How Ecology and Evolution Shape Species Distributions and Ecological Interactions Across Time and Space

Home , Amplexus, Branchinecta lynchi, Mosquitofish, Pelobatoidea, Scaphiopus, Scaphiopus holbrookii

HOW ECOLOGY AND EVOLUTION SHAPE SPECIES DISTRIBUTIONS AND ECOLOGICAL INTERACTIONS ACROSS TIME AND SPACE

IULIAN GHERGHEL

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Advisor: Ryan A. Martin

Department of Biology

CASE WESTERN RESERVE UNIVERSITY

January, 2021

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of

Iulian Gherghel

Candidate for the degree of Doctor of Philosophy*

Committee Chair

Dr. Ryan A. Martin

Committee Member

Dr. Sarah E. Diamond

Committee Member

Dr. Jean H. Burns

Committee Member

Dr. Darin A. Croft

Committee Member

Dr. Viorel D. Popescu

Date of Defense

November 17, 2020

* We also certify that written approval has been obtained for any proprietary material contained therein

TABLE OF CONTENTS

List of tables ...... v

List of figures ...... vi

Acknowledgements ...... viii

Abstract ...... iix

INTRODUCTION...... 1

CHAPTER 1. POSTGLACIAL RECOLONIZATION OF NORTH AMERICA BY

SPADEFOOT TOADS: INTEGRATING NICHE AND CORRIDOR MODELING TO

STUDY SPECIES’ RANGE DYNAMICS OVER GEOLOGIC TIME ...... 5

Abstract ...... 5

Introduction ...... 6

Materials and Methods ...... 9

Study organism ...... 9

Species occurrence data ...... 9

Climatic variables ...... 10

Ecological niche modeling ...... 11

Estimating dispersal patterns and migration corridors ...... 13

Testing the dispersal and migration corridor estimates with independent data ...... 14

Results ...... 15

Ecological niche models performance and estimated present distributions ...... 15

Past distributions and glacial refugia ...... 16 i

Identifying the likely migration corridors used to migrate from LGM refugia to extant ranges ...... 17

Testing dispersal patterns using genetic data: a case study in Spea hammondii ...... 17

Discussion ...... 18

Species distribution models: past and present distributions ...... 19

Using population genetic data to test our results ...... 20

The role of dispersal ...... 21

CHAPTER 2. BIOTIC INTERACTIONS VARY ACROSS SPECIES’ RANGE AND

ARE LIKELY CONSERVED THROUGH GEOLOGICAL TIME ...... 27

Abstract ...... 27

Introduction ...... 28

Material and Methods ...... 31

Model system ...... 31

Species occurrences ...... 32

Environmental data ...... 34

Modeling approach ...... 35

Post-processing methodology ...... 36

Results ...... 37

Model performance summary ...... 37

Fairy shrimp (prey) distribution patterns in the context of predator species ranges . 38

Correlation between the distribution of fairy shrimp and spadefoot toads across spatial scales...... 39

Discussion ...... 40

Presence of prey resources in time are important for patterns of predator co- occurrence ...... 41

Spatial variation in prey-predator relationships ...... 43

Biotic interactions in species distribution modeling ...... 44

CHAPTER 3. THE ECOLOGICAL AND SELECTIVE EFFECTS OF RESOURCE

POLYPHENISM ON LOWER TROPHIC LEVELS ...... 51

Abstract ...... 51

Introduction ...... 52

Materials and Methods ...... 54

Study system ...... 54

Effects of resource polyphenism ...... 57

Microcosm experiment ...... 57

Mesocosm experiment ...... 59

Statistical analyses ...... 61

Results ...... 63

Microcosm selection experiment ...... 63

Mesocosm experiment ...... 65

iii

Discussion ...... 66

Effects of trophic plasticity on the strength of selection ...... 67

Effects of trophic plasticity on the reproductive traits ...... 67

Effects of trophic plasticity on ecosystem functioning ...... 69

Appendix for Chapter 3 ...... 78

CONCLUSIONS ...... 82

BIBLIOGRAPHY ...... 84

LIST OF TABLES

Chapter 1

Table 1.1…………………………………………………………………….23

Chapter 2

Table 2.1…………………………………………………………………….46

LIST OF FIGURES

Introduction

Figure 1……………………………………………………………………..…2

Chapter 1

Figure 1.1…………………………………………………………………….24

Figure 1.2…………………………………………………………………….25

Figure 1.3…………………………………………………………………….26

Chapter 2

Figure 2.1…………………………………………………………………….47

Figure 2.2…………………………………………………………………….48

Figure 2.3…………………………………………………………………….49

Figure 2.4…………………………………………………………………….50

Chapter 3

Figure 3.1……………………………………………………………………71

Figure 3.2…………………………………………………………………….72

Figure 3.3…………………………………………………………………….73

Figure 3.4…………………………………………………………………….74

Figure 3.5…………………………………………………………………….75

Figure 3.6…………………………………………………………………….76

Figure 3.7…………………………………………………………………….77

vii

ACKNOWLEDGEMENTS

There are many people to which I would like to thank for supporting and helping me. First, I must thank to my advisor Ryan Martin, that helped, guided and believed in me along the five-year tenure that I was in his lab. I would also want to express my appreciation to my past academic advisors, Monica Papes and Stefan Zamfirescu for sharing with me their expertise, friendship and time that allowed me to become a better person and researcher. I also want to thank to my professors, my thesis committee, colleagues, my family, parents (Elena and Ioan) and friends that provided the support that allowed me to continue with my studies. During the summer field work at Southwestern Research Station

I met very supportive staff (here I must mention Mark Cooper, Geoff Bader and Garry

Wisdom) and volunteers (special thanks to John Cole, Colleen Meidt, Charles Baudry and

Alice Gadau) for their help in the field in Arizona and New Mexico. I would also like to thank the members of Diamond and Martin Labs and the Ecology and Evolution group at

Case Western for the much appreciated feedback on my research. Many thanks to

Alexandru Sotek, Florin Lauca, Michael Moore, Tiberiu Sahlean, Riley Tedrow,

Alexandru Strugariu, Sarah Diamond, Dan Cogalniceanu, Viorel Popescu, Costin Ion, Ana

Codita, Matt Dugas for their support and many advices they shared with me these many years. Finally, I would like to thank my finance, Raluca, for being supporting, loving and understanding during these years and for her help in the field.

viii

ABSTRACT

How Ecology and Evolution Shape Species Distributions and Ecological Interactions

Across Time and Space

IULIAN GHERGHEL

Understanding the factors that shape species’ distributions is a key topic in biogeography. As climates change, species can either cope with these changes through evolution, plasticity or by shifting their ranges to track the optimal climatic conditions.

Recent work suggests that biotic interactions, together with abiotic factors, are important in shaping patterns of species’ distribution and co-occurrence. Also, interspecific variation of traits has been found to have ecological consequences at the level of community and ecosystem dynamics, particularly in case of resource polyphenism, that allows in certain system differential use of resources by shifting the phenotype plastically. Ecological niche modeling (ENM) is a widespread technique in biogeography that estimates the niche of the organism by using occurrences and environmental data to estimate species’ potential distributions. ENMs are often criticized for failing to take species’ dispersal abilities into consideration. In this thesis we attempt to fill this gap by combining ENMs with dispersal and corridor modeling to study the range dynamics of North American spadefoot toads

(Scaphiopodidae) over the Holocene. Furthermore, we seek to understand the covariation between predator (Spea bombifrons and Spea multiplicata) and prey (North American fairy

shrimp species - Crustacea: Anostraca) range shifts in response to climate change oscillations, and how interactions are conserved across time and hence be used to project species distribution models on different time scales. Finally, we aim to evaluate whether the differential use of resources of two Spea multiplicate morphs impose different selection pressures on fairy shrimp traits and whether this also has feedbacking effects at the level of ecosystem.

INTRODUCTION

There are multiple explanations for the distribution of species across geography.

Most common are those associated with the concept of ecological niche, for example, from the early work of Grinnell (1917) which viewed the ecological niche more from the relationship between the broad abiotic (climatic) conditions and species occurrences. Elton

(1927) later proposed a slightly different view of the niche, where resources and species interactions would be more important than abiotic conditions. A half of century later,

Hutchinson (1957) proposed a more comprehensive view of the ecological niche, where the niche is seen as a hypervolume of abiotic and biotic conditions that shape species distributions, and in his work he also distinguished between the realized and fundamental niches. Pulliam (2000), Peterson and Soberon (2005), Soberon and Nakamura (2009) have expanded this framework by developing a conceptual framework known as BAM (Biotic,

Abiotic, Movement) (Figure 1). The BAM framework describes the simultaneous effects the environment, biotic interactions and dispersal have in shaping species distributions.

This framework also allows all components of the framework to be dynamic in their manifestation on geography, resulting in dynamic geographic distributions depending on the three components acting simultaneously. This later work has important implications for developing modern tools of ecological research (combined with advances in computer science and geographical information systems), known as ecological niche models or species distribution models (the terminology is determined by the aim and assumptions used when the models are run) (Peterson and Soberon 2012).

Figure 1. – The BAM framework inspired from Peterson and Soberon (2005). The intersection of the three circles represent the areas where the three conditions are met simultaneously for the range of the species to be occupied. The intersection between abiotic and biotic conditions represent invadable areas, where, if future dispersal allows, the species could potentially invade that supplemental area.

Climate changes across geography and time (Hewitt 2000). This has been best seen by the recent climatic oscillations characterizing the Quaternary period, where cold and warm periods of time are common and can last for multiple millennia (Hewitt 2000).

Another example is the modern-day climate change caused by anthropogenic activity

(IPCC 2016). These climatic changes induced dynamic processes in population dynamics which in some cases could lead to extinction. Species generally use two general strategies to cope with these changes, either to stay in place geographically and locally adapt to the novel conditions, or to track the suitable climatic and biotic conditions in areas where dispersal allows (Peterson et al. 2011). This leads to niche conservatism which allows 2

ecological niche models to investigate how species track their niches, taking into account modern climatic and species occurrence data to project the niche on climatic datasets corresponding to other times (either via hindcasting or forecasting these effects) (Peterson et al. 2011). Understanding how abiotic, biotic and movement act simultaneously in species’ responses is still an ongoing task in ecology and evolutionary biology (Peterson et al. 2011).

The importance of abiotic and biotic conditions to species distributions seem to be highly affected by scale and recent work suggests that biotic interactions are important in shaping species distribution more than previously was thought.

Interspecific variation offers a great opportunity to understand how biotic and abiotic factors affect species’ distributions by combining field, experimental and modeling work. Often, abiotic conditions can trigger dynamics in natural populations that would trigger plastic responses, altering the frequency of phenotypes in the population. One example are trophic polyphenisms triggered by environmental cues (Pfennig 1990, 1991).

Trophic polyphenisms may create feed backs with the environment making them an excellent model system to explore the role of plasticity in driving eco-evolutionary dynamics.

In my thesis I aim to explore these topics into the following three chapters. The first chapter intends to investigate how the ecological niche and dispersal influences species’ distributions across long time scales; the second chapter investigates the influence of biotic interactions (via predation) at different parts of the range and how the prey-predator relationship are influenced by past climate; and the third chapter investigates how interspecific variation in predators act on the traits of their prey, potentially changing

ecosystem productivity and functioning, using a series of microcosm and microcosm experiments.

CHAPTER 1. POSTGLACIAL RECOLONIZATION OF NORTH AMERICA BY

SPADEFOOT TOADS: INTEGRATING NICHE AND CORRIDOR MODELING

TO STUDY SPECIES’ RANGE DYNAMICS OVER GEOLOGIC TIME

Published in Ecography Vol. 43, No. 10, pp. 1499-1509, 2020. DOI: 10.1111/ecog.04942

Authors: Iulian Gherghel & Ryan A. Martin

Department of Biology, Case Western Reserve University, Cleveland, OH, USA

(Accepted 25 June 2020)

Abstract

Ecological niche modeling (ENM) is a widespread technique in biogeography that estimates the niche of the organism by using occurrences and environmental data to estimate species’ potential distributions. ENMs are often criticized for failing to take species’ dispersal abilities into consideration. Here, we attempt to fill this gap by combining ENMs with dispersal and corridor modeling to study the range dynamics of

North American spadefoot toads (Scaphiopodidae) over the Holocene. We first estimated the current and past distributions of spadefoot toads and then estimated their past distributions from the Last Glacial Maximum (LGM) to the present day. Then, we estimated how each taxon recolonized North American by using dispersal and corridor 5

modeling. By combining these two modeling approaches we were able to 1) estimate the

LGM refugia used by the North American spadefoot toads, 2) further refine these projections by estimating which of the putative LGM refugia contributed to the recolonization of North America via dispersal, and 3) estimate the relative influence of each LGM refugium to the current species’ distributions. The models were tested using previously published phylogeographic data, revealing a high degree of congruence between our models and the genetic data. These results suggest that combining ENMs and dispersal modeling over time is a promising approach to investigate both historical and future species’ range dynamics.

Introduction

Understanding the factors that influence and shape species’ distributions is a central topic in biogeography. Ecological niche modeling (ENM) estimates the niche of the organism by using occurrences and environmental data to estimate species’ potential distributions (Franklin 2009, Peterson et al. 2011). ENMs are a commonly used technique to study a wide array of topics in conservation (Gómez-Ruiz et al. 2017, Taylor et al. 2018), climate change (Dyderski et al. 2018, Wauchope et al. 2017), evolution (Loera et al. 2012,

Sillero et al. 2014), species invasion (Lantschner et al. 2017, Mutascio et al. 2017), and global health (Sehgal et al. 2011, Wang et al. 2018). ENMs are also used to study range shifts in response to both historical (Gray et al. 2017, Huang et al. 2017) and contemporary climate change (Boria et al. 2017, Veloz et al. 2012) by projecting the estimated niche on climate scenarios. Nonetheless, while ENMs contribute to our fundamental understanding 6

of how species are distributed now and into the future, there is a need to further develop and improve these methods.ENMs are often criticized for failing to take into consideration species’ dispersal abilitities. In a recent review, Miller and Holloway (2015) found just a handful of attempts that account for dispersal in ENMs (Dullinger et al. 2012, Engler and

Guisan 2009, Sahlean et al. 2014). Failing to account for dispersal can lead to substantial errors in projected species’ distribution estimates (Engler and Guisan 2009), especially if the models are being projected on areas that are inaccessible due to species’ dispersal abilities (Dullinger et al. 2012, Sahlean et al. 2014). Engler and Guisan (2009) introduced the MigClim to run simulations of species’ dispersal, similar to another approach proposed by Sahlean et al. (2014) which uses the landscape characters’ resistance estimates to estimate dispersal. When species distribution estimates are not informed by species’ dispersal abilities, the estimates often overestimate species’ range shifts, and dispersal estimates using statistical methods are highly desired to study species’ range shifts especially in the context of changing climates (Miller and Holloway 2015). To fill this gap, we combine ENMs with dispersal models over time to explore the dispersal patterns and corridors used by species to disperse from their glacial refugia. Combining the two methods

(ENMs and dispersal modeling) offers the advantage of using these well-established methods to simulate how species shift their ranges across time.

Past climate change events, such as the cycles between glaciated (cold) and interglacial (warm) periods through the Pleistocene, offer opportunities to study species’ range shifts as species track favorable climates. Postglacial migrations are usually sourced from a few glacial refugia (Normand et al. 2011, Svenning et al. 2008), and these refugia

do not contribute equally to the postglacial colonization process (Tzedakis et al. 2013).

Consequently, species do not often occupy their full potential distributions (Bai et al. 2018,

Dullinger et al. 2012, Normand et al. 2011, Svenning et al. 2008). Moving forward, examining the dispersal patterns and migration routes from LGM refugia is important for understanding how current species ranges have been shaped by dispersal and niche-related processes. A common way of determining these dispersal patterns is through population and landscape genetic tools, for example via haplotype networks combined with environmental data. However, these approaches can be expensive in time and money because they require dense genetic sampling across the range (i.e. Psonis et al. 2018, Psonis et al. 2017, Skourtanioti et al. 2016, Stöck et al. 2012, Wielstra and Arntzen 2012). In our study, we explore whether combining ENMs with dispersal and corridor modeling can provide reliable insights on the dispersal patterns, range dynamics and corridors used by species to disperse from their putative LGM refugia at the end of last Ice Age. To model dispersal and corridors, we use least-cost paths, a tool developed in spatial analysis to find the path between two locations that travels thru the most cost-effective route between the locations (as a function of time, or distance) also commonly used in population genetics

(Chan et al. 2011). We first estimate the current distributions for North American spadefoot toads and then estimate their past distributions during the Last Glacial Maximum, 21000 years ago. We then estimate how each taxon recolonized North American by using dispersal and corridor modeling. By combining these two modeling approaches, we 1) project which LGM refugia were used by the North American spadefoot toads, and 2) further refine these projections by estimating which of the putative LGM refugia have

contributed to the recolonization of North America via dispersal. Finally, we test the predictive ability of our models against previously published molecular phylogeography data (Neal et al. 2018). With this exercise, we aim to demonstrate a new approach for studying species’ range dynamics by combining ENMs and dispersal modeling.

