Climate drives genetic diversity loss in American Pika (Ochotona princeps) populations in the Great Basin

Abstract

American pikas (Ochotona princeps) are small, thermally sensitive mammals that primarily live in montane and alpine environments. The Sierra Nevada lineage (O. p. schisticeps) has experienced numerous local extinctions, most of which occurred in hotter, drier regions such as the Great Basin. Few genetic studies have assessed these at-risk populations. This study aims to fill that gap by conducting fine-scale genetic analyses on populations in low-elevation Great Basin habitat in northwestern Nevada. Specifically, we: (i) quantified genetic diversity and structure among populations within O. p schisticeps, with particular focus on northwestern Nevada; and (ii) assessed the influence of primary productivity and climate-related variables on genetic diversity within O. p schisticeps, as well as at the broader species level. Great Basin populations exhibited the lowest levels of genetic diversity. Within O. p. schisticeps, population genetic diversity was positively correlated with annual precipitation—while at the species level temperature explained the most variation in genetic diversity. These results provide insight into climate-driven range contractions predicted for this species and inform conservation and management decisions.

Genetic diversity and structure among populations of American pikas is described, with emphasis on the Sierra Nevada lineage. Genetic diversity is positively correlated with precipitation within the Sierra Nevada lineage and negatively correlated with temperature at the species range.

The negative impacts of rapid, contemporary climate change are becoming increasingly apparent. While the effects are widespread, they have proven to be especially dire for montane species (Parmesan 2006; Beever et al. 2011; Santos et al. 2014). Rapid environmental changes leave species with few options including moving to more suitable habitat, adapting to the changes, or going extinct (Hewitt and Nichols 2005). Species with limited dispersal ability, especially those that live at elevational extremes and have fragmented populations, are unlikely to shift their range to more suitable conditions—unless they exhibit significant phenotypic plasticity, they must adapt to new environmental conditions in situ (Sgrò et al. 2011). One such species potentially facing dire consequences due to rapid climate change is the American Pika (Ochotona princeps, Richardson 1828).

American pikas have been described as harbingers of the effects of climate change due to their low thermal tolerance, fragmented habitat, and limited dispersal ability (Smith 1974; Castillo et al. 2016). Pikas have a high resting body temperature of 40.4 °C and an estimated upper lethal body temperature of 43.1 °C, making them vulnerable to even modest increases in ambient temperature (Smith 1974). While typically restricted to high elevation, they may persist at lower elevations in fractured-rock thermal refugia such as talus at elevations ranging from sea level to 3,000 m (Smith and Weston 1990). Pikas use cooler ambient temperatures within talus to escape the heat, but this may come with a cost of decreased foraging and dispersal opportunities (Castillo et al. 2016; Wright and Stewart 2018). Reliance on this specialized habitat may also result in small population sizes and genetic isolation due to patchy distribution of climatically suitable talus and their low dispersal abilities (Castillo et al. 2016).

In recent years, research efforts have sought to determine what factors influence the distribution and vulnerability of American Pika populations (Beever et al. 2003, 2011; Jeffress et al. 2013; Castillo et al. 2016; Schwalm et al. 2016; (Smith et al. 2019). Previous studies have documented pikas adapting to rising temperatures by behavioral thermoregulation and adjusting their foraging strategies, such as limiting activity during the hottest times of the day (Smith 1974; Camp et al. 2020). However, the potential of behavioral plasticity to adequately mitigate the effects of climate change is unclear for this species, especially in the hotter parts of its range. Variable environmental conditions across the broad geographic range of the species—as well as differences in effective population size—may further impact pika adaptive potential, i.e., the potential for populations to adapt to selective pressure via genetic changes (Henry and Russello 2013). While the effects of climate change on American Pika populations have been widely studied, the mechanisms underlying extirpations and changes in distribution are still not well understood, especially for populations in the Great Basin (the region of internal drainage in the western United States bounded by the Sierra Nevada (SN) and Cascade mountain ranges on the west and Rocky Mountain range on the east; Smith 2020). To predict pika persistence, understanding of local differences in forage quality, population connectivity, microclimate, and adaptive potential is needed (Henry and Russello 2013; Castillo et al. 2016; Millar et al. 2018; Wright and Stewart 2018).

The SN lineage (O. p. schisticeps) was widespread throughout the Great Basin during the Pleistocene, but became restricted to mountaintops of the Great Basin during the Holocene with numerous known historic and recent extirpations and high levels of current isolation and population structure (Grayson 2005; Beever et al. 2011; Castillo et al. 2016; Nichols et al. 2016; Stewart et al. 2017; Millar et al. 2018). Occupancy across diverse habitats including atypical habitats such as hot and dry sites within the Great Basin; lava tubes in Craters of the Moon National Monument and Lava Beds National Monument; low-elevation ore dumps in Bodie, California; and low-elevation sites in the Columbia River Gorge suggests that pikas may be more ecologically plastic than previously understood (Millar et al. 2014; Jeffress et al. 2017; Millar and Westfall 2021). Yet despite numerous extirpations in the Great Basin ecoregion, some populations still exist outside of the expected climatic tolerance limit for the species. Jeffress et al. (2017) reported a recent discovery of O. p. schisticeps populations in previously undocumented areas in northwestern Nevada (NWNV) that exist at the edge of the known temperature tolerance limits for American pikas (Smith and Weston 1990). Since their discovery, those populations have been closely monitored by the Nevada Department of Wildlife (NDOW).

