University of Nevada, Reno
Population Ecology of the Pale Kangaroo Mouse (Microdipodops pallidus) and Community Diversity at Crescent Dunes, in the Lower Smoky Valley of Central Nevada, USA.
A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Natural Resources and Environmental Science
By Sarah J. Hegg
Dr. Marjorie D. Matocq/Advisor
May 2019
Copyright by Sarah J. Hegg © 2019 All Rights Reserved
THE GRADUATE SCHOOL We recommend that the thesis prepared under our supervision by SARAH J. HEGG Entitled Population Ecology of the Pale Kangaroo Mouse (Microdipodops pallidus) and Community Diversity at Crescent Dunes, in the Lower Smoky Valley of Central Nevada, USA. be accepted in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE
Marjorie D. Matocq, Ph.D., Advisor
James S. Sedinger, Ph.D., Committee Member
Peter J. Weisberg, Ph.D., Committee Member
Mary M. Peacock, Ph.D., Graduate School Representative
David W. Zeh, Ph.D., Dean, Graduate School
May 2019
i
ABSTRACT
Several species of granivorous rodents inhabit the unique semi-stabilized dune habitats of the Lower
Smoky Valley of Nevada, including the pale kangaroo mouse (Microdipodops pallidus). The pale kangaroo mouse is a sand-obligate species endemic to the Great Basin Ecosystem, is historically rare and thought to be in decline yet locally common in this study area. It has been suggested as an indicator species for sparse sand dune habitats in the Great Basin yet little is known about the basic ecology, habitat preferences and community dynamics of this specialized and species. I conducted capture-mark-recapture trapping surveys at 18 grids over 3 years with coincident vegetation and soil surveys. My objectives were to establish baseline information on the basic ecology, demographics, and fine-scale habitat preferences of M. pallidus. I used spatially explicit capture-recapture methods to estimate demographic parameters for
M. pallidus and generalized mixed-models to identify habitat associations. Additionally, I investigated the nocturnal rodent community structure, and how rodent diversity is influenced by environmental heterogeneity. I calculated 4 diversity indices and used linear regression to investigate the relative effects of biotic and abiotic heterogeneity on rodent diversity. My results show a negative relationship between
M. pallidus and gravelly soils, and a negative relationship between dune variation and community diversity. These results establish important baselines to fill gaps in knowledge and help guide conservation efforts for species and habitats in decline.
ii
ACKNOWLGDEMENTS Funding for this research was provided by Nevada Department of Wildlife, the Bureau of Land
Management, and the Nevada Wildlife Record Book Foundation. For access to the facility grounds and coordination, I thank B. Painter and Solar Reserve LLC.
I would like to thank my advisor M. Matocq: for giving me this opportunity, for your guidance, and patience, and for your trust and flexibility in unconventional circumstances. I also thank J. Sedinger for endless phone call meetings supporting my baby-steps of forward progress; your quiet reassurance and well-timed words encouraged me to persist and carry on when there didn’t seem to be a light at the end of the tunnel. Thank you to P. Weisberg for your always thoughtful comments and advice. I feel incredibly fortunate to have a committee of such high caliber scientists and mentors to have had this experience with.
For hosting us in Tonopah and letting us camp out in your front yard, I thank R. and S. Carpenter, and all the ladies at the Northern Nye and Esmeralda County Co-op Extension Office. Thank you to my field technicians: K. Gielow, B. Stallings, K. Wiggins, and especially J. Gable and E. Gordon. Thanks to each for your good attitudes even on cold, windy mornings or after endless vegetation surveys in the hot sun, for making living in a small trailer in Tonopah an enjoyable place to be, thank you for the exploring adventures by both foot or car, for Downton Abbey marathons, endless Its-Its in every flavor, the Mizpah and milkshake company.
For last minute help in the field, advice, and basic mental and emotional support because every grad student needs to know that they are not alone in the journey, I thank my labmates: J. Keehn, J.
Gansberg, J. Malaney, and A. Hornsby.
For assistance and advice with analyses, I thank T. Dilts, P. Murphy, and K. Shoemaker. For help in setting up the soil sorting procedure and letting me take over the greenhouse shop, I thank B. Blank and
S. Karam. Thanks to the many undergrads that volunteered with various tedious tasks.
iii
Thank you to my many mentors and teachers on my path. For inspiring and directing my passion for science and ecology: G. Brown, J. Poff, S. Saupe, and S. Thomas. Thank you to J. Stephenson and D.
Gustine for your support and belief in my abilities. And lastly, thank you to my friends and family: to
Chelsea, Krystin and Sheldon for being patient and still believing that I’m a fun person even when all I had time for was “thesising.” And thank you to my Mom and Tony – that list is endless and I’m forever grateful for your love.
iv
TABLE OF CONTENTS Abstract……………………………………………………………………………………………….……i Acknowledgments…………………………………………………………………………………………ii Table of Contents………………………………………………………………………………………….iv List of Tables………………………………………………………………………………………………v List of Figures……………………………………………………………………………………………..vi Chapter One: Fine-scale soil texture determines density distribution of a sand-dune endemic heteromyid, the Pale Kangaroo Mouse (Microdipodops pallidus)..……………………………..…….1 Introduction…………………………………………………………………………..………..…1 Methods………………………………………………………………………………..…………3 Results…………………………………………………………………………………..….…….9 Discussion…………………………………………………………………………….………….10 Chapter Two: Environmental Heterogeneity and Community Diversity at Crescent Dunes, Lower Smoky Valley, Nevada ………………………………….………………………………………………22 Introduction………………………………………………………..……………………………..22 Methods…………………………………………………………..………………………………24 Results……………………………………………………………………………………………28 Discussion……………………………………………….………………………………………..31 Literature Cited………………………………………………………………………………………….44 Supplemental Materials…………………………………………………………………………………49
v
LIST OF TABLES Chapter One Table 1.1 Covariate definitions………………………………………….….………….….……..18 Table 1.2 Habitat model AIC results…………………………………………………..…..….…19 Chapter Two Table 2.1 Summary of captures…………………………………………………….……………38 Table 2.2 Community abundance patterns with unique individuals by species….………...……39 Table 2.3 Community diversity model AIC results…………………….………………………..41 Supplemental Materials Table S.1 Trap session dates………………………………………………….…………………49 Table S.2 Chapter 1 environmental variables with sources……………………….…………….50 Table S.3 Summary of weather data…………………………………………….………………53 Table S.4 Summary of soil texture data……………………………………………..…………..54 Table S.5 Summary of shrub data………………………………………………….……………55 Table S.6 Summary of cover plot data……………………………………………..……………56 Table S.7 Chapter 2 covariate definitions………………………………….……………………57 Table S.8 Community nestedness based on presence/absence only……….……………………58
vi
LIST OF FIGURES
Chapter One Figure 1.1 Map of Crescent Dunes and trapping area…………………….…….…….……..…..17 Figure 1.2 Pale Kangaroo Mouse demographic parameter estimates….…….……….……..…..20 Figure 1.3 Predicted density of Pale Kangaroo Mouse……………………….….….…………..21 Chapter Two Figure 2.1 Variation and spread of diversity indices………...…………….………………...…..40 Figure 2.2 Predicted rodent diversity……………………………………………………………43 Supplemental Materials Figure S.10 Dune variation and counts versus mammal diversity………....…………..……….59
1
CHAPTER ONE
Fine-scale soil texture determines density distribution of a sand-dune endemic heteromyid, the Pale
Kangaroo Mouse (Microdipodops pallidus).
