Prioritizing Regions for the Conservation of Amphibians With

PRIORITIZING REGIONS FOR THE CONSERVATION OF AMPHIBIANS WITH

SPECIAL EMPHASIS ON THE RED HILLS SALAMANDER

(PHAEOGNATHUS HUBRICHTI)

JOSEPH J. APODACA

A DISSERTATION

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Biological Sciences in the Graduate School of The University of Alabama

TUSCALOOSA, ALABAMA

2010

ABSTRACT

Amphibians are the most threatened vertebrate group in the world, and are experiencing rapid species declines and numerous extinctions. The most effective way to stem these losses is through the establishment of protected areas. The limited amount of funding available to such efforts requires that conservation agencies and biologists must find a way to properly focus their efforts and resources. Yet, there is no clear-cut method to prioritize areas for biological reserves.

In fact, the identification of biologically important regions is one of the most debated topics in the field of conservation biology. As this debate wages on and as species continue to decline at an unprecedented rate, conservation biologists have come to rely on increasingly sophisticated methods for the identification of these areas. In this dissertation I focus on recently developed techniques for prioritizing reserve selection from macro to micro-scales for amphibians in the southeastern United States. For chapters one and two I focus on broad scale issues for wide taxonomic groups. Chapter one focuses on testing whether using environmental niche models rather than extent of occurrence maps to create richness patterns is a valid approach. I found that environmental niche models could be useful for generating richness patterns for understudied regions or taxa if proper precautions are taken. Chapter two focuses on implementing evolutionary data into richness and endemism patterns using all members of the family

Plethodontidae found in the southeastern United States. I found that using evolutionary data in conjunction with traditional biodiversity metrics provides a unique and valuable perspective.

Chapters three and four narrow the focus to a single taxon, the Red Hills salamander

(Phaeognathus hubrichti). The Red Hills salamander is a federally threatened species whose

conservation has been hampered by their secretive and fossorial nature. To circumvent this problem, I conduct a conservation genetics study in chapter three and combine the data with spatial and life history data in order to make habitat acquisition recommendations in chapter four. I found that there are five distinct and well supported populations of P. hubrichti.

Additionally, each population has extremely low levels of gene flow and high levels of inbreeding. I recommend that 21 sites are acquired and that attempts are made to restore habitat in-between known populations.

iii

DEDICATION

This dissertation is dedicated to Sharon and John Watson, who have provided an incredible amount of support throughout the years and who never stopped believing in me.

LIST OF ABBREVIATIONS AND SYMBOLS

U.S. United States km2 square kilometers km Kilometer

IUCN International Union for Conservation of Nature

PAs Protected Areas

EOO Extent of Occurrence

ENM Environmental Niche Modeling

GBIF Global Biodiversity Information Facility

ROC Receiver Operating Characteristic

AUC Area Under the Curve

LPT Least Point Threshold

OLS Ordinary Least Squares

AIC Akaike Information Criterion

SAR Simultaneous AutoRegression

SAM Spatial Analysis in Macroecology v. Version r Correlation coefficient

P Probability associated with the occurrence under the null hypothesis of a value as

extreme or more extreme than the observed value.

% Percent

= Equals

> Greater than

< Less than

& And e.g. Exempli gratia (for example) i.e. id est (that is) et al. et alii (and others)

- Negative

PD Phylogenetic Diversity

PE Phylogenetic Endemism

WE Weighted Endemism

EDGE Evolutionary Distinctive and Globally Endangered

Cyt b Cytochrome b

POMC Pro-opiomelanocortin

ND2 Nicotinamide adenine dinucleotide subunit 2

ND4 Nicotinamide adenine dinucleotide subunit 4

RAG-1 Recombination Activating Gene 1

BDNF Brain-derived Neurotrophic Factor

Fig. Figure

° Degree

λ Lambda

FST Fixation Index

FIS Inbreeding coefficient of an individual relative to the population m migration rate

ESU Evolutionarily Significant Unit

USGS United States Geological Society

DNA Deoxyribonucleic acid

PCR Polymerase Chain Reaction

MCMC Markov Chain Monte Carlo

AMOVA Analysis of Molecular Variance

HO Observed heterozygosity

HE Expected heterozygosity

µ Mutation rate

Ne Effective population size

IAM Infinite alleles model

TPM Two phase model

SMM Stepwise mutation model

AL-GAP Alabama Gap Analysis Project

ETM+ Enhanced Thermatic Mapper Plus

HWE Hardy Weinberg Equilibrium

vii

t-test a statistical test in which the test statistic follows a Student’s t distribution if the

null hypothesis is true.

Θ 4 times the effective population size times the mutation rate

USFWS United States Fish and Wildlife Service

HCP Habitat Conservation Plan

Pers. Comm. Personal Communication

NED National Elevational Dataset

ALNHP Alabama Natural Heritage Program

Etc. et cetera (and other things)

viii

ACKNOWLEDGMENTS

This dissertation would not be possible without the support and effort of several people.

First off I would like to thank the members of my dissertation committee: Drs. Leslie Rissler,

Joseph Travis, Jonathan Benstead, Phil Harris, and Katrina Ramonell. Their advice and expertise over the years are greatly appreciated. I am especially grateful to my advisor, Dr. Leslie Rissler.

Without her advice, support, and knowledge this project would not have been possible.

I would also like to thank Christina Anderson, who has done more for me than she will ever realize. J Gailbreath has always supported me and has been a huge influence throughout my life. I owe tremendous gratitude to Jim Godwin, who provided an incredible amount of effort and knowledge on the Red Hills salamander. Heather Cunningham has been an integral part of my graduate school career and provided an infinite amount of support. Walter Smith has provided a great deal of fieldwork help and support. Christoph Thawley, Horace Downer, and Nicole

Mattheus provided a lot of useful feedback on my writings. Shannon Hoss, Michael Wines, and

Will Callans provided much needed help in the field. Rachel Crawford, Jessica Mitchell,

Catherine Tucker, and Alicia Waggoner provided assistance in the lab. Piper Apodaca has provided an enormous amount of support and has been an incredible friend throughout my graduate school experience.

Financial support was provided by the University of Alabama Department of Biological

Sciences and by the State of Alabama Division of Wildlife & Freshwater Fisheries.

CONTENTS

ABSTRACT...... ii

DEDICATION...... iv

LIST OF ABBREVIATIONS AND SYMBOLS ...... v

ACKNOWLEDGMENTS ...... ix

LIST OF TABLES...... xi

LIST OF FIGURES ...... xiv

1. CHAPTER 1: INTRODUCTION...... 1

2. CHAPTER 2: CHAPTER TWO: USING ENVIRONMENTAL NICHE MODELING TO REFINE SPECIES RICHNESS MAPS FOR AMPHIBIANS IN THE SOUTHEASTERN U.S…………………6

3. CHAPTER 3: COMPARING APPROACHES FOR IDENTIFYING BIODIVERSITY HOTSPOTS: A CASE STUDY FEATURING PLETHODONTIDS IN THE SOUTHEASTERN UNITED STATES...... 32

4. CHAPTER 4: ESTIMATING THE EFFECTS OF HABITAT MODIFICATION ON GENETIC PATTERNS AND POPULATION CONNECTIVITY; A CASE STUDY USING THE FEDERALLY THREATENED RED HILLS SALAMANDER (PHAEOGNATHUS HUBRICHTI)………………………………59

5. CHAPTER 5: RECOMMENDATIONS FOR THE RECOVERY OF THE THREATENED RED HILLS SALAMANDER (PHAEOGNATHUS HUBRICHTI): INCLUDING HABITAT-PURCHASING GUIDELINES, A PROPOSAL FOR HABITAT RECOVERY, AND IDENTIFYING LOCATIONS OF UNKNOWN POPULATIONS……………………………………………….....92 6. CHAPTER 6: OVERALL CONCLUSION...... 122

7. LITERATURE CITED ...... 126

LIST OF TABLES

TABLE 2.1 Regression models evaluating the ability of Anuran environmental niche models

to predict extent of occurrence richness values…………………………………25

TABLE 2.2 Regression models evaluating the ability of Caudata environmental niche models

to predict extent of occurrence richness values…………………………………26

TABLE 3.1 Covariance and correlation between biodiversity metrics for southeastern U.S.

plethodontids. Covariance values are in the upper diagonal. Correlation values are

in the lower diagonal. PE = Phylogenetic endemism, PD= Phylogenetic diversity,

WE= weighted endemism………………………………………………………..48

TABLE 3.2 Null model regression analysis between non-genetic metrics (WE & Richness)

and genetic metrics (PE & PD) for southeastern U.S. plethodontids. PE =

Phylogenetic endemism, PD= Phylogenetic diversity, WE= weighted endemism.

……………………………………………………………………………………49

TABLE 3.3 Three Highest contribution index values for range size in southeastern U.S.

plethodontids…………………………………………………………….……….50

TABLE 4.1 Results of the Analysis of Molecular Variance (AMOVA). P-values were

determined using 10 000 permutations…………………………………………..82

TABLE 4.2 Pairwise FST values for P. hubrichti populations. * indicates statistically

significant to the < 0.05 level. Significance determined using 10 000 permutations

and a sequential Bonferroni correction……………………………………..……83

TABLE 4.3 Summation of allelic information for each of the five populations

of P. hubrichti……………………………………………………………………84

TABLE 4.4 Rates of migration (m) as inferred from a coalescent method (MIGRATE), and

from a Bayesian assignment (BayesAss+). All rates are represented as m……....85

TABLE 4.5 Estimates of Θ and Ne for all populations of P. hubrichti. Ne was estimated

from Θ using a well-accepted vertebrate mutation rate of 5.4 X 10-4 (Goldstein

et al. 1995; Howes et al. 2009)……………………………………………….…86

TABLE 4.6 Results of bottleneck analysis for all populations of P. hubrichti. All values are

presented as P-values. Bold values indicate statistically significant……………87

TABLE 4.7 Results of Mantel tests for correlations between geographic/model distance and

genetic distance. * indicates statistically significant to the level of < 0.05 using

10 000 permutations……………………………………………………………..88

xii

TABLE 5.1 An example of the typical habitat management guidelines for a habitat

conservation plan designed to minimize and mitigate habitat losses for the Red

Hills salamander (Phaeognathus hubrichti)...... 112

TABLE 5.2 Estimated probabilities of the potential for identified sites to harbor unknown

populations of the Red Hills salamander (Phaeognathus hubrichti). Estimates are

based on: 1.) Amount of available habitat, 2.) Proximity to known populations,

and 3.) is the area separated from other populations by a major barrier (i.e.

Alabama or Conecuh Rivers, etc.)...... 113

TABLE 5.3 Terrestrial vertebrates of the Red Hills physiographic province of high

conservation concern. P1=endangered; P2=threatened...... 114

xiii

LIST OF FIGURES

FIGURE 2.1 Physiographic provinces of the southeastern United States……………….…….27

FIGURE 2.2 Maps of Anuran richness in the southeastern United States created using (a.)

extent of occurrence (EOO) maps, (b.) presence/absence ENM maps with a least

point threshold(LPT), (c.) ENM maps with Maxent values and a LPT, and (d.)

ENM hybrid maps. Each map is displayed with the lowest values represented by

the darkest colors and the highest values represented by the lightest colors. Each

color change in this scale represents one half of a standard deviation within the

data………………………………………………………………………………28

FIGURE 2.3 Scatter plots of Anuran richness values (x axis) and (y axis) values of (a.)

presence/absence ENMs with a least point threshold(LPT), (b.) ENMs with

Maxent values and a LPT, or (c.) ENMs using a hybrid approach………………29

FIGURE 2.4 Maps of Caudata richness in the southeastern United States created using (a.)

extent of occurrence (EOO) maps, (b.) presence/absence ENM maps with a least

point threshold(LPT), (c.) ENM maps with Maxent values and a LPT, and (d.)

ENM hybrid maps. Each map is displayed with the lowest values represented by

the darkest colors and the highest values represented by the lightest colors. Each

color change in this scale represents one half of a standard deviation within the

data……………………………………………………………………………….30

xiv

FIGURE 2.5 Scatter plots of Caudata richness values (x axis) and (y axis) values of (a.)

presence/absence ENMs with a least point threshold(LPT), (b.) ENMs with

Maxent values and a LPT, or (c.) ENMs using a hybrid approach………………31

FIGURE 3.1 Maximum-likelihood phylogram of plethodontids found in the southeastern U.S.

with a summary of divergence times. Tree modified from Kozak et al. 2009…..51

FIGURE 3.2 Richness patterns for plethodontids in the southeastern U.S. calculated for 0.05°

cells………………………………………………………………………………52

FIGURE 3.3 Weighted endemism (WE) patterns for plethodontids in the southeastern U.S.

calculated for 0.05° cells…………………………………………………………53

FIGURE 3.4 Phylogenetic diversity (PD) patterns for plethodontids in the southeastern U.S.

calculated for 0.05° cells……………………………………………………...... 54

FIGURE 3.5 Phylogenetic endemism (PE) patterns for plethodontids in the southeastern U.S.

calculated for 0.05° cells………………………………………………………....55

FIGURE 3.6 Scatter plot of richness and phylogenetic diversity (PD) values for southeastern

U.S. plethodontids……………………………………………………….……….56

FIGURE 3.7 Scatter plot of weighted endemism (WE) and phylogenetic endemism (PE) values

for southeastern U.S. plethodontids……………………………………………...57

FIGURE 3.8 Physiographic provinces of the southeastern United States……………………..58

FIGURE 4.1 Range map of P. hubrichti with an outline of geologic layers required by the

species. Circles are known populations; stars represent tissue collection localities

……………………………………………………………………………...……8

FIGURE 4.2 Populations of P. hubrichti as determined by Bayesian clustering techniques.

Populations are numbered from left to right and represented by unique symbols

(1, circles; 2 triangles; 3, stars; 4, squares; 5, diamonds)……………………….90

FIGURE 4.3 Slope analysis and populations of P. hubrichti. Areas of steepest slope are

represented by darker colors……………………………………………………91

FIGURE 5.1 Geographic distribution of the Red Hills salamander (Phaeognathus hubrichti) as

it is currently known...... 115

FIGURE 5.2 Outline of the three geologic layers known to contain populations of the Red Hills

salamander (Phaeognathus hubrichti)...... 116

FIGURE 5.3 Known localities of the Red Hills salamander grouped into their respective

populations as determined by genetic data. Populations are from left to right:

population 1- black crosses, Population 2- black circles, Population 3- grey

squares, Population 4- dark-grey diamonds, Population 5- black octagons...... 117

FIGURE 5.4 Distribution model for the Red Hills salamander (Phaeognathus hubrichti).

xvi

FIGURE 5.5 Areas suggested for surveys to determine the presence of unknown Red Hills

salamander (Phaeognathus hubrichti) populations...... 118

FIGURE 5.6 Red Hills salamander (Phaeognathus hubrichti) localities suggested for

acquisition by Dodd (1988)...... 119

FIGURE 5.7 Sites suggested for acquisition to promote the recovery of the Red Hills

salamander (Phaeognathus hubrichti)...... 120

xvii

CHAPTER 1

INTRODUCTION

“…. The worst thing that will probably happen- in fact is already well under way- is not energy depletion, economic collapse, conventional war, or even the expansion of totalitarian governments. As terrible as these catastrophes would be for us, they can be repaired within a few generations. The one process now ongoing that will take millions of years to correct is the loss of genetic and species diversity by the destruction of natural habitats. This is the folly our descendants are least likely to forgive us for.”

E.O. Wilson- Biophilia (1984)

Anthropogenic activities have elevated global extinction rates well above historic levels, as shown by the fossil record (Pimm et al. 1995). If drastic actions are not taken rates may climb another order of magnitude in the near future (Pimm et al. 1995; Mace et al. 2005). The unifying goal of the field of conservation biology is to stem the blitz of species losses and to preserve the viability and persistence of functioning ecosystems (Soulé 1985). In fact, conservation biology has been described as a “crisis discipline” whose goal is to preserve biodiversity by providing the proper principles and tools (Soulé 1985). These principles and tools are adopted from a wide variety of fields, including but not limited to: ecology, systematics, population genetics, physiology, historical biogeography, behavior, wildlife and fisheries management, and forestry.

As the loss of species and destruction of habitat continue to increase at an exponential rate the field of conservation biology must quickly evolve to meet these challenges. The increasing complexity and difficulty of such problems coupled with a limited amount of conservation funds has led conservation biologists to adopt progressively more sophisticated

approaches (Margules and Pressey 2000; Sarkar et al. 2006). These novel methods provide critical information for state and federal agencies, non-governmental conservation organizations, and international treaties from macro to micro-scales. For example, the recently developed technique of neo-endemism (Rosauer et al. 2009) is helping to identify areas that contain a disproportionate amount of the world’s endemic species at a global scale; while the emerging field of landscape genetics is providing local agencies with critical data on the effects of landscape on gene flow between populations of endangered species at a local scale (e.g. Spear et al. 2005; Straub and Doyle 2009; Wang 2009; Wang and Summers 2010).

Regardless of scale, the outcome of conservation efforts ultimately comes down to three factors: do we have the resources, the time, and the data needed to save a threatened species or ecosystem? Unfortunately, we rarely possess an adequate amount of any of these factors, let alone all three. Therefore, conservation biologists and agencies are faced with the difficult decision of where to focus resources and effort (Myers et al. 2000; Orme et al. 2005; Wilson et al. 2006; Isaac et al. 2007; Leroux and Schmiegelow 2007), a dilemma that has come to be known as the “agony of choice” (Vane-Wright et al. 1991) or the “Noah’s Ark problem”(Weitzman 1998). The question of how to prioritize areas for conservation purposes is perhaps the most important question in conservation biology and also one of the most debated

(e.g. Faith 1992; Balmford et al. 1996; Caro and O’Doherty 1999; Andelman and Fagan 2000;

Arujo and Williams 2000; Myers et al. 2000; Brooks et al. 2006; Ceballos and Ehrlich 2006;

Isaac et al. 2007; Rosauer et al. 2009).

In the chapters of this dissertation I focus on recently developed techniques for prioritizing reserve selection from macro to micro-scales for amphibians in the southeastern

United States (U.S.). Conservation reserves (or protected areas) are by far the most efficient way

of preserving biodiversity (Rodrigues and Gaston 2002), and are therefore fundamental to conservation efforts. I begin by investigating how patterns of amphibian richness in the southeastern U.S. are affected by the method used to define the species’ ranges. Using nearly all amphibian species found in the southeastern U.S. I compare the traditional method of creating richness patterns (using extent of occurrence maps) with patterns created using environmental niche modeling. Patterns of richness are the most commonly used metric for identifying biodiversity hotspots and therefore exploring new methods with which to create them is important for global conservation efforts.

Chapter three focuses on implementing phylogenetic data into richness and endemism patterns using all members of the family Plethodontidae found in the southeastern U.S.

Plethodontidae is by far the most speciose amphibian family in the southeastern U.S., and contains three completely endemic genera (Phaeognathus, Desmognathus, and Pseudotriton).

Chapters four and five narrow the focus to a single species, the federally threatened Red Hills salamander (Phaeognathus hubrichti). The Red Hills salamander is a monotypic and fossorial plethodontid salamander whose existence is severely threatened by habitat destruction and fragmentation. Because of its secretive existence we know very little about this species.

Therefore, in chapter three I use genetic techniques to elucidate P. hubrichti population structure, gene flow, and effective population sizes. In chapter four I combine these genetic data with GIS techniques and life history data to recommend habitat-purchasing guidelines to ensure the survival of this imperiled salamander.

The southeastern U.S. is an ideal location to focus on amphibian conservation. It is a global hotspot of amphibian diversity (Duellman 1999), particularly for members of the order

Caudata. In fact, the southeastern U.S. contains nine completely endemic salamander genera

(Phaeognathus, Desmognathus, Amphiuma, Gyrinophilus, Haideotriton, Pseudotriton,

Pseudobranchus, Siren, and Urspelerpes). However, this region has seen an immense amount of habitat destruction, which threatens the vast amphibian diversity in the region. For example, historically the most dominant habitat type in the southeastern U.S. was longleaf pine (Pinus palustris) savanna. This habitat type once contained one of the world’s most diverse assemblages of flora and fauna (Straub and Doyle 2009). Since European settlement, longleaf pine habitat has been reduced to less than 2% of the original extent, and it is estimated that only 0.2% of the remaining habitat is healthy enough to support its native assemblage of species (Noss 1989; Frost

2006; Means 2006). Despite this shocking loss of habitat there is a relative shortage of protected areas in the southeastern U.S.

The total area covered by current reserves in the Southeast is approximately 22,975 km2, which is equal to 8.5% of the total land area considered (Apodaca unpublished data). However, the number of conservation reserves occurring in the Appalachian Highlands largely drives this pattern. When the Appalachian Highlands province is removed the amount of protected area drops to 11,780 km2 (5.31% of the remaining area). This falls well below the goal of the Fourth

World Congress on National Parks and Protected Areas, which laid out a goal of 10% for each biome (Brooks et al, 2004). Furthermore, the vast majority of protected areas in the Southeast are

IUCN category IV – VI (Apodaca unpublished data). These areas are afforded the lowest protection for biodiversity and are generally managed for natural resources (IUCN 1994). In fact, the total area of IUCN category Ia- III (reserves that afford the greatest degree of protection for biodiversity) in the Southeast is only 1,430 km2, a mere 0.5% of total area.

A shortage of protected areas is particularly troubling for amphibians, which, at a global scale, are significantly less covered by preserves than any other vertebrate group (Rodrigues et al. 2004). This lack of protected area coverage could help to explain the amphibian decline phenomenon (Stuart et al. 2004). In fact, amphibians are the most threatened vertebrate group, with over 32% of all species threatened, a number that will certainly continue to rise without significant intervention (Stuart et al. 2004). Therefore, it is vital that conservation research and efforts continue to increase for amphibian species’ and unique assemblages worldwide.

CHAPTER TWO: USING ENVIRONMENTAL NICHE MODELING TO REFINE SPECIES

RICHNESS MAPS FOR AMPHIBIANS IN THE SOUTHEASTERN U.S.