Materials and Methods

Study organism

To study whether combining ENMs with dispersal and corridor modeling provide a reliable method for studying species’ range dynamics during the Holocene, we used the spadefoot toads (Scaphiopodidae) as a case study because their distributions were affected by the Ice Age (Garcıa-Parıs et al. 2003). Scaphiopodidae is a group of spadefoot toads found only in North America, primarily in southwestern Canada, and throughout the United

States and Mexico (Duellman and Trueb 1986). There are seven extant species of North

American spadefoot toads, belonging to two genera, Scaphiopus (Sc. couchii, Sc. holbrookii, and Sc. hurterii) and Spea (Sp. bombifrons, Sp. hammondii, Sp. intermontana, and Sp. multiplicata) (Garcıa-Parıs et al. 2003). These frogs are fossorial, spending most of the time burrowed in the sand, emerging en masse to reproduce on the first night of seasonal rains (Duellman and Trueb 1986). All species of both Scaphiopus and Spea have extremely rapid larval development, as fast as 7 days in Sc. couchii, 12-15 days in Sc. holbrookii, Sc. hurterii, Sp. bombifrons and Sp. multiplicata, and under 30 days in Sp. hammondii and Sp. intermontana (Duellman and Trueb 1986, Zeng et al. 2014).

Species occurrence data 9

We collected occurrence data for the North American spadefoot toads from published literature and online databases (HerpMapper, Global Biodiversity Information

Facility). Occurrence data without Latitude and Longitude were manually geo-referenced using Global Gazetteer Version 2.1 (available at http://www.fallingrain.com/world). We then manually searched and georeferenced locations not found in the gazetteer using

Google Earth (available at https://earth.google.com/) following the instructions on the museum collection label. In total, the dataset comprised over 8000 georeferenced records, the entire range of the studied taxa, providing a good representation of their distribution.

We then tested for spatial bias in the occurrence dataset using a global Moran I test and found that the species occurrence data were heavily clumped (Moran I index=0.38, z=440, p=<0.001). To mitigate this spatial bias of the data, we used the function “trim duplicate” in the Pearl application ENMTools 1.3 (Warren et al. 2010), then we used SDMtoolbox

(Brown and Anderson 2014) (available at http://sdmtoolbox.org/) to further rarefy the occurrence points based on the environmental datasets (see Environmental data section) and the distance between points (20 km). This allowed us to avoid spatial bias and resulted into 2741 unique, spatially and environmentally unbiased occurrence records for the studied taxa (Sc. couchii (703), Sc. holbrookii (369), Sc. hurterii (130), Sp. bombifrons

(605), Sp. hammondii (138), Sp. intermontana (257), Sp. multiplicata (539)).

Climatic variables

We ran the ENMs using climatic data downloaded from WorldClim 1.4 database at 5 km resolution (Hijmans et al. 2005, available at: www.worldclim.org). The WorldClim database is comprised of 19 bioclimatic variables (Margules and Austin 1990, sensu Nix 10

and Busby 1986), which are a combination of seasonal and monthly temperature and precipitation variables that have proved to be good predictors for estimating species distributions (Beaumont et al. 2005, Hijmans and Graham 2006, Nix and Busby 1986).

Variables with a high degree of multicollinearity (Pearson correlation, r>0.75) were eliminated from the analysis (Dormann et al. 2013). This resulted in eight bioclimatic variables used to create the ENMs (mean diurnal temperature range, isothermality, minimum temperature of the coldest month, mean temperature of the wettest quarter, mean temperature of the driest quarter, mean temperature of the warmest quarter, annual precipitation, precipitation seasonality) (Table 1.1). For estimating the past species’ distributions of North American spadefoot toads, we used climatic reconstruction models corresponding to the Last Glacial Maximum (LGM) and mid-Holocene. The LGM data used in this study come from two global circulation model simulations, the Model for

Interdisciplinary Research on Climate (MIROC) produced by the Center for Climate

System Research (Hasumi 2007), and the Community Climate System Model (CCSM) produced by University Corporation for Atmospheric Research (Collins et al. 2006), available at www.worldclim.org (Hijmans et al. 2005).

Ecological niche modeling

To estimate the potential distributions of the North American spadefoot toads we used Classiﬁcation Tree Analysis, Multivariate Adaptive Regression Splines, Generalized

Linear Models, Generalized Additive Models, Artificial Neural Networks,

Maxent, and Support Vector Machines using the ‘SSDM’ R package (Schmitt et al. 2017).

To estimate the past species’ distributions of the North American spadefoot toads, the niche 11

models trained with present climate data were projected over the past climate reconstructions corresponding to the LGM (see Climatic variables section). Since significant inter-model variation is present when projecting models in time or space, we followed Araujo et al.’ (2019) recommendations to run multiple algorithms, stacking all the results of the algorithms into one AUC weighted stacked projection for each species that were used in all of our analyses. Since accounting for dispersal is an important prior of ENMs (Barve et al. 2011), we used as the accessible area the current range of the North

American spadefoot toads (IUCN 2016). Models were built using a random subset of occurrence points (75%, calibration dataset) and model performance was evaluated using the remaining occurrence points (25%, validation dataset) (Peterson et al. 2011, Phillips et al. 2006). Variable contribution to the model was calculated based on the Pearson correlation coefficient between the model with all variables and models where each variable was omitted in turn, using the SSDM package (Schmitt et al. 2017). The resulting models were evaluated using Receiver Operating Characteristic (ROC) Area Under the

Curve (AUC) and Cohen’s Kappa test (Mouton et al. 2010). AUC’s and Cohen’s Kappa are metrics that range from 0 (no model fit) to 1 (perfect model fit) (Hand and Till 2001) and are frequently used to evaluate ENMs (Franklin 2009, Phillips et al. 2006) (Jiménez-

Valverde 2012). In addition to the AUC and Cohen’s Kappa, we also used 10 percentile omission error as a measure of the models’ power to discriminate between suitable and unsuitable climatic conditions for the North American spadefoot toads. This metric is more reliable for evaluating the fit of the models because it takes into consideration only the omission error, which is independently calculated from a validation data set corresponding

to known species distributions (Jiménez-Valverde 2012, Mouton et al. 2010). To calculate the 10-percentile omission error, we took the thresholded binary present/absent prediction and calculated the ratio of occurrences predicted absent of the occurrences set aside for testing the models (the validation dataset) (see above) (Franklin 2009, Peterson et al. 2011).

Estimating dispersal patterns and migration corridors

For estimating the dispersal patterns (here used to describe the trends observed in the models) and migration corridors (here used to discuss the estimated corridors resulted from the dispersal analysis) used by each North American spadefoot toad species from their estimated glacial refugia to their current known distribution (represented by the occurrence points rarefied at 50 km to reduce spatial clumping and bias for each species), we first created a friction layer for each species. The friction layer assumes that movement can occur across a landscape in response to other landscape elements besides distance (the ease of movement is greatest where resistance/friction is least) (McRae et al. 2008). For creating the friction layer, we first summed the continuous SSDMs suitability’s of the present and mid-Holocene conditions, and then inverted these summed values so that lower values in the final friction layer represent higher suitability SSDMs values using

SDMToolbox (Brown and Anderson 2014). We choose this method due to the overwhelming evidence that species distribution models’ suitability scores correlate directly with functional connectivity and gene flow (Chan et al. 2011, Wang et al. 2008).

Using this technique to estimate the friction layer in corridor and dispersal analyses has advantages over methods relying on expert opinion where values are chosen subjectively

(Brown and Anderson 2014, Chan et al. 2011). Next, we used the friction layer to inform 13

movement across the landscape from the estimated LGM refugia (used as sources) to present-day occurrence records using least-cost paths (LCPs) (McRae et al. 2008). This method allowed us to estimate the most likely migration corridors (LCPs) that spadefoot toads followed to spread across North America to result in their present-day distributions.

For estimating the corridors, we combined all the LCPs and calculated the density of the

LCPs across North America. We define corridors used by the species to migrate from their putative LGM refugia to inhabit the extant range as the areas where most LCPs are passing

(high corridor = 66-100% of the LCPs pass the area; intermediate corridor = 33-65% of the

LCPs pass the area; minimum corridor = 5-32% of the LCPs pass the area, and the areas that are being passed by less than 10% of the LCPs were masked due to low congruence of

LCPs). Next, to estimate how many of the putative LGM refugia contributed to current distributions, we examined the association of the LCP of the extant occurrence points to the putative LGM refugia. Following this methodology, if an extant point is directly linked with a putative LGM refugium by an LCP, then we assume that a) the population associated with the occurrence point has sourced from the putative LGM refugium linking the two by an LCP, and b) the population is within the influence of the putative glacial refugium. If a putative glacial refugium is linked to no extant occurrence points by an LCP, then we assume that it was not used. These analyses were repeated for each species using Spatial

Analyst in ArcGIS 10.3.

Testing the dispersal and migration corridor estimates with independent data

Lastly, we tested our dispersal models by comparing our results to recent independent molecular phylogeographic data (Neal et al. 2018). This was done in ArcGIS 14

10.3, using the Spatial Analyst toolbox where we overlapped the previously published population genetic data (Neal et al. 2018) for one of the species (Sp. hammondii) used in our study with the dispersal and corridor estimates to test their reliability. In this and previous studies, Sp. hammondii was found to have a high degree of genetic divergence between the southern and northern populations (Garcıa-Parıs et al. 2003, Neal et al. 2018).

Therefore, we tested the degree of overlap between our estimated corridor model results and the results obtained from molecular phylogeography (Neal et al. 2018) as an independent method to test the predictions of our dispersal models. Population genetic data are currently insufficient to perform a similar test for the other North American spadefoot toad species, although we make more informal comparisons where possible.

Results

Ecological niche models performance and estimated present distributions

All ENMs performed very well (train AUC: 0.91) and adequately predicted the current distribution of the species (the average 10% omission test: 8%) (Table 1.1). Spea hammondii had the best model fit (where only 4% of the occurrences left for testing the models were omitted under the threshold), whereas the model for Spea bombifrons and

Spea multiplicata had a good AUC (train AUC: 0.88), but more occurrences used for testing were omitted under the threshold (Table 1.1). The most important predictor for estimating the distributions of North American spadefoot toads were the mean temperature of the wettest quarter (19.84%), precipitation seasonality (18.51%), and annual precipitation (17.46%). However, species differed in how important each variable was to

their distributions (Table 1.1). For example, Sc. holbrookii’s distribution is highly affected by the mean temperature of the wettest quarter, whereas Sp. multiplicata is less affected by this variable (Table 1.1). Generally, the models identified the core of the known distributions of the North American spadefoot toads (Table 1.1, Figure 1.1), with most predictions being within the known ranges of the studied species (Figure 1.1). However, the models for Sp. intermontana and Sp. multiplicata identified suitable areas outside of their known present range, most notably in California and Mexico.

Past distributions and glacial refugia

The two LGM circulation models identified key areas for the past distributions of the North American spadefoot toads. The past distributions of Sc. holbrookii, Sp. hammondii, and Sp. multiplicata highly overlap their current ranges (Figure 1, Figure 2).

Most changes occured in the northern part of the species’ ranges. In contrast, Sc. couchii,

Sc. hurterii, Sp. intermontana and Sp. bombifrons likely saw a drastic change in their ranges in the past compared with today (Figure 1.1, Figure 1.2). From the putative LGM refugia initially estimated, we have selected only the LGM refugia that were connected by a least cost path to a known presence occurrence of the spadefoot toad species (Figure 1.3).

Among all species, only 40% of all potential LGM refugia were found to have contributed to the extant ranges of the species. Overall, we identified two different patterns during

LGM: a) a continuous, large refugium, oftentimes used by many spadefoot toad species

(Sc. couchii, Sc. hurterii, Sp. bombifrons, Sp. multiplicata), in the Rio Negro river basin, western coast of Mexico, or by single species in the northern Rocky Mountains (in case of

Sp. intermontana) or in Florida and South-Eastern USA (in case of Sc. holbrookii) (Figure 16

2); and b) refugia sparsely distributed in the Rocky Mountains, and along the California coast (in case of Sp. intermontana and Sp. hammondii) (Figure 1.2).

Identifying the likely migration corridors used to migrate from LGM refugia to extant ranges

From the selected LGM refugia, we estimated the corridors used by each species to migrate to their extant ranges (Figures 1.1-1.3). We found that each species has a unique number of high probability corridors and that few of these corridors are overlapping (Figure

1.2). Scaphiopus couchii, Sp. bombifrons, and Sp. multiplicata had overlapping corridors in western Texas on their way north into the Great Plains (Figure 2). Scaphiopus holbrookii used major migration corridors northwards from the glacial refugium located in the southeastern USA along the Atlantic and Mexican Gulf Coast (Figure 1.2). Other minor corridors used by the Sc. holbrookii to pass the Appalachian Mountains into the Midwest are also found along major rivers (Figure 1.2). Scaphiopus hurterii used a few corridors from its refugium in eastern Texas to inhabit the rest of the range in the southern Great

Plains (Figure 1.2). Spea intermontana had multiple small corridors from multiple glacial refugia. Spea hammondii is clustered into two groups; the northern group likely never met the southern group as the two had separate glacial refugia and distinct corridors along the

Pacific coast (Figure 1.2). The northern group likely dispersed from one large refugia, whereas the southern group likely come from multiple small refugia (Figure 1.2).

Testing dispersal patterns using genetic data: a case study in Spea hammondii

We found that our predictions (LGM potential distributions, corridor, and dispersal modeling and the influence of each LGM) agreed with the existing population genetic data for S. hammondii (Figure 1.3). More specifically, our models predicted that S. hammondii’s distribution is split in two, generally characterized by a northern and a southern cluster that each used different glacial refugia and recolonized their range after glaciation using different corridors resulting in minimal overlap between the two groups, in agreement with genetic data (Figures 1.2, 1.3).

Discussion

The role of dispersal is often overlooked in ecological niche models (Barve et al.

2011, Guisan and Zimmermann 2000, Holloway et al. 2016, Peterson et al. 2011, Qiao et al. 2017, Waltari et al. 2007). Currently, methods for estimating ENMs are rarely combined with dispersal analyses (Holloway et al. 2016), leading to an inaccurate estimation of the factors that shape species distributions. Hence, combining the two techniques—ENMs and dispersal analyses—are crucial in developing more realistic models (Barve et al. 2011,

Qiao et al. 2017). Here, we used ENMs and dispersal analyses to study the range dynamics of species following the Last Glacial Maximum by estimating the current and past distributions for the North American spadefoot toads and the corridors used by these species to disperse and inhabit their current range (Figures 1.2, 1.3). For one species (Sp. hammondii), we tested these predictions against previously published phylogeography data

(Neal et al. 2018). This comparison revealed a high degree of agreement between the current population genetic structure of Sp. hammondii and the past range dynamics predicted by our models (Figure 1.3). Therefore, we suggest that combining ENMs and 18

dispersal modeling represents a promising approach for studying species’ range dynamics and for generating hypotheses of the biogeographic past.

Species distribution models: past and present distributions

Our models identified the core of the current distributions of North American spadefoot toads (Figure 1). Despite all of our models fitting the data well (AUC >0.75), the model for Sc. hurterii and Sp. intermontana exhibited lower fit compared with other top- performing models (such as the model for Sc. couchii) (Table 1.1). This often happens when the occurrence data for the species is scarce in some parts of the species range

(Franklin 2009, Peterson et al. 2011). However, together these models recovered the general patterns of North American spadefoot species distributions. Interestingly, we found that, by the mid-Holocene, the climate was suitable for all of the North American spadefoot toads to an extent similar to that of the current species’ ranges. For Sc. holbrookii, our models suggest the species inhabited a large Atlantic and Mexican Gulf coast refugium located in the southeastern USA and then expanded north along Atlantic Coast, the

Mississippi river and other major rivers (Figure 1.2). In the case of Sc. hurterii, the glacial refugium was probably in current day northern Mexico, and the species migrated north along the Great Plains. Scaphiopus couchii on the other hand likely had multiple glacial refugia, isolated by the Sierra Madre Occidental and Oriental Mountains during the LGM, and another glacial refugium in Baja California. This scenario suggests that the range of this species was fragmented by the LGM, leading to the isolation of these populations.

