArmstrongDM,FitzgeraldJP,MeaneyCA(2011) MammalsofColorado .University PressofColorado,Denver,CO. BarnoskyAD,KochPL,FeranecRS,WingSL,ShabelAB(2004)Assessingthecauses oflatePleistoceneextinctionsonthecontinents. Science , 306 ,70 – 75. BeeverEA,RayC,WilkeningJL,BrussardPF,MotePW(2011)Contemporaryclimate changealtersthepaceanddriversofextinction. GlobalChangeBiology , 17 ,2054 – 2070. Bininda-EmondsORP,CardilloM,JonesKE etal. (2007)Thedelayedriseofmodern mammals. Nature , 444 ,93 – 96. BloisJL,McguireJL,HadlyEA(2010)Smallmammaldiversitylossinresponseto late-Pleistoceneclimaticchange. Nature , 465 ,771 – 774. BlombergSP,GarlandT(2002)Tempoandmodeinevolution:phylogeneticinertia, adaptationandcomparativemethods. JournalofEvolutionaryBiology , 15 ,899 – 910. BlombergSP,GarlandT,IvesAR(2003)Testingforphylogeneticsignalincomparativedata:behavioraltraitsaremorelabile. Evolution , 57 ,717 – 745. BuckleyLB(2008)Linkingtraitstoenergeticsandpopulationdynamicstopredict lizardrangesinchangingenvironments. AmericanNaturalist , 171,E1 – E19. BulovaSJ(2002)Howtemperature,humidity,andburrowselectionaffectevaporativewaterlossindeserttortoises. JournalofThermalBiology , 27 ,175 – 189. BurnhamKP,AndersenDR(2002) Modelselectionandmulti-modelinference:apractical information-theoreticapproach .Springer-Verlag,Heidelberg. CardilloM,MaceGM,JonesKE etal. (2005)Multiplecausesofhighextinctionriskin largemammalspecies. Science , 309 ,1239 – 1241. ChenIC,ShiuHJ,BenedickS etal. (2009)Elevationincreasesinmothassemblages over42yearsonatropicalmountain. ProceedingsoftheNationalAcademyofSciences oftheUnitedStatesofAmerica , 106 ,1479 – 1483. DuffyPB,TebaldiC(2012)Increasingprevalenceofextremesummertemperaturesin theU.S. ClimaticChange , 111 ,487 – 495. FeldhamerGA,DrickamerLC,VesseySH,MerrittJF,KrajewskiC(2007) Mammalogy: adaptation,diversity,ecologyBaltimore .TheJohnsHopkinsUniversityPress,Maryland. GingerichPD(2006)EnvironmentandevolutionthroughthePaleocene – Eocenethermalmaximum. TrendsinEcology&Evolution , 21 ,246 – 253. GleasonJS,RodeKD(2009)Polarbeardistributionandhabitatassociationreect long-termchangesinfallseaiceconditionsintheAlaskanBeaufortSea. Arctic , 62 , 405 – 417. GrabherrG,GottfriedM,PauliH(1994)Climateeffectsonmountainplants. Nature , 369 ,448. HallER(1981) MammalsofNorthAmerica .JohnWiley&Sons,NewYork. HicklingR,RoyDB,HillJK,FoxR,ThomasCD(2006)Thedistributionsofawide rangeoftaxonomicgroupsareexpandingpolewards. GlobalChangeBiology , 12 , 450 – 455. HueyRB,KearneyMR,KrockenbergerA,JaMHoltum,JessM,WilliamsSE(2012) Predictingorganismalvulnerabilitytoclimatewarming:rolesofbehaviour,physiologyandadaptation. PhilosophicalTransactionsoftheRoyalSocietyB:Biological Sciences , 367 ,1665 – 1679. IUCN(2011)IUCNRedListofThreatenedSpecies.Version2011.2.http://www. iucnredlist.org(accessedFeb–Aug2012). IvesAR,GarlandT(2010)Phylogeneticlogisticregressionforbinarydependentvariables. SystematicBiology , 59,9 – 26. JonesKE,BielbyJ,CardilloM etal. (2009)PanTHERIA:aspecies-leveldatabaseoflife