Abstract. The identification and preservation of biodiversity hotspots is one of the most pressing issues in conservation biology. Distinguishing such areas is complicated and problematic given our imperfect understanding of species’ geographic ranges. New methods in environmental niche modeling may provide a novel approach for the delimitation of biologically important regions at a broad scale. In this study we compare and contrast the resulting patterns of amphibian richness maps created using environmental niche models with those created using traditional extent of occurrence maps for amphibians in the southeastern United States. Richness maps were created for amphibians using two approaches to estimate species geographic distributions: environmental niche modeling and extent of occurrence maps. Environmental niche models were created via a maximum entropy approach. Models were modified for analysis using three techniques: a presence/absence method, a model value method, and a hybrid method. All models were subjected to a least point threshold. Models were then compared to extent of occurrence maps using scatter plots and spatial regression analysis. We found that when compared to traditional richness maps (extent of occurrence), all three methods using environmental niche modeling yielded richness maps that were generally similar in identifying the areas of highest richness, but also had a high amount of variation. Regression analyses indicated that the presence/absence niche modeling method had the highest correlation with extent of occurrence maps (Anuran r =

0.69, Caudata r = 0.70). The use of environmental niche models in the creation of biodiversity composite variables appears to be a valid approach as long as precautions are taken while

analyzing the results. The methods presented here can help further our understanding of species distributions in lesser studied taxa and regions.

INTRODUCTION

The past three decades of conservation science have seen a cascade of research devoted to the creation of a global network of wildlife reserves or protected areas (PAs) (e.g. Myers

1990; Dobson et al. 1993; Prendergast et al. 1993; Myers et al. 2000; Bonn et al. 2002; Ceballos

& Ehrlich 2006; Davis et al. 2008). Global assessments of biodiversity patterns have been crucial in identifying the most imperiled regions and expanding the current network of PAs (Brooks et al. 2004). Despite the improvement and application of conservation science, there still exists a major discordance between PAs and the biodiversity of a region (Pressey et al. 1993; Scott et al.

1993; Rodrigues et al. 1999; Margules and Pressey 2000; Maiorano et al. 2006). The effort to expand and improve the global network of PAs has been supported by the development of many novel approaches designed for the identification of more economically efficient reserve networks, such as the use of umbrella and flagship species, the incorporation of phylogenetic diversity, irreplaceability, complementarity, threat, vulnerability, and flexibility (reviewed in

Margules and Pressey 2000; Cabeza and Moilanen 2001; Margules and Sarkar 2007).

Extent of occurrence (EOO) maps have traditionally formed the basis of biodiversity estimates (Jetz et al. 2008). This common representation of a species distribution is usually created based on expert opinion, whereby the range is presented as a simple polygon (Graham and Hijmans 2006). The fundamental problem in using EOO maps is that they inherently overestimate the total area inhabited by a species, and therefore distort perceived biodiversity patterns (Hurlbert and White 2005; Graham and Hijmans 2006; Jetz et al. 2008). In particular,

EOO maps do not include information on geographic variation in viability across a species’ range. Most species only occupy a small, environmentally appropriate subset of the overall EOO.

For example, recent studies indicate that the average bird species only occupies around 40-70% of their identified EOO range (Hurlbert and White 2005; Jetz et al. 2008).

Overestimating a species area of occupancy introduces important sources of error when considering local conservation strategies. One possible result is an inflated estimation of the biodiversity value of a region (Hurlbert and White 2005). The second, and perhaps more troublesome error, is that taxa may appear to be covered under the global network of PAs, when in fact they lack coverage (i.e. a “gap” species). This concern is elevated given the recent findings of Jetz et al. (2008) who found that at risk species of birds were much more likely to have their ranges overestimated than common widespread species. These results are particularly applicable to rare species with narrow environmental tolerances. Niche theory suggests that pervasive species generally have a much greater tolerance for varying ecological conditions and are therefore more likely to occupy more area within their range (Grinnell 1917; Brown 1995;

Holt 2003; Jetz et al. 2008). On the other hand, species with small geographic ranges tend to occupy a more narrow niche space and thus will occupy much less area within their estimated

EOO range (Wilson et al. 2004; Jetz et al. 2008). Unfortunately, the same traits that may lead to a species range being overestimated are also those that may cause a species to become threatened or endangered (i.e. small ranges, specialized niches, etc.) (Purvis et al. 2000; Jetz et al. 2008)

Ultimately, the successful mapping of biodiversity patterns will depend on complete distributional information for all taxa and information on geographic variation in population viability across a species’ range. Too often evaluations of biodiversity patterns do not explicitly consider the inherent variation among populations within a species and how such data could be

used to refine conservation priorities. Such knowledge is rarely known and difficult to obtain for even one species, let alone entire assemblages. However, new spatially explicit modeling tools have the potential to improve our understanding of species’ distributional patterns and the variation of habitat quality within each range (reviewed in Graham et al. 2004a; Elith et al.

2006). One such novel method that has seen increased application in conservation studies is environmental niche modeling (ENM) (e.g. Ferrier 2002; Raxworthy et al. 2003; Domínguez-

Domínguez et al. 2006; Garcia 2006; Rissler et al. 2006).

In general, environmental niche models (ENMs) create a prediction of a species’ geographic range by relating a species known locality data to environmental parameters. ENMs have been successfully integrated into a diverse set of ecological (e.g., Araújo and Williams

2000; Ferrier et al. 2002; Mac Nally and Fleishman 2004; Cunningham et al. 2009), evolutionary

(e.g. Graham et al. 2004b; Wiens and Graham 2005; Rissler and Apodaca 2007), and conservation (e.g. Ferrier 2002; Raxworthy et al. 2003; Domínguez-Domínguez et al. 2006;

Garcia 2006; Rissler et al. 2006) studies. However, one field where ENMs have not received adequate attention is in the assessment of biodiversity patterns (i.e. patterns of richness, endemism, etc.) at a regional level (but see Garcia 2006).

The integration of ENMs into biodiversity assessments has the potential to greatly improve our knowledge of the distributional patterns of taxonomic groups or geographic regions that lack sufficient data on species’ ranges. For the vast majority of species on earth we have very little, if any, distributional data, a dilemma that has been dubbed the “Wallacean shortfall”

(Tognelli 2005; Whittaker et al. 2005; Bini et al. 2006). Proper distributional modeling from species occurrence data could help alleviate this predicament, although it is vital that this approach be properly tested before implementation into a conservation strategy. For example,

there are many cases where ENMs can introduce a large amount of error (e.g. under sampled taxa, strong biotic interactions or species whose range limits are set by factors other than climate). These types of error could lead to a gross over or underestimation of the biodiversity value of a region. Therefore, it would be a valuable endeavor to explore the behavior of biodiversity patterns identified using ENM methods in a well-studied region.

Despite a plethora of papers discussing the implementation and caveats of ENMs in conservation there have been relatively few papers that explore how patterns (such as richness and endemism) created by niche modeling techniques differ from those created with traditional methods (but see Graham and Hijmans 2006). This study aims to compare and contrast the resulting patterns of richness maps created using ENMs with those created using traditional EOO maps for amphibians in the southeastern United States. Furthermore, we test whether ENM richness maps are able to accurately predict EOO richness values using a series of regression analyses. In doing so we hope to identify areas important for the long-term conservation of these taxa. By focusing on a fairly well studied area we aim to investigate whether using modeling techniques, such as ENMs, is appropriate when creating composite biodiversity variables (e.g. species richness, endemism, etc.) for lesser studied taxa and geographic regions.

METHODS

2.1. Study Group

Amphibians are the most threatened vertebrate group in the world and are declining at an alarming rate (Global Amphibian Assessment; IUCN et al. 2008). The identification and development of methods that can be used to accurately portray the distributional patterns of amphibians worldwide is paramount to the conservation of this group. The southeastern United

States provides a choice platform for such investigations. The region is a world hotspot for amphibians, with roughly 136 species and 23 genera (following the taxonomy presented in

Lannoo (2005)), including a high number of endemics. Additionally, the region has a rich history of amphibian research combined with museum collections. Consequently, the southeastern U.S. has a rare combination of high species richness and a fairly accurate understanding of species range limits.

2.2. Study Area

For the purposes of this study we chose to define the southeastern U.S. as the outline of the following states: Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi,

North Carolina, South Carolina, Tennessee, Virginia, and West Virginia. This region contains several unique physiographical provinces, including: Peninsular Florida, Atlantic Coastal Plain,

Gulf Coastal Plain, Piedmont, Appalachian Highlands, Interior Highlands, and Interior Plains

(Fig. 1).

2.3. Species Distribution Data

2.3.1. EOO Distributions

Most species range maps were downloaded from the NatureServe server (Global

Amphibian Assessment; IUCN et al., 2008). Species ranges that were not available for download were created using the county range maps based on Lannoo (2005). All range maps were converted to a grid size of 0.5° using ArcGIS (9.1). These maps were then overlaid in order to create EOO richness patterns.

2.3.2. ENM Distributions

Niche models were created using Maxent Version 3.2.1 (Phillips et al. 2006). Collection localities were downloaded from Herpnet (www.herpnet.org) and GBIF (www.GBIF.org).

Fourteen biologically relevant climate variables, at a 1-km2 resolution, were obtained from

Daymet (www.daymet.org). These included 18 year averages of the following variables: daily maximum air temperature, daily minimum air temperature, daily average air temperature, day to day variability in maximum temperature, day to day variability in minimum temperature, day to day variability in average temperature, number of frost days, precipitation frequency, average precipitation event size, total precipitation, daily total shortwave radiation, day to day variation in total shortwave radiation, daily average water vapor pressure, and day to day variability in water vapor pressure.

Maxent is a modeling program that uses a machine learning approach (maximum entropy) to predict a species range based on the association of climatic variables with known occurrence data (Phillips et al. 2006). The maximum entropy approach is one of many recent advances in machine learning approaches for ecological applications. This family of statistical techniques typically outperforms traditional statistical approaches (e.g. generalized linear models) in complex ecological situations (Olden et al. 2008). In fact, in a recent review of niche modeling techniques (Elith et al. 2006) Maxent performed as well or better than other environmental niche modeling techniques.

The accuracy of each model was tested by using the common model validation approach of creating a receiver operating characteristic (ROC) curve and measuring the area under the curve (AUC) (Hanley and McNeil 1982). AUC scores can be thought of as the probability that the ENM has correctly differentiated presence and absence points for a given species. Because

Maxent uses presence-only data rather than presence/absence data, Maxent calculates AUC scores using pseudo-absence points. AUC scores above 0.75 indicate adequate model performance (Pearce and Ferrier 2000). ROC curves were created by separating locality data into

training (80%) and test (20%) points. Test data that are withheld from known locality data are not completely independent of training data (Araújo et al. 2005). However, this method presents a more conservative approach when compared to independent training data (McPherson et al.

2004).

The logistic output of Maxent (Phillips et al. 2006) can be thought of as a continuous surface of the probability of a species occurrence, assuming a similar level of sampling effort across all collection localities (Phillips and Dudik 2008). We decided to use three different approaches to creating richness maps using the ENM models. The first is a basic threshold method, where each model either indicates presence (above the threshold) or absence (below the threshold). We chose the least point threshold (hereafter referred to as LPT) because of its conservative nature and straightforward ecological interpretation (Pearson et al. 2007). This LPT can be interpreted as identifying areas that are at least as suitable as areas where a species’ presence has been reported (Pearson et al. 2007). Another important advantage of the LPT is that it eliminates the possibility of omission error. For the second method we again applied the LPT, but above the threshold value we used the probability values assigned by Maxent (hereafter referred to as the ModelLPT). We chose this approach to investigate whether the probability values provide data that would be otherwise lost in a presence/absence approach. The third method we used is the hybrid distribution map (hereafter referred to as the hybrid approach) of

Graham and Hijmans (2006). In this method the ENM is clipped to the EOO map. This method requires the most data, and therefore might not be as useful to lesser studied regions or taxa.

However, for scenarios where this method is plausible it represents a “best of both worlds” option that could provide valuable data for conservation planning.

2.3.3 Statistical analysis

In order to test the degree to which each of the modeling techniques is able to predict

EOO richness we ran a series of regression analyses. For each regression the ENM method was the predictor and the EOO method was the response variable. First we used ordinary least squares (OLS) regression, and estimated how well the model performed using Akaike information criterion (AIC) (Burnham and Anderson 2002). Since spatial autocorrelation can exaggerate Type I errors in regression analyses (Legendre 1993; Diniz-Filho et al. 2003; Rangel et al. 2006), we also used a simultaneous autoregression (SAR) model. The SAR model enhances the standard linear regression model by adding an additional term that accounts for spatial autocorrelation in a given data set (Kissling and Carl 2008). Thus, this method helps explain patterns in the response variable that are not expected based on the predictor variable alone. All regression analyses were carried out in the program Spatial Analysis in Macroecology (SAM) v.

3.1 (Rangel et al. 2006), using 2000 randomly extracted points within the study area.

RESULTS

3.1. Anuran species

Anuran richness peaked at 27 species for any single EOO richness grid cell. The highest richness areas were found in the Atlantic and Gulf Coastal Plains (Fig. 2(a.)). The EOO richness patterns highlighted the majority of both the Atlantic and Gulf Coastal Plains. In order to create the Anuran ENMs, we obtained a total of 12,107 occurrence localities for the 34 species found in the region. The mean number of unique collection localities for a species was 403.57 and the median number of points was 190. The most unique localities for any single species was for

Rana catesbeiana with 2821, and the least for any included species was 21 for Hyla andersonii.

The mean AUC score for Anurans was 0.9423 with a median score of 0.9835.

In comparison to EOO richness patterns, the LPT method shifted some of the hotspots from the Atlantic Coastal Plain to the Piedmont region (Fig. 2(b.)). The LPT method also did not place as much emphasis on the Gulf Coastal Plain (specifically in the region of the Florida panhandle) as the EOO method. In general the LPT method appears to broaden the region highlighted as areas of high richness. In contrast, the ModelLPT appears to contract the highlighted areas to a slim belt along the Gulf and Atlantic Coastal Plains, taking emphasis away from inland areas (Fig. 2(c.)). The hybrid method also highlights this narrow band, but is more similar to the EOO method in that it emphasizes more of the Coastal Plain (Fig. 2(d.)).

All ENM methods significantly predicted EOO richness (Table 1.). The ModelLPT had the lowest predictive ability for both the OLS (r = 0.484, AIC = 11241.544) and SAR (r= 0.566,

AIC = 10986.954) models. The hybrid method had the highest predictive ability for the SAR model (r = 0.728, AIC = 10258.978), and the LPT method had the highest predictive ability for the OLS model (r= 0.688, AIC = 10484.968). Scatter plots for each method revealed that each of the methods had a high variance from the predicted slope (Fig. 3).

3.2. Caudata species

Caudata EOO richness was centered in the Appalachian Highlands, with several other small pockets of high richness in other physiographic regions (Fig.4(a.)). The highest number of species found in a single grid cell for the EOO method was 28 (Appalachian Highlands). Caudata

ENMs were created using a total of 13,163 unique localities for 84 species. Eighteen species had either insufficient locality data or poor AUC scores, resulting in the assignment of a standard value to their entire range. The mean number of unique collection localities for a species was

156.7 and the median number of points per species was 56. The most unique localities for any single Caudata species was for Plethodon glutinosus with 1135. The least for any included

species was 10, for both Plethodon kiamichi and Plethodon hubrichti. The mean AUC score for

Caudata species was 0.988, with a median of 0.994.

The LPT method focused on similar areas as the EOO method, highlighting the

Appalachian Highlands and several other regions within the Appalachians Mountains (Fig.

4(b.)). The ModelLPT broadened the Appalachian hotspot to cover more of the Appalachian

Highlands and a portion of the Piedmont (Fig. 4(c.)). Additionally, the ModelLPT highlighted an area along the Gulf Coastal Plain that was not identified by the EOO method. The hybrid method highlighted areas similar to the ModelLPT (Fig. 4(d.)). However, it did not emphasize the Gulf

Coastal Plain as strongly.

EOO patterns were once again significantly predicted by all ENM methods (Table 2). In general, regression models performed more strongly in the Caudata study system than in the

Anuran system. The ModelLPT had the lowest predictive ability for both the OLS (r = 0.638,

AIC = 10437.878) and SAR (r= 0.651, AIC = 10381.753) model. The hybrid method had the highest predictive ability for both the OLS (r= 0.737, AIC= 9926.039) and the SAR (r = 0.763,

AIC = 9750.27) models. Scatter plots revealed that each method had a high variance from the predicted slope, albeit much less so than the Anuran models (Fig. 5).

DISCUSSION

4.1. Environmental niche modeling vs. extent of occurrence maps

Techniques such as ENMs can help advance our knowledge of the distribution patterns of lesser-studied taxa (Raxworthy et al. 2003; Garcia 2006; Pearson et al. 2007) and as a result improve biodiversity assessments. One area of research that has the potential to benefit from the use of ENMs, but has seen limited research, is the identification of biologic hotspots. It has been demonstrated that Maxent (Phillips et al. 2006) performs admirably even with limited sample

sizes (Pearson et al. 2007). If ENMs could be accurately applied to the creation of biological composite variables for lesser studied taxa or regions this method would be extremely useful to conservation efforts. This is an exciting prospect, yet it is important to establish a baseline understanding of the strengths and weaknesses of such techniques in a relatively well-studied area before they can be applied to taxa and regions where our knowledge and data are lacking.

Our study shows that there are significant differences between richness patterns created using ENMs and EOO maps for Anuran species in the southeastern U.S. (Fig. 2) – a well-studied taxonomic group and geographic region. However, each ENM method significantly predicted

EOO values (Table 1; Fig. 3). For Anurans, the method that most accurately predicted EOO values was the hybrid method of Graham and Hijmans (2006) (Table 1). Interestingly, the r- values for this method were only 0.650 and 0.728 despite the fact that models are clipped to

EOO distributions for this method. These data further support the notion that EOO maps inherently overestimate the total area inhabited by a species (Hurlbert and White 2005; Graham and Hijmans 2006; Jetz et al. 2008), though in well-studied areas they provide a reasonably accurate estimate of species’ distributional limits. We found that for Anurans the LPT method was a more powerful predictor of EOO richness than the ModelLPT method (Table 1). This would most likely not be the case if the collecting localities met the assumptions of Maxent (a similar level of sampling at all sites). Because this assumption will rarely be met in a conservation setting, we suggest that the LPT method is a more appropriate method for creating composite variables.

There were also discrepancies between ENM and EOO methods for salamanders in the southeastern U.S. (Fig. 4) – another taxonomic group with detailed distribution data. However,

ENMs were more closely correlated to EOO patterns for this group than for Anurans. Once more

the hybrid method was the strongest model of the three (Table 2). The LPT model also performed strongly (Table 2), again suggesting that this method is more reliable than the

ModelLPT for the creation of biodiversity composite variables. It should be noted that the comparisons we made are based on the assumption that EOO richness values for amphibians in the southeastern U.S. are accurate. Of course, this assumption will rarely be met for any natural system. Yet, we feel that the strong history of amphibian research in this region provides at least an accurate picture of species’ distributional limits, and therefore a basis on which to accurately compare these methods.

One of the most urgent problems in preserving the biodiversity of a region is that species are distributed unevenly (Gaston 2000). The ongoing loss of habitat and current extinction crisis paired with a scarcity of resources and time have forced the creation of conservation strategies that necessarily focus on select, small and isolated areas (Olson and Dinerstein 1998; Myers et al. 2000; Lamoreux et al. 2006). Because we know very little about the geographic ranges of the majority of described species throughout the world (Raven and Wilson 1992; Brooks et al. 2004;

Lamoreux et al. 2006), conservation science must establish robust methods to map patterns of biodiversity. The results of our study demonstrate that ENMs provide a promising technique for the identification of such biodiversity patterns.

Our regression analyses (Tables 1 and 2) and scatter plots (Figs. 3 and 5) reveal that the

LPT method is suited for use in the creation of biodiversity composite variables in lesser-studied taxa or regions. Of course, we urge the use of caution in doing so, as the use of ENMs has the potential to introduce a large amount of error if assumptions are not met. Therefore, it is imperative that any estimates of biodiversity richness based on ENMs should be heavily supplemented with field surveys. However, this is true of any richness estimate, as EOO maps

also contain large sources of error (Hurlbert and White 2005; Graham and Hijmans 2006; Jetz et al. 2008). We highly recommend that studies such as the one we have presented are repeated across a broad scale of taxa and regions in order to better understand how ENM biodiversity patterns respond to varying conditions. Additionally, it would be valuable to explore how ENM patterns differ from EOO methods for other biodiversity metrics (such as endemism, threat, and phylogenetic diversity).

The use of ENMs for the identification of biodiversity hotspots is certainly not ideal.

However, given the rapid loss of species and habitats, conservationists do not have adequate time or resources to properly assess the geographic distribution of even a small portion of the world’s taxa, the so called “Wallacean shortfall” (see Tognelli 2005; Whittaker et al. 2005; Bini et al.

2006). Consequently, there are two options in the identification of biodiversity hotspots for the majority of the world’s species. The first is to assume that the relatively few groups for which we do have basic worldwide distributional data (such as mammals and birds) are proper surrogates for lesser-known taxa (Simberloff 1998; Caro and O’Doherty 1999; Andelman and Fagan 2000;

Moreno et al. 2007). The use of surrogates in conservation planning has received a great deal of debate (reviewed in Rodrigues and Brooks 2007), and it has been demonstrated that surrogates are an inadequate strategy (Leroux and Schmiegelow 2007; Ceballos and Ehrlich 2006). The second option is to further develop techniques that are able to fairly accurately estimate a species range based on known features of the species (such as collection localities, elevation constraints, habitat restrictions etc.). It appears that current ENM techniques are heading in the right direction to meet this challenge. Though, it is imperative that such methods continue to be thoroughly tested and improved upon if they are to become a mainstay for conservation biology.

LITERATURE CITED

Andelman SJ, Fagan WF (2000) Umbrellas and flagships: Efficient conservation surrogates or expensive mistakes? Proceedings of the National Academy of Sciences of the United States of America 97, 5954-5959.