Similarly, Sp. hammondii and Sp. intermontana had a series of multiple refugia, which should result in substantial genetic structure among populations (Neal et al. 2018, Wiens 19

and Titus 1991) (Figures 1.2, 1.3). Indeed, in the case of Sp. hammondii, our predictions of alternative refugia match well with the limited historic gene flow between the northern and southern populations, which have been recently proposed as distinct species (Garcıa-

Parıs et al. 2003, Neal et al. 2018). Similarly, we might predict that the multiple refugia identified in the Great Basin, as well as the Rocky Mountains, led to limited gene flow within Sp. intermontana (Figures 1.2, 1.3). This finding is also supported by earlier work which found high genetic divergence between eastern and western Sp. intermontana populations (Garcıa-Parıs et al. 2003). In the case of Sp. bombifrons, the predicted glacial refugium of the species was in Arizona. While previous molecular work suggests that the glacial refugium of Sp. bombifrons was in Kansas and Oklahoma (Rice and Pfennig 2008), these areas correspond with the most intensive sampling in their study, which could miss other centers of high diversity indicative of past refugia. Moreover, our models suggest that the isolated population of Sp. bombifrons from the southern part of Texas have been isolated from the rest of its range for at least 20k years (Figures 1.2, 1.3). This suggests that those populations might have used a different glacial refugium compared with the rest of the range. Our models for Sp. multiplicata on the other hand, found that the species inhabited the same range during the LGM as the extant range of the species.

Using population genetic data to test our results

Incorporating multiple lines of evidence to test estimates from ENMs is highly desirable (Franklin 2009, Peterson et al. 2011). Our model estimates identified glacial refugia used by the North American spadefoot toads during the LGM (Figure 1.2). By incorporating dispersal patterns and corridor modeling, we were able to create predictions 20

of the paths these species used to colonize the continent after the climate become suitable

(Figure 1.2). We used previously published molecular phylogeography results (Neal et al.,

2018) to test our models using independent genetic data. Previously, Sp. hammondii has been found to have a high degree of genetic divergence between their southern and northern populations (Garcıa-Parıs et al. 2003, Neal et al. 2018) (Figure 1.3). Estimated corridor models confirm the presence of the two clusters and suggest that they used different migration corridors to disperse from their respective (distinct) glacial refugia. The comparison between our estimates and the published population genetic results showed a high degree of congruence, suggesting that our models are reliable and that our technique accurately predicts within-range species patterns associated with dispersal and niche dynamics (Figure 1.3).

The role of dispersal

The distribution of any species is a complex interaction between its ecological niche

(including physiological tolerances) and its evolutionary history (Brown et al. 1996).

Dispersal also shapes species’ distributions, as they must colonize suitable, available geographic space (MacArthur 1984, Pulliam 2000, Pulliam 1988). As a consequence, incorporating dispersal into the estimates of species distributions generated using ENMs projected into geographical space is important (Barve et al. 2011, Holloway et al. 2016).

Nobis and Normand (2014) introduced a method to model accessible areas over time that further developed the field. However, estimating migration routes used by species during range shifts over time is still a topic that needs to be further studied in species distribution modeling. Another application performed by Gherghel and Papeş (2015) based solely on 21

dispersal models (no ENMs or genetic data were considered) estimated the connectivity and migration corridors between all known populations of Danube Crested Newt (Triturus dobrogicus). These models have since proven to be consistent with the genetic population structure of the species (Vörös et al. 2016, Wielstra et al. 2016). Based on our results and the lines of evidence from other studies, we can conclude that combining niche modeling with dispersal modeling can improve range dynamics that are influenced by climate and dispersal. When available, these models should be supported and tested against genetic and/or fossil data to avoid potential biases in predictions (Davis et al. 2014), and we encourage further research into combining ENMs and dispersal models.

Table 1.1 Summary of the average performance and variable contributions for the generated ENMs

Species AU 10% Coh Mean Annual Precipit Mean Isother Min Mean Mean C omis en's Temper Precipit ation Diurnal mality Temper Temper Temper train sion Kap ature of ation Season Temper ature of ature of ature of ing error pa Warme ality ature Coldest Wettest Driest st Range Month Quarter Quarter Quarter Scaphiopus 0.91 8% 0.82 11.47 24.25 7.20 5.99 13.85 9.56 22.65 5.04 couchii Scaphiopus 0.89 8% 0.75 10.56 14.62 25.13 0.67 4.88 5.84 16.35 21.96 holbrookii Scaphiopus 0.92 8% 0.75 25.85 13.74 11.24 5.70 15.79 7.21 14.32 6.15 hurterii Spea 0.91 9% 0.79 4.54 39.41 21.35 0.76 2.86 3.51 24.55 3.02 intermontana Spea 0.88 9% 0.74 13.29 5.39 25.24 13.99 8.36 14.95 7.81 10.97 bombifrons Spea 0.95 4% 0.87 6.71 11.32 34.60 7.57 3.53 3.87 25.20 7.19 hammondii Spea 0.88 10% 0.76 6.57 13.54 4.80 23.84 11.56 7.22 28.02 4.46 multiplicata

Figure 1.1 Current predicted distributions of the extant North American spadefoot toads.

Figure 1.2. Past distributions of the extant North American spadefoot toads during the Last Glacial Maximum, and the corridors used by the species to expand their ranges during the Holocene. Dispersal corridors are mapped according to the density of least-cost paths in a given area. Intermediate and high corridors show that the majority of least-cost paths are passing a given area and were likely used as a corridor during the dispersal from the Last Glacial Maximum to their current extant range. 25

Figure 1.3. The congruence between current population genetic structure and predicted dispersal from refugia after the LGM for Sp. hammondii. The purple area is the predicted distribution during the last glacial maximum, the black polylines are the individual least- cost paths from the glacial refugia to the known present occurrences of the species. Filled yellow points are the present species occurrence points used for constructing the ENMs. Population genetic sampling data identifying a northern clade (blue filled points) and southern clade (pink filled points) are from Neal et al. 2018. 26

CHAPTER 2. BIOTIC INTERACTIONS VARY ACROSS SPECIES’ RANGE

AND ARE LIKELY CONSERVED THROUGH GEOLOGICAL TIME

Abstract

How do abiotic and biotic factors shape species’ distributions? While historically, abiotic factors received more attention, recent work suggests that biotic interactions, together with abiotic factors, are important in shaping patterns of species’ distribution and co-occurrence. We seek to understand the covariation between predator and prey range shifts in response to climate change oscillations, and how biotic interactions between these species are conserved to be projected on species distribution models on different time scales.

We first estimated the potential distribution of central and western North American fairy shrimp species (Crustacea: Anostraca) and the potential distribution of two western spadefoot toad species (Spea bombifrons and Spea multiplicata) using multiple modeling techniques. We then produced a shrimp species richness layer by summing the individual species binary estimates (presence/absence) to produce a continuous species richness map.

Third, we studied the relationship between the probability of spadefoot toad presence and the fairy shrimp species richness during the present and past (Last Glacial Maximum) periods. Finally, we estimated the strength and direction of this co-occurrence between spadefoot toads and fairy shrimp at the range, regional, and occupancy levels.

We found that spadefoot toads and fairy shrimp species from central and western

North America are shaped by similar abiotic environmental variables. Furthermore, we 27

found that the areas of sympatry of S. bombifrons and S. multiplicata correspond with higher levels of shrimp richness and with arid and dry conditions. Finally, we found that spatial patterns in the direction and strength of predator-prey co-occurrence are highly variable across geography, forming a spatial mosaic over the species’ ranges.

Our results suggest that predator-prey relationships form a spatial mosaic across geography and the ranges of species. When species distribution estimates are built for organisms with dietary specialization, relevant prey information increases the robustness of the models.

Introduction

Understanding how abiotic and biotic factors shape species’ distributions is an ongoing research topic in biogeography (Gaston 2009). While it has been argued that biotic interactions are less important in shaping species distributions at larger scales (Soberon

2007, Soberón and Nakamura 2009, Whittaker, et al. 2001) and secondary to abiotic factors at the range level (Soberon 2007, Soberón and Nakamura 2009) recent work instead suggests that biotic interactions, together with abiotic factors, are important in shaping patterns of species’ distribution and co-occurrence, and including these interactions would improve species distribution models (Anderson, et al. 2002, Anderson 2017, Araujo and

Rozenfeld 2014, De Araújo, et al. 2014, Dormann, et al. 2018, Engelhardt, et al. 2020,

Gherghel, et al. 2018, Jenkins, et al. 2020, Mod, et al. 2015). However, including biotic interactions in the species distribution modeling process has been difficult due to the complexity and dynamic nature of biotic interactions (Anderson 2017, Araujo and 28

Rozenfeld 2014). Previous studies have successfully included biotic interactions via species richness as a proxy for prey species assemblages available for predators and their interactions (assuming that if prey and predators occupy the same geographic area, they might interact locally) (Gherghel, Brischoux and Papeş 2018, Kosicki, et al. 2016, Mod, le

Roux, Guisan and Luoto 2015). Using these methods, the effects of biotic interactions across broad spatial scales can be inferred from patterns of co-occurrence. However, previous studies have generally considered only current species’ distributions, resulting in a limited understanding of how biotic interactions are conserved through time.

When closely related species live in sympatry, their shared ecological characteristics may force them to compete for the available resources (Pfennig and Pfennig

2009). As a result of this competition, selection may favor divergence in traits associated with resource acquisition between the related species, potentially allowing for species co- existence (Pfennig, et al. 2006, Pfennig and Pfennig 2009). The resulting geographic pattern, in which closely related species diverge in resource use and phenotype where they co-occur, is known as ecological character displacement (Brown and Wilson 1956, Pfennig and Pfennig 2012, Schluter and McPhail 1992). Character displacement is also known to play a key role in creating and maintaining biodiversity (Pfennig and Pfennig 2012,

Schluter 2000). However, geographic patterns of species co-occurrence are likely often dynamic; both the existence of sympatric ranges, and the geographic locations of sympatry can shift in response to changing climates (Rice and Pfennig 2008).

Quaternary glaciation oscillations have played a major role in shaping the distribution of species (Hewitt 1999, Hewitt 2004, Ordonez and Svenning 2017, Svenning,

et al. 2015, Svenning, et al. 2011). During these times of severe climatic instability, cycles of glacial (characterized by cold climates) and inter-glacial (characterized by warm climates) periods caused species to shift their ranges or stay in place and locally adapt to the new conditions (Hewitt 1999, Hewitt 2004, Svenning, Eiserhardt, Normand, Ordonez and Sandel 2015). However, the role of past climates and range dynamics on present patterns of character displacement receives little attention. Yet, the process of character displacementevolutionary divergence due to selection for reduced competitioncannot occur if the closely related species and their shared resources do not inhibit the same geographic space. As such, investigating the range dynamics of closely related species and their shared resources is important in understanding how climate change has and will affect species co-existence and co-evolutionary dynamics.

Spadefoot toads (Spea bombifrons and Spea multiplicata) are a model system for understanding character displacement (e.g., Pfennig and Murphy 2000, 2002, 2003;

Pfennig and Martin 2009) and the interaction of abiotic conditions and biotic interactions in shaping patterns of species co-occurrence (Chunco, et al. 2012, Pfennig, Rice and Martin

2006). Moreover, recent work (Gherghel and Martin 2020) suggests that S. bombifrons and

S. multiplicate have likely co-existed in areas of sympatry through the Holocene, making spadefoot toads an excellent system to study whether past range dynamics might affect the relationship between interacting prey, predator and competitors across time. Here we aim to understand how prey (resources) and their competing predators track each other’s distributions in response to climate change oscillations. Furthermore, we aim to investigate how an important mechanism of co-occurrence (character displacement) was influenced

by the historical range dynamics of competing species and their shared resource. Initially, we tested whether the environmental factors that shape the prey species distributions are the same as the environmental factors that shape the predators’ potential distributions. We were then interested in studying geographical variation in how fairy shrimp species richness (the shared prey) relates to spadefoot toad species distributions’ during the present and Last Glacial Maximum. Next, we aimed to estimate the strength and direction of the biotic relationship between spadefoot toads and fairy shrimp at the range, regional, and occupancy geographic scales. Finally, we discuss the implications of including biotic interactions in species distribution models and of projecting them to different time contexts to study future or past effects of biotic interactions on species range dynamics.

Material and Methods

Model system

Spea bombifrons and Spea multiplicata have the widest ranges among the western spadefoot toads, distributed across the arid region of western and central North America

(Figure 1). They breed during the summer rainy season in ephemeral pools that are highly vulnerable to drying, and the larvae of spadefoot toads can speed up their larval development to avoid desiccation (for Spea, metamorphosis can occur in as little as 14 days

(Bragg 1967)). The larvae have a diet based primarily on algae and detritus, and when present, tadpoles will consume large amounts of fairy shrimp (Crustacea: Anostraca).

Consuming fairy shrimp or other tadpoles can induce a fast developing, alternative resource-use morph in Spea larvae, with large jaw muscles, serrated beaks, and shortened

intestines, specialized for feeding on fairy shrimp and other tadpoles (hereafter the

“carnivore morph”) from the default “omnivore morph”) (Pfennig 1990, Pomeroy 1982).

Where the two species live in sympatry (Figure 1), and both detritus and fairy shrimp resources are abundant, S. bombifrons larvae develop primarily as carnivores, and S. multiplicata develop primarily as omnivores, resulting in ecological character displacement (Pfennig and Murphy 2000, 2002). However, in ponds where one or both resources are limited, local competitive exclusion can instead occur (Pfennig, Rice and

Martin 2006, Pfennig and Pfennig 2012). A recent study on spadefoot toad biogeography found that the geographic area of sympatry remained relatively unchanged despite range shifts through Holocene (Gherghel and Martin 2020).

Fairy shrimp (Crustacea: Anostraca) occur in inland ephemeral wetlands that dry up a part of the year (killing the adults) leaving only the desiccation resistant eggs to lay dormant in the dry wetland until it floods again, generally during the summer rainy season in our study area (Brendonck, et al. 2008, Rogers 2014a, Rogers 2014b). Fairy shrimp eggs hatch during the seasonal rains and adults are generally a few centimeters long (Belk 1975,

Rogers 2014a, Rogers 2014b) and they often inhabit the same temporary wetlands where spadefoot toads reproduce (Pfennig 1990, Pomeroy 1982). Fairy shrimp are an important food source for S. bombifrons and S. multiplicata, and are necessary for the maintenance of the carnivore morph where each species occurs alone (Martin and Pfennig 2010).

Moreover, as described above, without the presence of fairy shrimp, local co-existence of the two species is also not possible (Pfennig et al 2006).

Species occurrences 32

For predicting the potential distribution of the two spadefoot toad species in our study, we used the same dataset as Gherghel and Martin (2020) which previously produced reliable current and past distributions estimates for the North American spadefoot toads

(Figure 1). The distribution records for the spadefoot toads were gathered from literature and online databases (HerpMapper (www.herpmapper.org), iNaturalist

(www.inaturalist.org), GBIF (GBIF.org 20 April 2016)) (Figure 2.1). For the fairy shrimp occurrence records, since the degree of false species identification in this group is high

(Belk 1975), we preferred to use data from the literature as a primary source of occurrence records for our models (Figure 2.2). As such, we georeferenced all records from a recent review on the distribution of fairy shrimp in North America published by Rogers (2014a),

Rogers (2014b). Locations without longitude and latitude were georeferenced using Global

Gazetteer Version 2.1 (available at http://www.fallingrain.com/world) or were manually searched and georeferenced using Google Earth (available at https://earth.google.com/).

The occurrence dataset was also tested for spatial bias, and to avoid regions of heavy bias we rarefied all the points at a distance of 20 km using SDMtoolbox (available at http://sdmtoolbox.org/) (Brown and Anderson 2014) to ensure that no region is oversampled. We produced models of the species with more than 20 occurrence records to ensure that our models are robust and create an accurate model of species distribution of the species. The cutoff technique of modeling species based on occurrence numbers is commonly used in the field (Hernandez, et al. 2006, Wisz, et al. 2008) and previously it has been used to make decisions on the structure of biotic species interactions when using species distribution modeling (Gherghel, Brischoux and Papeş 2018, Gherghel, et al. 2020).

The final occurrence dataset comprised a total of two spadefoot toad species (Spea bombifrons (605 records), and Spea multiplicata (539 records)) and 12 species of fairy shrimp (Artemia franciscana (25 records), Branchinecta coloradensis (23 records),

Branchinecta lindahli (64 records), Branchinecta lynchi (22 records), Branchinecta mackini (37 records), Branchinecta packardi (24 records), Streptocephalus dorothae (37 records), Streptocephalus mackini (129 records), Streptocephalus similis (32 records),

Streptocephalus texanus (36 records), Thamnocephalus mexicanus (29 records),

Thamnocephalus platyurus (63 records)) (Figure 2.1, Figure 2.2).