Introduction
Successful strategies for wildlife conservation require an understanding of how animals use their environment. The range of environmental conditions existing at the landscape level can affect presence or absence of a species (Guisan and Zimmermann 2000; Franklin 1995). At finer scales, habitat composition can affect reproductive rates, survival and fitness (Hutchinson and MacArthur 1959; Brown 1996; Price
1978). Understanding the habitat needs and tolerances of species of concern is crucial to guide where and how protection or preservation and restoration efforts are implemented (Thomas 1982).
The pale kangaroo mouse (Microdipodops pallidus) is endemic to sand-dune habitats within the
Great Basin Ecosystem and rare throughout much of its small and fragmented range. It is a highly stenotopic species and typically found only on loose, sandy soils in the lower Sonoran life zone hall 1941
(Hafner 1981; Hall 1941). The genus Microdipodops belongs to the seed-eating, desert-dwelling
Heteromyidae family of rodents and the only other species currently recognized in the Microdipodops genus is M. megacephalus – the dark kangaroo mouse, which is also rare with a relatively small geographic distribution (Hall 1941; Hafner 1981).
While the overall extent of the range of pale kangaroo mice seems stable, there are several localities where extirpation has been documented (Hafner et al. 2008), and their overall abundance appears to be in decline (Hafner and Upham 2011). There is substantial concern about their persistence due to a highly fragmented distribution, and continued habitat loss. Threats to Microdipodops habitat include increasing livestock grazing, invasive plants, fire frequency and intensity, and habitat loss due to agriculture and human development (Hafner et al. 2008). Recent phylogenetic studies of both Microdipodops species have found distinct and cryptic lineages with low (<500) effective population sizes and little to no gene
2 flow since divergence (Andersen et al. 2013; Light et al. 2013). The pale kangaroo mouse has been proposed as both a phylogeographical model for other species with similar distributions (Hafner et al.
2008), as well as an indicator species for healthy dune habitats within the Great Basin Ecosystem
(Andersen et al. 2013; Hafner et al. 2008).
The Bureau of Land Management in Nevada lists M. pallidus as a sensitive species, and the Nevada
Department of Wildlife lists it as moderately vulnerable (Wildlife Action Plan Team 2012), but it is not a federally protected species. The recent discovery of cryptic lineages, along with indications of extirpation and decreasing abundance, call for greater attention and protection for this rare and at-risk species (Hafner et al. 2008; Light et al. 2013; Andersen et al. 2013).
Early naturalists gave broad descriptions of the habitat associations of Microdipodops. Hall and
Linsdale (1929) were among the first to do so and described finding kangaroo mice in “neither the coarse sand with vegetation nor the fine sand without vegetation, but only that strip of fine sand with vegetation.”
Studies of fine-scale habitat associations of M. pallidus are scarce. Kotler (1984) found that M. pallidus prefers foraging in open rather than bushy areas, but change their foraging behavior in response to different combinations of increasing light and seed enrichment, likely due to both predation and competition or aggression from larger species. There is also evidence that pale kangaroo mice prefer to scatter-hoard their seeds in caches underneath but near the edge of shrub canopies (Swartz, Jenkins, and
Dochtermann 2010).
Two studies which involved the habitat preferences of the pale kangaroo mouse were aimed at identifying a mechanism for divergence between the two Microdipodops species, and focused largely on the role of soil texture. Both papers found differences in the soil preferences of the sibling species. The earlier paper concluded dark kangaroo mice were more stenotopic and preferred coarser soil types, and that pale kangaroo mice could tolerate a wider range of sandy soils (Ghiselin 1970). The results of the
3 second study matched more closely with Hall (1941), finding that M. pallidus preferred and was more restricted to finer, sandier soils and M. megacephalus was found more often in coarser, gravelly sandy soils, but had a broader tolerance for soil textures (Hafner, Hafner, and Hafner 1996).
Baseline information about many aspects of the basic ecology and habitat needs of M. pallidus is needed to direct land management efforts and successfully conserve this highly specialized species. The objectives of this paper are to 1) estimate basic demographic and ecological parameters for a local M. pallidus population, including quantifying detection probability, density, and activity center size, and 2) identify fine-scale habitat associations within the population, as defined by vegetation, soil, and habitat structure. This is the first study to quantify densities, detection probabilities and activity center sizes in this species and to integrate this information into the estimation of fine-scale habitat associations.
Methods
Study area
I conducted the study in the central Great Basin ecosystem, in the southern portion of the Lower
Smoky Valley, approximately 30 kilometers northwest from the town of Tonopah, Nevada (Figure 1.1).
The valley is broad, flat to rolling with lake plains, alluvial fans, ephemeral washes and sand dunes. The average elevation of the study area is 1530 meters (5020 feet). Mean annual precipitation is 3-9 inches, with most precipitation falling as snow between November and May. The average minimum and maximum temperatures in January are -5.3 and 3.3 °C and in July 16.3 and 31 °C, respectively (Western
Regional Climate Center 2015). Vegetation is primarily open shrublands comprised of fourwing saltbush
(Atriplex canescens), Bailey’s greasewood (Sarcobatus baileyi), Nevada dalea (Psorothamnus polydenius), littleleaf horsebrush (Tetradymia glabrata), winterfat (Krascheninnikovia lanata), Shockley wolfberry (Lycium shockleyi), and shadscale (Atriplex confertifolia).