Araujo MB, Whittaker RJ, Ladle RJ, Erhard M (2005) Reducing uncertainty in projections of extinction risk from climate change. Global Ecology and Biogeography 14, 529-538.

Araujo MB, Williams PH (2000) Selecting areas for species persistence using occurrence data. Biological Conservation 96, 331-345.

Bini LM, Diniz JAF, Rangel TFLVB, Bastos RP, Pinto MP (2006) Challenging Wallacean and Linnean shortfalls: knowledge gradients and conservation planning in a biodiversity hotspot. Diversity and Distributions 12, 475-482.

Bonn A, Rodrigues ASL, Gaston KJ (2002) Threatened and endemic species: are they good indicators of patterns of biodiversity on a national scale? Ecology Letters 5, 733-741.

Brooks TM, Bakarr MI, Boucher T, et al. (2004) Coverage provided by the global protected-area system: Is it enough? Bioscience 54, 1081-1091.

Brown JH (1995) Macroecology University of Chicago Press, Chicago.

Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach. 2nd edition Springer-Verlag, New York

Cabeza M, Moilanen A (2001) Design of reserve networks and the persistence of biodiversity. Trends in Ecology and Evolution 16, 242-248.

Caro TM, O'Doherty G (1999) On the use of surrogate species in conservation biology. Conservation Biology 13, 805-814.

Ceballos G, Ehrlich PR (2006) Global mammal distributions, biodiversity hotspots, and conservation. Proc Natl Acad Sci U S A 103, 19374-19379.

Cunningham HR, Rissler LJ, Apodaca JJ (2009) Competition at the range boundary in the slimy salamander: using reciprocal transplants for studies on the role of biotic interactions in spatial distributions. J Anim Ecol 78, 52-62.

Davis EB, Koo MS, Conroy C, Patton JL, Moritz C (2008) The California Hotspots Project: identifying regions of rapid diversification of mammals. Molecular Ecology 17, 120-138.

Diniz JAF, Bini LM, Hawkins BA (2003) Spatial autocorrelation and red herrings in geographical ecology. Global Ecology and Biogeography 12, 53-64.

Dobson AP, Rodrigues JP, Wilcove DS (1993) Geographic distribution of endangered species in the United States Science 275, 550-553.

Dominguez-Dominguez O, Martinez-Meyer E, Zambrano L, De Leon GP (2006) Using ecological-niche modeling as a conservation tool for freshwater species: live-bearing fishes in central Mexico. Conserv Biol 20, 1730-1739.

Elith J, Graham CH, Anderson RP, et al. (2006) Novel methods improve prediction of species' distributions from occurrence data. Ecography 29, 129-151.

Ferrier S, Watson G, Pearce J, Drielsma M (2002) Extended statistical approaches to modeling spatial patterns in biodiversity in north-east New South Wales. Species-level modeling. Biodiversity and Conservation 11, 2275-2307.

Garcia A (2006) Using ecological niche modelling to identify diversity hotspots for the herpetofauna of Pacific lowlands and adjacent interior valleys of Mexico. Biological Conservation 130, 25-46.

Gaston KJ (2000) Global patterns in biodiversity. Nature 405, 220-227.

Graham CH, Ferrier S, Huettman F, Moritz C, Peterson AT (2004) New developments in museum-based informatics and applications in biodiversity analysis. Trends in Ecology and Evolution 19, 497-503.

Graham CH, Hijmans RJ (2006) A comparison of methods for mapping species ranges and species richness. Global Ecology and Biogeography 15, 578-587.

Graham CH, Ron SR, Santos JC, Schneider CJ, Moritz C (2004) Integrating phylogenetics and environmental niche models to explore speciation mechanisms in dendrobatid frogs. Evolution 58, 1781-1793.

Grinnell J (1917) The niche-relationship of the California Thrasher. Auk 34, 427-433.

Hanley JA, McNeil BJ (1982) The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143, 29-36.

Holt RD (2003) On the evolutionary ecology of species' ranges. Evolutionary Ecology Research 5, 159-178.

Hurlbert AH, White EP (2005) Disparity between range map- and survey-based analyses of species richness: patterns, processes and implications. Ecology Letters 8, 319-327.

IUCN CI, and NatureServe (2008) An Analysis of Amphibians on the 2008 IUCN Red List. Downloaded on 5 June 2008.

Jetz W, Sekercioglu CH, Watson JEM (2008) Ecological correlates and conservation implications of overestimating species geographic ranges. Conservation Biology 22, 110- 119.

Kissling WD, Carl G (2008) Spatial autocorrelation and the selection of simultaneous autoregressive models. Global Ecology and Biogeography 17, 59-71.

Lamoreux JF, Morrison JC, Ricketts TH, et al. (2006) Global tests of biodiversity concordance and the importance of endemism. Nature 440, 212-214.

Lannoo MJE (2005) Declining amphibians: Conservation status of United States species University of California Press, Berkely.

Legendre P (1993) Spatial autocorrelation: trouble or new paradigm? . Ecology, 1659-1673.

Leroux SJ, Schmiegelow FKA (2007) Biodiversity concordance and the importance of endemism. Conservation Biology 21, 266-268.

Mac Nally R, Fleishman E (2004 ) A successful predictive model of species richness based on indicator species. Conservation Biology 18, 634-646.

Maiorano L, Falcucci A, Boitani L (2006) Gap analysis of terrestrial vertebrates in Italy: Priorities for conservation planning in a human dominated landscape. Biological Conservation 133, 455-473.

Margules CR, Pressey RL (2000) Systematic conservation planning. Nature 405, 243-253.

Margules CR, Sarkar S (2007) Systematic Conservation Planning Cambridge University Press, New York

McPherson JM, Jetz W, Rogers DJ (2004) The effects of species' range sizes on the accuracy of distribution models: ecological phenomenon or statistical artefact? Journal of Applied Ecology 41, 811-823.

Moreno CE, Pineda E, Escobar F, Sanchez-Rojas G (2007) Shortcuts for biodiversity evaluation: a review of terminology and recommendations for the use of target groups, bioindicators and surrogates. International Journal of Environment and Health 1, 71-86.

Myers N (1990) The biodiversity challenge: expanded hot-spots analysis. Environmentalist 10, 243-256.

Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403, 853-858.

Olden JD, Lawler JJ, Poff NL (2008) Machine learning methods without tears: A primer for ecologists. Quarterly Review of Biology 83, 171-193.

Olson DM, Dinerstein E (1998) The global 200: A representation approach to conserving the Earth's most biologically valuable ecoregions. Conservation Biology 12, 502-515.

Pearce J, Ferrier S (2000) Evaluating the predictive performance of habitat models developed using logistic regression. Ecological Modelling 133, 225-245.

Pearson RG, Raxworthy CJ, Nakamura M, Peterson AT (2007) Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34, 102-117.

Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190, 231-259.

Phillips SJ, Dudik M (2008) Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31, 161-175.

Prendergast JR, Quinn RM, Lawton JH, Eversham BC, Gibbons DW (1993) Rare species, the coincidence of diversity hotspots and conservation strategies. Nature (London) 365, 335- 337.

Pressey RL, Humphries CJ, Margules CR, Vanewright RI, Williams PH (1993) Beyond Opportunism - Key Principles for Systematic Reserve Selection. Trends in Ecology and Evolution 8, 124-128.

Purvis A, Gittleman JL, Cowlishaw G, Mace GM (2000) Predicting extinction risk in declining species. Proceedings of the Royal Society B-Biological Sciences 267, 1947-1952.

Rangel TFLVB, Felizola Diniz-Filho JA, Bini LM (2006) Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Global Ecology and Biogeography 15, 321-327.

Raven PH, Wilson EO (1992) A 50-Year Plan for Biodiversity Surveys. Science 258, 1099- 1100.

Raxworthy CJ, Martinez-Meyer E, Horning N, et al. (2003) Predicting distributions of known and unknown reptile species in Madagascar. Nature 426, 837-841.

Rissler LJ, Apodaca JJ (2007) Adding more ecology into species delimitation: ecological niche models and phylogeography help define cryptic species in the black salamander (Aneides flavipunctatus). Syst Biol 56, 924-942.

Rissler LJ, Hijmans RJ, Graham CH, Moritz C, Wake DB (2006) Phylogeographic lineages and species comparisons in conservation analyses: a case study of california herpetofauna. Am Nat 167, 655-666.

Rodrigues AS, Brooks TM (2007) Shortcuts for biodiversity conservation planning: the effectiveness of surrogates. Annual Review of Ecology, Evolution, and Systematics 38, 713-737.

Rodrigues ASL, Tratt R, Wheeler BD, Gaston KJ (1999) The performance of existing networks of conservation areas in representing biodiversity. Proceedings of the Royal Society of London Series B-Biological Sciences 266, 1453-1460.

Scott JM, Davis F, Csuti B, et al. (1993) Gap Analysis - a Geographic Approach to Protection of Biological Diversity (Vol 123, Pg 17, 1993). Journal of Wildlife Management 57, U673- U673.

Simberloff D (1998) Flagships, umbrellas, and keystones: Is single-species management passe in the landscape era? Biological Conservation 83, 247-257.

Tognelli MF, Silva-Garcia C, Labra FA, Marquet PA (2005) Priority areas for the conservation of coastal marine vertebrates in Chile. Biological Conservation 126, 420-428.

Whittaker RJ, Araujo MB, Paul J, et al. (2005) Conservation Biogeography: assessment and prospect. Diversity and Distributions 11, 3-23.

Wiens JJ, Graham CH (2005 ) Niche conservatism: integrating evolution, ecology, and conservation biology. Annual Review of Ecology, Evolution and Systematics 36, 519- 539.

Wilson RJ, Thomas CD, Fox R, Roy DB, Kunin WE (2004) Spatial patterns in species distributions reveal biodiversity change. Nature 432, 393-396.

Table 1. Regression models evaluating the ability of Anuran environmental niche models to predict extent of occurrence richness values.

Predictor Model r P value AIC LPT OLS 0.688 <0.001 10484.968 LPT SAR 0.639 <0.001 10714.231 ModelLPT OLS 0.484 <0.001 11241.544 ModelLPT SAR 0.566 <0.001 10986.954 Hybrid OLS 0.65 <0.001 10666.525 Hybrid SAR 0.728 <0.001 10258.978

Table 2. Regression models evaluating the ability of Caudata environmental niche models to predict extent of occurrence richness values.

Predictor Model r P value AIC LPT OLS 0.7 <0.001 10153.727 LPT SAR 0.708 <0.001 10096.362 ModelLPT OLS 0.638 <0.001 10437.878 ModelLPT SAR 0.651 <0.001 10381.753 Hybrid OLS 0.737 <0.001 9926.039 Hybrid SAR 0.763 <0.001 9750.27

Figure 1. Physiographic provinces of the southeastern United States.

Figure 2 Maps of Anuran richness in the southeastern United States created using (a.) extent of occurrence (EOO) maps, (b.) presence/absence ENM maps with a least point threshold(LPT), (c.) ENM maps with Maxent values and a LPT, and (d.) ENM hybrid maps. Each map is displayed with the lowest values represented by the darkest colors and the highest values represented by the lightest colors. Each color change in this scale represents one half of a standard deviation within the data.

Figurue 3. Scatter plots of Anuran richness values (x axis) and (y axis) values of (a.) presence/absence ENMs with a least point threshold(LPT), (b.) ENMs with Maxent values and a LPT, or (c.) ENMs using a hybrid approach.

Figure 4. Maps of Caudata richness in the southeastern United States created using (a.) extent of occurrence (EOO) maps, (b.) presence/absence ENM maps with a least point threshold(LPT), (c.) ENM maps with Maxent values and a LPT, and (d.) ENM hybrid maps. Each map is displayed with the lowest values represented by the darkest colors and the highest values

represented by the lightest colors. Each color change in this scale represents one half of a standard deviation within the data.

Figure 5. Scatter plots of Caudata richness values (x axis) and (y axis) values of (a.) presence/absence ENMs with a least point threshold (LPT), (b.) ENMs with Maxent values and a LPT, or (c.) ENMs using a hybrid approach.

CHAPTER THREE: COMPARING APPROACHES FOR IDENTIFYING BIODIVERSITY

HOTSPOTS: A CASE STUDY FEATURING PLETHODONTIDS IN THE SOUTHEASTERN

UNITED STATES.

Abstract. Prioritizing geographic areas for conservation purposes is one of the most important and complicated challenges facing conservation biology. Traditionally, prioritization has been based on species richness or on endemic areas. However, these methods tell us nothing about the evolutionary history of the taxa present. The incorporation of evolutionary data (such as phylogenetic diversity) into hotspot identification provides a method to evaluate the amount of unique evolutionary history in a region. Yet, it has been suggested that species richness is a good surrogate for other biodiversity metrics. In this study we compare patterns of traditional biodiversity metrics (richness and weighted endemism) with patterns of biodiversity metrics that integrate phylogenetic information (phylogenetic diversity and phylogenetic endemism) for members of the family Plethodontidae in the southeastern United States. We found that the phylogenetic metrics were weakly correlated with the non-phylogenetic methods. Though, the patterns differentiate at the highest values. Therefore, each method identifies unique and important areas for conservation. We found that none of the methods are a proper surrogate for the others. We strongly urge the incorporation of phylogenetic information into biodiversity metrics whenever possible.

INTRODUCTION

One of the most important and complicated challenges facing conservation biology is the prioritization of geographic areas for conservation. Anthropogenic activities have drastically

elevated extinction rates when compared to background averages, and we are currently experiencing a loss of biodiversity matched only by prehistoric mass extinction events (Pimm et al. 1995; McCallum 2004), with rates of species losses expected to increase in the near future

(Pimm et al. 1995; Mace et al. 2005). The most efficient way to slow losses is through the establishment of a well-supported system of protected areas (Brooks et al. 2004). However, limited funds and time make it imperative that resources are utilized to their full potential in terms of preserving the maximum amount of biodiversity while minimizing losses (James et al.

1999; Bottrill et al. 2008). Yet, this is certainly easier said than done, as oftentimes conservation decisions must be made quickly and without complete information (Soulé 1985).

As the preservation of biodiversity continues to increase in importance, the science of prioritizing areas for conservation has become increasingly sophisticated (Margules and Pressey

2000; Sarkar et al. 2006). Several methods have been proposed that aim to guide decision makers in the selection of protected areas. These methods range from using biodiversity metrics (such as richness and endemism) to focusing on small representative groups in hopes of saving entire assemblages through the preservation of a few (e.g. flagship, umbrella, keystone, or indicator species) (Myers et al. 2000; Mace and Collar 2002; Margules and Sarkar 2007).

The utility of incorporating evolutionary data into conservation planning has been widely recognized (Ehrlich and Wilson 1991; Vane-Wright et al. 1991; Faith 1992; Humphries et al.

1995; Moritz and Faith 1998; Isaac et al. 2007; Davis et al. 2008; Rosauer et al. 2009), and has even been incorporated into several national and international conventions (Moritz and Faith

1998). There have been several proposed methods that take into account evolutionary information: such as phylogenetic diversity (PD; Faith 1992), highlighting areas that may generate or maintain diversity (Neo-endemism; Davis et al. 2008), combining evolutionary

distinctiveness and endangerment (EDGE; Isaac et al. 2007) and phylogenetic endemism (PE;

Rosauer et al. 2009). The incorporation of such methods has been shown to drastically alter which areas are ranked as a high priority for conservation purposes (e.g. Barker 2002; Redding and Mooers 2006; Forest et al. 2007; Isaac et al. 2007; Rosauer et al. 2009).

Phylogenetic diversity (PD) is perhaps the most widely accepted method that explicitly incorporates evolutionary data in the identification of important biodiversity areas. PD is a simple measure that considers the branch lengths of a phylogenetic tree as a surrogate for the underlying diversity of unique or shared evolutionary features (Faith 1992; Moritz and Faith

1998). It is based on the concept that a deeply divergent species represent more evolutionary history and genetic diversity than a species with many recently diverged relatives (Rodrigues and

Gaston 2002).

One of the core strengths of PD is that diversity is measured based on phylogenetic variation, thereby shifting the focus away from species and the inherent taxonomic uncertainty that accompanies them (Mace et al. 2003). It has been demonstrated that conservation priorities based on genotypic data are discordant with those based on existing taxonomic practices (Rissler et al. 2006; Forest et al. 2007). However, PD does not explicitly take into account the threat of severely restricted ranges (But see Faith et al. 2004).

Areas of endemism, or clusters of species with severely restricted ranges, are a focal point of conservation (Pressey et al. 1994; Sechrest et al. 2002; Faith et al. 2004; Brooks et al.

2006), and should be considered a key component in the identification of biologically important areas. Phylogenetic endemism (Rosauer et al. 2009) is a recently developed method that combines elements of weighted endemism (WE) and PD. Weighted endemism circumvents the traditional problem of assigning an arbitrary range size threshold to define an endemic. Instead

WE applies a simple continuous weighting technique, where high values are assigned to species with small ranges and vice versa (Slayter et al. 2007). Other studies have incorporated WE into evolutionary analyses (e.g. Soutullo et al. 2005; Rissler et al. 2006), though PE is alone in using the combination of unbiased branch lengths and equal spatial units (Rosauer et al. 2009). The combination of PD and WE provides a useful tool that can be used to identify areas where high amounts of PD are restricted (Mooers and Redding 2009; Rosauer et al. 2009).

Clearly, the use of phylogenetic data in conservation planning is not faultless or its inclusion would have been quick to sweep into conservation practice. In reality complete or nearly complete phylogenies for large taxonomic groups are rarely available (Isaac et al. 2007), especially for lesser studied taxa or regions. Therefore, it would be valuable to gain an understanding of how patterns of PD and PE are correlated to metrics that do not require specific evolutionary data, such as richness and WE. Previous studies have shown taxonomic richness to be a good surrogate for PD (Polasky et al. 2001; Rodrigues and Gaston 2002; Rodrigues et al.

2005; Torres and Diniz 2004; Brooks et al. 2006), yet others found the two metrics to be decoupled (Forest et al. 2007; Rosauer et al. 2009). Obviously there is not a simple relationship between the two, thus more studies are needed to help clarify what scenarios may cause PD and richness to diverge. Additionally, because PE is a recent proposal we know very little about how it correlates to other metrics, though initial findings for Australian tree frogs and pea-flowered shrubs show that it correlates less with species richness than either PD or WE (Mooers and

Redding 2009). If this finding holds true in other regions and for other biota then focusing on areas of high endemism would not effectively conserve the tree of life (Mooers and Redding

2009, Rosauer et al. 2009).

In this study we document patterns of richness, PD, PE, and WE for members of the family Plethodontidae in the southeastern United States. By doing so we hope to identify areas important to the long-term conservation of eastern plethodontids and to examine the relationship between richness, PD, PE, and WE. Additionally, because factors such as phylogenetic niche conservatism (Peterson et al. 1999) may predispose particular clades to having a disproportionate influence on PE (Rosauer et al. 2009), we explore whether range size has a phylogenetic signal for southeastern plethodontids. The relationship between phylogeny and range size does not appear to be a simple one. It has been argued that within-group range size is not phylogenetically conserved (Gaston 2003), though there is some evidence that taxa with small ranges are more likely to be closely related (Jones et al. 2005).

Plethodontidae is the most speciose family of salamanders, representing approximately

68% of extant caudates (Amphibiaweb 2010). It is one of the most recently diverged salamander families (Wiens et al. 2005) and is also the most differentiated in morphology, ecology, and behavior (Vieites et al. 2007). The southeastern United States is a global hotspot of plethodontid diversity and contains several narrow endemics (Duellman 1999), and three endemic genera

(Desmognathus, Gyronophilus, and Phaeognathus). Plethodontid salamanders are also a key component of many ecosystems in the southeastern U.S., often comprising the majority of the vertebrate biomass in a habitat and impacting ecosystem services (Burton and Likens 1975;

Chippindale et al. 2004). Furthermore, increased conservation research on amphibians is imperative, as they are the most threatened vertebrate group in the world, and populations are declining at an alarming rate (Global Amphibian Assessment; IUCN et al., 2008).

METHODS

Species data

Species range data were downloaded from the NatureServe server (Global Amphibian

Assessment; IUCN et al., 2008) for 75 species in the family Plethodontidae (~20% of the worldwide plethodon diversity). For analysis purposes range maps were rasterized to a grid size of 5 km2 using ArcGIS (9.2). We used a recent time-calibrated phylogeny from Kozak et al.

(2009), which included 3 mitochondrial (Cyt b; ND2; and ND4) and 3 nuclear (BDNF; POMC; and RAG-1) genes. The Kozak et al. (2009) original tree included 184 ingroup taxa and 5590 characters. The final tree, which was used for our analysis, is a maximum likelihood tree with support evaluated with nonparametric bootstrapping. We pruned the final tree to include only the

75 eastern U.S. species (Fig. 1). Details of the phylogenetic analysis for the final maximum likelihood tree can be found in Kozak et al. (2009).

Creation of biodiversity maps

All maps were created using the species rasterized ranges and the Spatial Analyst extension in ArcGIS (9.2) at a size of 0.05°. Richness maps were created by summing all rasterized maps in the raster calculator of the Spatial Analyst toolbox. Range sizes were obtained from the number of occupied grid cells in rasterized ranges. WE values were calculated as the sum of the inverse of the range size of each clade (known as clade range size). PD values were calculated using Phylocom v. 4.1 (Webb et al. 2008). PD is calculated as the sum of the branch lengths on a path linking a clade to the root of the tree, as a proportion of the total length of the tree (Faith 1992). PE values were calculated using equation 3 from Rosauer et al. (2009).

Essentially PE is the sum of branch length divided by clade range for each branch on the path that leads a clade to the root of the tree (Rosauer et al. 2009). WE, PD, and PE values were input into species rasterized ranges for analysis in ArcGIS (9.2). Basic statistics for each map as well

as correlation and covariance between layers were calculated using the Spatial Analyst extension in ArcGIS (9.2). The spatial correlation between two layers is a measure of dependency between the layers. It is calculated as the ratio of the covariance between the two layers divided by the product of their standard deviations, and is therefore a unitless number ranging from -1 to 1.