Environmental data

To estimate the species distributions of the spadefoot toads and fairy shrimp species we used the climatic layers from the WorldClim 1.4 database at 5 km resolution (Hijmans, et al. 2005, ). The WorldClim database consists of 19 climatic variables that represent different combinations of temperature and precipitation at the month or season level (Beaumont, et al. 2005, Hijmans, Cameron, Parra, Jones and Jarvis

2005) and since its development, it has been widely used to estimate species distributions

(Broennimann, et al. 2007, Elith, et al. 2006, Sahlean, et al. 2014). To reduce the effect of multicollinearity, we eliminated the variables that were highly correlated (Pearson correlation, r > 0.75) (Dormann, et al. 2013). After the variable reduction, eight climatic variables were used to run all the species model estimates (mean diurnal temperature range, isothermality, minimum temperature of the coldest month, mean temperature of the wettest quarter, mean temperature of the driest quarter, mean temperature of the warmest quarter, annual precipitation, precipitation seasonality) (Table 2.1). To estimate past species’ 34

distributions of the studied taxa, we used two global circulation model simulations (Model for Interdisciplinary Research on Climate (MIROC) (Hasumi 2007) and the Community

Climate System Model (CCSM) (Collins, et al. 2006)) of the Last Glacial Maximum available on WorldClim database (Hijmans, Cameron, Parra, Jones and Jarvis 2005).

Modeling approach

To reduce model inter-variation we followed the recommendations of Araujo, et al.

(2019) and produced consensus models using seven modeling approaches (including

Artificial Neural Network, Classification Tree Analysis, Generalized Linear Model,

Generalized Additive Model, MaxEnt, Multivariate Adaptive Regression Splines, and

Support Vector Machines) (Elith, et al. 2007, Phillips, et al. 2004, Schmitt, et al. 2017).

The background area was chosen to ensure the models were tuned based on species’ accessible area (Barve, et al. 2011) and we used minimum convex polygons in lack of a better proxy for shrimp dispersal. For spadefoot toads, we used their specific distribution ranges (IUCN 2016) (Figure 1). Background ensures the relatively accessible area of the studied species is taken into account when species distribution models are built, as recommended by previous literature (Barve, Barve, Jimenez-Valverde, Lira-Noriega,

Maher, Peterson, Soberon and Villalobos 2011, Cooper and Soberon 2018, Gherghel, et al.

2019). These settings were also used in a previous paper on spadefoot toads biogeography

(Gherghel and Martin 2020). Past distributions were estimated by transferring the model generated using the current climatic data over the climatic reconstructions corresponding to the Last Glacial Maximum (see Environmental data section). For each species studied

(see Model system and Species occurrences section above), the seven models were 35

assembled into one prediction using a weighted AUC approach, using a performance cutoff of AUC=<0.75 as a rule for models to be included in the final ensemble model for each species. As such only models with an AUC higher than 0.75 are included in the final species ensemble model. All species distribution models were produced with the SSDM package

(Schmitt, Pouteau, Justeau, Boissieu, Birnbaum and Golding 2017) in R. Individual models were evaluated using Kappa, Omission error, and AUC metric (the latter metric should be interpreted with greater skepticism – see Lobo, et al. (2008), Peterson, et al. (2008)) (Table

2.1). These metrics are calculated based on a confusion matrix (Fielding and Bell 1997) that looks at the relationship between the correctly predicted absences and presences based on a thresholded (binary, presence-absence) map (Fielding and Bell 1997, Peterson, Papes and Soberon 2008). In our study, all binary maps were produced at a threshold that maximizes the sensitivity to specificity equity as suggested by Liu, et al. (2016).

Post-processing methodology

To study the relationship between the probability of the presence of the spadefoot toads (predator) and the fairy shrimp (prey) species richness, the final models of species distribution of the studied species were post-processed in ArcGIS 10.6. The shrimp species richness map was created by summing the binary response (see Modelling approach section) of all 12 species of fairy shrimp modelled (Table 2.1). Binary stacks of species distribution models perform well in the context of predicting species assemblages (Zurell, et al. 2016, Zurell, et al. 2020) including for the task of assembling prey species richness in species distribution modeling studies (Gherghel, Brischoux and Papeş 2018). To study the whole range effects between the spadefoot toads and the fairy shrimp species richness, 36

we sampled every pixel predicted within the IUCN range of the two spadefoot toad species

(accessible at https://www.iucnredlist.org/resources/spatial-data-download). To study the occupancy level effects between the spadefoot toads and the fairy shrimp species richness, we sampled only the pixels corresponding to known occurrence records (see Species occurrence for details). All relationships between fairy shrimp species richness and spadefoot toad’ distributions were calculated using Pearson correlations coefficients in

JMP 14 and ArcGIS 10.6.

Results

Model performance summary

Our models predicted the potential distribution of both the fairy shrimp (AUC average

= 0.89, SD = 0.04; Omission average = 10%, SD = 4%; Kappa average = 0.56, SD = 0.12) and the spadefoot toad species (AUC average = 0.88, SD = 0.01; Omission average = 10%, SD =

0.005%; Kappa average = 0.75, SD = 0.01) very well (Table 2.1). The models for two fairy shrimp species performed poorly from the point of view of Cohen’s Kappa metric (0.37 for Artemia franciscana, and 0.44 for Thamnocephalus platyurus, respectively). However, we choose to keep these species in the final species richness model because they did well in predicting their known occurrences (83%) (Table 2.1). The models for the other fairy shrimp species and of the two spadefoot toad species are very robust (Table 2.1). The environmental variables that are influential to the distribution of fairy shrimp in North

America are the mean temperature of the wettest quarter (avg = 18.05%), the annual precipitation (avg = 21.3%) level, and the precipitation seasonality (avg = 16.61%) (Table

2.1); whereas the variables that seem to be the most influential to the distribution of Spea bombifrons and Spea multiplicata are the mean temperature of the wettest quarter (avg =

18.92%) and the annual precipitation (28.02%) level for S. multiplicata and minimum temperature of the coldest month (25.24%) for S. bombifrons (Table 2.1).

Fairy shrimp (prey) distribution patterns in the context of predator species ranges

The fairy shrimp species richness map reveals that the highest diversity of fairy shrimp in the studied area is Baja Peninsula, Mexico, California, along the Rio Grande river and along with the Sierra Oriental in Mexico (Figure 2.2), whereas the lowest diversity of shrimp diversity is along the Mississippi River floodplain (Figure 2.2). When we overlapped the binary (presence/absence) map of the predator species (S. multiplicata and S. bombifrons) we found that in areas where the two species of spadefoot toads are in sympatry, shrimp species richness is very high especially in areas such as New Mexico, southeastern Arizona, northern Mexico and western Texas (Figure 2.2). Overall, S. multiplicata has a complete range overlap with fairy shrimp species, whereas S. bombifrons lacks overlap with fairy shrimp as it gets closer to its eastern range limit (Figure 2.2).

The patterns during the Last Glacial Maximum are fairly similar, although the two climatic models produced some idiosyncratic answers. Namely, in the case of the prediction based on CCSM, areas of sympatry between the two spadefoot toads were also areas with high shrimp species richness, whereas the MIROC model shows a moderate number of shrimp species in the areas of sympatry (Figure 2.3). The two climatic models agree that a high number of shrimp species were available in the range of the two spadefoot toad species during the Last Glacial Maximum (Figure 2.3). 38

Correlation between the distribution of fairy shrimp and spadefoot toads across spatial scales

We found a generally strong correlation between the predicted potential distribution of spadefoot toads and fairy shrimp across all levels of their geographic range (Figure 2.4).

At the level of the entire range, we found a positive correlation between the probability of presence of spadefoot toads and shrimp species richness (S. bombifrons: whole range r =

0.27, p < 0.001, allopatry r = 0.17, p <0.001; S. multiplicata: whole range r = 0.5, p < 0.001, allopatry r = 0.48, p < 0.001). Although in sympatry the correlation between the probability of presence of S. bombifrons and shrimp species richness is negative and significant (r = -

0.13, p <0.001) the species can be found in areas with higher shrimp richness (Figure 4).

In the case of S. multiplicata, at the level of the species range, the relationship between probability of presence of the toad species and shrimp species richness was positive and stronger in areas of sympatry (r = 0.52, p < 0.001) (Figure 2.4). At the level of occupancy, we only found a strong correlation between the probability of presence of S. multiplicata and shrimp species richness (whole range: r = 0.45, p < 0.001, allopatry: r = 0.52, p <

0.001), whereas S. bombifrons is present at overall higher shrimp richness than S. multiplicata. However, the correlation between shrimp richness and the probability of the presence of S. bombifrons is overall negative (whole range: r = -0.24, p < 0.001, allopatry: r = -0.11, p = 0.13) (Figure 2.4). In areas of sympatry, similarly, with the patterns seen at the regional level, at the occupancy level we found that S. bombifrons inhabits areas with a higher number of shrimp species, and the overall direction of the correlation is negative

(r = -0.21, p <0.01).

We found a strong effect of geography on the direction and strength of the relationship between the probability of the presence of the predators (S. bombifrons or S. multiplicata) and fairy shrimp species richness (Figure 2.4). Generally, in areas where shrimp richness was low, the probability of presence for S. bombifrons was weak (Figure

2.4). In the case of S. multiplicata the relationship was stronger especially in the southernmost part of its range and more localized relationships were observed in the northern part of its range (Figure 5). In the case of S. bombifrons, the strongest relationships between the probability of the presence of the spadefoot toads and shrimp species richness appear to be present more at the western and southern edges of its geographic range (Figure

2.4). Also, in areas of sympatry between the two spadefoot toad species, the relationship between the probability of the presence of the toad species and shrimp species richness are highly variable spatially, with little clustering (Figure 2.4).

Discussion

Species distribution models are frequently used to study the importance of abiotic and biotic factors for determining species distributions (Araujo and Rozenfeld 2014).

Abiotic factors are often used to estimate species distribution models because many abiotic environmental datasets (Fick and Hijmans 2017, Hengl, et al. 2017, Hijmans, Cameron,

Parra, Jones and Jarvis 2005, Karger, et al. 2017, Tyberghein, et al. 2012) are available for such studies but no global datasets of species interactions exist, and so datasets of biotic interactions have to be constructed for each studied species (Gherghel, Brischoux and

Papeş 2018). Moreover, biotic interactions have been previously thought as less important for shaping species’ distributions at larger scales (Soberon 2007, Soberón and Nakamura 40

2009, Whittaker, Willis and Field 2001), and have been harder to include in species distribution models because of their dynamic nature (Araujo and Rozenfeld 2014, Soberon

2007, Soberón and Nakamura 2009). However, recent work found that biotic interactions are indeed important in shaping patterns of species distribution and co-occurrence

(Anderson, Peterson and Gómez-Laverde 2002, Anderson 2017, Araujo and Rozenfeld

2014, De Araújo, Marcondes-Machado and Costa 2014, Dormann, et al., 2020, Gherghel,

Brischoux and Papeş 2018, Jenkins, Lecomte, Andrews, Yannic and Schaefer 2020, Mod, le Roux, Guisan and Luoto 2015).

The effects of biotic interactions (e.g. predation, competition, etc.) on interacting species ranges across broad spatial scales can be inferred from patterns of co-occurrence between closely related species (Anderson 2017, Dormann, et al., 2018). However, we have a limited understanding of how biotic interactions are conserved through time. Here, we studied how the distributions of competitors and their prey covary in response to climate change oscillations at continental and regional scales using the closely related spadefoot toads Spea bombifrons and Spea multiplicata and fairy shrimp of central and western North

America (Figure 2.1, 2.2).

Presence of prey resources in time are important for patterns of predator co-occurrence

Abiotic niches of spadefoot toad and fairy shrimp species are very similar. We found that the amount of annual precipitation and the mean temperature of the wettest quarter are extremely important in determining the distributions of spadefoot toads and fairy shrimp (Table 2.1). Co-occurrence between the two spadefoot toad species has also been tied to the presence of fairy shrimp, allowing for character displacement and 41

preventing competitive exclusion (Pfennig and Murphy 2000, Pfennig, Rice and Martin

2006). Chunco, Jobe and Pfennig (2012) previously investigated patterns of the abiotic niche in shaping areas of sympatry and allopatry between S. multiplicata and S. bombifrons.

Chunco, Jobe and Pfennig (2012) found that the areas of sympatry between the two spadefoot toad species correspond to the most environmentally extreme areas (hotter and drier) of their range. Our results confirm this finding using similar modeling techniques

(see Table 2.1 and Figure 2.1).

Biotic interactions are important in shaping regional and local processes within the species fundamental niche. How these interactions vary spatially is still an open question, and by using spadefoot toads and fairy shrimp species we could study these regional effects

(Figure 2.3, Figure 2.4). In areas of allopatry, the presence of shrimp is likely the driver for maintaining the resource polymorphism in the spadefoot toad species (Figure 2.3 and

Figure 2.4). Areas of sympatry between S. bombifrons and S. multiplicata (western Texas,

New Mexico, and Arizona) are characterized by high suitability values for both spadefoot toad species and high shrimp species richness (Figure 2.1 and Figure 2.2), this creates the ecological opportunity for the two species to exploit alternative resources. Moreover, when we projected the models over the Last Glacial Maximum conditions, we found a similar pattern, where areas of sympatry between S. bombifrons and S. multiplicata are found in western Texas, New Mexico, Arizona, and northern Mexico, and are characterized by high shrimp richness (Figure 2.3). These patterns are also confirmed by phylogeographic analysis of the two species that also suggest that S. bombifrons is the species with a higher rate of range shift while S. multiplicata is more stable (Rice and Pfennig 2008). This

process of co-occurrence between fairy shrimp and spadefoot toads is most likely the driver for the patterns of co-occurrence in spadefoot toads via character displacement (Pfennig,

Rice and Martin 2006, Rice and Pfennig 2008). Hence, we can conclude that the area of sympatry of S. bombifrons and S. multiplicata likely remained geographically stable in time allowing for the potential of long-term biotic interactions between the species and on with other interacting members of the community (e.g. fairy shrimp). Alternatively, if conserved, the niches of the two predator species (S. bombifrons and S. multiplicata) would track in space the distribution of the fairy shrimp. Either of the two biotic interactions scenarious would occur inside the accessible fundamental niche, sensu Soberón and Peterson (2005), of the predator and their prey.

Spatial variation in prey-predator relationships

Mapping the spatial variation of predator-prey relationships can provide important insights into the local and regional processes that drive species distributions and co- occurrence (DeGregorio, et al. 2016, Hague, et al. 2020). Our results show that prey- predator relationships are highly variable across geography (Figure 2.4) which suggest that processes at the regional and local level are extremely important (Chunco, Jobe and

Pfennig 2012, Pfennig, Rice and Martin 2006). This is consistent with the resource polymorphism known to be present in spadefoot toads, wherein the presence of shrimp, both species are capable of producing a carnivorous morph to exploit this alternative high- protein resource (Pfennig 1990, Pfennig 1992). On the other hand, if shrimp abundance is low, then the default omnivorous tadpoles occur at higher frequencies than the carnivorous morph (Pfennig 1990, Pfennig 1992). However, our models do not include other resources 43

exploited by spadefoot toad larvae, such as algae or organic detritus, and consequently cannot evaluate their role in the patterns of species distributions modelled here.

Considering strong mosaic pattern at the level of the landscape over the entire range of the two species (Figure 2.1 and Figure 2.4) and the resource polymorphism present in Spea

(Pfennig 1990, Pfennig 1992), this indicates that co-evolution between spadefoot toads and their prey may also facilitate co-occurrence. The correlation between the predator probability of occurrence and prey species richness showed strong spatial variation in the strength of the correlation (Figure 2.4) which offers important insight that will allow the design of future field or lab experiments testing patterns of co-evolution between spadefoot toads and fairy shrimp. Although we know how the spadefoot toads respond to the presence of shrimp (Martin and Pfennig 2009, Martin and Pfennig 2010, Martin and Pfennig 2012,

Pfennig 1990, Pfennig 1992, Pomeroy 1982), we have no information on how the fairy shrimp may evolve in response to the presence of spadefoot toads. Future studies are required to better understand the targets of selection (if any) on the shrimp traits that might be hiding behind the mosaic variation present in the spadefoot toad – fairy shrimp system

(Figure 2.4).

Biotic interactions in species distribution modeling

Biotic interactions, together with abiotic conditions and dispersal are among the factors that determine the ecological niche, determining species distributions (Soberon

2007, Soberón and Nakamura 2009). However, prey-predator (including herbivory) relationships have only recently been included in species distribution modeling (Araujo and Rozenfeld 2014, De Araújo, Marcondes-Machado and Costa 2014, Gherghel, 44

Brischoux and Papeş 2018, Jenkins, Lecomte, Andrews, Yannic and Schaefer 2020).

Previous studies have shown that predators' dietary specialization is an important natural history trait likely to improve species distribution models of predators when prey data (such as prey species richness) are included (Gherghel, Brischoux and Papeş 2018). Our results add to these studies by demonstrating that studying prey-predator relationships on spadefoot toads, which evolved environmentally induced dietary specialization to consume fairy shrimp, produce robust and transferable models. Spadefoot toads and fairy shrimp overlap temporally and spatially during key developmental stages of both species, facilitating their co-occurrence (Bragg 1967, Pfennig 1990, Pomeroy 1982). This shared environmental niche increases the likelihood that predator-prey interactions between fairy shrimp and Spea tadpoles are conserved through geologic time, suggesting that species richness maps can be a good proxy for studying biotic interactions over time (Figure 2.3).