I trapped in an area approximately 4.5 km north-south by 1 km east-west and 500 hectares in area
(Figure 1.1). Directly east of the trapping area is a large, active sand dune system, the Crescent Dunes.
4
Directly west of the trapping area is a 100-megawatt concentrated-solar energy facility encompassing approximately 630 hectares. The construction was completed in November 2015. The study area and much of the Lower Smoky Valley is comprised of land owned and managed by the US Bureau of Land
Management. Cattle grazing is permitted and occurs within the trapping area as well as throughout most of the valley. Recreational use near the sand dunes includes camping and riding all-terrain vehicles and motorcycles.
Trapping and handling procedures
I conducted mark-recapture trapping from 2012 through 2014, for approximately 3-4 months each of those years, starting in March or April and continuing through June or July (Table S.1 in appendix lists specific trapping dates for each session). I trapped nightly but suspended trapping for the 7 days around a full moon as previous studies have suggested that nocturnal desert rodents reduce or alter their activity during full moon periods (Upham and Hafner 2013). I established 18 survey grids throughout the study area; 16 grids were established in 2012 with 2 additional grids established in 2013. Over the course of the study, I trapped at each grid for 6 sessions (with the exception of the 2 grids established in 2013 which had 4 sessions each). Each trapping session consisted of 3 consecutive nights of trapping and I trapped at each grid for 2 sessions per year. For the safety of any animals inside traps, I did not conduct trapping efforts if significant, steady rain occurred for more than 30 minutes during the evening. In those situations, I resumed trapping the next dry evening.
Trapping grids consisted of 4 transect lines, each line 10 meters apart, and running east-west for 150 meters. Each line was comprised of 16 trap stations, one station every 10 meters. I set two traps at each trap station, for a total of 128 traps per grid. I used extra-long aluminum Sherman traps (3” wide by 3 ¾” tall by 9” long) and baited them with bird seed. I also added 1-2 cotton balls to the traps in the early part of the season when overnight temperatures were low to help provide insulation.
5
I set the traps between 6pm and sunset, or whenever the temperature dropped below 90 F°, and I left the traps open overnight. I then checked and closed traps in the morning, starting at first light. I sexed, weighed, and identified to species each individual animal caught. I recorded additional measurements, including ear, foot and tail length, reproductive status, parasite load and age group. I biopsied small tissue samples from each animal’s ear and applied individually marked metal ear tags at the time of first capture.
I followed the guidelines of the American Society of Mammologists (Gannon and Sikes 2007), and used protocols approved by Institutional Animal Care and Use committee at the University of Nevada, Reno
(protocol 00543) for all animal trapping and handling procedures.
Vegetation and habitat surveys
To collect fine-scale habitat data, I conducted 2-part vegetation surveys at each trap station once throughout the study duration, which occurred with a 3-meter radius circle centered at the trap station. I first identified to species and life stage each shrub, and measured distance from station center, height, canopy area and distance to nearest neighbor. I then estimated ground cover type and amount using 4 quadrat frames that were randomly placed within strata reflecting increasing distance from plot center.
The ground cover variables documented in each quadrat survey were sand, gravel, coarse gravel, cobble, rock, boulder (<0.2 cm = sand, 0.2-1 cm = gravel, 1-3 cm = coarse gravel, 3-6 cm = cobble, 6-25 cm = rock, >25 cm = boulder), shrub, annual grass, annual forb, perennial grass, perennial forb, litter, soil crust, and bare ground. I estimated and recorded the percent cover as one of 6 classes (0 = absent, 1 = present-
5%, 2 = 5-25%, 3 = 25-50%, 4 = 50-75%, 5 = 75-95%, 6 = 95-100%). For data analysis, I converted each cover class to the median cover value of that category, and then calculated the mean over all 4 quadrat placements. Lastly, I recorded total counts of bunchgrass, small and mini dunes (<0.3 m tall and forming around vegetation, rocks or other features that cause drifts = mini dune; 0.3-0.5 m tall = small dune). I collected a soil sample from inside each quadrat frame and combined them together into one sample per station for texture analysis. I compiled weather data both locally from a weather station located at the
6 solar facility and from the closest NOAA weather station located at the Tonopah Airport (approx. 28 km
SE of study area).
To further evaluate surface soil texture, I used a Ro-Tap Testing Sieve Shaker machine (C-E Tyler
Company - Model B, serial# 10014) to process soil samples. I poured each sample through a stack of 8 sieves (US Standard sieve sizes #4, 10, 35, 45, 60, 100, 140, 200; 4.76, 2.00, 0.500, 0.354, 0.250, 0.149,
0.105, 0.074 mm, respectively) and sifted the samples in the Ro-Tap machine for 5 minutes. I then weighed the soil from each sieve to determine percent of each size class within the sample. For initial data exploration, I combined several of the soil sieve size classes together into the following size categories: cobble (>4.75 mm), gravel (<4.75-2.0 mm), sand (<2.0 mm-0.15 mm), clay (<0.15-0.075 mm) and silt
(<0.075 mm). I defined the coarse soil covariate as cobble and gravel combined.
Demographic estimation
To estimate density and population parameters of M. pallidus, I used likelihood-based spatial capture-recapture methods, as implemented in the R package “secr” (Efford 2015). I ran a model that did not include any environmental covariates and only allowed for density to vary by trap session, while keeping capture probability and activity center (sigma) parameters constant to avoid forcing any trends in density over trap sessions. This model was run separately for each grid because the lack of animal movement and fine-scaled spatial data between grids precluded analyzing all the grids together. I assumed population closure within each trapping session, chose “multi” for the detector type because I had 2 traps at each station, and set a buffer of 40 meters around my trapping grids.