In order to further explore the statistical properties between maps we ran a series of regression analyses using extracted values from each grid cell. We performed a regression analysis and a null model hypothesis test between each layer using EcoSim v. 7.72 (Gotelli and Entsminger

2009), with 50 000 permutations. Because we were interested in whether or not non-genetic methods (i.e. richness and WE) are proxies for genetic methods (i.e. PD and WE) all regressions were run with the non-genetic method as the predictor variable and the genetic method as the dependent variable.

Phylogenetic correlation of range size

In order to assess to what degree range size is determined by phylogenetic history we used two approaches. First, to test for phylogenetic signal in the variation of range sizes for southeastern plethodontids we calculated Blomberg’s K (Blomberg et al. 2003). Blomberg’s K uses Brownian motion to measure phylogenetic signal. It is computed using two mean square area ratios, one from the phylogeny and data, and one from the phylogeny and Brownian motion

(Revell et al. 2008). We used K as opposed to the similar statistic λ (Pagel 1999) because the likelihood of λ when phylogenetic signal may be greater than expected under Brownian motion is undefined (Freckleton et al. 2002). K is measured from zero to infinity, where low values indicate low phylogenetic signal and vice versa. Because K has an expected value of 1 under

Brownian motion, values less than 1 indicate that closely related individuals resemble each other less than expected under this model. A value > 1 indicates that close relatives are more similar

than expected under Brownian motion (Blomberg et al. 2003). In order to test the null hypothesis of no pattern of similarity for range size between closely related species we used the randomization procedure of Blomberg et al. (2003) using 10 000 permutations. Calculating K and the randomization procedure were carried out using the program Picante (Kembel et al.

2009). We also used the analysis of traits module in Phylocom (Webb et al. 2008) to investigate trait conservatism by node. Phylocom uses Moles’ et al. (2005) contribution index to calculate the contribution of a particular divergence in a phylogeny to the overall variance of a trait across the tree.

RESULTS

Biodiversity maps

We identified four areas that have a high amount of richness for eastern plethodontids

(Fig. 2). The areas of highest richness are centered in the southern Blue Ridge and Valley and

Ridge physiographic provinces. There are also smaller sections of high richness in the

Cumberland Plateau and southern portion of the Appalachian Plateaus. The highest values of WE

(Fig. 3) are in large sections of the southern Blue Ridge and Ouachita Mountains. Additionally, the East Gulf Coastal Plains, the Cumberland Plateau, and the Allegheny Mountains all contain small pockets of very high WE. PD values were clustered in large swaths of the Blue Ridge,

Valley and Ridge, and Cumberland Plateau (Fig. 4). The Appalachian Plateaus, Highland Rim, and Allegheny Mountains also contain narrow sections of high PD. There are five areas that contain small sections with high amounts of PE (Fig. 5): the East Gulf Coastal Plain, the

Ouachita Mountains, the Cumberland Plateau, the Blue Ridge, and the Allegheny Mountains.

Spatial statistics indicate that all biodiversity indices considered are correlated at some level

(Table 1). The highest spatial correlation is between PD and Richness, followed closely by the correlation between PE and WE. These sets of variables also had a high covariance (Table 1).

The least correlated indices are PD and WE (r = 0.19). Regression analyses found a strong correlation between PE and WE and PD and richness (Table 2). Null models indicated that each biodiversity index was significantly more correlated than expected by chance (Table 2).

Phylogenetic correlation of range size

Our results indicated that range size does not exhibit significant phylogenetic signal for eastern plethodontids. We found a K (Blomberg et al. 2003) value of 0.2578, with a P value of

0.021. The highest contribution index (Moles et al. 2005) value (0.51) was for the deep divergence between the clades of Hemidactylium, Eurycea,-Gyrinophilus, Pseudotriton and

Plethodon, Desmognathus, Aneides (Table 3). The next highest value (0.11) was for the divergence of Gyrinophilus, Pseudotriton and Eurycea. Every other divergence within the tree had a value of < 0.01.

DISCUSSION

In this study we have analyzed four different approaches to quantifying the biodiversity patterns for plethodontids in the southeastern U.S., a world hotspot for salamander diversity

(Duellman 1999). Though each index generates a somewhat unique pattern (Figs. 2-5), they are all correlated to some degree (Table 1). Thus it is important to investigate the relationship between each method thoroughly before assuming that correlated biodiversity patterns portray equivalent information. Given that statistical analyses indicated a strong correlation between the patterns of a) richness and PD and b) WE and PE, we will discuss these methods together.

Richness values for southeastern plethodontids are highest in the southern Appalachians.

The areas of highest richness are centered in the Blue Ridge and Valley and Ridge physiographic provinces. Richness is strongly correlated with PD in our study (Tables 1 & 2; Figs. 2 & 4), and multiple studies also have demonstrated a correlation between richness and PD (e.g. Polasky et al. 2001; Rodrigues et al. 2002; Torres et al. 2004; Brooks et al. 2006). Thus, it has been suggested that richness could be a valid surrogate for PD. However, it has been shown that a strong correlation does not necessarily translate into a functional relationship in terms of conservation applicability (Forest et al. 2007). Our analysis is in accord with this conclusion, as there are broad similarities in the patterns created using these methods but a decoupling at key localities. For example, richness patterns do not highlight any high priority areas within the east

Gulf Coastal Plain. However PD recognizes the range of Phaeognathus hubrichti as an area of conservation importance. Phaeognathus hubrichti is a federally listed (IUCN: Endangered) highly endemic species that represents nearly 37 million years of unique evolution (Fig. 1). This example is only one of several that could be made, yet it demonstrates the importance of additional data for real world conservation implications, as conservation planning is not based solely on total numbers but also gains in biodiversity at all levels, including genetic (Forest et al.

2007). Additionally, a scatter plot of richness and PD values for each grid cell (Fig. 6) reveals that although there is a general trend of higher PD values at high richness values there is also a great deal of variation, indicating that richness is not an ideal surrogate for PD.

WE and PE patterns for plethodontids in the southeastern U.S. are broadly similar and highlight a great deal of the southern Blue Ridge province. The Ouachita Mountains,

Cumberland Plateau, East Gulf Coastal Plain, and Allegheny Mountains also harbor areas of high WE and PE (Figs.3 &5). However, similar to the relationship between richness and PD, the

association between WE and PE dissipates rapidly as values increase (Fig. 7). These results corroborate trends noted elsewhere (Rosauer et al. 2009), and thus further enforces the notion that including genetic data into conservation planning enhances the ability to identify unique areas important to conservation planning (Vane-Wright et al. 1991; Crozier 1992; Faith 1992;

Witting and Loeschcke 1995; Moritz and Faith 1998; Isaac et al. 2007; Rosauer et al. 2009).

Our analysis of range size evolution in eastern plethodontids revealed that there is not a strong phylogenetic signal for this trait (K= 0.2578, p= 0.0210). A K value of 0.2578 indicates that range sizes of closely related individuals resemble each other less than expected under Brownian motion. Therefore we do not believe that factors such as niche conservatism (Peterson et al.

1999) influence patterns of PE in this study. However, the contribution index (Table 3) indicates that there may be a phylogenetic signal at deeper levels of the salamander phylogeny. Therefore, niche conservatism should not be ignored in future investigations of PE as it could significantly shape PE patterns at a larger scale or in studies that consider a broad taxonomic assemblage.

The use of biodiversity indices is vital to the identification of areas important to global conservation efforts (e.g. Myers, 1990; Vane-Wright et al. 1991; Faith 1992; Dobson et al., 1993;

Prendergast et al., 1993; Myers et al., 2000; Bonn et al., 2002; Brooks et al. 2004; Ceballos and

Ehrlich, 2006; Forest et al. 2007). Interpreting the relationships and differences between these metrics is fundamental to identifying and prioritizing biodiversity hotspots. Basic statistical correlations between methods can be deceiving and could potentially shift conservation priority settings away from biologically valuable regions (Forest et al. 2007). For example, correlations at low values of metrics have been show to drive statistical patterns, while high end values show a great deal of variation (Forest et al. 2007; Rosauer et al. 2009).

In this study we have demonstrated that each of the four metrics considered add unique data and provide a valuable perspective to the identification of areas important to conservation for plethodontids in the southeastern U.S. Furthermore; we demonstrate that none of these methods can serve as an adequate surrogate for another, despite being highly correlated.

Although each method has its own merits, we advocate the use of genetic data, especially PE, in the prioritization of areas important to plethodontid conservation. PE has the unique ability to highlight species with currently restricted ranges that represent a great deal of evolutionary history, and whose long-term survival is reliant on current conservation actions (Rosauer et al.

2009).

LITERATURE CITED

Barker GM (2002) Phylogenetic diversity: a quantitative framework for measurement of priority and achievement in biodiversity conservation. Biological Journal of the Linnean Society 76, 165-194.

Blomberg SP, Garland T, Jr., Ives AR (2003) Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57, 717-745.

Bonn A, Rodrigues ASL, Gaston KJ (2002) Threatened and endemic species: are they good indicators of patterns of biodiversity on a national scale? Ecology Letters 5, 733-741.

Bottrill MC, Joseph LN, Carwardine J, et al. (2008) Is conservation triage just smart decision making? Trends Ecol Evol 23, 649-654.

Brooks TM, Bakarr MI, Boucher T, et al. (2004) Coverage provided by the global protected-area system: Is it enough? Bioscience 54, 1081-1091.

Brooks TM, Mittermeier RA, da Fonseca GAB, et al. (2006) Global biodiversity conservation priorities. Science 313, 58-61.

Burton TM, Likens GE (1975) Salamander populations and biomass in the Hubbard Brook Experimental Forest, New Hamphshire Copeia 1975, 541-546.

Ceballos G, Ehrlich PR (2006) Global mammal distributions, biodiversity hotspots, and conservation. Proc Natl Acad Sci U S A 103, 19374-19379.

Chippindale PT, Bonett RM, Baldwin AS, Wiens JJ (2004) Phylogenetic evidence for a major reversal of life-history evolution in plethodontid salamanders. Evolution 58, 2809-2822.

Crozier RH (1992) Genetic Diversity and the Agony of Choice. Biological Conservation 61, 11- 15.

Davis EB, Koo MS, Conroy C, Patton JL, Moritz C (2008) The California Hotspots Project: identifying regions of rapid diversification of mammals. Molecular Ecology 17, 120-138.

Dobson AP, Rodrigues JP, Wilcove DS (1993) Geographic distribution of endangered species in the United States Science 275, 550-553.

Duellman WE (1999) Patterns of distributions of amphibians The John Hopkins University Press, Baltimore, MD

Ehrlich PR, Wilson E (1991) Biodiversity Studies: Science and Policy. Science 253, 758-762.

Faith DP (1992) Conservation Evaluation and Phylogenetic Diversity. Biological Conservation 61, 1-10.

Faith DP, Reid CAM, Hunter J (2004) Integrating phylogenetic diversity, complementarity, and endemism for conservation assessment. Conservation Biology 18, 255-261.

Forest F, Grenyer R, Rouget M, et al. (2007) Preserving the evolutionary potential of floras in biodiversity hotspots. Nature 445, 757-760.

Freckleton RP, Harvey PH, Pagel M (2002) Phylogenetic analysis and comparative data: a test and review of evidence. Am Nat 160, 712-726.

Gaston KJ (2003) The structure and dynamics of geographic ranges Oxford University Press, Oxford

Gotelli NJ, Entsminger GL (2009 ) Ecosim: Null models software for ecology. Version 7 Acquired Intelligence Inc. & Kesey-Bear. Jericho, VT 05465. http://garyentsminger.com/ecosim.htm.

Humphries CJ, Williams PH, Vanewright RI (1995) Measuring Biodiversity Value for Conservation. Annual Review of Ecology and Systematics 26, 93-111.

Isaac NJB, Turvey ST, Collen B, Waterman C, Baillie JEM (2007) Mammals on the EDGE: Conservation Priorities Based on Threat and Phylogeny. Plos One 2, -.

IUCN CI, and NatureServe (2008) An Analysis of Amphibians on the 2008 IUCN Red List. Downloaded on 5 June 2008.

James AN, Gaston KJ, Balmford A (1999) Balancing the Earth's accounts. Nature 401, 323-324.

Jones KE, Sechrest W, Gittleman JL (2005 ) Age and area revisited: identifying global patterns and implications for conservation. In: Phylogeny and Conservation (eds. Purvis A, Gittleman JL, Brooks TM), pp. 141-165. Cambridge University Press Cambridge, UK.

Kembel SW, Ackerly DD, Blomberg SP, et al. (2009 ) picante: R tools for integrating phylogenies and ecology. R package version 1.0-0. http://picante.r-forge.r-project.org.

Kozak KH, Mendyk RW, Wiens JJ (2009) Can Parallel Diversification Occur in Sympatry? Repeated Patterns of Body-Size Evolution in Coexisting Clades of North American Salamanders. Evolution 63, 1769-1784.

Mace GM, Collar NJ (2002) Conserving bird biodiversity: General principles and their application Cambridge University Press, Cambridge

Mace GM, Gittleman JL, Purvis A (2003) Preserving the Tree of Life. Science 300, 1707-1709.

Mace GM, Masundire H, Baillie J (2005) Biodiversity. Chapter 4 in: Millenium ecosystem assessment, 2005. Current state and trends: Findings of the condition and trends working group ecosystems and human well-being Island Press, Washington, DC.

Margules CR, Pressey RL (2000) Systematic conservation planning. Nature 405, 243-253.

Margules CR, Sarkar S (2007) Systematic Conservation Planning Cambridge University Press, New York

McCallum ML (2007) Amphibian decline or extinction? Current declines dwarf background extinction rate. Journal of Herpetology 41, 483-491.

Moles AT, Ackerly DD, Webb CO, et al. (2005) A brief history of seed size. Science 307, 576- 580.

Mooers AO, Redding DW (2009) Where the rare species are. Molecular Ecology 18, 3955-3957.

Moritz C, Faith DP (1998) Comparative phylogeography and the identification of genetically divergent areas for conservation. Molecular Ecology 7, 419-429.

Myers N (1990) The biodiversity challenge: expanded hot-spots analysis. Environmentalist 10, 243-256.

Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403, 853-858.

Pagel M (1999) Inferring the historical patterns of biological evolution. Nature 401, 877-884.

Peterson AT, Soberon J, Sanchez-Cordero V (1999) Conservatism of ecological niches in evolutionary time. Science 285, 1265-1267.

Pimm SL, Russell GJ, Gittleman JL, Brooks TM (1995) The Future of Biodiversity. Science 269, 347-350.

Polasky S, Csuti B, Vossler CA, Meyers SM (2001) A comparison of taxonomic distinctness versus richness as criteria for setting conservation priorities for North American birds. Biological Conservation 97, 99-105.

Prendergast JR, Quinn RM, Lawton JH, Eversham BC, Gibbons DW (1993) Rare species, the coincidence of diversity hotspots and conservation strategies. Nature (London) 365, 335- 337.

Pressey RL, Humphries CJ, Margules CR, Vanewright RI, Williams PH (1993) Beyond Opportunism - Key Principles for Systematic Reserve Selection. Trends in Ecology & Evolution 8, 124-128.

Redding DW, Mooers AO (2006) Incorporating evolutionary measures into conservation prioritization. Conserv Biol 20, 1670-1678.

Revell LJ, Harmon LJ, Collar DC (2008) Phylogenetic signal, evolutionary process, and rate. Syst Biol 57, 591-601.

Rissler LJ, Hijmans RJ, Graham CH, Moritz C, Wake DB (2006) Phylogeographic lineages and species comparisons in conservation analyses: a case study of California herpetofauna. Am Nat 167, 655-666.

Rodrigues AS, Brooks TM, Gaston KJ (2005) Integrating phylogenetic diversity in the selection of priority areas for conservation: does it make a difference? . In: Phylogeny and Conservation (eds. Purvis A, Gittleman JL, Brooks TM), pp. 101-119. Cambridge University Press, Cambridge, UK

Rodrigues ASL, Gaston KJ (2002) Maximising phylogenetic diversity in the selection of networks of conservation areas. Biological Conservation 105, 103-111.

Rosauer D, Laffan SW, Crisp MD, Donnellan SC, Cook LG (2009) Phylogenetic endemism: a new approach for identifying geographical concentrations of evolutionary history. Molecular Ecology 18, 4061-4072.

Sarkar S, Pressey RL, Faith DP, et al. (2006) Biodiversity conservation planning tools: present status and challenges for the future. Annual review of Environment and Resources 31, 123-159.

Sechrest W, Brooks TM, da Fonseca GAB, et al. (2002) Hotspots and the conservation of evolutionary history. Proceedings of the National Academy of Sciences of the United States of America 99, 2067-2071.

Slatyer C, Rosauer D, Lemckert F (2007) An assessment of endemism and species richness patterns in the Australian Anura. Journal of Biogeography 34, 583-596.

Soulé ME (1985) What is conservation biology? Bioscience 35, 727-734.

Soutullo A, Dodsworth S, Heard SB, Mooers AO (2005) Distribution and correlates of carnivore phylogenetic diversity across the Americas. Animal Conservation 8, 249-258.

Torres NM, Diniz-Filho JAF (2004) Macroecology of New World carnivores: constraint envelopes and analysis of phylogenetic patterns. Iheringia Serie Zoologia 94, 155-161.

Vane-Wright R, Humphries C, Williams PH (1991) What to protect-systematics and the agony of choice. Biological Conservation 55, 235-254.

Vieites DR, Min MS, Wake DB (2007) Rapid diversification and dispersal during periods of global warming by plethodontid salamanders. Proc Natl Acad Sci U S A 104, 19903- 19907.

Webb CO, Ackerly DD, Kembel SW (2008) Phylocom: software for the analysis of phylogenetic community structure and trait evolution. Bioinformatics (Oxford) 24, 2098-2100.

Wiens JJ, Bonett RM, Chippindale PT (2005) Ontogeny discombobulates phylogeny: Paedomorphosis and higher-level salamander relationships. Systematic Biology 54, 91- 110.

Witting L, Loeschcke V (1995) The Optimization of Biodiversity Conservation. Biological Conservation 71, 205-207.

Table 1. Covariance and correlation between biodiversity metrics for southeastern U.S. plethodontids. Covariance values are in the upper diagonal. Correlation values are in the lower diagonal. PE = Phylogenetic endemism, PD= Phylogenetic diversity, WE= weighted endemism.

Layer PE PD WE Richness

PE 1.00 1.22 7.66 1.54

PD 0.20 1.00 2.10 8.57

WE 0.88 0.19 1.00 2.84

Richness 0.22 0.94 0.21 1.00

Table 2. Null model regression analysis between non-genetic metrics (WE & Richness) and genetic metrics (PE & PD) for southeastern U.S. plethodontids. PE = Phylogenetic endemism, PD= Phylogenetic diversity, WE= weighted endemism.

Predictor Dependent

variable variable r2 Permutations p value

WE PE 0.723037 50 000 <0.0000

WE PD 0.086962 50 000 <0.0000

Richness PE 0.082790 50 000 <0.0000

Richness PD 0.898583 50 000 <0.0000

Table 3. Three Highest contribution index values for range size in southeastern U.S. plethodontids.

Contribution #

Split Index Age taxa

Hemidactylium,Eurycea/Plethodon,Desmognathus,

Aneides 0.516334 61 88

Hemidactylium,Eurycea/Gyrinophilus,

Pseudotriton 0.110414 55.947815 18

Gyrinophilus, Pseudotriton/Eurycea 0.003489 37.570267 17

Aneides,Desmognathus/Plethodon 0.003025 48.13681 70

P. cinerus group/ P. glutinosus group 0.009722 27.089523 41

Figure 1. Maximum-likelihood phylogram of plethodontids found in the southeastern U.S. with a summary of divergence times. Tree modified from Kozak et al. 2009.

Figure 2. Richness patterns for plethodontids in the southeastern U.S. calculated for 0.05° cells.

Figure 3. Weighted endemism (WE) patterns for plethodontids in the southeastern U.S. calculated for 0.05° cells.

Figure 4. Phylogenetic diversity (PD) patterns for plethodontids in the southeastern U.S. calculated for 0.05° cells.

Figure 5. Phylogenetic endemism (PE) patterns for plethodontids in the southeastern U.S. calculated for 0.05° cells.

Figure 6. Scatter plot of richness and phylogenetic diversity (PD) values for southeastern U.S. plethodontids.

Figure 7. Scatter plot of weighted endemism (WE) and phylogenetic endemism (PE) values for southeastern U.S. plethodontids.

Figure 8. Physiographic provinces of the southeastern United States.

CHAPTER FOUR: ESTIMATING THE EFFECTS OF HABITAT MODIFICATION ON

GENETIC PATTERNS AND POPULATION CONNECTIVITY: A CASE STUDY USING

THE FEDERALLY THREATENED RED HILLS SALAMANDER (PHAEOGNATHUS

HUBRICHTI).

Abstract. Estimating gene flow and population connectivity is a fundamental goal of conservation genetics, especially for imperiled species. Many factors can influence dispersal patterns and therefore gene flow in a natural landscape. However, these patterns can be substantially distorted by the impacts of anthropogenic habitat modification. Studies of genetic divergence can reveal the influence of altered landscape on gene flow; yet it is difficult to differentiate whether such patterns are due to historical or contemporary factors and whether they are influenced by life history characteristics. In this study we investigate the spatial genetic patterns of the federally threatened Red Hills salamander (Phaeognathus hubrichti) using 10 microsatellite markers. Our results indicate that there are 5 well-supported populations (FST =

0.13009-0.1879) across the entire range of P. hubrichti, and that current migration rates between populations are low (m = 0.0025-0.0687). We also estimate that P. hubrichti habitat has been reduced by as much as 86% when compared to historical levels. By accounting for history and species characteristics we demonstrate that this loss and fragmentation of habitat has had an immense impact on P. hubrichti in the form of reduced migration, bottlenecks, and high levels of inbreeding. We discuss the results of this study in terms of the direct impact to P. hubrichti and in the broader context of conservation genetics.