history,ecology,andgeographyofextantandrecentlyextinctmammals. Ecology , 90 ,2648 – 2648. KayFR(1977)Environmentalphysiologyofthebanner-tailedkangarooratsII.Inuencesoftheburrowenvironmentonmetabolismandwaterloss. ComparativeBiochemistryandPhysiologyPartA:Physiology , 57 ,471 – 477. KearneyMR,WintleBA,PorterWP(2010)Correlativeandmechanisticmodelsof speciesdistributionprovidecongruentforecastsunderclimatechange. ConservationLetters , 3 ,203 – 213. LawlerJJ,ShaferSL,WhiteD,KareivaP,MaurerEP,BlausteinAR,BartleinPJ(2009) Projectedclimate-inducedfaunalchangeintheWesternHemisphere. Ecology , 90 , 588 – 597. LenoirJ,GegoutJC,MarquetPA,DeRuffrayP,BrisseH(2008)Asignicantupward shiftinplantspeciesoptimumelevationduringthe20thcentury. Science , 320 , 1768 – 1771. LiowLH,ForteliusM,LintulaaksoK,MannilaH,StensethNC(2009)Lowerextinctionriskinsleep-or-hidemammals. AmericanNaturalist , 173 ,264 – 272. LloydGT,DavisKE,PisaniD etal. (2008)Dinosaursandthecretaceousterrestrial revolution. ProceedingsoftheRoyalSocietyB:BiologicalSciences , 275 ,2483 – 2490.© 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–17691768 C.M.MCCAIN&S.R.B.KING

PAGE 10

LundyM,MontgomeryI,RussJ(2010)Climatechange-linkedrangeexpansionof Nathusius’pipistrellebat, Pipistrellusnathusii (Keyserling&Blasius,1839). Journal ofBiogeography , 37 ,2232 – 2242. LyonsSK(2003)AquantitativeassessmentoftherangeshiftsofPleistocenemammals. JournalofMammalogy , 84 ,385 – 402. MatocqMD,MurphyPJ(2007)Fine-scalephenotypicchangeacrossaspeciestransitionzoneinthegenusNeotoma:disentanglingindependentevolutionfromphylogenetichistory. Evolution , 61 ,2544 – 2557. McCainCM,ColwellRK(2011)Assessingthethreattomontanebiodiversityfrom discordantshiftsintemperatureandprecipitationinachangingclimate. Ecology Letters , 14 ,1236 – 1245. MoritzC,PattonJL,ConroyCJ,ParraJL,WhiteGC,BeissingerSR(2008)Impactofa centuryofclimatechangeonsmall-mammalcommunitiesinYosemiteNational Park,USA. Science , 322 ,261 – 264. MoyesK,NusseyDH,ClementsMN etal. (2011)Advancingbreedingphenologyin responsetoenvironmentalchangeinawildreddeerpopulation. GlobalChange Biology , 17 ,2455 – 2469. MyersP,LundriganBL,HoffmanSMG,HaraminacAP,SetoSH(2009)ClimateinducedchangesinthesmallmammalcommunitiesoftheNorthernGreatLakes Region. GlobalChangeBiology , 15 ,1434 – 1454. NowakRM(1991) Walker’sMammalsoftheWorld .TheJohnHopkinsUniversityPress, Baltimore&London. OzgulA,ChildsDZ,OliMK etal. (2010)Coupleddynamicsofbodymassandpopulationgrowthinresponsetoenvironmentalchange. Nature , 466 ,482 – 485. ParmesanC(2006)Ecologicalandevolutionaryresponsestorecentclimatechange. AnnualReviewofEcologyEvolutionandSystematics , 37 ,637 – 669. ParmesanC,YoheG(2003)Agloballycoherentngerprintofclimatechangeimpacts acrossnaturalsystems. Nature , 421 ,37 – 42. PoroshinEA,PollyPD,WojcikJM(2010)Climateandmorphologicalchangeondecadalscales:Multiannualvariationinthecommonshrew Sorexaraneus innortheast Russia. ActaTheriologica , 55 ,193– 