Exploratory Data Analysis
I chose covariates that represented different aspects of the habitat that have been previously suggested as important for microhabitat selection in other heteromyid species and communities (Price and
Waser 1985; Upham and Hafner 2013; Hafner 1981; Price 1978). Covariates included competition, canopy and structure-related variables, variables that represented food resources, soil texture variables, and ground cover related variables. I also included weather and temporal-related variables to account for
7 possible changes in rodent activity levels. The source, description and definition of all environmental variables collected is found in the appendix (Table S.2). I used the basic statistics functions in R to generate histograms, boxplots, dotcharts, correlation matrices, and co-plots and implemented standard data exploration methods to gauge independence, collinearity, and normality of covariates and environmental variables (Zuur, Ieno, and Elphick 2010).
To reduce the number of variables used for model fitting, I chose variables that would represent a wide range of environmental categories (e.g. overstory and understory vegetation, and soil related), as well as habitat features that have been suggested as important in studies with other heteromyids (e.g. canopy cover, soil texture). I also considered data structure, distribution, and correlation levels.
Models incorporated two of the vegetation-related variables – percent cover of shrub canopy, and average cover of forbs and grasses. All of the shrub variables were highly correlated (all >0.538). I used percent cover of shrub canopy because it accounts for the total amount of cover at the station, which has been considered important in many other heteromyid habitat studies (Price 1978; Upham and Hafner
2013). To include aspects of the understory vegetation, I combined both forbs and grasses by using the sum of the average cover of each.
Two soil variables were used in the habitat models - the percent of sandy soil and the percent of coarse soil (sum of gravel and cobble) texture from the soil samples. The sand variable was included because it is considered a consistent feature of habitat associated with M. pallidus (Hall 1941; Hafner
1981). Despite high negative correlation of sand with the other soil variables collected (correlation levels ranged from -0.550 to -0.704; correlation with cobble = -0.550, correlation with gravel = -0.553), I also chose to include the coarse soil variable for its significance in previous research (Ghiselin 1970; Hafner,
Hafner, and Hafner 1996).
Variables that described habitat structure included mini dune count, small dune count, average cover of rocks and boulders, average distance to nearest shrub neighbor and average shrub height. Very few
8 stations had any rocks and boulders present, so I did not bring those variables forward. The shrub height and nearest neighbor variables were highly correlated with all other shrub variables and therefore were not used in model building. Of the two dune variables, I brought forward the mini dune variable as it had a more normal distribution compared to the small dune counts.
I also included the average percent cover of litter, as this variable was not highly correlated with other variables (correlation levels ranged from -0.177 to 0.163) and is indicative of overall site productivity. Lastly, I included a competition/mammal-related variable - the total number of captures of all other (non-M. pallidus) species. See Table 1.1 for a list and basic definition of covariates used in the habitat model.
Estimating Habitat Associations
I used generalized linear mixed effects analysis (GLMM) as implemented in the R package “lme4”
(Bates et al. 2015) to estimate the relationship between density of pale kangaroo mice and habitat. To avoid over-parameterizing the model, I applied an iterative approach to model selection. I first evaluated models including only temporal environmental variables and then models using only spatial covariates.
The covariates included in both model sets were identified from my initial data exploration. I assessed fit of these models using AIC scores according to recommendations by Anderson and Burnham (2002) as well as standard diagnostic plots including residuals vs fitted values, histograms of the residuals to verify homogeneity of residuals and residuals vs explanatory values to verify independence (Zuur, Ieno, and
Elphick 2010). When running the GLMM’s, I weighted the response variable (pale kangaroo mouse density) by 1/se^2 to give more weight to more precise density estimates, and converted predictor variables to standard normal variates using the default scale() function in R (Bates et al. 2015). All models also included random intercepts by grid. I used the gamma error distribution family, as it typically allows for some overdispersion (Zuur, Ieno, and Elphick 2010).
9
I manually calculated 95% confidence intervals for the results of the top 2 performing models. To do this I took an evenly distributed sample of 15 values from the data set for each covariate (percent sandy soil texture, and percent coarse soil texture), including the minimum and maximum values. I then used the
PopTools extension for Microsoft Excel (Hood 2010), to generate 1000 random numbers with distributions based on both the intercept and beta estimates and standard errors (from the GLMM model output of fixed effects). These random numbers were then used to back-transform and generate M. pallidus density predictions for each of the 15 values. Each of the random density estimate sets were then sorted from smallest to largest. The 25th and 975th values were used as the confidence intervals, and the density was based on the actual covariate value from the sample set.
Results
I trapped 3 nights per session, for 6 total sessions (or 4 sessions for grids 19 and 20), which totaled
39,936 trap nights. I recorded 5985 overall captures (including recaptures - 14.9% trap success), 3048 unique individuals (50.9% recapture rate), and 10 species. The pale kangaroo mouse comprised 22.7% of total captures (1358 overall captures) and 3.4% trap success for all M. pallidus captures. For unique individuals (not including recaptures), pale kangaroo mice comprised 22.9% (699 unique individuals).
The minimum number alive for M. pallidus per grid-session ranged from 0 to 21.
Over all sessions for each trapping grid, the percent of unique individuals caught that were M. pallidus ranged from 11.8% at grid 13 to 38.6% at grid 12. The trapping success rates for unique M. pallidus individuals over all sessions for each grid ranged from 0.65% at grid 14 to 3.08% at grid 1.
Minimum and maximum temperatures ranged from -9.0 to 38.2 °C (the lowest average minimum to highest average maximum), and the mean temperature was 16.5 °C over each of the 6 trapping sessions.
Average maximum wind speeds at 1-meter height was 11.8 meters per second, and the average total precipitation was 43.7 mm (Table S.3). Summary tables of cover plot and habitat count data, as well as shrub and soil data are included in the supplemental materials (Tables S.4, S.5, and S.6).
10
Density estimates for M. pallidus for each grid-session ranged from 0 to 21.7 individuals per hectare
(Figure 1.2). The average density over all grids for each session ranged from 1.7 to 10.3 individuals per hectare and the overall average density for all grids and sessions was 6.5 individuals per hectare.
Detection probabilities (g0) for each grid ranged from 0.04-0.26, with a mean of 0.15. Activity center radii ranged from 9.0 to 21.1 meters, with a mean of 14.2 meters. The average mean maximum distance moved (MMDM) was 21.4 meters, and the average mean distance between consecutive capture locations
(dbar) was 17.2 meters.