INTRODUCTION

The identification of intra-population genetic structure and diversity is fundamental to many fields within evolutionary biology and ecology. From the perspective of conservation biology this information has become increasingly important in conservation planning for imperiled species (e.g. Geist and Kühn 2005; Pabijan et al. 2005; Dixon et al. 2006; Schwartz et al. 2007; Marshall et al. 2009; Matern et al. 2009; Straub and Doyle 2009). Maintaining genetic diversity, and therefore evolutionary potential, is fundamental to the long-term survival and recovery of at-risk species (Avise 2004; Morgan et al. 2008). Theoretical models have demonstrated that the maintenance of genetic diversity and population viability is critically dependent on gene flow among local populations (Gilpin and Hanski 1991; Harrison 1991). This transfer of genes between populations influences a number of evolutionary processes, such as population persistence, release from inbreeding depression, and adaptive response (Frankham et al. 2002). Therefore, it has become a crucial goal of conservation genetics to not only identify the overall genetic structure, diversity and connectivity between populations, but also to understand the factors that shape such patterns.

Estimating migration and gene flow between populations is an increasingly important task for conservation biologists as habitat fragmentation continues to isolate populations and leave them more susceptible to the effects of genetic stochasticity and inbreeding depression

(Frankham et al. 2002). It has been well documented that a heterogeneous landscape can impede gene flow (Ricketts 2001; Funk et al. 2005; Spear et al. 2005; Wang 2009; Zellmer and Knowles

2009). However, landscape features are not static through time, and a quickly changing landscape (as in the case of anthropogenic modification) can make it difficult to interpret current patterns (Zellmer and Knowles 2009). In order to gain a more complete picture of the effects of anthropogenic habitat modification on gene flow, we must first establish a baseline

understanding of natural migration patterns. Of course certain life history traits, such as low dispersal ability, will affect the overall impact of fragmentation. This is especially true in the case of habitat specialists, which may already face a myriad of dispersal barriers via the lack of suitable habitat, or when dealing with secretive, rare, fossorial, or naturally fragmented organisms. Yet, one area of conservation research that has been largely overlooked is the effect of anthropogenic fragmentation on naturally fragmented populations (Mayer et al. 2009).

In the case of naturally fragmented populations it may be difficult to differentiate whether genetic structuring and gene flow patterns are due to their natural fragmentation, anthropogenic fragmentation, or a combination of both. Though it has become abundantly clear that fragmentation reduces gene flow in a wide variety of species (Epps et al. 2005; Proctor et al.

2005; Coulon et al. 2006; Cushman et al. 2006; Vandergast et al. 2007; Zellmer and Knowles

2009), it is still difficult to disentangle historical and contemporary effects. Distinguishing the roles that natural and anthropogenic fragmentation have on species’ genetic structure should be an important consideration in any management or recovery plan. Such considerations would allow for a more effective identification of gene flow corridors and would aid in the identification of biologically important sub-populations (such as ESUs; Ryder 1986). Despite the inherent difficulty of addressing such questions, new methods in genetic analysis and GIS are emerging that have allowed investigators to begin tackling such problems (See Reed et al. 2009;

Zellmer and Knowles 2009).

The major goals of this study are to analyze the link between spatial genetic patterns with past and present landscape features and to gain insight into the basic population statistics of the federally threatened (IUCN: Endangered) Red Hills salamander (Phaeognathus hubrichti),

Specifically we aim to: (I) define the population structure and population parameters across the

range of P. hubrichti; (II) compare estimates of recent and historical patterns of migration; (III) establish the role that landscape plays in shaping genetic patterns; and (IV) to investigate the effect that anthropogenic habitat modification has had on recent patterns of migration and demographic events. P. hubrichti has an extremely restricted range (Fig. 1) due to their reliance on a set of specific geologic formations within the Red Hills region of southern Alabama (Dodd

1991). These geologic layers are naturally patchy and create a fragmented landscape for the species. Additionally, because of their fossorial nature and life history characteristics they require steep ravines (generally with a slope greater than 35°), north facing slopes, outcrops of siltstone, a mature hardwood canopy, and a loamy topsoil (Schwaner and Mount 1970; Jordan and Mount 1975; Dodd 1991). In addition to their naturally intermittent range, they also face severe fragmentation from anthropogenic habitat modification. Large timber companies own the vast majority of their range, which has resulted in a great deal of habitat being negatively impacted by pine plantations (Dodd 1991). The results of this study will help to understand basic, yet crucial, data on an endangered and secretive species that we know surprisingly little about. We hope these data will not only shed light on the biology of the species but also lead to a comprehensive recovery plan.

METHODS

Population sampling:

Tissue samples were collected following U.S. Fish and Wildlife Threatened Species

Permit TE136961-0. For each individual we collected a small tail-clip following standard protocols. Tissue was immediately preserved in 95% ethanol and transferred to a -80°C freezer at the University of Alabama Herpetology Collection. Sampling occurred across the entire known range of P. hubrichti, but was focused on sites that had been surveyed by Dodd (1991). In total

we collected 105 individuals from 21 unique localities (Fig. 1). In view of the fact that P. hubrichti burrows are difficult to find and may be distributed patchily across even the best habitat, we considered a unique locality to consist of one USGS quadrangle. However, unique decimal degree coordinates were taken for each individual.

Microsatellite analysis:

We used Qiagen DNeasy tissue kit and protocol (Qiagen Inc, Valencia, CA) to extract

DNA from all tissue samples. We amplified 10 microsatellite markers using polymerase chain reaction (PCR). Each primer was developed specifically for use in P. hubrichti (Lance et al.

2009). Information on each primer as well as PCR conditions can be found in Lance et al.

(2009). Each Locus was amplified individually and labeled PCR products were run on an ABI

3730 Genetic Analyzer. Samples were genotyped using GeneMapper 3.7 software (Applied

Biosystems, Inc.). Scoring and quality control of data were done using GeneMarker V. 1.7

(Softgenetics, LLC). Microchecker V2.2.3 (Van Oosterhout et al. 2004) was used to check for possible null alleles, linkage disequilibrium and scoring errors. Data quality and repeatability were tested by re-genotyping 10 individuals per locus. This test resulted in a repeatability success rate of 98%. Hardy-Wienberg equilibrium proportions for each population and locus were tested using an approach analogous to Fisher’s exact test with a Markov chain of 1,000,000 iterations and 10,000 dememorization steps (Guo and Thompson 1992) in Arlequin version 3.11

(Excoffier et al. 2005).

Population genetic structuring and diversity:

We used two Bayesian methods to investigate the genetic structuring of populations. The first was implemented in TESS V. 2.3 (Chen et al. 2007). TESS uses hidden Markov random

fields in order to model spatial dependence among individuals (Chen et al. 2007). This approach has the advantage that it incorporates the a priori assumption that nearby individuals are more likely to have similar allele frequencies than individuals from distant localities. TESS was run for 100,000 simulations with a burn-in of 20,000. To estimate K we ran 100 replicates each for K values ranging from 3 to 9. For each K we averaged the 10 best DIC values and plotted them.

Once we established the proper K value the 10 runs with the lowest DIC values for that K were exported to CLUMPP V. 1.1.2 (Jakobsson and Rosenberg 2007). CLUMPP uses three algorithms in order to properly match cluster membership over multiple runs. We repeated the above procedure using both the admixture and the no admixture models implemented in TESS V. 2.3

(Chen et al. 2007). Since the models did not vary significantly we used the results from the no admixture model as recommended by the authors. We also studied the spatial genetic patterns by using STRUCTURE 2.2 (Pritchard et al. 2000). Structure uses a Bayesian framework to assign individuals to populations based on their multilocus genotypes and is one of the most commonly used structuring programs. The ∆K method of Evanno et al. (2005) was used to assess the best value of K. For each run of STRUCTURE, the program was run for 1,000,000 MCMC cycles, with a burn-in of 100,000 and default settings. We also used Arlequin version 3.11 (Excoffier et al. 2005) in order to determine the ability of each Bayesian clustering method to assign the genotypes of individuals to the populations from which they came (Paetkau et al. 1995). Log- likelihoods were calculated using the allele frequencies from the observed data, and therefore providing the global individual likelihood. The global individual likelihood is the likelihood of that individual coming from the predetermined population for each locus of the individual’s genotype (Cabe et al. 2006).

We investigated patterns of intra and inter population diversity among assigned populations using Arlequin V. 3.11. We performed a locus-by-locus AMOVA (Excoffier et al.

1992) using 10,000 permutations and computed pairwise and global values of FST using 1000 permutations for significance followed by a sequential Bonferroni correction. Observed heterozygosity (HO), expected heterozygosity (HE) (Nei 1987), allelic diversity, and fixation index (FIS) for all populations were also calculated using Arlequin V. 3.11.

Estimates of recent and historical gene flow:

In order to estimate the historical patterns of gene flow we used the program MIGRATE

V. 3.0.3 (Beerli and Felsenstein 1999,2001). MIGRATE uses coalescent theory and Markov chain Monte Carlo techniques to estimate historical pairwise migration rates (M=m/µ, where m= migration rate and µ= mutation rate) and effective population sizes (Θ=4Neµ, where Ne is effective population size). Distributions were estimated using the Bayesian implementation of

Migrate based on the accuracy of the Bayesian approach under a wider variety of conditions

(Beerli 2006). Following the recommendations of the author, we did an initial run on our data set using FST to find the start parameters, and the results of this run were used as start parameters for subsequent runs. Results from different runs were stable, indicating that the Markov chains had likely converged on the stationary distribution. For each run we used the continuous Brownian mutation model and ran 5,000,000 generations per long chain with a burn-in of 100,000. We also used BayesAss+ V.1.3 (Wilson and Rannala 2003) to estimate more recent migration rates.

Although both programs use a Bayesian MCMC approach, BayesAss+ uses a genetic assignment method rather than a coalescent method to estimate gene flow. Genetic assignment methods tend to estimate more recent dispersal rates, as compared to coalescent methods that are closer to long-term averages (Berry et al. 2004; Paetkau et al. 2004). We performed 5 runs (each with

different seed values) of 5 million generations with a 2 million generation burn-in and sampled the chain every 2,000 generations. Estimates of M from MIGRATE were converted to proportion of migrants (m) for populations by using the formula m = Mµ, where µ = 5.4 X 10-4 (Goldstein et al. 1995; Howes et al. 2009). Recent and historical gene flow estimates (m) were compared using a Wilcoxon matched pairs test in STATISTICA V. 6.0.

Recent demographic events:

To test for the signature of recent bottleneck events we used the program Bottleneck

V1.2.02 (Cornuet and Luikart 1996; Piry et al. 1999). Bottleneck is based on the theory that in a population bottleneck the allelic diversity will decline more rapidly than the heterozygosity.

That is to say, because of the loss of rare alleles the observed heterozygosity is larger than the heterozygosity expected based on the number of alleles present if the locus was at mutation-drift equilibrium. We ran the infinite alleles model (IAM) because it is most likely to fit the large number of interrupted and compound microsatellites in our study (Cornuet and Luikart 1996).

Because interrupted and compound microsatellites tend to fit the IAM much more closely they are among the most useful markers for detecting bottlenecks using this method (Cornuet and

Luikart 1996). We also tested the robustness of the model by varying the parameters using the two-phase model (TPM) with varying degrees of stepwise mutation model (SMM; 10 to 40% in steps of 10). All tests were run using 10,000 permutations. One-tailed Wilcoxon tests were used to determine the significance of heterozygosity excess or deficiency for each population.

Landscape analyses:

To examine the effect that landscape features and specific habitat requirements of P. hubrichti have on population structure we used a series of Mantel tests (Wright 1943; Mantel

1967; Smouse et al. 1986) to test for correlations between geographic distance and genetic

distance. As true Euclidian distances are often not the best indicator of genetic distance (Funk et al. 2005; Spear et al. 2005; Storfer et al. 2006; Wang et al. 2009; Wang 2009) we also tested alternative distance measures. We tested if the distance along steep slope (Fig. 3) was more closely associated with genetic distance. Additionally, we created two resistance layers in

ARCGIS V 9.2 using the spatial analyst extension. The first distance was calculated using a resistance layer created by assigning steep slopes as the most conductive to migration. This is a valid assumption due to the fact that P. hubrichti requires steep slopes to burrow and most likely cannot migrate long distances without suitable habitat. The second distance was a combination of slope conductance and assigning stiff penalties to areas of unsuitable geology. Each of these layers was imported into Circuitscape 2.2 (McRae 2006). Circuitscape uses circuit theory to evaluate a total landscape resistance over multiple paths. The advantage here being that in most biological systems it is highly unlikely that a single path (as in least cost path analysis) will explain gene flow over many generations (Lee-Yaw et al. 2009). All Mantel tests were done using FST genetic distances and were conducted using the Isolation By Distance Web Service V.

3.16 (Jensen et al. 2005) using 10,000 permutations.

Fragmentation analysis:

We used the land cover data from the Alabama Gap Analysis Project (AL-GAP; Kleiner et al. 2007) to investigate the amount of anthropogenic fragmentation that has taken place throughout the range of P. hubrichti. The AL-GAP data is a 71-class land cover data-base, where vegetative classes are mapped to NatureServe’s terrestrial ecological systems as defined in

Comer et al. (2003). The data source for the layer is Enhanced Thermatic Mapper Plus (ETM+) satellite imagery, refined at a 30-meter resolution. Land cover within the distribution range was quantified as: 1) habitat that is currently viable for P. hubrichti and 2) habitat that was most

likely conducive to either habitation or migration prior to modification. From these data we made two estimates of current viable habitat as a percentage of historical habitat: a conservative estimate that includes both mesic slope and current longleaf pine (Pinus palustris) habitat, and a liberal estimate that uses only mesic slope habitat. The rationale for including longleaf pine is that historically populations would have been separated by stands of P. palustris, therefore this habitat may be more conducive to gene flow than typical pine plantations.

RESULTS

Genotypic data:

A total of 100 alleles were observed for the 10 loci sampled across all populations of P. hubrichti. The mean number of alleles per locus was 5.94 with a standard deviation of 0.641. We did not find the presence of linkage disequilibrium, scoring errors, or null alleles. However, some of the missing data (especially locus 62) could be due to null alleles causing a failure to identify the alleles as loci. There were deviations from HWE in some loci in some of the populations (no more than 2 for any single population), but since the deviations were not consistent across all populations we did not exclude them from the analyses.

Population genetic structuring and diversity:

TESS (Chen et al. 2007) and STRUCTURE (Pritchard et al. 2000) each identified 5 populations throughout the range of P. hubrichti (Fig. 2). Both programs identified the same populations whether or not admixture was used (for TESS), lending support that these populations are a biological reality. Furthermore, a likelihood assignment test (Paetkau et al.

1995) as performed in Arlequin (Excoffier et al. 2005) showed overwhelming support for each of the populations. Consistent with these results, the AMOVA (Table 1) indicated a high amount

(15.58%) of variation between populations, as well as a high global fixation index (0.1558).

Pairwise FST values were also high (Table 2), with each population pair having a value above

0.13. All FST value were highly significant after Bonferroni correction. We found large differences in FIS values between populations (Table 3), although each population did exhibit a heterozygosity deficiency.

Estimates of recent and historical gene flow:

Recent estimates of gene flow using BayesAss+ were highly variable between populations

(Table 4). Estimates ranged from a mean of 0.002 to a mean of 0.06. We found no asymmetric gene flow (non-overlapping 95% confidence intervals) for any of the population pairs. For one population (Pop. 1) the analysis indicated that the results were close to those being generated by uninformative data, so we urge the use of caution when considering BayesAss+ estimates for that population. However, apart from one seemingly high estimate of migration from population 2 into population 1, the remaining estimates concerning that population are not outside the realm of normalcy nor are they biologically unfeasible. The estimates for the 3 runs of BayesAss+ did not vary statistically (t-tests; P < 0.05), indicating that the method produced consistent results pertaining to this data set.

Historical estimates of gene flow using MIGRATE (Beerli 2006) were overall more consistent than recent estimates (Table 4) but still showed a fair amount of variation between population pairs. One population pair did exhibit a pattern of historically asymmetric migration; population 3 showed a much higher rate of migration into population 2 than vice versa. Overall, the migration rates from MIGRATE fall in line with what one would normally expect (i.e. neighboring populations seem to have the highest rate of migration), and at least at a range-wide scale it does not appear that any of the populations are genetic sinks. Estimates of Θ across all populations were nearly identical (Table 5), with all values between 0.09 and 0.1. The Wilcoxon

test indicated that the BayesAss+ estimates for recent m (migration rate) were significantly lower

(P < .001) than historical rates from MIGRATE, with most of the values being a magnitude lower.

Recent demographic events:

We found evidence for recent genetic bottlenecks in up to 3 of the 5 populations (Table

6). The IAM model indicates that 3 populations have experienced a bottleneck in the recent past

(Table 6). The sensitivity analysis indicates that two of the populations lose support for a recent bottleneck as more of the SMM model is introduced into the algorithm (Table 6). However, because the IAM model is more appropriate given our data (Cornuet and Luikart 1996) we feel that there is strong support for at least two of these populations (1&4). Population 3 loses support quickly as SMM is added into the model and therefore is considered the least likely of the 3 to have experienced a recent bottleneck.

Landscape analyses:

Simple isolation by distance analyses (Mantel tests; Mantel 1967) indicated that there is not a significant relationship between Euclidian distance and genetic distance for P. hubrichti

(Table 7). There is also not a significant relationship between the distance measured along suitable slope and genetic distance, an approach more analogous to how the salamander crawls

(rather than how the crow flies). We did however find a significant relationship between geographic and genetic distance when we used a model that incorporated slope as a conducting

(compared to the lack of resistance for other habitat) factor in Circuitscape (McRae 2006).

However, adding geologic layers into this model weakened the explanatory power (Table 7).

Fragmentation analysis:

The fragmentation analysis indicated that the habitat which P. hubrichti requires has been severely disturbed via anthropogenic alteration. Using ALGAP data (Kleiner et al. 2007) we created a conservative and liberal estimate. The true amount most likely lies within this range of estimates. At the lower end, we estimate that 69.5% of the once available habitat has been altered. Conversely, our liberal approximation puts the upper bound at 86.1% of all available habitat having been lost or altered significantly.

DISCUSSION

This study demonstrates the importance of adopting a multifaceted approach, including historical and landscape data, when analyzing the genetic patterns of a species and the inherent conservation implications of such results. Recent advances in Landscape genetics (Manel et al.

2003) have demonstrated that it is a useful endeavor to incorporate multiple sources of data (such as landscape resistance) into a population genetics framework (e.g. Storfer et al. 2007; Lee-Yaw et al. 2009; Wang 2009; Wang and Summers 2010). Here we demonstrate that in addition to considering landscape, it is valuable to explicitly account for the effect that life history characteristics and the history of habitat alteration have on genetic structuring and gene flow between populations. In this study we were able to show evidence that P. hubrichti naturally exhibits a low migration rate and therefore strong population structuring. Using these baseline data we demonstrate that recent migration rates between populations are minute in addition to, not because of, the species’ low vagility. By taking this approach we can more definitely state that habitat modification has had a negative effect on the species, thereby removing doubt that such results are due to the life-history characteristics of a naturally fragmented species.

Results from our study demonstrate that the federally threatened Red Hills salamander exhibits a strong pattern of genetic structuring across its entire geographic range. Bayesian

clustering identified five distinct and well supported populations (Fig. 2). These groupings were highly supported by an abundance of evidence, including high FST values between populations

(Table 2). It is not surprising that a species with the characteristics of P. hubrichti would exhibit such a high degree of spatial structuring across their range. However, based on our previous knowledge of the species distribution and genetic history (Shwaner and Mount 1970; McKnight et al. 1991) the geographic nature of the population structuring was unexpected. Mitochondrial data has suggested that there were only two main lineages within the species (McKnight et al.

1991, thus our results are novel in that they identify five strongly supported populations.

Previous surveys have shown that P. hubrichti is closely tied to three geologic formations, all of which are a “claystone” that help retain moisture and sustain burrow shape (Shwaner and Mount

1970; French and Mount 1978; Dodd 1991). Nonetheless, our results demonstrate that the absence of these geologic formations does not necessarily translate to an absence of gene flow

(Fig. 2; Table 2; Table 4; Table 7). Additionally, previous research utilizing mtDNA has suggested that large rivers may impede gene flow (McKnight et al. 1991), but our results show little support for this hypothesis (Fig.2; Table 2; Table 4). In fact, populations seem to be maintained by the continuous presence of suitable slope (Fig. 3;), which often accompanies river drainages. This conclusion is also supported by our landscape genetic analysis. Mantel test results indicated that the availability of steep (> 35°) slope between populations best predicts genetic distance (Table 7). Interestingly, our landscape genetic analysis also provides evidence that geology does not play a large role in shaping genetic patterns, a result that supported by the

Bayesian clustering analysis.

Estimates of historical rates of migration were fairly low (Table 4), but are similar to values found for other low vagility amphibians (e.g. Wang 2009). However, estimates of recent

migration rates were significantly lower than historical rates (Wilcoxon matched pairs Test; P <

0.001). By comparing these rates (Table 4) we can infer that migration rates in the recent past have declined a great deal. There is little doubt that this decline can be attributed to anthropogenic habitat destruction and modification. In fact, other analyses corroborate strongly with this conclusion. For example, our fragmentation analysis indicated that P. hubrichti has lost between 69.5 and 86.1% of their original habitat. It is unlikely that losing this amount of habitat would not affect a species’ migration rates, especially in a naturally fragmented species with low dispersal ability.

The effects of fragmentation can also be seen through an examination of recent demographic events, which reveal that multiple populations have undergone a recent bottleneck

(Table 6). This number is most likely a low-end estimate since the two populations that were not identified as having undergone a recent bottleneck (pops. 2& 5; Table 6) break one of the key assumptions of bottleneck analyses, the absence of sub-population structure (Busch et al. 2007;

Marshall et al. 2009). Estimates of very small effective population sizes (Ne) for all populations

(Table 5) support the evidence for recent bottlenecks. However, we urge caution, as low effective population sizes can also result from other factors that remain unknown about P. hubrichti, such as a strong reproductive skew or uneven sex ratios (Beebee 2005). Effective population sizes this low are extremely rare in natural populations and testify to the amount of fragmentation this species has experienced. In fact, other amphibians that display a similar Ne are either threatened species or found in populations that have experienced heavy fragmentation (e.g.