This suggests that the power of species distribution estimates, when transferred on different time slices (in our case, Last Glacial Maximum), are very robust when the predator’s exhibit dietary specialization. This insight further provides evidence on the importance of including biotic interactions in species distribution models by suggesting that these interactions can be projected in time and space using existing correlative species distribution modeling tools.

Table 2.1 – Model performance summary and variable contributions for the generated models

Mean Min Mean Mean 10% Mean Tempe AU Coh Tempe Tempe Tempe Precipi traini Diurnal rature Annual C en's Isother rature rature rature tation Species ng Tempe of Precipi train Kap mality of of of Season omis rature Warme tation ing pa Coldest Wettest Driest ality sion Range st Month Quarter Quarter Quarter

Artemia 0.84 0.17 0.37 7.99 5.70 12.76 8.61 5.76 10.05 37.29 11.85 franciscana Branchinecta 0.91 0.07 0.52 9.91 11.11 5.15 41.00 11.78 4.39 10.16 6.50 coloradensis Branchinecta 0.89 0.11 0.64 7.09 8.31 4.25 36.51 5.66 7.12 21.82 9.24 lindahli Branchinecta 0.93 0.08 0.68 9.01 6.69 8.45 18.64 16.36 10.37 10.00 20.48 lynchi Branchinecta 0.95 0.06 0.71 6.37 7.17 7.34 25.45 14.21 9.32 21.19 8.95 mackini Branchinecta 0.88 0.05 0.46 23.92 10.51 11.64 9.75 5.19 9.97 5.55 23.48 packardi Streptocephalus 0.85 0.10 0.46 6.26 6.05 5.96 12.02 6.59 5.35 17.97 39.81 dorothae Streptocephalus 0.92 0.07 0.73 18.56 11.13 10.21 10.66 13.59 9.94 10.67 15.25 mackini Streptocephalus 0.90 0.07 0.63 9.99 10.73 7.65 11.20 8.33 9.68 21.25 21.17 similis Streptocephalus 0.90 0.08 0.51 14.00 6.47 7.77 10.10 4.38 6.19 37.21 13.88 texanus Thamnocephalus 0.91 0.13 0.63 8.16 12.82 5.35 25.79 5.32 3.97 21.32 17.26 mexicanus Thamnocephalus 0.84 0.16 0.44 4.74 7.75 7.09 6.89 16.53 4.42 41.13 11.44 platyurus Spea bombifrons 0.88 0.09 0.74 13.29 5.39 25.24 13.99 8.36 14.95 7.81 10.97 Spea 0.88 0.10 0.76 6.57 13.54 4.80 23.84 11.56 7.22 28.02 4.46 multiplicata

Figure 2.1 – Spea bombifrons and Spea multiplicata species distribution and the occurrence records (right panel) used to analyze and estimate the potential distribution of the two species during present climatic conditions (right panel).

Figure 2.2 – Distribution and the estimated species richness of the fairy shrimp species used in our study overlapped over the predicted binary estimated distributions of the two spadefoot toads. Species occurrence data is adapted from (Rogers 2014a, Rogers 2014b).

Figure 2.3 – Potential distribution of Spea bombifrons and Spea multiplicata during the Last Glacial Maximum (~20,000 years ago), and the distribution patterns of the estimated fairy shrimp species richness during over the range of the two species. 49

Figure 2.4 – Geographic variation in the strength of interaction between fairy shrimp and the predicted suitability of the two predator species (Spea bombifrons and Spea multiplicata)

CHAPTER 3. THE ECOLOGICAL AND SELECTIVE EFFECTS OF RESOURCE

POLYPHENISM ON LOWER TROPHIC LEVELS

Abstract

Interspecific variation of traits has been found to have ecological consequences at the level of community and ecosystem dynamics. Particularly in case of resource polyphenism, that allows in certain system differential use of resources by shifting the phenotype plastically. Mexican spadefoot toads (Spea multiplicata) tadpoles have the plastic ability to shift their phenotype to a carnivorous morph as a response to the breeding pond community structure (higher shrimp densities, more tadpoles with carnivore morph), while the default, omnivorous morph, remains in environments if shrimp is less abundant or detritus is plenty in the breeding ponds. In this chapter, I evaluated whether the differential use of resources of the two morphs impose different selection pressures on fairy shrimp traits. We found a significant effect of trophic polyphenism on the body size of fairy shrimp, which differs if tadpoles are present, and have a positive selection pressure on the body sizes but that is generally only marginal. Selection on the traits involved in reproduction have conflicting directions, where males with smaller tusks and females with larger ovisac length are favored in tadpole treatments, the magnitude being higher in case of the omnivore treatment. At the level of the ecosystem, we found that the productivity was significantly affected by the presence of the polyphenic tadpoles. Our experiments suggest that the omnivore morph, because it competes with the fairy shrimp and also eats

smaller sized fairy shrimp, would produce a higher ecosystem level change, than the carnivore morph. These co-evolutionary effects at the level of the ecosystem are particularly important to be understood especially in the context of fast changing environments, especially as the process of global change is accelerating.

Introduction

The hypothesis that evolutionary change can influence ecological dynamics, just as ecological change influences evolution (i.e. eco-evolutionary dynamics) has gained considerable support since its conception, due in part to changing perceptions regarding

(1) the speed of evolutionary change in nature; and (2) the ecological importance of intraspecific variation (Hendry 2017, 2019). Evolution was long thought to be a gradual process, generally unfolding too slowly to affect the ecological dynamics of populations and communities (Hendry 2017, Reznick et al 2019). This perception has now shifted, and examples of rapid (or contemporary) evolution are now commonplace (reviewed in

Thompson 1998, Hendry and Kinneson 1999, Reznick et al 2019). Similarly, the effects of trait variation within species on ecological interactions was once largely overlooked

(Bolnick et al. 2003). However, we now know that variation within species and population can affect ecological, community and ecosystem dynamics to the same extent as variation between species (reviewed in Bolnick et al. 2003, Des Roches et al. 2018, Hughes et al

2008). Together, these facts suggest that evolutionary change in phenotypic and genetic variation often occurs at similar timescales as those of ecological dynamics, allowing the

two processes to influence and feedback on one another (Hendry 2017). And, as predicted, we now have many empirical examples of evolutionary and ecological change interacting in natural populations (reviewed in Fussmann et al. 2007, Post and Palkovacs 2009, Hendry

2017, Pelletier et al 2009; Reznick and Travis 2019).

While eco-evolutionary dynamics has grown into a thriving area of research, another source of intraspecific phenotypic variation and change, phenotypic plasticity, has received little attention in the context of driving eco-evolutionary dynamics (Hendy 2016,

Skúlason et al 2019). Phenotypic plasticity — the expression of multiple phenotypes by an individual genotype in response to environmental variation — can itself evolve, and is likely a fundamental consequence of organismal development (West Eberhard 2003).

Phenotypic plasticity can also manifest at very short timescales, much faster than the generational limit of evolution, but can also cause phenotypic changes that rival those seen between long-diverged species (Pfennig et al 2010). Due to the magnitude and rapid pace of phenotypic change it can induce, plasticity is likely an important cause of eco- evolutionary dynamics itself (Hendry 2016, Skúlason et al 2019) but empirical examples are scarce. For one example, the foraging preferences of benthic and limnetic whitefish species were altered when reared on alternative diets in mesocosms, affecting prey abundance at lower trophic levels (Lundsgraard-Hansen et al 2014). And in red alder, induced defenses caused by herbivory (e.g. secondary metabolite production) suppresses leaf litter decomposition in adjacent streams (Jackrel and Morton 2018).

The developmental plasticity in resource specialization of spadefoot toad tadpoles

(Spea) is likely another example of plasticity led eco-evolutionary dynamics. In brief,

within their ephemeral pond habitats, Spea tadpoles can develop into either of two resource-use morphs (i.e. a resource polyphenism): a default omnivorous form or a carnivorous form induced by the consumption of fairy shrimp or other tadpoles (Pfennig

1990, 1992, Levi et al 2015). Although previous research in this system was not conducted in the context of eco-evolutionary dynamics, there is already strong evidence that this developmental plasticity drives ecological effects at lower trophic levels, which may feedback to influence evolutionary change. For example, morph frequency and morph fitness within ponds is determined by ecological factors (i.e., tadpole density, the relative availability of fairy shrimp and detrital resources, and pond ephemerality [Pfennig1992,

Martin and Pfennig 2010, Pfennig and Pfennig 2019]). Moreover, the tadpoles’ morphs differ substantially in their resource-use and competitive ability. Carnivore morphs are much more efficient foragers for fairy shrimp (Martin and Pfennig 2009), and indeed fairy shrimp abundance declines much faster in ponds with higher frequencies of the carnivore morph (Pfennig 1992). In contrast, omnivores are much better competitors for detrital resources (Martin and Pfennig 2009). In this study, we aim to build on this earlier work to evaluate (1) if the alternative resource-use morphs impose different selection pressures on fairy shrimp traits; and (2) if the resource-use morphs also influence lower trophic levels

(e.g., primary productivity). We use manipulative experiments in microcosm and mesocosm habitats to investigate these questions.

Materials and Methods

Study system

Mexican spadefoot toads (Spea multiplicata) inhabit arid areas across the

Southwestern United States and central Mexico (Gherghel and Martin 2020) and breed during the summer monsoons that fill depressions, ditches and cattle tanks, forming ephemeral pools that are highly exposed to solar radiation and are generally hot, fast drying habitats, often lasting for only a few weeks (Pomeroy 1982, Pfennig 1990). Consequently, spadefoot toad tadpoles have evolved rapid development rates (they can reach metamorphosis in two weeks from hatching) (Pomeroy 1982, Pfennig 1990). Spea tadpoles are characterized by a larval polymorphism where two morphs co-occur in wild populations, the carnivore morph and the omnivore morph (Pomeroy 1982, MacKay et al.

1990). Pfennig (1992) showed experimentally that Mexican spadefoot toad tadpoles (Spea multiplicata) that do not ingest fairy shrimp only develop into the omnivore morph; the carnivore morph requiring ingesting shrimp (or other tadpoles Levis et al 2015). Later,

Martin and Pfennig (2009) showed that the carnivore morph is more efficient at eating shrimp than the omnivore morph, while the omnivore morph achieves higher growth on a detritus diet. The alternative, carnivore morph, is different from the default omnivore morph, in a few key traits: carnivores develop an exaggerated orbitohyoideus muscle increasing facilitating capture of large prey; develop shorter intestines compared to long intestines of the omnivore morph; develop augmented keratin buccal plates; and; in the wild, carnivores are generally more active swimmers, while omnivores forage more often at the shallow edge of ponds (Pomeroy 1982, Pfennig 1990, Martin and Pfennig 2009,

Levis et al. 2018). Moreover, carnivores also develop faster than omnivores, but store less fatty reserves post-metamorphosis (Pfennig et al 1991). This divergent morphology leads

to divergent resource-use and performance between the morphs. The omnivore morph is a generalist, primarily consuming detritus, algae and small zooplankton, while the carnivore morph specializes on fairy shrimp and other tadpoles (Pomeroy 1982, Paul et al 2012).

Fairy shrimp (Crustacea: Anostraca) commonly found in the breeding pools used by Mexican spadefoot toads are generally from genera Streptocephalus sp., and

Thamnocephalus sp (Pomeroy 1982, MacKay et al. 1990, Pfennig 1990). They are filter feeders and occur high abundances. In the natural ponds in our study (Figure 3.1),

Streptocephalus texanus is the most common species that are also regularly consumed by tadpoles. Thamncephalus sp. is often eaten too, but their densities lower (MacKay et al.

1990). Like spadefoot toad larvae, fairy shrimps exhibit adaptive phenotypic plasticity and can alter the timing of hatching and their rate of growth in response to environmental cues

(e.g. pond formation and drying Belk 1975, MacKay et al. 1990).

The Greater Plains fairy shrimp (Streptocephalus texanus) is extremely abundant in our study area, and we also use this species for all of our experiments. Males are larger in this species, and in the wild, the sex ration is skewed with females being more abundant than males (MacKay et al. 1990). Breeding is done via amplexus which lasts for a few seconds, where the male approaches the female from above and then drops below the female and uses their tusks to amplex the female (Belk 1975, Rogers 2002). Generally, males with larger tusks have higher breeding success, and these appendages are also a signal of male quality (Belk 1975, MacKay et al. 1990, Rogers 2002). Females lay eggs that after the wetland dries, become stored in the topsoil and enter a dormancy period during the time that the wetland is dry. This period of dormancy is often required for the

eggs to hatch when the pond gets filled again by the seasonal rains when eggs of multiple years’ eggs generation will hatch (Belk 1975, Rogers 2002). The density of fairy shrimps in wetlands can vary upon environmental conditions, however, like the spadefoot toad tadpoles, the fairy shrimps are highly sensitive to water level and water temperature (Belk

1975, Rogers 2002).

Effects of resource polyphenism

Spea multiplcata developmental plasticity may cause significant ecological and evolutionary consequences for their ecosystem. For example, the resource polyphenism present in the Mexican spadefoot toads may lead to differential selection on fairy shrimp body size, female ovisac size and male tusk size. These traits might face contrasting selective pressures by the carnivores and omnivores which can in turn might have cascading consequences for the ephemeral pond community. To evaluate whether these traits are being affected by trophic polyphenism we designed a series of microcosm and mesocosm experiments where we tested whether a) carnivore morph frequency (itself controlled by the density of fairy shrimp) could increase the cost of and tradeoffs of reproductive traits b) tusk size may face contrasting selection pressures where the trait has a natural selection cost, c) ovisac size can also tradeoff between fecundity (females’ larger ovisacs lay more eggs), and d) how selection on body size could influence ecosystem level functioning where shrimp of smaller size could potentially lead to less phytoplankton consumption.

Microcosm experiment 57

To evaluate if Spea multiplicata tadpoles impose selection on fairy shrimp traits, and if the tadpole morphs impose different selective pressures, we conducted shoebox, microcosm experiments in the laboratory. We collected 378 tadpoles (186 of carnivore morph, and 192 of omnivore morph) of Mexican spadefoot toad (Spea multiplicata) from ephemeral ponds near Portal, Arizona in July 2016 (Figure 3.1). We chose tadpoles for the experiment that were unambiguous in their morphotype (i.e. unambiguous omnivores and carnivores through visual inspection). The fairy shrimp used in the microcosm experiment were Greater Plains fairy shrimp (Streptocephalus texanus) collected from an ephemeral pond along Highway 9, seven miles west of Animas, New Mexico (Figure 3.1).

The tadpoles and fairy shrimp for each days’ experiments were collected the evening before to allow the tadpoles and fairy shrimp to acclimate to the laboratory setting and to ensure they were healthy before the trials. The experiments were held at the

Southwestern Research Station (SWRS) in Portal Arizona. The experiments consisted of selection feeding trials with three treatments, where 6 tadpoles (either 6 carnivores, 6 omnivores, and a mix of 3 carnivores 3 omnivores) were placed in 6-liter volume plastic arenas filled with dechlorinated well water, together with 100 Greater Plains fairy shrimp.

The arenas were then placed on a wooden rack, randomly with regards to treatment to avoid any effect of position on the behavior of the tadpoles or fairy shrimp. The experiment was terminated each day when 50% of the shrimp were eaten by the tadpoles in each of the arenas. A control treatment consisting of 100 Greater Plains fairy shrimp was kept each day to calculate selection differentials of each treatment on body size and tusk size of the

fairy shrimp. Replicates in the control treatment were treated identically to the experimental treatment, with the important exception that no tadpoles were added.

A total of 84 replicates of the three treatments and controls were run (20 replicates for carnivore treatment, 22 replicates for the carnivore-omnivore treatment, 21 replicates of omnivore treatment and 21 controls). At the end of each feeding trials, the remaining fairy shrimp were preserved in 95% Ethanol for later measurement. The tadpoles used in the experiments were euthanized by immersion in buffered 0.1% solution of tricane methanesulfonate (MS 222), and then preserved in 95 % Ethanol.

To measure the shrimp body size, we placed each shrimp on one side along with a

1 mm scale for reference. We then used a Nikon D610 with a Tamron 100mm f2.8 on a copy stand in the lab, we took the pictures of the shrimps prior measuring in Image J. We measured shrimp body size (from the forehead, along the thorax and tail to the end of organism’s furca), and we take into account the sex of the individual (male or female). A total of 6138 shrimp individuals were measured from microcosm experiments. To measure the male shrimp tusks, we separated out the males from the experiment, dissected the tusks of each male using a Leica Stereomicroscope, and we photograph the tusk alongside the body of the male and then in Image J we measured the total length of the tusks of the male fairy shrimp. A total of 1626 male shrimp were dissected and their body and tusk sizes were measured.