None of the temporal variables improved model performance relative to a constant density model, and so none of the temporal variables were used in the assessment of habitat variables. The best performing model in the suite of habitat models had only coarse soil as a predictor variable (∆AICc = 0), and the model with only sandy soil as the predictor resulted in a close 2nd (∆AICc = 0.2). The next several models included combinations of the shrub canopy variable but coefficients for this variable only differed from zero when shrub canopy occurred by itself (∆AICc = 2.20).
Using the beta estimates from the model including coarse soil, there is a predicted maximum M. pallidus density of more than 10 individuals per hectare when coarse soil is close to 0.1% of the overall soil sample (Figure 1.3), and then density is predicted to decline to just above 1 individual per hectare as coarse soil approaches 22%. Inversely, the second ranked model shows a slight increase in M. pallidus density as the percent of sandy soil increases – at just approximately 51% sandy soil, M. pallidus density is estimated at just below 20 individuals per hectare, with an estimated increase to just above 20 individuals per hectare at 100% sandy soil. The predicted relationship between density and sandy soil is flatter and has larger confidence intervals than the predicted relationship with coarse soil (Figure 1.3).
Discussion
Accurately estimating population size and other demographic parameters is crucial for the management and conservation of any species, and no previous studies have reported density or
11 abundance, detection probabilities or activity center sizes for M. pallidus. Non-spatial capture-recapture models consider individuals with home ranges completely within the trapping array to have the same detection probability as individuals whose home range is at the edge of a trapping array. This inability to account for the lack of geographic closure around trapping arrays can bias the estimation of demographic parameters (Gerber and Parmenter 2015). Spatial capture-recapture methods (SCR), such as the secr package in R, estimate density by defining the effective sampling area based on trap locations and animal movement distances, and then adjusting capture probabilities relative to animal density and use (Borchers and Efford 2008; Efford 2004). Accurate estimates of detection probabilities and activity center sizes are also vital as they are used to estimate density and furthermore, are independently fundamental components in understanding species ecology and population dynamics (Williams, Nichols, and Conroy
2002).
While there are not any previously reported estimates of population size for Microdipodops, there are a small number of studies reporting statistics such as minimum number alive or trap success rates (the number of individuals caught divided by the number of trap nights). Trap success rates are often biased indicators of abundance as they do not take into account capture probabilities (Skalski, Robson, and
Simmons 1983; Thompson 2002; Nichols et al. 2000). Hafner et al. (2008) reported a mean trapping success rate of 2.88% for M. pallidus throughout its range, and when Brown (1973) studied desert rodents in sand dune habitats throughout the Great Basin Ecosystem, he reported trap success rates ranging from
2.5-10.36% and 4.29-6.07% for M. pallidus and M. megacephalus, respectively. Other studies of both
Microdipodops species reported mean trap success rates ranging from 5.2-5.3% (Hafner, Hafner, and
Hafner 1996; Hafner 1981), and 0.38-1.18% over 3 study areas (Ghiselin 1970). I found a mean trap success rate of 3.4% for pale kangaroo mice, which is within the range of previously reported trap success rates.
There have not been any published detection probability estimates for either Microdipodops species. In a meta-analysis of mark-recapture data for several families of small mammals, Heteromyid
12 capture probabilities ranged from 0.31-0.55 (Hammond and Anthony 2006). My estimated detection probabilities for M. pallidus are quite a bit lower, with the mean at 0.15 and ranging from 0.04 to 0.26 by grid (confidence intervals = 0.026-0.414). Future analyses to investigate how, when and why detection probabilities change are important to better understand factors affecting detection in M. pallidus, however, estimates obtained through the current study provide an important baseline and step towards dealing with sources of errors in population estimates (Thompson 2002).
One demographic attribute that has been published specifically for Microdipodops sp. is home range and movement distances. Ghiselin (1970) reported a maximum distance between trapping locations for an individual Microdipodops sp. as 237 meters for males and 132 meters for females. Using the secr package in R, I calculated mean maximum distance moved (MMDM) and dbar. The MMDM is the average maximum distance between any detections of each individual observed (Otis et al. 1978). The dbar statistic is the mean distance between consecutive capture locations, pooled over all individuals
(Efford 2004). In this study MMDM ranged from 11.3 meters at grid 2, to 36.1 meters at grid 3 (this is the average MMDM at each grid over all sessions). The highest MMDM detected was 66.6 meters, at grid 1 during session 3 (the first session in 2013). The dbar values ranged from 9.4 meters at grid 20, to 20.1 meters at grid 3 (averages for each grid over all sessions). The maximum dbar values was 66.6 meters, at grid 1 during session 3. I did not differentiate between males and females in the analysis. All of the values reported in this study are less than half of the maximum distances reported by Ghiselin.
O’Farrell studied annual composite home ranges of a community of small mammals (including
M. megacephalus) in sagebrush habitats of west-central Nevada. He found annual composite home range areas to vary from 0.33-0.49 ha for all Heteromyids, and that M. megacephalus had mean annual home ranges of 0.66 and 0.39 ha for males and females, respectively (O’Farrell 1974, 1978). These results would correspond to an average of 0.063 ha, with a range of 0.025-0.14 ha (based on the activity center radii estimates, and the assumption of a circular bivariate normal shape for M. pallidus activity centers).
Similar to the maximum distances, my estimated activity center sizes for the pale kangaroo mouse would
13 be relatively smaller than O’Farrell’s estimates for its sibling species. However, different field and analysis methods were used by O’Farrell (1974, 1978) and this study, and so may not be directly comparable.
Density estimates varied both between grids within the same session as well as between sessions at the same grid. The largest single density estimate was 21.73 individuals/hectare (se = 5.41, lcl = 13.44, ucl = 35.13) at grid 5 during the later session in 2013 (session 4). In general, grids with higher mean densities also had higher variation in density between trapping sessions. The grids located in the northern section of the study area (grids 11-16) had more consistent density estimates between sessions, yet lower mean density estimates than grids in the southern portion of the study area. There were 17 grid-session instances (out of 116 total grid-sessions) where no M. pallidus were trapped, and 11 of those 17 instances were during the later session in 2012 (session 2). During the session 2, M. pallidus was only caught at 4 grids (1, 2, 16, and 18) and with the exception of grid 2, density estimates for session 2 were notably lower than their overall mean at each grid.