Rowe and Beebee 2004; Funk et al. 1999; Wang 2009).

Our study has several important implications for the recovery and management of P. hubrichti. Protection of continuous slope habitat is seemingly the most important element in any

recovery plan for this species. We have shown through the use of a landscape genetics approach that a steep slope habitat is key to gene flow and ultimately the connectivity of populations.

Surveys for the species (French and Mount 1978; Dodd 1991) have clearly demonstrated that high-quality slope habitat (including a mature hardwood over-story) is essential for a healthy local population. We have shown here that similar habitat is also the key to the overall evolutionary stability and therefore the long-term survival of the species. Additionally, we show evidence for the negative effects of anthropogenic habitat fragmentation by showing a dramatic decline in recent migration rates and the presence of bottlenecks. Although we estimate that as much as 86.1% of the current habitat has been lost or altered, there is no way to estimate how much of the habitat has been negatively affected by forestry practices such as clear cutting to the edge of a the hardwood canopy, or chemical preparation of surrounding areas. Therefore, we urge the adoption of the habitat guidelines laid out by Dodd (1991).

The future status of P. hubrichti is highly dependent on the establishment of protected areas. During the nearly 35 years that P. hubrichti has been listed as a USFWS threatened species the only protection afforded them has been in the form of habitat conservation plans

(HCPs), generally with large timber companies. Although HCPs are a viable tool for the conservation of a species, they do not always incorporate sufficient scientific data to ensure species persistence (Harding et al. 2001). In fact, most current P. hubrichti HCPs allow for selective harvest on any habitat with slope less than 28 degrees. This type of habitat may not contain the highest number of individuals, but it is likely vital to gene flow between populations.

The results of our study have broad conservation and population genetic implications. We have demonstrated that the genetic patterns of a naturally fragmented and low vagility species are heavily impacted by the acts of anthropogenic habitat destruction. Recent work (Zellmer and

Knowles 2009) on another amphibian species (Rana sylvatica) has revealed a correlation between recent human landscape alteration and genetic differentiation. These two studies are examples that current genetic patterns can be misleading if recent landscape changes are not considered. However, it can be difficult to unravel the effects of habitat modification and the natural ability of a species to disperse. Through the methods employed here we were able to demonstrate that the natural dispersal ability of a species can be controlled for while trying to gauge the impacts of recent events. Although there have been numerous studies that have established an effect of recent demographic events (such as bottlenecks) on genetic differentiation (e.g. Bouzat et al. 1998; Rowe et al. 1998; Wang 2009), there have been relatively few that are able to successfully disentangle such factors as historical conditions or the impacts of life-history traits (but see: Austin et al. 2004; Howes et al. 2009; Zellmer and Knowles 2009).

We contend that the integration of such data into a population or conservation genetics program allows for a more thorough understanding of the synergistic factors that create current patterns of genetic structure and gene flow. These data are crucial to the long-term evolutionary stability and survival of threatened and imperiled species, especially with the ever-growing intrusion of habitat alteration.

LITERATURE CITED

Austin JD, Lougheed SC, Boag PT (2004) Controlling for the effects of history and nonequilibrium conditions in gene flow estimates in northern bullfrog (Rana catesbeiana) populations. Genetics, 168, 1491-1506.

Avise JC (2004) Molecular Markers, Natural History and Evolution (Second Edition) Sinauer, Sunderland, MA.

Beebee TJC (2005) Conservation genetics of amphibians. Heredity, 95, 423-427.

Beerli P (2006) Comparison of Bayesian and maximum-likelihood inference of population genetic parameters. Bioinformatics, 22, 341-345.

Beerli P, Felsenstein J (1999) Maximum-likelihood estimation of migration rates and effective population numbers in two populations using a coalescent approach. Genetics 152, 763- 773.

Beerli P, Felsenstein J (2001) Maximum likelihood estimation of a migration matrix and effective population sizes in n subpopulations by using a coalescent approach. Proceedings of the National Academy of Science U S A, 98, 4563-4568.

Berry O, Tocher MD, Sarre SD (2004) Can assignment tests measure dispersal? Molecular Ecology, 13, 551-561.

Bouzat JL, Lewin HA, Paige KN (1998) The ghost of genetic diversity past: Historical DNA analysis of the greater prairie chicken. American Naturalist, 152, 1-6.

Busch JD, Waser PM, DeWoody JA (2007) Recent demographic bottlenecks are not accompanied by a genetic signature in banner-tailed kangaroo rats (Dipodomys spectabilis). Molecular Ecology, 16, 2450-2462.

Cabe PR, Page RB, Hanlon TJ, et al. (2006) Fine-scale population differentiation and gene flow in a terrestrial salamander (Plethodon cinereus) living in continuous habitat. Heredity, 98, 53-60.

Chen C, Durand E, Forbes F, Francois O (2007) Bayesian clustering algorithms ascertaining spatial population structure: a new computer program and a comparison study. Molecular Ecology Notes, 7, 747-756.

Comer P, Faber-Langendoen D, Evans R, et al. (2003 ) Ecological systems of the United States: A working Classification of U.S. Terrestrial Systems. NatureServe, Arlington, VA.

Cornuet JM, Luikart G (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics, 144, 2001-2014.

Coulon A, Guillot G, Cosson JF, et al. (2006) Genetic structure is influenced by landscape features: empirical evidence from a roe deer population. Molecular Ecology, 15, 1669- 1679

Cushman SA (2006) Effects of habitat loss and fragmentation on amphibians: A review and prospectus. Biological Conservation, 128, 231-240.

Dixon JD, Oli MK, Wooten MC, et al. (2006) Effectiveness of a regional corridor in connecting two Florida black bear populations. Conservation Biology, 20, 155-162.

Dodd CK (1991) The Status of the Red Hills Salamander Phaeognathus-Hubrichti, Alabama, USA, 1976-1988. Biological Conservation, 55, 57-75.

Epps CW, Palsboll PJ, Wehausen JD, et al. (2005) Highways block gene flow and cause a rapid decline in genetic diversity of desert bighorn sheep. Ecology Letters, 8, 1029-1038.

Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology, 14, 2611-2620.

Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): An integrated software package for population genetics data analysis. Evolutionary Bioinformatics, 1, 47-50.

Frankham R, Ballou JD, Briscoe DA (2002) Introduction to conservation genetics Cambridge University Press, Cambridge.

French TW, Mount RH (1978) Current status of the Red Hills salamander Phaeognathus hubrichti Highton, and factors affecting its distribution. . Journal of the Alabama Academy of Science 49, 172-179.

Funk WC, Blouin MS, Corn PS, et al. (2005) Population structure of Columbia spotted frogs (Rana luteiventris) is strongly affected by landscape. Molecular Ecology, 483-496.

Funk WC, Tallmon DA, Allendorf FW (1999) Small effective population size in the long-toed salamander. Molecular Ecology, 8, 1633-1640.

Geist J, Kuehn R (2005) Genetic diversity and differentiation of central European freshwater pearl mussel (Margaritifera margaritifera L.) populations: implications for conservation and management. Molecular Ecology, 14, 425-439.

Gilpin ME, Hanski IA (1991) Metapopulation Dynamics: Empirical and Theoretical Investigations Academic Press, London.

Goldstein DB, Linares AR, Cavallisforza LL, Feldman MW (1995) Genetic Absolute Dating Based on Microsatellites and the Origin of Modern Humans. Proceedings of the National Academy of Sciences of the United States of America, 92, 6723-6727.

Guo SW, Thompson EA (1992) Performing the Exact Test of Hardy-Weinberg Proportion for Multiple Alleles. Biometrics, 48, 361-372.

Harding EK, Crone EE, Elderd BD, et al. (2001) The scientific foundations of habitat conservation plans: a quantitative assessment. Conservation Biology, 15, 488-500.

Harrison RG (1991) Molecular-Changes at Speciation. Annual Review of Ecology and Systematics, 22, 281-308.

Howes BJ, Brown JW, Gibbs HL, et al. (2009) Directional gene flow patterns in disjunct populations of the black ratsnake (Pantheropis obsoletus) and the Blanding's turtle (Emydoidea blandingii). Conservation Genetics, 10, 407-417.

J R Jordan J, Mount RH (1975 ) The status of the Red Hills salamander, Phaeognathus hubrichti Highton. Journal of Herpetology, 211-215.

Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics, 23, 1801-1806.

Jensen JL, Bohonak AJ, Kelley ST (2005) Isolation by distance, web service. Bmc Genetics 6:13 v.3.16 http://ibdws.sdsu.edu/

Kleiner K, Mackenzie M, Silvano A, et al. (2007) GAP land cover map of ecological systems for the state of Alabama (Provisional). Alabama Gap Analysis Program Acessed 8-2009 from www.auburn.edu/gap.

Lance SL, Hagen C, Glenn TC, Apodaca JJ, Rissler LJ (2009) Development and characterization of twelve polymorphic microsatellite loci in the threatened Red Hills salamander, Phaeognathus hubrichti. Conservation Genetics, 10, 1919-1921.

Lee-Yaw JA, Davidson A, Mcrae BH, Green DM (2009) Do landscape processes predict phylogeographic patterns in the wood frog? Molecular Ecology, 18, 1863-1874.

Manel S, Schwartz MK, Luikart G, Taberlet P (2003) Landscape genetics: combining landscape ecology and population genetics. Trends in Ecology & Evolution, 18, 189-197.

Mantel N (1967) The detection of disease clustering and a generalized regression approach. Cancer Research, 27, 209-220.

Marshall JCJ, Kingsbury BA, Minchella DJ (2009) Microsatellite variation, population structure, and bottlenecks in the threatened copperbelly water snake. Conservation Genetics,10, 465-476.

Matern A, Desender K, Drees C, et al. (2009) Genetic diversity and population structure of the endangered insect species Carabus variolosus in its western distribution range: Implications for conservation. Conservation Genetics, 10, 391-405.

Mayer C, Schiegg K, Pasinelli G (2009) Patchy population structure in a short-distance migrant: evidence from genetic and demographic data. Molecular Ecology, 18, 2353-2364.

Mcknight ML, Dodd CK, Spolsky CM (1991) Protein and Mitochondrial-DNA Variation in the Salamander Phaeognathus-Hubrichti. Herpetologica, 47, 440-447.

McRae BH (2006) Isolation by resistance. Evolution, 60, 1551-1561.

Morgan MJ, Hunter D, Pietsch R, Osborne W, Keogh JS (2008) Assessment of genetic diversity in the critically endangered Australian corroboree frogs, Pseudophryne corroboree and Pseudophryne pengilleyi, identifies four evolutionarily significant units for conservation. Molecular Ecology, 17, 3448-3463.

Pabijan M, Babik W, Rafinski J (2005) Conservation units in north-eastern populations of the Alpine newt (Triturus alpestris). Conservation Genetics, 6, 307-312.

Paetkau D, Calvert W, Stirling I, Strobeck C (1995) Microsatellite Analysis of Population- Structure in Canadian Polar Bears. Molecular Ecology, 4, 347-354.

Paetkau D, Slade R, Burden M, Estoup A (2004) Genetic assignment methods for the direct, real-time estimation of migration rate: a simulation-based exploration of accuracy and power. Molecular Ecology, 13, 55-65.

Piry S, Luikart G, Cornuet JM (1999) BOTTLENECK: A computer program for detecting recent reductions in the effective population size using allele frequency data. Journal of Heredity, 90, 502-503.

Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics, 155, 945-959.

Proctor MF, McLellan BN, Strobeck C, Barclay RMR (2005) Genetic analysis reveals demographic fragmentation of grizzly bears yielding vulnerably small populations. Proceedings of the Royal Society B-Biological Sciences, 272, 2409-2416.

Reed DH, Teoh V-H, Stratton GE, Hataway RA (2009) Levels of gene flow among populations of wolf spider in a recently fragmented habitat: current versus historical rates. Conservation Genetics, In Press.

Ricketts TH (2001) The matrix matters: effective isolation in fragmented landscapes. American Naturalist, 158, 87-99.

Rowe G, Beebee TJC (2004) Reconciling genetic and demographic estimators of effective population size in the Anuran amphibian Bufo calamita. Conservation Genetics, 5, 287- 298.

Rowe G, Beebee TJC, Burke T (1998) Phylogeography of the natterjack toad Bufo calamita in Britain: genetic differentiation of native and translocated populations. Molecular Ecology, 7, 751-760.

Ryder AO (1986 ) Species conservation and systematics: the dilemma of subspecies Trends in Ecology and Evolution, 1, 9-10.

Schwaner TD, Mount RH (1970 ) Notes on the distribution, habits, and ecology of the salamander Phaeognathus hubrichti Highton. Copeia, 1970, 571-573

Schwartz MK, Luikart G, Waples RS (2007) Genetic monitoring as a promising tool for conservation and management. Trends in Ecology and Evolution, 22, 25-33.

Smouse PE, Long JC, Sokal RR (1986) Multiple regression and correlation extensions of the Mantel test for matrix correspondance. Systematic Zoology, 35, 627-632.

Spear SF, Peterson CR, Matocq MD, Storfer A (2005) Landscape genetics of the blotched tiger salamander (Ambystoma tigrinum melanostictum). Molecular Ecology, 14, 2553-2564.

Storfer A, Murphy MA, Evans JS, et al. (2007) Putting the "landscape" in landscape genetics. Heredity, 98, 128-142.

Straub SC, Doyle JJ (2009) Conservation genetics of Amorpha georgiana (Fabaceae), an endangered legume of the Southeastern United States. Molecular Ecology, 18, 4349- 4365.

Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes, 4, 535-538.

Vandergast AG, Bohonak AJ, Weissman DB, Fisher RN (2007) Understanding the genetic effects of recent habitat fragmentation in the context of evolutionary history: phylogeography and landscape genetics of a southern California endemic Jerusalem cricket (Orthoptera : Stenopelmatidae : Stenopelmatus). Molecular Ecology, 16, 977-992.

Wang IJ (2009) Fine-scale population structure in a desert amphibian: landscape genetics of the black toad (Bufo exsul). Molecular Ecology, 18, 3847-3856.

Wang IJ, Summers K (2010) Genetic structure is correlated with phenotypic divergence rather than geographic isolation in the highly polymorphic strawberry poison-dart frog. Molecular Ecology, 19, 447-458.

Wilson GA, Rannala B (2003) Bayesian inference of recent migration rates using multilocus genotypes. Genetics, 163, 1177-1191.

Wright S (1943) Isolation by Distance. Genetics, 28, 114-138.

Zellmer AJ, Knowles LL (2009) Disentangling the effects of historic vs. contemporary landscape structure on population genetic divergence. Molecular Ecology, 18, 3593-3602.

ACKNOWLEDGEMENTS

We would like to thank J. Godwin, H. Cunningham, W. Smith, and S. Hoss for help with collections, and R. Crawford, J. Mitchell, and C. Tucker for help in the lab. We are also grateful to J. Travis, K. Dodd, M. Gunzburger, C. Thawley, N. Mattheus, R. Downer, and K. Bakkegard for various assistance and comments. This study was funded by a grant from the State of Alabama Division of Wildlife and Freshwater Fisheries.

Table 1. Results of the Analysis of Molecular Variance (AMOVA). P-values were determined using 10 000 permutations. Source of Sum of Variance Percentage Variation d.f. Squares Components of Variation P-value Among < Populations 4 105.717 0.60845 Va 15.58 0.00001 Within < Populations 203 669.149 3.2963 Vb 84.42 0.00001 Total 207 774.865 3.90475 Fixation Index

(FST) 0.15582

Table 2. Pairwise FST values for P. hubrichti populations. * indicates statistically significant to the < 0.05 level. Significance determined using 10 000 permutations and a sequential Bonferroni correction.

Population 1 2 3 4 5 1 0 * * * * 2 0.14231 0 * * * 3 0.17663 0.1313 0 * * 4 0.13516 0.13723 0.1399 0 * 5 0.16755 0.1879 0.1846 0.13009 0

Table 3. Summation of allelic information for each of the five populations of P. hubrichti.

Population HO(mean) HE(mean) FIS 1 0.57976 0.73427 0.210426682 2 0.56026 0.64981 0.137809514 3 0.56695 0.67289 0.157440295 4 0.59962 0.67236 0.108186091 5 0.61053 0.69587 0.122637849 HO, heterozygosity observed; HE, heterozygosity expected; FIS, fixation index.

Table 4. Rates of migration (m) as inferred from a coalescent method (MIGRATE), and from a Bayesian assignment (BayesAss+). All rates are represented as m.

95% Confidence 95% Confidence Populations MIGRATE Interval BayesAss+ Interval 2 -> 1 0.02835 0.0108-0.0378 0.0687679 0.000004-0.2963 3 -> 1 0.03105 0.0108-0.0459 0.0124238 0.000003-0.0698 4 -> 1 0.02835 0.0081-0.0405 0.0122148 0.000003-0.0641 5 -> 1 0.01215 0-0.0189 0.0108126 0.000004-0.0632 1 -> 2 0.01485 0-0.0243 0.00639563 0.000001-0.0377 3 -> 2 0.05265 0.0297-0.0756 0.00756993 0.000001-0.0450 4 -> 2 0.02565 0.0054-0.0350 0.00618354 0.000001-0.0358 5 -> 2 0.02565 0.0081-0.0351 0.00596068 0.000001-0.0376 1 -> 3 0.01215 0-0.0189 0.00291852 0.000004-0.0150 2 -> 3 0.00945 0-0.0162 0.00594981 0.000051-0.0231 4 -> 3 0.01755 0-0.0243 0.00288727 0.000002-0.0149 5 -> 3 0.01215 0-0.0189 0.00255547 0.000001-0.0147 1 -> 4 0.02565 0.0054-0.0351 0.0053779 0.000001-0.0298 2 -> 4 0.02025 0.0027-0.0270 0.00622125 0.000001-0.0339 3 -> 4 0.01755 0-0.0270 0.00981886 0.000003-0.0475 5 -> 4 0.03375 0.0162-0.0432 0.00547953 0.000001-0.0305 1 -> 5 0.01485 0-0.0243 0.00567046 0.000001-0.0324 2 -> 5 0.01755 0-0.0270 0.00549078 0.000002-0.0293 3 -> 5 0.02025 0-0.0297 0.013468 0.000105-0.0489 4 -> 5 0.02025 0.0027-0.0297 0.0061104 0.000003-0.0350

Table 5. Estimates of Θ and Ne for all populations of P. hubrichti. Ne was estimated from Θ using a well-accepted vertebrate mutation rate of 5.4 X 10-4 (Goldstein et al. 1995; Howes et al. 2009). 95% Confidence

Population Θ (4Neµ) Interval Ne 1 0.09725 0.0885 - 0.1 45.02314815 2 0.09775 0.0905 - 0.1 45.25462963 3 0.09875 0.0950 - 0.1 45.71759259 4 0.09825 0.0930 - 0.1 45.48611111 5 0.09825 0.0925 - 0.1 45.48611111

Table 6. Results of bottleneck analysis for all populations of P. hubrichti. All values are presented as P-values. Bold values indicate statistically significant.

Prob of H Prob of H Population Model % of SSM def excess 1 IAM 0 0.98779 0.01611 2 IAM 0 0.57715 0.46094 3 IAM 0 0.98779 0.01611 4 IAM 0 0.99756 0.00342 5 IAM 0 0.90332 0.11621 1 TPM 10 0.95801 0.05273 2 TPM 10 0.27832 0.75391 3 TPM 10 0.88379 0.1377 4 TPM 10 0.99658 0.00488 5 TPM 10 0.78418 0.24609 1 TPM 20 0.95801 0.05273 2 TPM 20 0.27832 0.75391 3 TPM 20 0.83887 0.1875 4 TPM 20 0.99658 0.00488 5 TPM 20 0.78418 0.24609 1 TPM 30 0.94727 0.06543 2 TPM 30 0.27832 0.75391 3 TPM 30 0.78418 0.24609 4 TPM 30 0.99512 0.00684 5 TPM 30 0.75391 0.27832 1 TPM 40 0.94727 0.06543 2 TPM 40 0.21582 0.8125 3 TPM 40 0.8623 0.16113 4 TPM 40 0.98779 0.01611 5 TPM 40 0.72168 0.3125

Table 7. Results of Mantel tests for correlations between geographic/model distance and genetic distance. * indicates statistically significant to the level of < 0.05 using 10 000 permutations.

2 Model Included Variables r Euclidian Distance None 0.306 Slope Distance None 0.287 Slope Conductance Slope 0.321* Slope, Geologic Slope & Geology formations 0.281

Figure 1. Range map of P. hubrichti with an outline of geologic layers required by the species. Circles are known populations; stars represent tissue collection localities.

Figure 2. Populations of P. hubrichti as determined by Bayesian clustering techniques. Populations are numbered from left to right and represented by unique symbols (1, circles; 2 triangles; 3, stars; 4, squares; 5, diamonds).

Figure 3. Slope analysis and populations of P. hubrichti. Areas of steepest slope are represented by darker colors.

CHAPTER FIVE: RECOMMENDATIONS FOR THE RECOVERY OF THE THREATENED

RED HILLS SALAMANDER (PHAEOGNATHUS HUBRICHTI): INCLUDING HABITAT-

PURCHASING GUIDELINES, A PROPOSAL FOR HABITAT RECOVERY, AND

IDENTIFYING LOCATIONS OF UNKNOWN POPULATIONS.

Abstract The Red Hills salamander (Phaeognathus hubrichti) is a federally threatened member of the family Plethodontidae that is endemic to a narrow band of geologic formations in southern

Alabama. Despite being federally listed for well over three decades the species continues to decline. Past conservation efforts have focused on the use of habitat conservation plans with large landowners. However, this method is focused on mitigating further habitat losses rather than promoting the recovery of the species. The goal of this paper is to devise effective measures that will lead to the stabilization and eventual recovery of the Red Hills salamander. We base all of our recommendations on data pertaining to the species’ habitat requirements, climatic suitability, available habitat, habitat fragmentation, gene flow patterns, genetic structure and genetic variability. We recommend the acquisition and proper management of twenty-one new conservation areas for the species. Additionally, we urge that large landowners with populations of P. hubrichti enter into safe harbor agreements, a process for which we provide guidelines here.