Mesocosm experiment

We used 800-liter wading pools that were filled with 20-liter volume each of surface topsoil from an ephemeral pond nearby along Highway 9, seven miles west of

Animas, New Mexico (Figure 3.1). We then filled the pools with dechlorinated well water

3 days before we started the experiment to ensure that shrimp would hatch from the eggs present in topsoil (see Study system above). In the same pond where we collected the topsoil used in our mesocosms, together with the Greater Plains fairy shrimp (Streptocephalus texanus), Beaver-tail fairy shrimp (Thamnocephalus platyurus) and the American tadpole shrimp (Triops longicaudatus) are also known to occur in the area (Belk 1975, Rogers

2014a, b, Personal observation) and were expected to hatch from the topsoil in the pools.

Before we started the experiment we supplemented the pools with a 5-liter volume of

Greater Plains fairy shrimp caught with a dip net from the same pond. In total we used 12 mesocosms with three treatments (3 mesocosms with carnivores, 3 mesocosms with omnivores, and 3 mesocosms with a half-mixture of carnivores and omnivores) and control

(no tadpoles). In each pool we added 32 tadpoles in total. For the mesocosm experiment we used a total of 288 tadpoles (144 of carnivore morph, and 144 of omnivore morph) of

Mexican spadefoot toad (Spea multiplicata) from an ephemeral pond in Price Canyon, near

Apache, Arizona on July 2017 (Figure 3.1). The tadpoles were collected the day before the mesocosm experiment started, and we selected only the tadpoles that had the strongest morphological traits corresponding to carnivore and omnivore morphs (See Model system above). Throughout the experiment we measured the Chlorophyll and oxygen concentration in the pools as well as turbidity using an YSI EXO1 Multiparameter Sonde, twice every day (morning and evening) during the experiment. The experiment was

terminated when the tadpoles reached metamorphosis (front legs developed). Once the experiment was terminated, the tadpoles were removed and euthanized by immersion in a buffered 0.1% solution of tricane methanesulfonate (MS 222), and then we in 95% Ethanol.

The remaining shrimp were also collected and preserved in 95% Ethanol for measurement.

To measure shrimp body size and ovisac size, we placed each shrimp on one side along with a 1 mm scale for reference. The pictures were taken using a Nikon D610 with a

Tamron 100mm f2.8 on a copy stand, and then the images were processed in Image J. In

Image J, we measured the body size of each shrimp from the forehead, along the thorax and tail to the end of the furca. A total of 1037 shrimp individuals were measured.

Statistical analyses

Microcosm selection experiment

We calculated variance standardized selection differentials to measure the magnitude and direction of selection imposed by Spea tadpole morphotypes on fairy shrimp. We aimed to estimate selection acting on three traits, fairy shrimp body size, male tusk size, and female ovisac size (all measured as length). We took the same approach to estimate selection acting on each trait. Specifically, we calculated standardized selection differentials by comparing the mean trait value of the control treatments to the mean of each experimental replicate, separately for each day the experiment was run.

For tusk and ovisac size, we first corrected for body size by taking the residuals of each trait separately regressed onto total body length using a mixed effects model ({lme4} library (Bates et al. 2015)) including day as a random effect. Tusk size was averaged across both tusks. For each experimental replicate we then calculated the mean trait value, its 61

standard deviation and its standard error. We did the same for each control replicate. For days with multiple control replicates we calculated an overall mean, and a pooled standard deviation and standard error. We then calculated the selection differentials for each trait as the mean for each experimental treatment minus the control mean for each day, variance standardized by dividing by the pooled standard deviation of the control.

We evaluated the hypothesis that the magnitude and direction of selection differed between our tadpole treatments by fitting mixed-effect meta-analytic models for each trait using the rma.mv function in the {metafor}library (Viechtbauer 2010). For each model we included the selection differentials for each experimental replicate as the dependent variable, tadpole treatment as our predictor variable, a random effect of day to account for shared environmental factors among the replicates run on the same day, and we included the standard errors of the selection coefficients sampling error. We excluded replicates with fewer than 6 measured individuals (the results did not qualitatively change if only replicates with at least 10 measured individuals were included).

Mesocosm experiment

Relative primary productivity. To test whether the larval polymorphism has consequences on primary productivity in the experimental mesocosms, we used a repeated- measurements linear model where the natural logarithm of relative chlorophyll concentration measured in each pool. Together with the fixed effect of treatment, we controlled for the effect of time of the day of each measurement (as a covariate) and for the non-independence of measures within the same pool by adding a random effect of pool identity. To evaluate whether the effect of the tadpoles (treatment) changed in the pools 62

across the duration of the experiment we converted our dates to Julian days, and included this term as a fixed effect in our model. To evaluate the significance of our fixed effects we used the Kenward-Roger denominator degrees of freedom approximation. All analyses were done in R 3.6.1 using “{me4}, {emmeans} and {lmerTest} library (Bates et al. 2014,

Kuznetsova et al. 2017, Lenth et al. 2018). The visualization of the results was done using the {visreg} library (Breheny et al. 2020).

Fairy shrimp traits. To evaluate if tadpoles exerted similar selective pressures on fairy shrimp traits in the larger, less artificial and longer duration mesocosms, as they did in the microcosm selection experiments, we fit separate linear mixed models for the two traits we had measurements for, body size (total length) and female ovisac size. For ovisac size, we fit a model with tadpole treatment, body size, and their interaction as fixed effects, and a random effect of mesocosm identity. For body size, we first fit a model with tadpole treatment as our fixed effect and mesocosm identity as a random effect. We then fit a second model evaluating the hypothesis that body size differed between the control treatment and mesocosms with tadpoles, considered as a single treatment. We fit our models using the {lme4} library (Bates et al. 2015) and evaluated the significance of fixed effects using the Satterthwaite denominator degrees of freedom approximation with the

{lmerTest} library (Kuznetsova, Brockhoff, & Christensen, 2017). We performed all statistical analyses using R 4.0.2 (R Core Team, 2020)

Results

Microcosm selection experiment

Selection on body size.

We found significant differences among all tadpole treatments in the magnitude of selection on body size (QM = 346.561, df = 3, P < 0.0001, Figure 3.2, Appendix 3 - Table

S3.1) with the largest magnitude of selection in the Omnivore treatment and the weakest in the Carnivore & Omnivore treatment. Selection was positive but not significantly different from zero for the Carnivore (S = 0.288 ± 0.24 se, z = 1.192, P = 0.233), Carnivore

& Omnivore (S = 0.253 ± 0.24 se, z = 1.047, P = 0.295) and Omnivore treatments (S =

0.374 ± 0.24 se, z = 1.550, P = 0.121).

Selection on male tusk size.

We found significant differences among all tadpole treatments in the direction and magnitude of selection on relative male tusk size (QM = 94.81, df = 3, P < 0.0001, Figure

3.3, Appendix 3 - Table S3.2). Selection was strongest in the Carnivore treatment, negative, and marginally different from zero (S = -0.209 ± 0.117 se, z = -1.786, P = 0.074), intermediate and not different from zero in the Carnivore & Omnivore treatment (S = -

0.151 ± 0.117 se, z = -1.285, P = 0.199) and weakest and not different from zero in the

Omnivore treatment (S = 0.0127 ± 0.117 se, z = 0.108, P = 0.914).

Selection on female ovisac size.

We found significant differences tadpole treatments in the magnitude of selection on relative female ovisac size (QM = 91.655, df = 3, P < 0.0001, Figure 3.4, Appendix 3 -

Table S3.3) with the largest magnitude of selection in the Omnivore and Carnivore &

Omnivore treatments and the weakest in the Carnivore treatment. Selection was positive

and marginally different from zero in the Omnivore treatment (S = 0.31 ± 0.169 se, z =

1.835, P = 0.067), and positive but not different from zero in the Carnivore & Omnivore

(S = 0.27 ± 0.169 se, z = 1.6, P = 0.11) and Carnivore treatments (S = 0.02 ± 0.169 se, z =

0.905, P = 0.351).

Mesocosm experiment

Relative primary productivity.

In our mesocosm experiment, we found that the productivity of the experimental ecosystem (relative chlorophyll concentration) was significantly influenced by the effect of the tadpole treatment (F1,8.275 = 6,4336, p = 0.014) (Figure 3.5), the effect of Julian date was significant as well (F1,100.38 = 27.7086, p < 0.001), and the interaction between Julian date and the effect of tadpoles (F1,100.05 = 2.1152, p = 0.005) was also significant. In the post-hoc analysis, we found statistically significant differences between control and the mixed (mesocosms with a mixture of carnivore and omnivore tadpoles) and control and omnivores (see Appendix 3, Table S3.4). Differences among treatment levels were not statistically significant, but there was a trend of increased relative chlorophyll levels in treatments with carnivores compared with the control (Figure 3.5). We found that 4 days after adding tadpoles to the mesocosms, chlorophyll concentrations diverged enough to be statistically significantly different between each other (Figure 3.6) (see Appendix 3, Table

S3.5). These differences were then maintained until the experiment was terminated (Figure

3.6).

Body size.

Body size did not significantly differ among treatments (F3,7.983 = 0.658, P = 0.6), nor did it significantly differ between the control treatment and the mesocosms with tadpoles, evaluated as a single treatment (F1,10.245 = 2.152, P = 0.172) (see Appendix 3,

Table S3.6). However, the estimated means suggest a trend for larger shrimp body sizes in treatments with tadpoles (Fig. 3.7).

Female ovisac size.

We found a significant interaction between treatment and body size, indicating the relationship between ovisac size and body size differed among treatments (F3,901.84 = 6.636,

P = 0.0.002). Specifically, the slope between ovisac and body size was greater in Omnivore, and in Omnivore & Carnivore treatments than in the tadpole free control treatments (see

Appendix 3, Table S3.7).

Discussion

Variation within species and populations can act on community and ecosystem dynamics to the same extent as variation between species (reviewed in Bolnick et al. 2003,

Des Roches et al. 2018, Hughes et al 2008). This suggests that within the same habitat, different frequencies of phenotypes would cause different ecosystem dynamics specific to the level of specialization of the morphotype in case of plastic trophic polymorphisms.

These different morphs might lead to different selection pressures on shrimp body size, male ornaments and female ovisac length. These could potentially feedback onto how the ecosystem is functioning as a function of frequency of the morphs (Hendy 2016, Skúlason et al 2019). In our paper we used a combination of microcosm and mesocosm experiments 66

to test whether the alternative resource-use morphs impose different section pressures fairy shrimp traits and whether the alternative carnivorous morph has effects on the productivity of the ecosystem.

Effects of trophic plasticity on the strength of selection

Alternative resource-use morphs impose different selection pressures on fairy shrimp traits. We found a significant effect on the magnitude of selection of the body sizes in fairy shrimps from the microcosm experiment. In the case of the mesocosm experiment, we found no significant differences between treatments. Generally, shrimp body sizes were larger in pools that had tadpoles than in the no tadpole control group. Congruently, the two experiments showed that the body size of fairy shrimp differs if tadpoles are present, and have a positive selection pressure on body sizes, but that effect is generally only marginal.

Previous work by MacKay et al. 1990 and by Pfennig (1992), found that higher frequencies of the carnivore morph in ponds, lead to significant decreases in shrimp densities, but in a novel finding, we show that shrimp body sizes shift in presence of spadefoot tadpoles due to selection by the predators. This opens a new opportunity to better understand the evolutionary dynamics between spadefoot toads and the fairy shrimp.

Effects of trophic plasticity on the reproductive traits

Intraspecific trait variation, especially sexual ornaments has been the focus of much evolutionary research (Jones and Ratterman 2009). In organisms with female choice, many males use ornaments to increase their chances of mating (Jones and Ratterman 2009).

Conflicting tradeoffs between selection on body size and sexual ornaments are often described in nature, where males with larger ornaments are more prone to predation

(Hernandez-Jimenez and Rios-Cardenas 2012). Alternative resource-use morphs can be expected to impose different selection pressures on fairy shrimp sexual traits. In case of the selection posed by the different tadpole morphotype treatments on the tusk size of the male fairy shrimp, we found that significant negative selection between the three tadpole treatments. The strongest positive selection was found in case of the carnivore treatment, although only marginally different from zero. In case of the mosquitofish (Gambusia sp.), predation regimes significantly affects the length of the gonopodium, favoring smaller gonopodia in males, while males with longer gonopodia have higher mating success in the absence of predators (Langerhans et al. 2005). Similarly, in our case, the tadpole morphs impose different selective pressures, and the carnivore morph favors smaller male tusk sizes. In contrast, for female ovisac size, which is correlated with the number of eggs that can be laid by a female (Prophet 1963), we found a trend where females in the omnivore treatment had larger ovisacs. This result suggests that omnivores may indirectly select for females with higher fecundity. Consequently, we found that selection on the male and female traits involved in reproduction have conflicting selection pressures where males with smaller tusks and females with larger ovisac length are favored in different tadpole treatments. Taken holistically, since we know that in the wild high shrimp densities induce higher carnivore morph frequencies in ponds, shrimp densities may then indirectly cause altered selection on fairy shrimp body size and sexually selected traits through carnivore morph induction. These different predation regimes, if stable through time in the breeding

ponds might lead to local adaptation as a response to the frequency of carnivores and omnivores on the length of the tusk sizes in males and the length of the ovisac in females.

Effects of trophic plasticity on ecosystem functioning

Phenotypic plasticity has been traditionally overlooked as a factor for driving eco- evolutionary dynamics (Hendry 2016; Skúlason et al. 2019). Phenotypic plasticity manifests at shorter time scales within population and has the ability to affect phenotypic shifts allowing rapid shifts in trophic resource use (sometimes, including carnivory, see

Levis et al. 2015), expanding their niche breadth and causing changes in community and ecosystem functioning. In our mesocosm experiments, we found that the productivity of the ecosystem was significantly affected by the presence of the tadpoles (Figure 3.5, Figure

3.6). On the community level, previous work found that Spea tadpoles have significant predatory effects on the fairy shrimp (McKay et al. 1990, Pfennig 1990, 1992). MacKay et al. (1990) suggested that both morphs eat shrimp, however the carnivorous morph has a greater ecological effect on fairy shrimp populations. In our results, we found that the omnivore and a mixture between omnivores and carnivores also cause cascading changes in productivity of the mesocosms. However, this result was also surprising, as we expected that the carnivore morph would have a greater effect on productivity than the omnivore morph, due to their trophic specialization (Pfennig 1990, 1992, Martin and Pfennig 2009).

On the other hand, considering that productivity in mesocosms with carnivores did not differ from those without tadpoles, and that in the body size selection experiments, tadpoles did not change body size distributions in fairy shrimp population, we suspect this is due to

the high efficiency at which the carnivore morph is consuming fairy shrimp. Our experiments suggest that the omnivore morph, caused a greater change in primary productivity. In the case of whitefish species, foraging preferences of benthic or lentic morphs were altered when reared on alternative diets, which had consequences on the lower trophic levels (Lundsgraard-Hansen et al 2004). It is also possible that local adaptation and coevolution between the fairy shrimp and the spadefoot toad tadpoles is also at play. One prediction that can be made is that in wild ponds where omnivore morph is consistently more frequent, shrimp locally adapt to better compete with this common morph. As such, it is puzzling that the frequency of the carnivore morph increases with fairy shrimp densities (Pfennig 1990, 1992), and high carnivore frequencies subsequently consume more shrimp (Martin and Pfennig 2009), but its presence in the pond seems to inflict little change to productivity. These co-evolutionary effects at the level of the ecosystem are particularly important for future study, especially in the context of fast changing environments, and as the process of global change is accelerating.

Figure 3.1: Presentation of the study area and the locations of the ephemeral ponds where the shrimp and tadpoles were sourced and the research station where the microcosm and mesocosm experiments have taken place.

Figure 3.2: - Directional selection on the total length of the fairy shrimp

Figure 3.3: Directional selection on the fairy shrimp male tusk sizes

Figure 3.4: Directional selection on the female ovisac length of the fairy shrimp

Figure 3.5: The effect of treatment on the productivity (Chlorophyll) of the mesocosm ecosystems.

Figure 3.6: Effect of treatment and Julian date on the evolution of productivity (Chlorophyll) of the mesocosm ecosystems.