Demographic and population parameters can be influenced by a variety of temporal or spatial factors. For example, it has been shown that when temperatures below certain levels are combined with limited food availability, pale kangaroo mice often enter a state of torpor for up to several days (Brown and Bartholomew 1969) or even weeks during winter months (French 1989), which can consequently reduce detection rates. Low detection rates can create bias in density estimates by increasing confidence intervals or lowering density estimates even if actual abundance is high (Williams, Nichols, and Conroy
2002). Home range, or activity center sizes can vary based on habitat quality, density of inter- or intra- specific competition, among other things (Wolff 1985). Age and sex can also influence behavior and how much exploration and movement an individual is able or willing to tolerate (Randall 1993; O’Farrell
1978).
Based on previous field work, spring “green up” season is likely the best time to detect pale kangaroo mice if they are present in an area. O’Farrell (1974) trapped for 2, 3-night sessions per month
14 for a year and found that M. megacephalus were only active from March-October. He also found that there were two peaks in M. megacephalus activity – one during March and April, and the other in
September, with relatively minimal detections for the time in between those peaks. While these estimates of density and capture probability are often much lower compared to similar species from different areas, they are likely to be on the high end of the range for pale kangaroo mice in this study area due to the late winter through early summer trapping schedule. However, variation in any combination of demographic parameters can differ based on locality, time of year, environmental conditions, field methods or based on characteristics inherent to different species such as behavior (Otis et al. 1978; Hammond and Anthony
2006). Long term ecological studies have shown complex patterns of population and community dynamics in desert rodents (Brown and Ernest 2002). Repeating similar studies using mark-recapture trapping over several years, and during different seasons is essential to gain a more complete understanding of baseline population dynamics, as well as trapping in other locations where pale kangaroo mice are found to explore the extent and factors affecting these population parameters.
I used a model that kept the both the detection probability (g0) and activity center size (sigma) parameters constant over trapping sessions to avoid overfitting. Therefore, it’s likely that there were some grid-session occasions when activity center sizes were higher and closer to previously published estimates. Exploring how detection probability and activity center size changes throughout the study and the factors affecting their estimates would be a good idea for future work to help gain a more complete ecological perspective and understanding of M. pallidus activity patterns.
This study shows that at a fine spatial scale and for this locality, the density of M. pallidus is most closely related to soil texture among temporal or spatial covariates explored. Specifically, M. pallidus is found at lower densities in coarser soils. These findings support previous research on pale kangaroo mice showing large-scale habitat preferences for habitats with fine sand soils (Hall 1941; Hafner 1981). It also aligns with research comparing the preferences of M. pallidus and M. megacephalus at a fine scale
(Ghiselin 1970; Hafner, Hafner, and Hafner 1996), but further confirms these observations in an allopatric
15 population of M. pallidus. The relationship between pale kangaroo mouse density and sandy soil is flatter and has larger confidence intervals than the coarse soil model, when you might expect a more equal
(while opposite) slope. This may be explained due to greater variation in the amount of sandy soil within most grids versus coarse soil variation.
With seeds as the primary source of food for the pale kangaroo mouse, as well as the numerous previous studies focusing on Heteromyid use of open vs covered habitats (Rosenzweig 1973), I was surprised that no vegetation-related covariates resulted in significant relationships with M. pallidus density. The shrub cover variable was included in the 3rd and 5th best models, but its relationship with M. pallidus density was not significant. In the 6th, 8th, and 9th best models, shrub canopy was significant but only at the lowest level (p>0.1). These results could suggest that shrubs are important, but perhaps that soil texture may mask the effect of shrubs. One possible explanation is that the grid random effects reduced the importance of covariates in the models because there was only one value for each covariate at each grid. Another possible explanation for the absence of a relationship is that Heteromyids often forage for seeds within the soil rather than on the plant itself (Reichman 1979), so perhaps if there are enough seed-producing plants within wind-dispersal distance to keep an adequate level of seeds in the soil, then the local amount of grasses, forbs, or shrubs is not as influential. Studies conducted in sand dunes of the
Negev Desert in Israel found that there is daily redistribution of soil and seeds by wind and that rodents spent less time foraging when redistribution was prevented by cloth barriers (Ben-Natan et al. 2004).
Many studies have examined Heteromyid foraging preferences of open vs closed canopy. One motive is hypothesized to be avoidance of predation (Kotler 1984) which can be influenced by the amount of moonlight (Brown et al. 1988), inter-specific competition (Davidson, Brown, and Inouye 1980), among other theories. Price and Waser (1985) suggested that desert heteromyids chose microhabitats based on the combination of two attributes – rate of resource harvest and risk of predation. They showed that microhabitats “differed in seed and soil characteristics that affect harvest rates, that heteromyids exhibit species-specific patterns of preference for artificial seed patches that differ in soil texture, soil density,
16 and seed distribution.” It is likely that pale kangaroo mice use a combination of factors beyond soil texture that vary spatially and temporally, at both fine and broad scales.
By using the mean values some of the environmental variation was undoubtedly lost, which may have masked possible habitat relationships that exist at a sub-grid level. A sub-grid level of analysis for spatial mark-recapture models may be possible in the Bayesian statistical framework. Another source of potential error in these analyses is that, despite conducting a thorough exploration of possible variables, I did not include covariates in the set of predictor variables with significant relationships to the pale kangaroo mouse.
I have estimated the first densities, detection rates and activity center sizes for the pale kangaroo mouse. These provide the beginning of a baseline for this rare species to help assess population trends and determination of conservation status going forward. These results contribute to the knowledge and understanding of which habitat characteristics are ideal for populations of pale kangaroo mice in an area where they are one of the most abundant species within the Heteromyid community and allopatric from
M. megacephalus. The environmental factors that create fine, loose sandy soils and dune systems are rare in the Great Basin deserts. This highlights the importance for conserving and protecting these areas in the face of current threats, including climate change, human development, invasive species, and more.
17
Figure 1.1. Map of the trapping area. Stars indicate the location of trapping grids labelled with the grid number. The circular array to the left of the trapping grids is the solar energy facility. The undulating lines shown to the right of the grids are the Crescent Dunes sand dune formation. The inset map in the upper corner shows the range of M. pallidus (outlined in black), confirmed occurrences (white circles), and study area location (red triangle). Predicted habitat suitability is shown with an 80 km buffer around all verified M. pallidus occurrences (T. Dilts, personal communication).