202. PorterWP,KearneyM(2009)Size,shapeandthethermalnicheofendotherms. ProceedingsoftheNationalAcademyofSciences , 106 ,19666 – 19672. PostE,ForchhammerMC(2008)Climatechangereducesreproductivesuccessofan Arcticherbivorethroughtrophicmismatch. PhilosophicalTransactionsoftheRoyal SocietyB:BiologicalSciences , 363 ,2367 – 2373. RitchieME(2010) Scale,Heterogeneity,andtheStructureandDiversityofEcologicalCommunities .PrincetonUniversityPress,Princeton,NewJersey. RobertsonDS,MckennaMC,ToonOB,HopeS,LillegravenJA(2004)Survivalinthe rsthoursofthecenozoic. GeologicalSocietyofAmericaBulletin , 116 ,760 – 768. RootTL,PriceJT,HallKR,SchneiderSH,RosenzweigC,PoundsJA(2003)Fingerprintsofglobalwarmingonwildanimalsandplants. Nature , 421 ,57 – 60. RoweRJ,FinarelliJA,RickartEA(2010)Rangedynamicsofsmallmammalsalongan elevationalgradientoveran80-yearinterval. GlobalChangeBiology , 16 ,2930 – 2943. ScheffersBR,BrunnerRM,RamirezSD,ShooLP,DiesmosA,WilliamsSE(2013)Thermalbufferingofmicrohabitatsisacriticalfactormediatingwarmingvulnerability offrogsinthePhilippinebiodiversityhotspot. Biotropica , 45 ,628 – 635. ScherrerD,K   ornerC(2011)Topographicallycontrolledthermal-habitatdifferentiationbuffersalpineplantdiversityagainstclimatewarming. JournalofBiogeography , 38 ,406 – 416. SchlossCA,Nu ~ nezTA,LawlerJJ(2012)Dispersalwilllimitabilityofmammalsto trackclimatechangeintheWesternHemisphere. ProceedingsoftheNationalAcademyofSciences , 109 ,8606 – 8611. SmithFA,BrowningH,ShepherdUL(1998)Theinuenceofclimatechangeonthe bodymassofwoodratsNeotomainanaridregionofNewMexico,USA. Ecography , 21 ,140 – 148. SmithFA,LyonsSK,ErnestSKM etal. (2003)Bodymassoflatequaternarymammals. Ecology , 84 ,3403. SuggittAJ,GillinghamPK,HillJK,HuntleyB,KuninWE,RoyDB,ThomasCD (2011)Habitatmicroclimatesdrivene-scalevariationinextremetemperatures. Oikos , 120 ,1 – 8. TrenberthKE,JonesPD,AmbenjeP etal. (2007)Observations:surfaceandatmo-sphericclimatechange.In: ClimateChange2007:ThePhysicalScienceBasis.ContributionofWorkingGroupItotheFourthAssessmentReportoftheIntergovernmentalPanel onClimateChange .(edsSolomonS,QinD,ManningM,ChenZ,MarquisM, AverytKB,TignorM,MillerHL),pp.235 – 336.CambridgeUniversityPress,Cambridge,UnitedKingdomandNewYork,NY,USA. USGCRP(2009) GlobalClimateChangeImpactsintheUnitedStates .(edsKarlTR, MelilloJM,PetersonTC).UnitedStatesGlobalChangeResearchProgram.CambridgeUniversityPress,NewYork,NY. WaltherGR,PostE,ConveyP etal. (2002)Ecologicalresponsestorecentclimate change. Nature , 416 ,389 – 395.SupportingInformationAdditionalSupportingInformationmaybefoundinthe onlineversionofthisarticle:DataS1. Methods. AppendixS1 .NorthAmerica(NA)databaseandliterature cited. AppendixS2 .OutsideNAdata. AppendixS3 .AllNAspeciesdataset. AppendixS4. BestNAsubsetdataset. FigureS1. Responsecategories. FigureS2. Sampledatacomparison. TableS1. Ordinaryandphylogeneticlogisticregression statistics. TableS2. Best-tAICmultivariatemodels. © 2014JohnWiley&SonsLtd, GlobalChangeBiology , 20 ,1760–1769MAMMALTRAITSDETERMINECLIMATECHANGERESPONSES 1769