Figure 3.7: Shrimp body total body size in the mesocosm pools

Appendix for Chapter 3

Table S3.1: Post-hoc tests for testing differences in differential coefficients between body- shrimp size in the tadpole treatments Treatment Estimate Std. Error z value p omnivores 0 -0.086 0.007 -11.685 <0.001 carnivores and omnivores 0 -0.121 0.007 -18.214 <0.001 carnivores 0 -0.035 0.007 -4.807 <0.001

Table S3.2: Post-hoc tests for testing differences in differential coefficients between male tusks sizes in the tadpole treatments Treatment Estimate Std. Error z value p omnivores 0 -0.235 0.0206 -11.4470 <0.001 carnivores and omnivores 0 -0.152 0.0210 -7.2440 <0.001 carnivores 0 0.083 0.0206 4.0250 <0.001

Table S3.3: Post-hoc tests for testing differences in differential coefficients between female ovisac length in the tadpole treatments Treatment Estimate Std. Error z value p omnivores 0 -0.290 0.033 -8.865 <0.001 carnivores and omnivores 0 -0.040 0.032 -1.267 0.414 carnivores 0 0.250 0.033 7.631 <0.001

Table S3.4: Post-hoc tests (estimates, standard errors, degrees of freedom, test ratio and p- values) for the treatment tadpole models of Chlorophyll Treatment estimate SE df t ratio p c - co -0.1573 0.0836 8.28 -1.881 0.3053 c - control 0.2274 0.0837 8.29 2.717 0.0969 c - o -0.0697 0.0836 8.26 -0.833 0.8376

co - control 0.3847 0.0837 8.29 4.597 0.007 co - o 0.0877 0.0836 8.26 1.048 0.7276

control - o -0.297 0.0837 8.28 -3.551 0.0294 c = carnivores; co = carnivores and omnivores; o = omnivores

Table S3.5: Post-hoc tests (estimates, standard errors, degrees of freedom, test ratio and p- values) for the treatment tadpole models of Chlorophyll thought each day of the mesocosm experiment.

Treatment estimate SE df t ratio p

c - co -0.0165 0.13 48.7 -0.128 0.9992

c - control 0.0479 0.13 48.7 0.37 0.9825

c - o -0.1322 0.13 48.7 -1.021 0.7383 Julian date 211 co - control 0.0645 0.13 48.7 0.498 0.9592

co - o -0.1157 0.13 48.7 -0.893 0.8085

control - o -0.1801 0.13 48.7 -1.391 0.5112

c - co 0.0512 0.13 48.7 0.395 0.9789

c - control 0.1702 0.13 48.7 1.314 0.5588

c - o -0.038 0.13 48.7 -0.293 0.9911 Julian date 212 co - control 0.119 0.13 48.7 0.919 0.795

co - o -0.0892 0.13 48.7 -0.688 0.901

control - o -0.2082 0.13 48.7 -1.607 0.3842

c - co 0.0694 0.13 48.7 0.536 0.9499

c - control 0.1347 0.13 48.7 1.04 0.727

c - o -0.0189 0.13 48.7 -0.146 0.9989 Julian date 213 co - control 0.0653 0.13 48.7 0.504 0.9577

co - o -0.0883 0.13 48.7 -0.682 0.9036

control - o -0.1536 0.13 48.7 -1.186 0.6386

c - co -0.1286 0.105 24.6 -1.229 0.6151

Julian date 214 c - control 0.2092 0.105 24.6 1.999 0.2156

c - o -0.1435 0.103 23.3 -1.39 0.5179

co - control 0.3377 0.105 24.6 3.228 0.0173 co - o -0.015 0.103 23.3 -0.145 0.9989

control - o -0.3527 0.103 23.3 -3.415 0.0117 c - co -0.2842 0.13 48.7 -2.194 0.1393

c - control 0.2443 0.13 48.7 1.886 0.2474

c - o -0.2912 0.13 48.7 -2.248 0.1248 Julian date 215 co - control 0.5285 0.13 48.7 4.08 0.0009 co - o -0.007 0.13 48.7 -0.054 0.9999

control - o -0.5355 0.13 48.7 -4.134 0.0008 c - co -0.2054 0.13 48.7 -1.586 0.396

c - control 0.1299 0.13 48.7 1.002 0.7487

c - o -0.2293 0.13 48.7 -1.77 0.2998 Julian date 216 co - control 0.3353 0.13 48.7 2.588 0.0592

co - o -0.0239 0.13 48.7 -0.184 0.9977

control - o -0.3592 0.13 48.7 -2.773 0.038 c - co -0.2475 0.105 24.6 -2.366 0.1106

c - control 0.2732 0.108 27.1 2.539 0.0762

c - o 0.0237 0.105 24.6 0.227 0.9958 Julian date 217 co - control 0.5207 0.108 27.1 4.839 0.0003 co - o 0.2713 0.105 24.6 2.593 0.0702

control - o -0.2495 0.108 27.1 -2.318 0.1188

c - co -0.2058 0.105 24.6 -1.967 0.2274

c - control 0.3055 0.103 23 2.974 0.0321 Julian date 218 c - o 0.0599 0.105 24.6 0.572 0.9394

co - control 0.5113 0.103 23 4.977 0.0003 80

co - o 0.2657 0.105 24.6 2.54 0.0783

control - o -0.2456 0.103 23 -2.391 0.1071

c - co -0.0992 0.13 48.7 -0.766 0.8694

c - control 0.2954 0.13 48.7 2.28 0.1168

c - o 0.1109 0.13 48.7 0.856 0.8272 Julian date 219 co - control 0.3946 0.13 48.7 3.046 0.0189

co - o 0.2101 0.13 48.7 1.622 0.3761

control - o -0.1845 0.13 48.7 -1.424 0.4907 c = carnivores; co = carnivores and omnivores; o = omnivores

Table S3.6: Post-hoc tests for testing differences in body size between in the mesocosm experiment Treatment Estimate Std. Error df t value p control 6.3167 0.2989 8.3948 21.131 <0.001 carnivores 0.4435 0.4205 8.2197 1.055 0.321 carnivores and omnivores 0.3752 0.4214 8.2871 0.89 0.398 omnivores 0.5581 0.4201 8.1933 1.329 0.22

Table S3.7: Slope estimates of the female ovisac length in the mesocosm experiment Treatment Mean Std. Error df lower CL upper CL control 0.861 0.0357 8.06 0.779 0.944 carnivores 0.87 0.0352 7.66 0.788 0.952 carnivores and omnivores 0.933 0.0357 8.07 0.851 1.015 omnivores 0.882 0.0355 7.93 0.8 0.964

CONCLUSIONS

In the three chapters of my thesis I attempt to study how ecology and evolution is influencing the distribution of species and their interactions across space and time. The first chapter was already published (Gherghel and Martin 2020) and aimed to investigate how the ecological niche and dispersal influences species distributions across long time scales. In this chapter we found that combining ENMs and dispersal modeling to investigate historic and future range dynamics is a promising approach. In the second chapter we investigated the influence of biotic interactions (via predation) at different parts of the range and how prey-predator relationship were influenced by past climate. In this chapter we found that prey-predator relationships are likely conserved through time and that during the Last Glacial Maximum the prey and their predator likely had overlapping ranges, as they do today. As such, this adds to the body of literature encouraging biotic interactions to be taken into account when studying species distributions. In the third chapter we investigated how interspecific variation in predators selects upon the traits of their prey and whether the interspecific variation in predators can affect ecosystem functioning using a series of microcosm and microcosm experiments. In this chapter we found that trophic specialization does have an effect on the ecosystem functioning but this has limited support across all treatments. Similarly, we found that the body size and the sexual traits of fairy shrimp were generally affected by the tadpoles. Again, the tadpole treatments had limited support across all treatments, omnivores generally eat smaller bodied shrimp, while carnivores eat shrimp with larger tusk sizes. Taken together, these

results suggest that the scaling of the biotic, abiotic and dispersal is important. Including both biotic and abiotic effects in attempts to understand how organisms cope with their environment is essential, as both the species and their traits and the abiotic and biotic conditions in which they live are changing. Expanding the study of these effects to also understand the dynamics and feedbacks of these interacting parts is essential to understand holistically ecological - evolutionary feedbacks on species’ ranges across spatial and temporal scales.

BIBLIOGRAPHY

Anderson, R. P. 2017. When and how should biotic interactions be considered in models of species niches and distributions? - J Biogeogr 44: 8-17.

Anderson, R. P., et al. 2002. Using niche-based GIS modeling to test geographic predictions of competitive exclusion and competitive release in South American pocket mice. - Oikos 98: 3-16.

Araujo, M. B. and Rozenfeld, A. 2014. The geographic scaling of biotic interactions.

- Ecography 37: 406-415.

Araujo, M. B., et al. 2019. Standards for distribution models in biodiversity assessments. - Sci Adv 5: eaat4858.

Bai, Y., et al. 2018. Distributional dynamics of a vulnerable species in response to past and future climate change: a window for conservation prospects. - Peerj 6: e4287.

Barve, N., et al. 2011. The crucial role of the accessible area in ecological niche modeling and species distribution modeling. - Ecol Model 222: 1810 - 1819.

Bates, D., et al. 2014. Fitting linear mixed-effects models using lme4. - arXiv preprint arXiv:1406.5823.

Beaumont, L. J., et al. 2005. Predicting species distributions: use of climatic parameters in BIOCLIM and its impact on predictions of species’ current and future distributions. - Ecol Model 186: 251-270.

Belk, D. 1975. Key to the Anostraca (Fairy Shrimps) of North America. - The

Southwestern Naturalist 20: 91-103.

Bolnick, D. I., et al. 2003. The ecology of individuals: incidence and implications of individual specialization. - The American Naturalist 161: 1-28.

Boria, R. A., et al. 2017. A single-algorithm ensemble approach to estimating suitability and uncertainty: cross-time projections for four Malagasy tenrecs. - Divers.

Distrib. 23: 196-208.

Bragg, A. N. 1967. Recent studies on the spadefoot toads. - Bios 38: 75-84.

Breheny, P., et al. 2020. Package ‘visreg’.

Brendonck, L., et al. 2008. Global diversity of large branchiopods (Crustacea:

Branchiopoda) in freshwater. - Hydrobiologia 595: 167-176.

Broennimann, O., et al. 2007. Evidence of climatic niche shift during biological invasion. - Ecol Lett 10: 701-709.

Brown, J. H., et al. 1996. The geographic range: size, shape, boundaries, and internal structure. - Annu. Rev. Ecol. Syst. 27: 597-623.

Brown, J. L. and Anderson, B. 2014. SDMtoolbox: a python-based GIS toolkit for landscape genetic, biogeographic and species distribution model analyses. - Methods Ecol

Evol 5: 694-700.

Brown, W. L. and Wilson, E. O. 1956. Character displacement. - Systematic zoology 5: 49-64.

Chan, L. M., et al. 2011. Integrating statistical genetic and geospatial methods brings new power to phylogeography. - Mol. Phylogenet. Evol. 59: 523-37.

Chunco, A. J., et al. 2012. Why Do Species Co-Occur? A Test of Alternative

Hypotheses Describing Abiotic Differences in Sympatry versus Allopatry Using Spadefoot

Toads. - Plos One 7:

Collins, W., et al. 2006. The Community Climate System Model version 3

(CCSM3). - J Climate 19: 2122 - 2143.

Cooper, J. C. and Soberon, J. 2018. Creating individual accessible area hypotheses improves stacked species distribution model performance. - Global Ecol Biogeogr 27: 156-

165.

Cornwallis, C. K., and Uller, T. 2010. Towards an evolutionary ecology of sexual traits. -Trends in Ecology & Evolution 25: 145-152.

Davis, E. B., et al. 2014. Ecological niche models of mammalian glacial refugia show consistent bias. -Ecography 37: 1133-1138.

De Araújo, C. B., et al. 2014. The importance of biotic interactions in species distribution models: A test of the Eltonian noise hypothesis using parrots. - J Biogeogr 41:

513-523.

DeGregorio, B. A., et al. 2016. Nest Predators of North American Birds:

Continental Patterns and Implications. - Bioscience 66: 655-665.

Des Roches, S., et al. 2018. The ecological importance of intraspecific variation. -

Nature Ecology & Evolution 2: 57-64.

Dormann, C. F., et al. 2013. Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. - Ecography 36: 27-46.

Dormann, C. F., et al. 2018. Biotic interactions in species distribution modelling:

10 questions to guide interpretation and avoid false conclusions. - Global Ecol Biogeogr

27: 1004-1016.

Duellman, W. E. and Trueb, L. 1986. - Biology of amphibians. JHU press.

Dugas, M. B., et al. 2016. Choosy Cannibals Preferentially Consume Siblings with

Relatively Low Fitness Prospects. - The American Naturalist 188: 124-131.

Dullinger, S., et al. 2012. Post-glacial migration lag restricts range filling of plants in the European Alps. - Global Ecol. Biogeogr. 21: 829-840.

Dyderski, M. K., et al. 2018. How much does climate change threaten European forest tree species distributions? - Global Change Biol. 24: 1150-1163.

Elith, J. and Leathwick, J. 2007. Predicting species distributions from museum and herbarium records using multiresponse models fitted with multivariate adaptive regression splines. - Divers Distrib 13: 265-275.

Elith, J., et al. 2006. Novel methods improve prediction of species' distributions from occurrence data. - Ecography 29: 129-151.

Elith, J., et al. 2010. The art of modelling range-shifting species. - Methods Ecol.

Evol. 1: 330-342.

Engelhardt, E. K., et al. 2020. Ignoring biotic interactions overestimates climate change effects: The potential response of the spotted nutcracker to changes in climate and resource plants. - J Biogeogr 47: 143-154.

Engler, R. and Guisan, A. 2009. MigClim: Predicting plant distribution and dispersal in a changing climate. - Divers. Distrib. 15: 590-601.

Fick, S. E. and Hijmans, R. J. 2017. WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. - Int J Climatol n/a-n/a.

Fielding, A. H. and Bell, J. F. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. - Environ Conserv 24: 38-49.

Franklin, J. 2009. Mapping species distributions: spatial inference and prediction.

- Cambridge University Press.

Fussmann, G. F., et al. 2007. Eco-evolutionary dynamics of communities and ecosystems. Functional Ecology 21: 465-477.

Garcıa-Parıs, M., et al. 2003. Phylogenetic relationships of Pelobatoidea re- examined using mtDNA. - Mol. Phylogenet. Evol. 28: 12-23.

Gaston, K. J. 2009. Geographic range limits of species. - Proceedings of the Royal

Society B: Biological Sciences 276: 1391-1393.

https://doi.org/10.15468/dl.6fpwlh Gherghel, I. and Papeş, M. 2015. Landscape as a determinant of dispersal patterns and population connectivity in a newt species. - Ecol.

Inform. 28: 1-6.

GBIF.org 20 April 2016. - GBIF Occurrence Download.

Gherghel, I. and Martin, R. A. 2020. Postglacial recolonization of North America by spadefoot toads: integrating niche and corridor modeling to study species’ range dynamics over geologic time. - Ecography 43: 1499-1509.

Gherghel, I., et al. 2018. Using biotic interactions in broad-scale estimates of species’ distributions. - J Biogeogr 45: 2216-2225.

Gherghel, I., et al. 2019. Refining model estimates of potential species’ distributions to relevant accessible areas. - Progress in Physical Geography: Earth and

Environment 44: 449-460.

Gherghel, I., et al. 2020. A simple framework for estimating potential distributions of amphibious marine species and implications for conservation. - Coral Reefs 39: 1081-

1090.

Gómez-Ruiz, E. P., et al. 2017. Modelling the potential geographic distribution of an endangered pollination corridor in Mexico and the United States. - Divers. Distrib. 23:

67-78.

Gray, L. K., et al. 2017. Potential effects of climate change on members of the

Palaeotropical pitcher plant family Nepenthaceae. - Plos One 12: e0183132.

Guisan, A. and Zimmermann, N. E. 2000. Predictive habitat distribution models in ecology. - Ecol. Model. 135: 147-186.

Hague, M. T. J., et al. 2020. The geographic mosaic of arms race coevolution is closely matched to prey population structure. - Evol Lett 4: 317-332.

Hand, D. J. and Till, R. J. 2001. A simple generalisation of the area under the ROC curve for multiple class classification problems. - Machine Learning 45: 171-186.

Hasumi, H. 2007. K-1 coupled GCM (MIROC) description. - Center for Climate

System Research, University of Tokyo.

Hendry, A. P. 2016. Key questions on the role of phenotypic plasticity in eco- evolutionary dynamics. - J Heredity 107: 25-41.

Hendry, A. P. 2017. Eco-evolutionary dynamics. - Princeton university press.

Hendry, A. P. 2019. A critique for eco‐evolutionary dynamics. - Functional

Ecology 33: 84-94.

Hendry, A. P., and Kinnison, M. T. 1999. Perspective: the pace of modern life: measuring rates of contemporary microevolution. - Evolution 53: 1637-1653.

Hengl, T., et al. 2017. SoilGrids250m: Global gridded soil information based on machine learning. - Plos One 12: e0169748.

Hernandez, P. A., et al. 2006. The effect of sample size and species characteristics on performance of different species distribution modeling methods. - Ecography 29: 773-

785.

Hernandez-Jimenez, A., and Rios-Cardenas, O. 2012. Natural versus sexual selection: predation risk in relation to body size and sexual ornaments in the green swordtail. - Animal Behav 84: 1051-1059.