INTRODUCTION

The year 2010 marks the 50-year anniversary of the discovery of the Red Hills salamander (Phaeognathus hubrichti) by Leslie Hubricht. Since the description of P. hubrichti

(Highton 1961) there has been a great deal of research on this species. Yet, there still remains a large gap in basic knowledge about the salamander due to its secretive and fossorial nature. This

lack of data has hampered conservation efforts for P. hubrichti despite it being federally listed as a threatened species since 1976. The original federal listing came in response to the concerns of local herpetologists based on the impact that forestry practices were having on the few known populations at the time (Schwaner 1970: Jordan and Mount 1975; Mount 1975). Since the listing it has become abundantly clear that timber practices have a strong negative effect on the persistence of P. hubrichti (Dodd 1991; Godwin 2008). Unfortunately, nearly the entirety of the range of P. hubrichti is owned by large timber conglomerates or managed for timber production

(Bailey and Means 2004).

The federal listing of P. hubrichti provided some degree of protection for the salamander through the employment of habitat conservation plans (HCPs) with large landholders. However,

HCPs do not offer a solid long-term option for the conservation of a species, as oftentimes they are based on limited data (Harding et al. 2001). Additionally, HCPs do not transfer with landownership when a piece of property changes possession (Bonnie 1999; Harding et al. 2001;

Godwin 2008), thereby exposing imperiled species to the threats of landscape alteration and fragmentation immediately following a land transaction.

The HCPs for P. hubricthi are particularly inadequate as the majority of them focus solely on preserving small patches of high-quality habitat, while ignoring how this habitat is impacted by forestry practices on surrounding habitat or the connectivity between patches (Table

1). Populations within these highly fragmented patches are extremely vulnerable to the loss of genetic variability and extinction due to inbreeding and the susceptibility to demographic, environmental, and genetic stochasticity (Conner and Hartl 2004; Frankham et al. 2007).

Currently it is believed that there are eight HCPs in place with large timber companies (J.

Smithem pers. comm.). These HCPs represent roughly 37% of the entire range of P. hubrichti.

However, there have been a large number of timber transactions within the last 5 years in the

Red Hills, possibly nullifying the benefits of these HCPs (E. Soehren pers. comm.).

For P. hubrichti current management via HCPs will only lead to increasingly fragmented populations with a greater degree of edge effects in remaining habitat. Edge effects can be detrimental to amphibians as edges are typically warmer, have a lower relative humidity, are more susceptible to frequent and longer frosts, have increased soil temperature, decreased soil moisture, and decreased amounts of leaf litter (Klein 1989; Parker 1989). A survey twelve years after the initial listing of P. hubrichti found that edge effects had a major impact on the distribution of P. hubrichti burrows on inhabited slopes and that there had been a large number of heavily impacted sites despite HCPs (Dodd 1991). This trend has continued as numerous sites surveyed for this study were highly degraded due to forestry practices. Undoubtedly, we must seek out more effective conservation measures if we wish to ensure the long-term persistence of

P. hubrichti

The acquisition of protected areas for conservation purposes is one of the most effective methods for assuring the persistence of a species (Soulé 1991; Balmford et al. 1996; Redford and

Richter 1999; Rodrigues and Gaston 2002). However, identifying proper land tracts for attainment is not a straightforward undertaking. There are several factors that should be considered when pursuing tracts of land for the conservation of a species, including but not limited to: habitat feasibility, climatic suitability, population connectivity, population viability, population size, acquisition cost, genetic distinctiveness, and genetic diversity. For P. hubrichti the attainment of such data is complicated by the fact that they spend the majority of their time underground within their burrow (Bakkegard 2002). Therefore, it is necessary to depend on indirect methods, such as genetic techniques and GIS approaches, to estimate these data.

In this study we aim to apply several sources of data in order to make both broad and specific recommendations for the purchase and management of protected areas for P. hubrichti and for the improvement of privately managed lands. We base these recommendations on habitat requirements, climatic suitability, available habitat, habitat fragmentation, gene flow patterns, genetic structure and genetic variability. We hope that these data can be applied to a large-scale habitat purchasing guideline for the recovery of P. hubrichti. Additionally, we use a GIS habitat modeling approach in order to identify areas that may harbor unknown populations of P. hubrichti.

METHODS

Species data

Phaeognathus hubrichti is a monotypic species that diverged from its nearest living relative over 35 million years ago (Highton 1961; Vieites et al. 2007; Kozak et al. 2009). It is restricted to a narrow geologic band that runs across Alabama known as the Red Hills (Fig. 1).

Within the Red Hills it is thought to be restricted to the Tallahatta and Hatchetigbee formations

(Dodd 1991), though a more recent discovery (Bailey and Miller 2006) indicates they may also be found in the Nanafalia (Fig. 2). These formations are Eocene in age and consist of claystone, siltstone and sandstone (Scott 1972). It is likely that these geologic layers provide a substrate that allows P. hubrichti burrows to be easily created and persist for months (Jordan 1975;

Gunzburger and Guyer 1998). Phaeognathus hubrichti has several habitat requirements in addition to their geologic restrictions.

Phaeognathus hubrichti are generally found on mature mesic slopes in the ravines of the

Red Hills (Jordan 1975). They are most common on steep (> 30°) north-facing slopes, with a full canopy of mature hardwood trees (Dodd 1991). However, individual burrows are not uncommon

in small isolated patches of microclimate, such as isolated ledges or at the base of large trees

(Dodd 1991; Pers. Obs.). A full hardwood canopy allows a greater degree of moisture retention and a vast array of invertebrate prey (Dodd 1991). These hardwood ravines were historically surrounded by mature longleaf (Pinus palustris)(or possibly a longleaf shortleaf (Pinus echinata) mixture) pine forests (Harper 1920; Mohr 1901). Though, longleaf may not have been completely dominant across the entire range (Mohr 1901). The clear cutting of these ridge tops followed by their replacement with pine plantations has had detrimental effects on P. hubrichti.

Timber harvest on the ridge tops increases the edge effects on the mesic slopes. This type of disturbance causes P. hubrichti to no longer use burrows on the top third of the slope, thereby effectively shrinking the amount of available habitat (Dodd 1991). Furthermore, it is likely that shifting the ridge tops from a mature open-canopy system (as in longleaf systems) to a densely packed canopy of growing pines drastically alters the hydrology of the system (Godwin 2008).

This can occur through two mechanisms. The first mechanism being that a dense canopy causes rainfall interception, and thus a loss of water through direct evaporation (Godwin 2008). The second mechanism is that vigorously growing plantation trees will remove a great deal more water from the soil than a mature forest (Vertessy 2001). This problem is exacerbated by the fact that a natural longleaf system would consist of widely spaced trees and an intervening grassland habitat. Godwin (2008) hypothesized that this reduction of water on the ridge tops leads to a reduction in moisture levels in the underlying Tallahatta, Hatchetigbee, and Nanafalia formations, and may lead to physiological stress, impaired reproduction, unsuccessful egg development, increased mortality, and a decrease in prey availability.

In addition to a restricted habitat, P. hubrichti also has a highly specialized life history.

They have evolved many adaptations to life within a burrow, including: elongated body,

numerous vertebrae, solidly constructed skull, small nostrils, modified eyelids, small limbs, absence of lateral line organs, and direct development within the burrow (Highton 1961; Jordan

1975; Means 2003). A specialized life history is one of the strongest predictors of extinction risk

(Hunter and Gibbs 2007).

Habitat modeling

In order to identify suitable areas for P. hubrichti we divided the range of P. hubrichti and the surrounding area into grid cells of 0.005°. The total size of the model was 215 km x 170 km, allowing for the inclusion of the known six counties that contain P. hubrichti and a 20 km buffer. Each cell was then assigned a value of suitability based on the addition of three variables:

1) slope, 2) proper habitat type, and 3) proper geologic layer. We then eliminated cells that were outside of the climatic envelope created using environmental niche modeling.

Slope was calculated using the national elevational dataset (NED) available from the

USGS. NED is available at 1/9 arc second (about 3 meters) resolution. Slope was calculated using the ARCGIS v9.3 spatial analyst extension. This slope calculation finds the maximum rate of change between each cell and its neighbors. Therefore, the maximum change in elevation between neighboring cells receives the highest value, and the lower the slope value the flatter the terrain. Since P. hubrichti are most commonly found on the steep slopes, this technique allows us to identify suitable tracts of slope. For the habitat suitability analysis, each cell was grouped into one of six groups using the Jenks optimization method (Jenks 1967). This method is also known as the goodness of variance fit, and is analogous to a one-way analysis of variance. Essentially the Jenks method seeks to maximize variance between natural breaks in the data by minimizing the squared deviations of the class means. Each category was then assigned a value, with the lowest slope values receiving a low number and vice versa.

We used the AL-GAP dataset (Kleiner et al. 2007) to identify habitat suitable for P. hubrichti. By using recently collected P. hubrichti points and life history knowledge of the species we determined that category 51 (East Gulf Coastal Plain Southern Mesic Slope Forest) was the only category that properly represented P. hubrichti habitat. This category was assigned a value of 1 and all others a 0. Though it is likely that there are small patches of habitat not identified by this method they are probably too small or infrequent to harbor a viable population.

For the geologic layers we used the Alabama geologic map made available by the USGS. From this data set we selected the three geologic layers required by P. hubrichti (Tallahatta,

Hatchetigbee, and Nanafalia).

We used ecological niche modeling (ENM) in order to identify areas that are climatically similar to known P. hubrichti populations. We chose to use Maxent (Phillips et al. 2006) because it has been shown to be one of the most reliable ENM methods (Elith et al. 2006; Pearson et al.

2007). In general, Maxent uses a machine learning approach (maximum entropy) to predict the probability of occurrence of a species given an equal effort of sampling across the locality data.

Maximum entropy modeling is part of a family of statistical approaches (machine learning) that typically outperforms traditional statistical approaches (e.g. generalized linear models) in complex ecological situations (Olden et al. 2008). We used a least point threshold (LPT) in order to determine the Maxent value that would serve as a cutoff point for what we would consider as suitable climatic conditions. Although there are several threshold methods available for niche modeling (see Pearson et al. 2007), the LPT is one of the most conservative and has the biological reality that the model is identifying area at least as suitable as areas where the species has been reported. All cells with values above the threshold were deemed as climatically suitable, and all of those below were categorized as unsuitable.

100

To evaluate the accuracy of our distributional model we tested for a statistically significant difference between the values of the cells that contain known P. hubrichti localities and an equal number of randomly selected cells using an unpaired two-tailed t-test. We used all known locality data from the Alabama Natural Heritage Program (ALNHP), which included 130 unique localities. We then generated two sets of randomly selected cells, one set was randomly selected from the six counties known to contain P. hubrichti and another set that was selected just from the area within the six known counties that contained the proper geologic layers. This process allowed us to evaluate if our modeling technique identified P. hubrichti habitat significantly better than random.

Genetic data

We incorporated the genetic data of Apodaca and Rissler (2010) into our recommendations for habitat purchasing. These data indicate that there are 5 major populations throughout the range of P. hubrichti (Fig. 3). Additionally, these populations have very little gene flow between each other and all display a high amount of inbreeding. Therefore, any conservation efforts for the species should consider each of these populations as essentially an independent management unit.

RESULTS

Our model included a total of 173,056 cells (Fig. 4). The highest two categories, representing the most suitable P. hubricthti habitat, contained 196 (0.11%) and 745 (0.43%) cells, respectively. The next two highest categories, which still represent high-quality habitat, contained 1925 (1.11%) and 3690 (2.13%) cells. The next to last category, which contains habitat that is still viable but not ideal, contained 10856 cells (6.27%). The lowest category, representing unsuitable habitat, contained the vast majority of cells (155,644, 89.9%). The ENM

101

(Appendix 1) indicated that all of the areas within our habitat model were climatically similar and therefore we do not feel that climate limits the distribution of P. hubrichti.

Statistical analysis indicated that our model performed significantly better than random when identifying suitable habitat for P. hubrichti. Known localities of P. hubrichti were found in cells that had significantly higher model values than locations drawn from random from the six counties that contain P. hubrichti (P < 0.0001). Additionally, known localities of P. hubrichti were found in cells that also had significantly higher model values than locations drawn from random from within the three geologic layers known to contain P. hubrichti (P = 0.0026).

We have identified 14 sites that we feel have a good possibility of containing unknown populations of P. hubrichti (Fig 5). We have based these predictions on the combination of our habitat model and our ENM. ENMs have been successfully employed for the discovery of unknown populations or closely related species in other settings (Raxworthy et al. 2003; Pearson et. al. 2007). We placed each predicted site in one of three categories based on how likely we felt they were to contain unknown populations (Table 2). The categories were based on the following features: 1.) Amount of available habitat, 2.) Proximity to known populations, and 3.) whether the area is separated from other populations by a major barrier (i.e. Alabama or Conecuh Rivers, etc.).

DISCUSSION

It has been well over thirty years since P. hubrichti was listed as a federally threatened species. Yet, there are no signs that there has been any significant recovery. The species was originally listed under concerns that habitat destruction in conjunction with a limited distribution, unique life history, and restricted habitat requirements could lead to the species becoming critically endangered (U.S. Fish and Wildlife Service 1976). Yet, significant habitat destruction

102

continues to occur across the range of the species. Since the listing the main tool for P. hubrichti conservation efforts has been the employment of HCPs with large landowners. In theory this practice would eliminate a large amount of habitat destruction. However, the uncertainty associated with HCPs make them less than ideal for the long-term conservation of P. hubrichti

(Godwin 2008). Furthermore, HCPs are not required to insure that they contribute to the recovery of a listed species, but rather to minimize and mitigate habitat losses (Bonnie 1999).

Therefore, it is imperative that recovery efforts are centered on a number of key land acquisitions and supplemented with an incentive-based recovery program such as safe harbor agreements or conservation banking (see Wilcove and Lee 2004).

The recovery of P. hubrichti has recently received a major enhancement. In 2009 the

Alabama Forever Wild Land Trust acquired 1 048 ha in four continuous parcels on the western edge of P. hubrichti’s range. This gives a total amount of 1 117 ha of protected land when combined with the Haines Island park operated by the Army Corps of Engineers. While this acquisition is a major accomplishment, ultimately several more tracts throughout the range of P. hubrichti must be secured to ensure their long-term survival.

Recommendations for habitat acquisition

The acquisition of viable P. hubrichti habitat followed by proper habitat management is by far the most effective method for ensuring their long-term survival. However, the effectiveness of this approach must be based on reliable data in order to maximize the conservation impact of limited funds. Dodd (1988) recommended 23 sites for this purpose based on his assessment of burrow abundance and status of the habitat (Fig. 6). He also made an attempt to select sites from across the range of the species in order to preserve genetic diversity.

However, the knowledge of the species’ genetic diversity at that time was limited to a small

103

amount of variation found in allozymes (McKnight et al. 1991). This data indicated that there were two major populations, separated by the Sepulga River (McKnight et al. 1991). Recent data using microsatellites (Apodaca and Rissler 2010) indicate that in fact there are 5 populations that should be considered as unique management units (Fig. 3). Based on these findings as well as the recent land purchases by state of Alabama Forever Wild Land Trust we have reassessed Dodd’s

(1988) recommendations and updated them to reflect the current state of knowledge of the species. We will frame and discuss recommendations based on preserving the genetic diversity of the five distinct populations (Fig. 3) and with the goal of increasing gene flow between them.

Additionally, since specific land tracts may not become available and unique land purchasing opportunities may arise we will tend to focus on general areas rather than specific tracts.

Population 1 is the smallest population in terms of geographic distribution and has approximately 69 ha in permanent habitat protection from the Haines Island Park. Therefore, we suggest a maximum of two more habitat acquisitions for this population. Dodd (1988) did not suggest any additional sites for this area, yet our model suggests that proper habitat conditions are readily available in the area (Fig. 4). Thus, we suggest one site to the southwest of Haines

Island (site 1) and one site to the northeast of Haines Island (site 2). Of these two sites, site 2 should be a higher priority, as it would help to facilitate gene flow between populations 1 and 2.

Population 2 includes a large amount of good habitat for P. hubrichti (Fig. 4). This area includes four of Dodd’s (1988) recommended sites (Dodd sites: 4, 125, 39, and 41). Three of these sites

(125, 39, and 41) are located very near the recently purchased Forever Wild Land tract. For that reason, we are removing those three sites from our recommendations. We suggest two sites (3 &

4) to the south of the Forever Wild Land Tract. Site 3 is placed in an effort to enhance gene flow between populations 2 and 3, and if possible should be purchased in close proximity to Big Flat

104

Creek (likely the major hindrance to migration between these two populations). Site 4 is Dodd’s

(1988) original site number 4. Population 2 also includes the recently discovered population in

Wilcox County (Bailey and Miller 2004). This area is in desperate need of surveys to discover the true nature of populations in this area. For the time being we suggest one site for Wilcox

County (site 5). It is highly possible that many new populations will be discovered in this area, thus increasing the need for habitat protection.

Population 3 consists of a long narrow band of P. hubrichti localities, often dispersed among fragmented habitat (Fig.4). This area includes 7 of Dodd’s (1988) recommended sites (44,

48, 96, 103, 104, 23, and 137). We agree with these recommendations and add an additional site

(6) that may help connect eastern and western localities within the population. These recommended sites are crucial to preserving the genetic diversity of P. hubrichti due to the fact that this area of the species’ range is composed of a very narrow band of habitat. Therefore, it would not take a significant amount of habitat loss in this area to completely sever the already minimal amount of gene flow between eastern (4 &5) and western populations (1 & 2) (see

Apodaca and Rissler 2010).

Population 4 is similar to population 3 in that it is very narrow. However, our model suggests that there may be additional unknown localities along the Sepulga River (Fig. 4). This area contains six of Dodd’s (1988) recommended sites (29, 138, 49, 52, 59, and 142). We do not recommend any additional sites for this area, though we do note that sites 52, 59, and 142 would most likely be included in a single land purchase. Therefore, the actual number of recommended sites for this population is four.

Population 5 is geographically large compared to the other populations. Suitable habitat is not as dense as in other areas of P. hubrichti’s range (Fig. 4). In fact, Mohr (1901) noted that

105

the hills in this region become less prominent, thereby reducing the amount of slope available as habitat. This area includes seven of Dodd’s (1988) original recommendations (60, 63, 90, 91,

121, 78, and 84), though sites 90 and 91 could be considered one site. We recommend one additional site (7) for this area. We note that our model suggests that there is a great deal of suitable habitat in the area that may harbor unknown populations (Fig. 5).

Recommendations concerning surveys for unknown populations

In order to fully protect P. hubrichti we must have a strong understanding of their geographic distribution. Unknown populations may provide important sources of genetic variation or be fundamental to the connectivity between other populations. Thus, it is important that we identify and protect such populations. We have identified 14 sites that we feel have a strong possibility of containing unknown populations of P. hubrichti. We suggest that at the very minimum the sites we have ranked as highly probable to contain populations (Table 2) are surveyed as soon as possible.

Recommendations for habitat improvement

While the acquisition of viable P. hubrichti habitat is the most effective method for stabilizing the species it may not lead to a return to adequate levels of gene flow between populations. Without adequate levels of gene flow populations are extremely vulnerable to the loss of genetic variability and extinction due to inbreeding and the susceptibility to demographic, environmental, and genetic stochasticity (Conner and Hartl 2004; Frankham et al. 2007). In fact, each of the five P. hubrichti populations were found to have an extremely low effective population size (> 46), indicating that the populations are even more susceptible to the perils of small population size than previously believed (Apodaca and Rissler 2010). Genetic data has revealed that recent levels of gene flow have been greatly reduced when compared to pre-timber

106

harvest conditions (Apodaca and Rissler 2010). It is likely that these reduction are a result of the loss of viable habitat between populations caused by timber practices such as the clear cutting of intervening habitat followed by the replanting of pine plantations. A full recovery of the species will depend not only on preserving currently recognized habitat but also restoring historical conditions. We recommend the initiation of a P. hubrichti safe harbor program. The goal of this program would be to convince major landowners to enter into safe harbor agreements with both state and federal agencies.

Safe harbor programs are most well known for their involvement in the recovery of the endangered Red-cockaded Woodpecker (Picoides borealis). There are over 35 safe harbor agreements in 16 states protecting a wide variety of species, including 3 mussel species within the state of Alabama (www.edf.org). Essentially a safe harbor agreement is a binding agreement where the landowners agree to protect a pre-established baseline population (similar to an HCP) and to enhance additional habitat for the species. For example, in the case of the Red-cockaded

Woodpecker several large landowners throughout the southeast have used prescribed fire, mid- story hardwood removal, drilling of artificial cavities, and other means to create additional habitat for the species (Bonnie 1997). In return for their efforts, landowners are not responsible for additional individuals not counted under the baseline population, thereby removing uncertainty in forestry management (Zhang and Mehmood 2002). One of the main goals of this program is to remove the fear of regulatory consequences that prevents some landowners from participating in habitat restoration (Wilcove and Lee 2004). The major advantage of this program for P. hubrichti is that it would protect currently known populations (similar to an HCP) and additionally it would add habitat and reduce population fragmentation.

107

Traditionally safe harbor programs do not include a financial reward to the owner.

Though landowners have often received financial aid from federal, state, or private sources to help offset some of the costs associated with habitat restoration (Wilcove and Lee 2004). We suggest that safe harbor agreements for P. hubricthi would not be cost-intensive, as most of the modifications could occur after a harvest and would not require land modification.

We recommend that safe harbor agreements for P. hubrichti implement the following recommendations:

- Any area that is currently managed for non-historical plant assemblages (i.e. loblolly

or evergreen pine plantations, etc), has a slope greater than 35°, is within 1 km of

known P. hubrichti habitat, and is found within the proper geologic formations

(Tallahatta, Hatchitigbee, or Nanafalia) should be restored to a native assemblage of

hardwood species (see Mohr 1901; Harper 1943; Jordan 1975; Diamond 1987;

Godwin 2008) after the next scheduled harvest.