18
Variable Definition
MoonScr Index created to broadly quantify the brightness of the moon.
Ppt Total precipitation in cm (mm?)
MinTmp Minimum temperature in °C for the 24 hour time period
MaxTmp Maximum temperature in °C for the 24 hour time period Maximum wind speed in meters per second for the 24 hour time period at 1 MaxWnd meter heights Average wind speed in meters per second for the 24 hour time period at 1 AvgWnd meter heights
ShbCpy Total percent cover of shrub canopy
FrbGrs Average percent cover of forbs and grasses
Total number of mini dunes at plot. Mini dunes were classified as substrate MDn <0.3m forming around vegetation, rocks, or other features that cause drifts. Snd Percent of surface soil sample in the sand size class (<2 mm but > 0.15mm)
Percent of surface soil sample in the gravel, cobble and larger size classes Crs (greater than 2mm). Total number of captures of any species (not including M. pallidus) at that Tcap trap location over all sessions
Ltr Average percent cover of dead plant/organic matter
Table 1.1. Definitions or description for the environmental variables used in analyses. Sources for each can be found in the appendix Table S.2.
19
Model ∆AICc wi k Temporal Covariates
D~(1|Gd) 0.00 0.40 2
D~AvgWnd+(1|Gd) 1.95 0.15 3
D~Ppt+(1|Gd) 1.95 0.15 3
D~MaxTmp+(1|Gd) 2.15 0.14 3
D~Ppt+AvgWnd+(1|Gd) 4.14 0.05 4
D~Ppt+MaxTmp+(1|Gd) 4.14 0.05 4
D~MaxTmp+AvgWnd+(1|Gd) 4.14 0.05 4
D~Ppt+MaxTmp+AvgWnd+(1|Gd) 6.40 0.02 5
D~MoonScr+Ppt+MinTmp+MaxTmp+MaxWnd+AvgWnd+(1|Gd) 12.70 0.00 8
Spatial Covariates
D~Crs*+(1|Gd) 0.00 0.22 3
D~Snd*+(1|Gd) 0.20 0.20 3
D~ShbCpy+Crs.+(1|Gd) 1.60 0.10 4
D~Snd+Crs(1|Gd) 1.80 0.09 4
D~ShbCpy+Snd+(1|Gd) 2.40 0.07 4
D~ShbCpy●+(1|Gd) 2.20 0.07 3
D~(1|Gd) 2.45 0.07 2
D~ShbCpy●+MDn+(1|Gd) 3.40 0.04 4
D~ShbCpy●+FrbGrs+(1|Gd) 4.20 0.03 4
D~MDn+(1|Gd) 4.00 0.03 3
D~FrbGrs+(1|Gd) 4.40 0.02 3
D~Tcap+(1|Gd) 4.60 0.02 3
D~Ltr+(1|Gd) 4.60 0.02 3
D~ShbCpy+FrbGrs+MDn+Snd+Crs+Tcap+Ltr+(1|Gd) 10.65 0.00 9
Table 1.2. Mixed-effects model description and results. Model notation refers to M. pallidus density estimates (D) for as response variable, the variable(s) included in the model as the predictor variables. All models were run with grid as the random effects parameter (1|Gd). Presented are differences in the corrected Akaike’s information criterion values (∆AICc) model weights (wi), and the number of model parameters (k). No covariate betas in either model set resulted in a significance level higher than 0.05. *- denotes a covariate beta significance = 0.05 ° - denotes a covariate beta significance = 0.1
20
Figure 1.2. Density (top), Detection probability (middle) and activity center size (bottom) of M. pallidus for each trapping grid. Density shown is the mean of each grid by session density estimated by SECR. Dashed lines are respective 95% confidence intervals.
21
Figure 1.3. Predicted density based on results from top 2 spatial models from the mixed effects analysis. The figures show the estimated density of M. pallidus with 95% confidence intervals as the covariate changes. The x-axis scale is based on the range of soil textures found at the trapping site.
22
CHAPTER TWO
Environmental Heterogeneity and Community Diversity at Crescent Dunes, Lower Smoky Valley,
Nevada
Introduction
Biological diversity is the fundamental source of ecosystem function; ecosystems with higher levels of diversity are more productive and stable (Willig 2011; Grace et al. 2016). A central concept in ecology is that greater environmental heterogeneity will be positively correlated with biological diversity
(Hutchinson 1959; Tilman 1986; Rosenzweig 1995). The basis of this theory is that heterogeneous environments have more ecological niches, thereby allowing coexistence of a greater number and diversity of taxa (Tews et al. 2004). Speciation events may be more common in areas of higher environmental heterogeneity by the opportunity to adapt to diverse conditions (Rosenzweig 1995; Hughes and Eastwood 2006; Antonelli and Sanmartin 2011). Additionally, in the uncertainty of climate change progression, areas with higher environmental heterogeneity may provide suitable refuge options to multiple species (Fjeldså, Bowie, and Rahbek 2012).
The link between environmental heterogeneity and species diversity has largely centered on two mechanisms or aspects of heterogeneity. The first is structural or abiotic heterogeneity, which I define here as variation in physical characteristics of the habitat. Examples of abiotic heterogeneity include edaphic qualities, topographic features, the variation in height and vertical structure of trees/shrubs/litter, and variation in open versus closed canopies. The other functional aspect of environmental heterogeneity that is often measured is resource or biotic heterogeneity, which I define here as variation in the amount and type of energy resources available. Biotic heterogeneity is typically measured as the amount and type of live plants or prey species. In early studies, the mechanism that was thought to link species diversity to environmental heterogeneity was vertical and horizontal habitat structure wherein greater structure from the ground through the tree canopy created more ecological niches, leading to higher bird diversity
23
(MacArthur and MacArthur 1961). Examples abound of studies showing a relationship between habitat structure or abiotic heterogeneity and species diversity in various taxons, systems, and regions (Horvath,
March, and Wolf 2001; Sessitsch et al. 2001; Ecke, Löfgren, and Sörlin 2002; Gratwicke and Speight
2005). Other early studies suggested that higher biological diversity was the result of greater variation in energy resources (e.g. biotic heterogeneity); a framework based on the competitive exclusion principle, which predicts that the number of coexisting species is determined by the number of limiting food resources (Gause 1934; Elmberg et al. 1994). While these two mechanisms are similar and often interrelated, it is likely that each of these mechanisms has varying importance in different habitats or for different taxa, and in those cases, distinguishing between the two mechanisms and their relative importance will enhance our understanding of the processes that shape community structure and diversity
(Stevens and Tello 2011).