HerpMapper 2019. HerpMapper - A global herp atlas and data hub. www.herpmapper.org

Hewitt, G. M. 1999. Post-glacial re-colonization of European biota. - Biol J Linn

Soc 68: 87-112.

Hewitt, G. M. 2004. Genetic consequences of climatic oscillations in the

Quaternary. - Philos Trans R Soc Lond B Biol Sci 359: 183-95; discussion 195.

Hijmans, R. J. and Graham, C. H. 2006. The ability of climate envelope models to predict the effect of climate change on species distributions. - Global Change Biol. 12:

2272-2281.

Hijmans, R. J., et al. 2005. Very high resolution interpolated climate surfaces for global land areas. - Int. J. Climatol. 25: 1965-1978.

Holloway, P., et al. 2016. Incorporating movement in species distribution models: how do simulations of dispersal affect the accuracy and uncertainty of projections? - Int. J.

Geogr. Inf. Sci. 30: 2050-2074.

Huang, J., et al. 2017. Projecting the range shifts in climatically suitable habitat for

Chinese Sea Buckthorn under climate change ccenarios. - Forests 9: 9.

Hughes, A. R., et al. 2008. Ecological consequences of genetic diversity. - Ecology

Letters 11: 609-623.

IUCN 2016. The IUCN Red List of Threatened Species. Version 2014.1. - http://www.iucnredlist.org. In: IUCN.

Jackrel, S. L., and Morton, T. C. 2018. Inducible phenotypic plasticity in plants regulates aquatic ecosystem functioning. - Oecologia 186: 895-906.

Jenkins, D. A., et al. 2020. Biotic interactions govern the distribution of coexisting ungulates in the Arctic Archipelago — a case for conservation planning. - Glob Ecol

Conserv e01239.

Jiménez-Valverde, A. 2012. Insights into the area under the receiver operating characteristic curve (AUC) as a discrimination measure in species distribution modelling.

- Global Ecol. Biogeogr. 21: 498-507.

Jones, A. G., and Ratterman, N. L. 2009. Mate choice and sexual selection: what have we learned since Darwin?. - Proceedings of the National Academy of Sciences 106:

10001-10008.

Karger, D. N., et al. 2017. Climatologies at high resolution for the earth’s land surface areas. - Scientific Data 4: 170122.

Kosicki, J. Z., et al. 2016. Are predators negative or positive predictors of farmland bird species community on a large geographical scale? - Ecol Indic

62: 259-270.

Kuznetsova, A., et al. 2017. lmerTest package: tests in linear mixed effects models.

-Journal of Statistical Software 82: 1-26.

Langerhans, R. B., Layman, C. A., and DeWitt, T. J. 2005. Male genital size reflects a tradeoff between attracting mates and avoiding predators in two live-bearing fish species.

-Proceedings of the National Academy of Sciences 102: 7618-7623.

Lantschner, M. V., et al. 2017. Predicting North American Scolytinae invasions in the Southern Hemisphere. - Ecol. Appl. 27: 66-77.

Lenth, R., et al.. 2018. Emmeans: Estimated marginal means, aka least-squares means. - R package version 1:3.

Levis, N. A., et al. 2015. An inducible offense: carnivore morph tadpoles induced by tadpole carnivory. - Ecology and Evolution 5: 1405-1411.

Liu, C., et al. 2016. On the selection of thresholds for predicting species occurrence with presence-only data. - Ecol Evol 6: 337-348.

Lobo, J. M., et al. 2008. AUC: a misleading measure of the performance of predictive distribution models. - Global Ecol Biogeogr 17: 145-151.

Loera, I., et al. 2012. Diversification in North American arid lands: niche conservatism, divergence and expansion of habitat explain speciation in the genus Ephedra.

- Mol. Phylogenet. Evol. 65: 437-50.

Lundsgaard-Hansen, B., et al. 2014. Ecological speciation and phenotypic plasticity affect ecosystems. - Ecology 95:2723–2735.

MacArthur, R. H. 1984. Geographical ecology: patterns in the distribution of species. - Princeton University Press.

MacKay, W. P., et al..1990. Population dynamics of a playa community in the

Chihuahuan desert. - The Southwestern Naturalist: 393-402.

Margules, C. and Austin, M. 1990. Nature conservation: cost effective biological surveys and data analysis. - CSIRO PUBLISHING.

Martin, R. A. and Pfennig, D. W. 2009. Disruptive selection in natural populations: the roles of ecological specialization and resource competition. - The American Naturalist

174: 268-281.

Martin, R. A. and Pfennig, D. W. 2010. Field and experimental evidence that competition and ecological opportunity promote resource polymorphism. - Biol J Linn Soc

100: 73-88.

Martin, R. A. and Pfennig, D. W. 2012. Widespread disruptive selection in the wild is associated with intense resource competition. - Bmc Evol Biol 12: 136.

Martin, R. A., and Pfennig, D. W. 2009. Disruptive selection in natural populations: the roles of ecological specialization and resource competition. - The American Naturalist

174: 268-281.

Martin, R. A., and Pfennig, D. W. 2010. Field and experimental evidence that competition and ecological opportunity promote resource polymorphism. - Biol J Linnean

Soc 100: 73-88.

Martin, R. A., and S. C. Garnett. 2013. Relatedness and resource diversity interact to influence the intensity of competition. - Biol J Linnean Soc 110: 689-695.

McRae, B. H., et al. 2008. Using circuit theory to model connectivity in ecology, evolution, and conservation. - Ecology 89: 2712-2724.

Miller, J. A. and Holloway, P. 2015. Incorporating movement in species distribution models. - Prog. Phys. Geog. 39: 837-849.

Mod, H. K., et al. 2015. Biotic interactions boost spatial models of species richness.

- Ecography 38: 913-921.

Mouton, A. M., et al. 2010. Ecological relevance of performance criteria for species distribution models. - Ecol. Model. 221: 1995-2002.

Mutascio, H. E., et al. 2017. Modeling relative habitat suitability of southern

Florida for invasive Burmese pythons (Python molurus bivittatus). - Landscape Ecol. 33:

257-274.

Neal, K. M., et al. 2018. Genetic structure and environmental niche modeling confirm two evolutionary and conservation units within the western spadefoot (Spea hammondii).- Conserv. Genet. 19: 937-946.

Nix, H. and Busby, J. 1986. BIOCLIM, a bioclimatic analysis and prediction system. Annual report CSIRO. -CSIRO Division of Water and Land Resources,

Canberra.

Nobis, M. P. and Normand, S. 2014. KISSMig - a simple model for R to account for limited migration in analyses of species distributions. - Ecography 37: 1282-1287.

Normand, S., et al. 2011. Postglacial migration supplements climate in determining plant species ranges in Europe. - Proc. Biol. Sci. 278: 3644-53.

Ordonez, A. and Svenning, J.-C. 2017. Consistent role of Quaternary climate change in shaping current plant functional diversity patterns across European plant

orders. - Sci Rep-Uk 7: 42988.

Pelletier, F., et al. 2009. Eco-evolutionary dynamics. - Philos Trans R Soc B

3364:1483-1489.

Peterson, A. T., et al. 2008. Rethinking receiver operating characteristic analysis applications in ecological niche modeling. - Ecol Model 213: 63-72.

Peterson, A. T., et al. 2011. Ecological niches and geographic distributions. - In:

Monographs in population biology. Princeton University Press, p. 328.

Peterson, A.T. and Soberón, J. 2012. Species distribution modeling and ecological niche modeling: getting the concepts right. Natur Conserv 10: 102-107.

Pfennig, D. 1990. The adaptive significance of an environmentally-cued developmental switch in an anuran tadpole. - Oecologia 85: 101-107.

Pfennig, D. et al. 2010. Phenotypic plasticity's impacts on diversification and speciation. -Trends in Ecology & Evolution 25: 459-467.

Pfennig, D. W. 1992. Polyphenism in spadefoot toad tadpoles as a locally adjusted evolutionarily stable strategy. - Evolution 46: 1408-1420.

Pfennig, D. W. and Murphy, P. J. 2000. Character displacement in polyphenic tadpoles. - Evolution 54: 1738-1749.

Pfennig, D. W. and Pfennig, K. S. 2012. Evolution's wedge: competition and the origins of diversity. - Univ of California Press.

Pfennig, D. W., and Collins, J. P. 1993. Kinship affects morphogenesis in cannibalistic salamanders. - Nature 362: 836-838.

Pfennig, D. W., et al. 2006. Ecological opportunity and phenotypic plasticity interact to promote character displacement and species coexistence. - Ecology 87: 769-779.

Pfennig, K. S. and Pfennig, D. W. 2009. Character displacement: ecological and reproductive responses to a common evolutionary problem. - The Quarterly review of biology 84: 253-276.

Phillips, S. J., et al. 2004. A maximum entropy approach to species distribution modeling. - In: Proceedings of the 21st International Conference on Machine Learning.

ACM Press, pp. 655–662.

Phillips, S. J., et al. 2006. Maximum entropy modeling of species geographic distributions. - Ecol. Model. 190: 231-259.

Pomeroy, L. V. 1982. Developmental polymorphism in the tadpoles of the spadefoot toad Scaphiopus multiplicatus. - In: University of California, Riverside.

Post, D. M., and Palkovacs, E. P. 2009. Eco-evolutionary feedbacks in community and ecosystem ecology: interactions between the ecological theatre and the evolutionary play. - Philos Trans R Soc B 364: 1629-1640.

Prophet, C. W. 1963. Egg production by laboratory-cultured anostraca. - The

Southwestern Naturalist 8: 32-37.

Psonis, N., et al. 2017. Hidden diversity in the Podarcis tauricus (Sauria,

Lacertidae) species subgroup in the light of multilocus phylogeny and species delimitation.

- Mol. Phylogenet. Evol. 106: 6-17.

Psonis, N., et al. 2018. Resolving complex phylogeographic patterns in the Balkan

Peninsula using closely related wall-lizard species as a model system. - Mol. Phylogenet.

Evol. 125: 100-115.

Pulliam, H. R. 1988. Sources, sinks and population regulation. - Am. Nat. 132: 652-

661.

Pulliam, H. R. 2000. On the relationship between niche and distribution. - Ecol.

Lett. 3: 349.

Qiao, H., et al. 2017. Accessible areas in ecological niche comparisons of invasive species: Recognized but still overlooked. - Sci. Rep. 7: 1213.

Reznick, D. N., and Travis, J. 2019. Experimental studies of evolution and eco-evo dynamics in guppies (Poecilia reticulata). - Annu Rev Ecol Evol Syst 50:, 335-354

Reznick, D. N., et al. 2019. From low to high gear: There has been a paradigm shift in our understanding of evolution. - Ecology Letters 22: 233-244.

Rice, A. and Pfennig, D. 2008. Analysis of range expansion in two species undergoing character displacement: why might invaders generally ‘win’during character displacement? - J. Evolution Biol. 21: 696-704.

Rogers, D. C. 2002. The amplexial morphology of selected Anostraca. -

Hydrobiologia 486:1-18.

Rogers, D. C. 2014a. Anostracan (Crustacea: Branchiopoda) zoogeography I. North

American bioregions. - Zootaxa 3838: 251-275.

Rogers, D. C. 2014b. Anostracan (Crustacea: Branchiopoda) zoogeography II.

Relating distribution to geochemical substrate properties in the USA. - Zootaxa 3856: 1-

49.

Sahlean, T. C., et al. 2014. Refining climate change projections for organisms with low dispersal abilities: a case study of the Caspian whip snake. - Plos One 9:

e91994.

Schluter, D. 2000. Ecological character displacement in adaptive radiation. - Am

Nat 156: S4-S16.

Schluter, D. and McPhail, J. D. 1992. Ecological character displacement and speciation in sticklebacks. - The American Naturalist 140: 85-108.

Schmitt, S., et al. 2017. SSDM: An r package to predict distribution of species richness and composition based on stacked species distribution models. - Methods Ecol

Evol 8: 1795-1803.

Sehgal, R. N., et al. 2011. Spatially explicit predictions of blood parasites in a widely distributed African rainforest bird. - Proc. Biol. Sci. 278: 1025-33.

Sillero, N., et al. 2014. Niche evolution and thermal adaptation in the temperate species Drosophila americana. - J. Evol. Biol. 27: 1549-61.

Skourtanioti, E., et al. 2016. A reinvestigation of phylogeny and divergence times of the Ablepharus kitaibelii species complex (Sauria, Scincidae) based on mtDNA and nuDNA genes. - Mol. Phylogenet. - Evol. 103: 199-214.

Skúlason, S., et al. 2019. A way forward with eco evo devo: an extended theory of resource polymorphism with postglacial fishes as model systems. - Biological Reviews 94:

1786-1808.

Soberon, J. 2007. Grinnellian and Eltonian niches and geographic distributions of species. - Ecol Lett 10: 1115-1123.

Soberón, J. and Nakamura, M. 2009. Niches and distributional areas: concepts, methods, and assumptions. - P Natl Acad Sci USA 106: 19644-19650.

Soberón, J. and Peterson, A. T. 2005. Interpretation of models of fundamental ecological niches and species' distributional areas. - Biodiversity Informatics 2: 1-10.

Stöck, M., et al. 2012. Cryptic diversity among Western Palearctic tree frogs: postglacial range expansion, range limits, and secondary contacts of three European tree frog lineages (Hyla arborea group). Mol. Phylogenet. - Evol. 65: 1-9.

Svenning, J. C., et al. 2008. Postglacial dispersal limitation of widespread forest plant species in nemoral Europe. - Ecography 31: 316-326.

Svenning, J. C., et al. 2011. Applications of species distribution modeling to paleobiology. - Quaternary Sci Rev 30: 2930-2947.

Svenning, J.-C., et al. 2015. The Influence of Paleoclimate on Present-Day Patterns in Biodiversity and Ecosystems. - Annual Review of Ecology, Evolution, and Systematics

46: 551-572.

Taylor, A. T., et al. 2018. Incorporating fragmentation and non-native species into distribution models to inform fluvial fish conservation. - Conserv. Biol. 32: 171-182.

Thompson, J. N. 1998. Rapid evolution as an ecological process. - Trends in

Ecology & Evolution 13: 329-332.

Tyberghein, L., et al. 2012. Bio-ORACLE: a global environmental dataset for marine species distribution modelling. - Global Ecol Biogeogr 21: 272-281.

Tzedakis, P. C., et al. 2013. Cryptic or mystic? Glacial tree refugia in northern

Europe. - Trends Ecol. Evol. 28: 696-704.

Veloz, S. D., et al. 2012. No-analog climates and shifting realized niches during the late quaternary: implications for 21st-century predictions by species distribution models. -

Global Change Biol. 18: 1698-1713.

Vörös, J., et al. 2016. Phylogeographic analysis reveals northerly refugia for the riverine amphibian Triturus dobrogicus (Caudata: Salamandridae). - Biol. J. Linn. Soc.

119: 974-991.

Waltari, E., et al. 2007. Locating pleistocene refugia: comparing phylogeographic and ecological niche model predictions. - Plos One 2: e563.

Wang, R., et al. 2018. Modeling and mapping the current and future distribution of

Pseudomonas syringae pv. actinidiae under climate change in China. - Plos One 13: e0192153.

Wang, Y.-H., et al. 2008. Habitat suitability modelling to correlate gene flow with landscape connectivity. - Landscape Ecol. 23: 989-1000.

100

Warren, D. L., et al. 2010. ENMTools: a toolbox for comparative studies of environmental niche models. - Ecography 33: 607-611.

Wauchope, H. S., et al. 2017. Rapid climate-driven loss of breeding habitat for

Arctic migratory birds. Glob. Chang. - Biol. 23: 1085-1094.

West-Eberhard, M. J. 2003. Developmental plasticity and evolution. - Oxford

University Press.

Whittaker, R. J., et al. 2001. Scale and species richness: towards a general, hierarchical theory of species diversity. - J Biogeogr 28: 453-470.

Wielstra, B. and Arntzen, J. W. 2012. Postglacial species displacement in Triturus newts deduced from asymmetrically introgressed mitochondrial DNA and ecological niche models. - BMC Evol. Biol. 12: 161.

Wielstra, B., et al. 2016. Is the Danube crested newt Triturus dobrogicus polytypic?

A review and new nuclear DNA data. - Amphibia-Reptilia 37: 167-177.

Wiens, J. J. and Titus, T. A. 1991. A phylogenetic analysis of Spea (Anura,

Pelobatidae). - Herpetologica 47: 21-28.

Wisz, M. S., et al. 2008. Effects of sample size on the performance of species distribution models. - Divers Distrib 14: 763-773.

Zeng, C., et al. 2014. Evolution of rapid development in spadefoot toads is unrelated to arid environments. - Plos One 9: e96637.

Zurell, D., et al. 2016. Effects of functional traits on the prediction accuracy of species richness models. - Divers Distrib 22: 905-917.

101

Zurell, D., et al. 2020. Testing species assemblage predictions from stacked and joint species distribution models. - J Biogeogr 47: 101-113.

102