- Mechanical preparation of restored sites should be avoided.

- A buffer should be placed at the top of any slope containing P. hubrichti, preferably

greater than 5 meters and comprised of a natural system (i.e. Long-leaf Pine).

- Chemical preparation of any kind should be avoided near any slope known to contain

P. hubrichti or on any restored site.

- Timber harvest should be completely avoided on any slope containing P. hubrichti

that has a slope greater than 18°.

- Harvest on any slope (< 18°) known to contain P. hubrichti should be limited to 35%

of the bottom third of the slope. Plantation style pine assemblages should not be

replanted after this harvest.

108

- We also suggest all of the recommendations of Dodd (1991) be followed.

CONCLUSIONS

We have outlined a basic plan that would lead to a substantial recovery of P. hubrichti.

This plan includes the purchase a total of 21 sites (Fig. 7) dedicated to preserving P. hubricthi and the other biodiversity of the region. We have based this plan on P. hubrichti habitat requirements, climatic suitability, available habitat, habitat fragmentation, gene flow patterns, genetic structure and genetic variability. We strongly believe that if followed this combination of habitat acquisitions and a series of safe harbor agreements will allow for the future recovery and possibly delisting of P. hubrichti.

The Red Hills of Alabama are a unique and biologically diverse region of the state. This physiographic province supports at least 24 vertebrates of conservation concern (Table 3) and a unique assemblage of flora (Diamond 1987), including a recently discovered azalea species that is endemic to the Red Hills (Zhou et al. 2008). Given the unique nature of the slope habitat occupied by P. hubrichti it is likely that there are other unidentified distinctive taxa endemic to the region. The conservation efforts put forth to conserve P. hubrichti will benefit this vast biodiversity, especially in the case of acquired and properly managed land where both ridge-tops and ravines can be returned to a natural state.

LITERATURE CITED

Apodaca JJ, Rissler LJ (In Review ) Estimating the effects of habitat modification on genetic patterns and population connectivity; a case study using the federally threatened Red Hills salamander (Phaeognathus hubrichti) Molecular Ecology.

109

Bailey MA, Means DB (2004 ) Red Hills Salamander Phaeognathus hubrichti Highton. . In: Alabama Wildlife.Volume 3. Imperiled amphibians, reptiles, birds, and mammals. (eds. Mirarchi RE, Bailey MA, Haggerty TM, Best TL), pp. 34-36. The University of Alabama Press Tuscaloosa, AL.

Bailey MA, Miller DA (2006) Phaeognathus hubricthti. Herpetological Review 37, 357. Bakkegard KA (2002) Activity patterns of Red Hills salamanders (Phaeognathus hubrichti) at their burrow entrances. Copeia, 851-856.

Balmford A, Green MJB, Murray MG (1996) Using higher-taxon richness as a surrogate for species richness .1. Regional tests. Proceedings of the Royal Society of London Series B- Biological Sciences 263, 1267-1274.

Bonnie R (1997) Safe harbor for the red-cockaded woodpecker. Journal of Forestry 95, 17-22.

Bonnie R (1999) Endangered species mitigation banking: promoting recovery through habitat conservation planning under the Endangered Species Act. Science of the Total Environment 240, 11-19.

Conner JK, Hartl DL (2004) A primer of ecological genetics Sinauer Associates Inc. , Sunderland, MA.

Diamond A ( 1987) A flora of the mesic ravines of the central Red Hills of Alabama Auburn University, Auburn Alabama.

Dodd CK (1988) A re-examination of the status of the Red Hills salamander, Phaeognathus hubrichti, Alabama, USA, 1976-1988. Report to the U.S. Fish and Wildlife Service, Jackson, M.S., 74 pp. .

Dodd CK (1991) The Status of the Red Hills Salamander Phaeognathus-Hubrichti, Alabama, USA, 1976-1988. Biological Conservation 55, 57-75.

Elith J, Graham CH, Anderson RP, et al. (2006) Novel methods improve prediction of species' distributions from occurrence data. Ecography 29, 129-151.

Frankham R, Ballou JD, Briscoe DA (2007) Introduction to conservation genetics Cambridge University Press, Cambridge, UK.

Godwin J (2008 ) Red Hills salamander habitat delineation, breeding bird surveys, and habitat restoration recommendations on commercial timberlands Unpublished report prepared for The Alabama Department of Conservation and Natural Resources.

Gunzburger MS, Guyer C (1998) Longevity and abandonment of burrows used by the Red Hills salamander (Phaeognathus hubrichti). Journal of Herpetology 32, 620-623.

110

Harding EK, Crone EE, Elderd BD, et al. (2001) The scientific foundations of habitat conservation plans: a quantitative assessment. Conservation Biology 15, 488-500.

Harper RM (1920) Resources of southern Alabama: a statistical guide for investors and settlers, with an exposition of some of the general principles of economic geography. Geological Survey of Alabama, Special Report 11, 152 pgs.

Highton R (1961 ) A new genus of Lungless salamander from the coastal plain of Alabama Copeia 1, 65-68.

Hunter MLJ, Gibbs J (2007 ) Fundamentals of Conservation Biology 3rd edn. Blackwell Publishing, Malden, MA.

Jordan JR, Mount RH (1975 ) The status of the Red Hills salamander, Phaeognathus hubrichti Highton. Journal of Herpetology, 211-215.

Jenks GF (1967) The data model concept in statistical mapping. International Yearbook of Cartography 7, 186-190.

Jordan JR (1975) Observations on the natural history and ecology of the Red Hills salamander Phaeoganathus hubricthi Highton (Caudata: Plethodontidae). M.S. Thesis, Auburn University, Auburn, AL. 59 pp.

Klein BC (1989) Effects of Forest Fragmentation on Dung and Carrion Beetle Communities in Central Amazonia. Ecology 70, 1715-1725.

Kleiner K, Mackenzie M, Silvano A, et al. (2007) GAP land cover map of ecological systems for the state of Alabama (Provisional). Alabama Gap Analysis Program Acessed 8-2009 from www.auburn.edu/gap.

Mcknight ML, Dodd CK, Spolsky CM (1991) Protein and Mitochondrial-DNA Variation in the Salamander Phaeognathus-Hubrichti. Herpetologica 47, 440-447.

Means DB (2003) Notes on the reproductive biology of the Alabama Red Hills salamander (Phaeognathus hubrichti). Contemporary Herpetology 2003, 1-5.

Mohr C (1901) Plant life of Alabama Contribution from the U.S. National Herbarium 4, 921 pgs.

Mount RH (1975) The Reptiles & Amphibians of Alabama Alabama Agricultural Experiment Station, Auburn University Auburn, AL.

111

Olden JD, Lawler JJ, Poff NL (2008) Machine learning methods without tears: A primer for ecologists. Quarterly Review of Biology 83, 171-193.

Parker CA (1989) Soil biota and plants in the rehabilitation of degraded agricultural soils. In: Animals in primary succession. The role of fauna in reclaimed lands (ed. Majer J). Cambridge University Press, Cambridge, UK.

Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190, 231-259.

Raxworthy CJ, Martinez-Meyer E, Horning N, et al. (2003) Predicting distributions of known and unknown reptile species in Madagascar. Nature 426, 837-841.

Redford KH, Richter BD (1999) Conservation of biodiversity in a world of use. Conservation Biology 13, 1246-1256.

Rodrigues ASL, Gaston KJ (2002) Maximising phylogenetic diversity in the selection of networks of conservation areas. Biological Conservation 105, 103-111.

Schwaner TD, Mount RH (1970 ) Notes on the distribution, habits, and ecology of the salamander Phaeognathus hubrichti Highton. Copeia 1970, 571-573.

Scott JC (1972 ) The geology of Monroe County, Alabama. Alabama Geological Survey Map 101.

US Fish and Wildlife Service (1976) Determination that the Red Hills salamander is a threatened species. . federal Registry 41, 53032-53034.

Soule ME (1991) Conservation - Tactics for a Constant Crisis. Science 253, 744-750. Vertessy RA (2001) Impacts of plantation forestry on catchment runoff. In: Plantations, Farm Forestry. and Water: Workshop Proceedings Publication No. 01/20 (eds. Nambiar EKS, Brown AG). Rural Industries Research and Development Corporation.

Vieites DR, Min MS, Wake DB (2007) Rapid diversification and dispersal during periods of global warming by plethodontid salamanders. Proc Natl Acad Sci U S A 104, 19903- 19907.

Wilcove DS, Lee J (2004) Using economic and regulatory incentives to restore endangered species: Lessons learned from three new programs. Conservation Biology 18, 639-645.

112

Zhang DW, Mehmood SR (2002) Safe Harbor for the red-cockaded woodpecker: Private forest landowners share their views. Journal of Forestry 100, 24-29.

Zhou W, Gibbons T, Goetsch L, Hall B (2008) Rhododendron colemanii: A New Species of Deciduous Azalea (Rhododendron section Pentanthera; Ericaceae) from the Coastal Plain of Alabama and Georgia. Journal of the American Rhododendron Society 75, 72-78.

Table 1. An example of the typical habitat management guidelines for a habitat conservation plan designed to minimize and mitigate habitat losses for the Red Hills salamander (Phaeognathus hubrichti).

113

Table 2. Estimated probabilities of the potential for identified sites to harbor unknown populations of the Red Hills salamander (Phaeognathus hubrichti). Estimates are based on: 1.) Amount of available habitat, 2.) Proximity to known populations, and 3.) is the area separated from other populations by a major barrier (i.e. Alabama or Conecuh Rivers, etc.).

114

Site Probability Of Number Occurrence

1 High 2 High 3 High

4 High 5 High 6 High

7 High 8 High 9 High

10 Medium

11 Medium 12 Medium

13 Low

14 Low

Table 3. Terrestrial vertebrates of the Red Hills physiographic province of high conservation concern. P1=endangered; P2=threatened.

115

Figure 1. Geographic distribution of the Red Hills salamander (Phaeognathus hubrichti) as it is currently known.

116

Figure 2. Outline of the three geologic layers known to contain populations of the Red Hills salamander (Phaeognathus hubrichti).

117

Figure 3. Known localities of the Red Hills salamander grouped into their respective populations as determined by genetic data. Populations are from left to right: population 1- black crosses, Population 2- black circles, Population 3- grey squares, Population 4- dark-grey diamonds, Population 5- black octagons. 118

Figure 4. Distribution model for the Red Hills salamander (Phaeognathus hubrichti).

119

Figure 5. Areas suggested for surveys to determine the presence of unknown Red Hills salamander (Phaeognathus hubrichti) populations.

120

Figure 6. Red Hills salamander (Phaeognathus hubrichti) localities suggested for acquisition by Dodd (1988).

121

122

Figure 7. Sites suggested for acquisition to promote the recovery of the Red Hills salamander (Phaeognathus hubrichti).

123

CHAPTER SIX

OVERALL CONCLUSION

One of the most fundamental problems in preserving the biodiversity of a region is that species are distributed unevenly (Gaston 2000). The ongoing loss of habitat and current extinction crisis paired with a scarcity of resources and time have forced the creation of conservation strategies that necessarily focus on select, small and isolated areas (Olson and

Dinerstein 1998; Myers et al. 2000; Lamoreux et al. 2006). The identification and prioritization of these areas is one of the most urgent goals in conservation biology. The overall objective in this dissertation was to explore novel and integrative approaches to identifying and prioritizing areas for the preservation of biodiversity at both the macro and micro scales.

Chapter two focused on comparing and contrasting the resulting patterns of richness maps created using environmental niche models with those created using traditional extent of occurrence maps for amphibians in the southeastern United States. I found that when compared to traditional richness maps (extent of occurrence), all three methods using environmental niche modeling yielded richness maps that were generally similar in identifying the areas of highest richness, but also had a high amount of variation. Regression analyses indicated that the presence/absence niche modeling method had the highest correlation with extent of occurrence maps (Anuran r = 0.69, Caudata r = 0.70), suggesting that the use of environmental niche models in the creation of biodiversity composite variables appears to be a valid approach as long as precautions are taken while analyzing the results.

Chapter three focused on comparing patterns of traditional biodiversity metrics (richness and weighted endemism) with patterns of biodiversity metrics that integrate phylogenetic information (phylogenetic diversity and phylogenetic endemism) for members of the family

124

Plethodontidae in the southeastern United States. I found that the phylogenetic metrics were weakly correlated with the non-phylogenetic methods. The patterns differentiate at the highest values, indicating that each method identifies unique and important areas for conservation.

Additionally, I found that none of the methods investigated are a proper surrogate for the others, and therefore it should not be assumed that species richness adequately represents all biodiversity metrics. I strongly urge the incorporation of phylogenetic information into biodiversity metrics whenever possible.

The combination of chapters two and three demonstrate that novel methods for identifying biologically important areas can provide a valuable and diverse set of data at a macro scale. The use of environmental niche models has the potential to further our knowledge of biodiversity hotspots for lesser known taxa or regions. This tool could be invaluable as the majority of our knowledge of biodiversity hotspots is drawn from the major vertebrate groups.

Undoubtedly, as we continue to explore the distribution patterns of lesser-studied taxa we will find that biodiversity hotspots for various taxa are widely disparate.

The incorporation of evolutionary data into conservation planning will be a much greater challenge for the majority of taxonomic groups. However, with the rapid advancement of sequencing technologies, this prospect will become an increasingly achievable objective. In fact, it is entirely possible that as we continue to sequence unknown taxa, via techniques such as environmental sequencing, methods such as phylogenetic diversity and phylogenetic endemism will be the only way to identify areas important to microscopic diversity. These methods will be especially important for many groups, such as microscopic algae, fungi, and bacteria, which are only recognized as genotypes rather than as formal taxonomic units.

125

The identification of biodiversity hotspots at macro scales is fundamentally important to global conservation efforts. However, efforts at this scale will inevitably fail to provide coverage for a number of species, especially in the case of narrow endemics. Therefore, at a local scale it is imperative to identify species outside of the global network of protected areas that face substantial risk of extinction. This is the case for the federally threatened Red Hills salamander.

Despite the listing of the Red Hills salamander in 1976 as a threatened species (U.S. Fish and

Wildlife Service 1976) their habitat has continued to decline. Clearly, new methods of conservation for this rare and endemic species are needed to ensure their long-term persistence.

Chapter four investigates the spatial genetic patterns of the federally threatened Red Hills salamander (Phaeognathus hubrichti) using 10 microsatellite markers. The identification of intra-population genetic structure and diversity is fundamental to many fields within evolutionary biology and ecology. From the perspective of conservation biology this information has become increasingly important in conservation planning for imperiled species (e.g. Geist and

Kühn 2005; Pabijan et al. 2005; Dixon et al. 2006; Schwartz et al. 2007; Marshall et al. 2009;

Matern et al. 2009; Straub and Doyle 2009). Maintaining genetic diversity, and therefore evolutionary potential, is fundamental to the long-term survival and recovery of at-risk species

(Avise 2004; Morgan et al. 2008). My results indicate that there are 5 well-supported populations (FST = 0.13009-0.1879) across the entire range of P. hubrichti, and that current migration rates between populations are low (m = 0.0025-0.0687). I also estimate that anthropogenic activities have reduced P. hubrichti habitat by as much as 86% when compared to historical levels. By accounting for history and species characteristics I demonstrate that this loss and fragmentation of habitat has had a strongly negative impact on P. hubrichti by reducing migration, increasing bottlenecks, and promoting high levels of inbreeding.

126

Chapter five combines the genetic data from chapter four with the species’ habitat requirements, climatic suitability, available habitat, habitat fragmentation, gene flow patterns, genetic structure and genetic variability in order to devise effective measures that will lead to the stabilization and eventual recovery of the Red Hills salamander. I recommend the acquisition and proper management of twenty-one new conservation areas for the species. Additionally, I urge that large landowners with populations of P. hubrichti enter into safe harbor agreements, for which we have provided guidelines.

The prioritization of areas for the conservation of biodiversity is not a straightforward science with definitive conclusions. It is ultimately a judgment call made by federal and local governments, conservation agencies, and private citizens who wish to conserve habitat.

However, one of the fundamental goals of conservation biology is to provide data and methods that can be used to make informed decisions on how to maximize the impact of conservation funds and effort. The chapters of this dissertation demonstrate that this field is rapidly evolving and that recently developed techniques are able to contribute a great deal of information to this effort. I have demonstrated that the addition of novel genetic techniques can help to elucidate otherwise indistinguishable patterns at a broad scale, thereby helping to conserve cryptic diversity. Furthermore, the use of genetic techniques at a more local scale can provide data on particularly important, yet difficult to obtain, species characteristics. The continued development and testing of novel conservation techniques, such as the ones in this dissertation, is essential to conservation efforts at a local and global scale.

127

LITERATURE CITED

Araujo MB, Williams PH (2000) Selecting areas for species persistence using occurrence data. Biological Conservation 96, 331-345.

Avise JC (2004) Molecular Markers, Natural History and Evolution (Second Edition) Sinauer, Sunderland, MA.

Brooks TM, Bakarr MI, Boucher T, et al. (2004) Coverage provided by the global protected-area system: Is it enough? Bioscience 54, 1081-1091.

Brooks TM, Mittermeier RA, da Fonseca GAB, et al. (2006) Global biodiversity conservation priorities. Science 313, 58-61.

Caro TM, O'Doherty G (1999) On the use of surrogate species in conservation biology. Conservation Biology 13, 805-814.

Ceballos G, Ehrlich PR (2006) Global mammal distributions, biodiversity hotspots, and conservation. Proceedings of the National Academy of Sciences U S A 103, 19374- 19379.

Dixon JD, Oli MK, Wooten MC, et al. (2006) Effectiveness of a regional corridor in connecting two Florida black bear populations. Conservation Biology 20, 155-162.

Duellman WE (1999) Patterns of distributions of amphibians The John Hopkins University Press, Baltimore, MD.

Faith DP (1992) Conservation Evaluation and Phylogenetic Diversity. Biological Conservation 61, 1-10.

Frost C (2006) History and future of the longleaf pine ecosystem In: The longleaf pine ecosystem ecology silviculture, and restoration (eds. Jose S, Jokela EJ, Miller DL). Springer Science+Business Media, New York, NY.

Gaston KJ (2000) Global patterns in biodiversity. Nature 405, 220-227.

128

Isaac NJB, Turvey ST, Collen B, Waterman C, Baillie JEM (2007) Mammals on the EDGE: Conservation Priorities Based on Threat and Phylogeny. Plos One 2.

IUCN (1994) Guidelines for protected areas management categories IUCN, Cambridge, UK.

Lamoreux JF, Morrison JC, Ricketts TH, et al. (2006) Global tests of biodiversity concordance and the importance of endemism. Nature 440, 212-214.

Leroux SJ, Schmiegelow FKA (2007) Biodiversity concordance and the importance of endemism. Conservation Biology 21, 266-268.

Margules CR, Pressey RL (2000) Systematic conservation planning. Nature 405, 243-253.

Marshall JCJ, Kingsbury BA, Minchella DJ (2009) Microsatellite variation, population structure, and bottlenecks in the threatened copperbelly water snake. Conservation Genetics 10, 465-476.

Means DB (2006) Vertebrate faunal diversity of longleaf pine ecosystems. In: The longleaf pine ecosystem ecology silviculture, and restoration (eds. Jose S, Jokela EJ, Miller DL). Springer Science+Business Media, New York, NY.

Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403, 853-858.

Noss R (1989) Longleaf pine and wiregrass: keystone components of an endangered ecosystem. Natural Areas Journal 9, 211-213.

Olson DM, Dinerstein E (1998) The global 200: A representation approach to conserving the Earth's most biologically valuable ecoregions. Conservation Biology 12, 502-515.

129

Orme CDL, Davies RG, Burgess M, et al. (2005) Global hotspots of species richness are not congruent with endemism or threat. Nature 436, 1016-1019.

Pabijan M, Babik W, Rafinski J (2005) Conservation units in north-eastern populations of the Alpine newt (Triturus alpestris). Conservation Genetics 6, 307-312.

Pimm SL, Russell GJ, Gittleman JL, Brooks TM (1995) The Future of Biodiversity. Science 269, 347-350.

Rodrigues AS, Andelman SJ, Bakarr MI, et al. (2004) Effectiveness of the global protected area network in representing species diversity. Nature 428, 640-643.

Rodrigues ASL, Gaston KJ (2002) Maximising phylogenetic diversity in the selection of networks of conservation areas. Biological Conservation 105, 103-111.

Sarkar S, Pressey RL, Faith DP, et al. (2006) Biodiversity conservation planning tools: present status and challenges for the future. Annual review of Environment and Resources 31, 123-159.

Schwartz MK, Luikart G, Waples RS (2007) Genetic monitoring as a promising tool for conservation and management. Trends Ecology and Evolution 22, 25-33.

Service UFaW (1976) Determination that the Red Hills salamander is a threatened species. . federal Registry 41, 53032-53034.

Soulé ME (1985) What is conservation biology? Bioscience 35, 727-734.

Spear SF, Peterson CR, Matocq MD, Storfer A (2005) Landscape genetics of the blotched tiger salamander (Ambystoma tigrinum melanostictum). Molecular Ecology 14, 2553-2564.

Straub SC, Doyle JJ (2009) Conservation genetics of Amorpha georgiana (Fabaceae), an endangered legume of the Southeastern United States. Molecular Ecology 18, 4349-4365.

Stuart SN, Chanson JS, Cox NA, et al. (2004) Status and trends of amphibian declines and extinctions worldwide. Science 306, 1783-1786.

Vane-Wright R, Humphries C, Williams PH (1991) What to protect-systematics and the agony of choice. Biological Conservation 55, 235-254.

Wang IJ (2009) Fine-scale population structure in a desert amphibian: landscape genetics of the black toad (Bufo exsul). Molecular Ecology 18, 3847-3856.

130

Weitzman ML (1998) The Noah's Ark Problem. Econometrica 66, 1279-1298.

Wilson KA, McBride MF, Bode M, Possingham HP (2006) Prioritizing global conservation efforts. Nature 440, 337-340.

131