Case study: the Heteromyid “players” and their ecosystem.
Desert rodent communities, particularly the Heteromyid family of North America, have long been targets of research examining factors that contribute to community structure, species coexistence, and diversity, as well studies of community ecology, population dynamics and evolution (Rosenzweig and
Winakur 1969; J. H. Brown 1973; J. H. Brown and Lieberman 1973; Price 1978; Kotler and Brown
1988). Heteromyids are nocturnal, granivorous rodents, are either bipedal or quadrupedal, are often found in complex communities of up to 6 species, and have many adaptations to endure the unpredictable fluctuations in climate and productivity of the arid habitats they occupy (Kenagy and Bartholomew 1985).
Heteromyids have been the subject of much research revolving around theories of how these co- occurring species partition their environment to allow coexistence. These include how foraging behavior is related to body size (Lemen 1978), how different means of locomotion allow differential use of open versus covered microhabitats (Price 1978), to how inter-specific competition may be minimized by different foraging or seed storage methods (Jenkins and Breck 1998; Swartz, Jenkins, and Dochtermann
2010). Much of this work has been done with Heteromyid communities in the Mojave and Sonoran
24 deserts, while fewer studies have been conducted in the northern, colder Great Basin Desert where resource availability and habitat structure are distinct from the warmer deserts, yet nearly the same cadre of heteromyids co-exist.
Great Basin sand dunes are unique, island-like habitats where currently little is known about the basic ecology of the distinct biota living there. Endemic flora and fauna throughout the Great Basin region are at risk and imperiled by habitat loss and alterations, primarily attributed to invasive species, livestock grazing and ever increasing human development and use (Brussard, Charlet, and Dobkin 1998).
Establishing baseline information for these native species and unique communities is essential for their conservation.
Here, I examine a community of desert rodents living in a sand dune habitat in the central Great
Basin Ecosystem to ask how and if community composition changes based on heterogeneity of both abiotic and biotic factors. Further, if and when fewer than the maximum number of species was detected in a portion of the study site, I ask if there is a pattern regarding which species are not present. I address these questions by conducting an intensive mark-recapture study of the entire rodent community over a three-year period. For a community of rodents that are well adapted to drought and living in an already sparse and unproductive habitat with low shrub species richness, I predict that abiotic heterogeneity will have a larger effect on mammal diversity than biotic heterogeneity. Due to many previous studies demonstrating that desert rodents use different microhabitats to partition their habitat use, I predict a positive relationship between mammal diversity and environmental heterogeneity.
Methods Study area I conducted the study in the central Great Basin ecosystem, in the southern portion of the Lower
Smoky Valley, approximately 30 kilometers northwest of the town of Tonopah, Nevada (see Figure 1.1).
I trapped in an area approximately 4.5 km north-south by 1 km east-west and 500 hectares in area.
Directly east of the trapping area is a large, active sand dune system, the Crescent Dunes. Directly west of
25 the trapping area is a 100-megawatt concentrated-solar energy facility encompassing approximately 630 hectares. The study area and much of the Lower Smoky Valley is comprised of land owned and managed by the US Bureau of Land Management. Further details pertaining to the study area can be found in
Chapter 1.
Trapping and handling procedures
I conducted mark-recapture trapping from 2012 through 2014, for approximately 3-4 months in the late spring and early summer (Table S.1 in appendix lists specific trapping dates for each session). I established 18 survey grids throughout the study area; 16 grids were established in 2012 with 2 additional grids established in 2013. I trapped at each grid for 6 sessions (with the exception of the 2 grids established in 2013 which had 4 sessions each) and each trapping session consisted of 3 consecutive nights of trapping and I trapped at each grid for 2 sessions per year.
Trapping grids consisted of 4 transect lines, each line 10 meters apart, and running east-west for 150 meters. Each line was comprised of 16 trap stations, one station every 10 meters. I set two traps at each trap station, for a total of 128 traps per grid. I used extra-long aluminum Sherman traps (3” wide by 3 ¾” tall by 9” long) and baited them with approximately 2 tablespoons of bird seed. I identified and marked each individual with an ear tag. Further details pertaining to my trapping and handling procedures can be found in Chapter 1.
Vegetation and habitat surveys
To collect fine-scale habitat data, I conducted 2-part vegetation surveys at each trap station once throughout the study duration, which occurred with a 3-meter radius circle centered at the trap station. I first identified to species and life stage each shrub, and measured distance from station center, height, canopy area and distance to nearest neighbor. For the second part, I estimated ground cover type and amount using 4 quadrat frames that were randomly placed within strata reflecting increasing distance
26 from plot center. I also collected a soil sample from each trap station. See Chapter 1 for more details of habitat data collected and procedures used.
Demographic and Diversity Estimation
To estimate abundance of each species in the rodent community, I first used program MARK
(White 2017) in order to incorporate detection probabilities and have more accurate estimates of abundance for each species-grid-session combination. However, the abundance estimates were often clearly incorrect for species that were rare in the community. The derived estimates for the rare species were sometimes exponentially larger than the number of unique individuals, or minimum number alive
(MNA), along with equally large confidence intervals. The erroneous abundances were likely due to the inability to estimate capture probability because of the small sample size for these rare species. Therefore,
I calculated the diversity statistics in two ways: the first used the derived estimates when the confidence intervals seemed reasonable and substituted with the MNA for the species-grid-sessions when capture rates were low and detection probabilities were not accurately estimated, and the second method used
MNA for all species-grid-sessions. When comparing the diversity statistics between the two methods, there were only slight differences, so the results reported here are based on MNA estimates to maintain consistency.
I calculated 4 diversity indices in order to capture a comprehensive representation of diversity at the Crescent Dunes (Morris et al. 2014). One index is a basic species richness (SR) index, which is calculated as the total number of species present. I also used the Shannon-Weaver Index (H), the
Simpson’s Dominance Index (D2), and the Simpson’s Evenness (E; see below for formulas).