Although American pikas have been surveyed extensively in the Great Basin portion of the range of O. p. schisticeps, little fine-scale population genetic work has been conducted there or in the southern SN (USFWS 2010; Beever et al. 2011; Nichols et al. 2016; Jeffress et al. 2017). To date, few genetic studies have focused on that subspecies outside of the Bodie Hills, California site in the eastern SN (Castillo et al. 2016; Klingler et al. 2021; Klingler et al. 2023), despite the acknowledged need to do so (USFWS 2010). A fine-scale genetic analysis on populations in the northern Great Basin, where many extirpations have been documented, is lacking. The SN and Great Basin ecoregions differ widely in climate (e.g., temperature, precipitation) and environment (e.g., vegetation, rock type, rock structure), presenting a unique opportunity to evaluate the effects of climatic and ecological regimes on genetic diversity and structure of populations, while controlling for differences among subspecies due to deeper phylogeographic history (Galbreath et al. 2010).

Here, we present the results of a fine-scale population genetic study of 23 populations of pikas, focusing on the O. p. schisticeps subspecies and the Great Basin populations in particular, with a comparison to the broader geographic range of the species. Genetic diversity within metapopulations of highly specialized taxa such as pikas has been found to be largely influenced by functional connectivity (Castillo et al. 2016; Schwalm et al. 2016; Rodhouse et al. 2018). This functional connectivity is highly dependent on both within-site-level characteristics and suitable dispersal habitat which is often limited by temperature for pikas (Smith et al. 1974; Castillo et al. 2016). Therefore, we hypothesized that pika populations in environmental extremes would be smaller compared to those in less extreme climates and, as a result, would have lower genetic diversity and increased genetic structure. Specifically, we predicted that populations experiencing more extreme temperatures and lower precipitation would have lower genetic diversity, particularly in more arid portions of the species range. Likewise, we predicted that populations in locations with extremely high precipitation may also have lower genetic diversity than those with more moderate precipitation levels due to snowpack persisting longer into the summer months, limiting pika foraging and dispersal (Castillo et al. 2016; Schwalm et al. 2016). While temperature is likely to affect pikas directly (Smith 1974), climatic variables also affect pikas indirectly through environmental effects such as primary productivity. We predicted that populations from areas with greater primary productivity, quantified as time-integrated NDVI (normalized difference vegetation index, a metric of greenness), would have greater genetic diversity. Genetic diversity is required if populations are to adapt to environmental change (Willi et al. 2006); thus, comparisons among sites allow inferences about evolutionary potential of populations and the relative level of threat currently imposed by climate change (Visser 2008; Dawson et al. 2011; Butt et al. 2016). Further, as some sites become hotter and drier, the Great Basin may provide insight into the future of those sites that are currently less extreme but becoming more so.

Materials and methods.

Study sites.

We included 20 sites to represent the environmental and geographic variation experienced by O. p. schisticeps within California, Oregon, and Nevada (Table 1; Fig. 1). We assigned sites to the following regions: northwestern Great Basin (NWGB); southern Great Basin (SGB); SN; and Cascade Range (CR). For some analyses, we included an additional categorization for sites only occurring in NWNV. Therefore, the NWGB data set includes all NWNV sites, but NWNV excludes sites in California and Oregon (Fig. 1). These sites range in mean elevation from 1,486 to 2,915 m (Table 1). NWGB and SGB are characterized as low-elevation and xeric with talus composed of basalt rock. The sagebrush steppe vegetation is mostly comprised of Big Sagebrush (Artemisia tridentata) with a grass and forb understory. SN is characterized as subalpine to alpine with large patches of granitic talus and alpine-specialized forbs and grasses. We included an additional 8 sites from 3 other subspecies found throughout the species range for genetic diversity comparison: 6 O. p. princeps; 2 O. p. saxatilis; and 1 O. p. fenisex (Table 1).

Table 1.

Study sites included in our data sets with 4-letter site codes, assigned region, elevation, number of individuals genotyped, mean observed heterozygosity (H_o), and mean observed gene diversity (H_s). Mean allelic richness (Ar) is included for populations with 7+ individuals genotyped.

Region	Subspecies	Study site	Site code	Elevation (m)	n	Mean Ho	Mean Hs	Mean Ar
Cascade Range	schisticeps	Lava Beds NM	LABE	1,486	47	0.55	0.62	3.01
		Lassen Volcanic NP	LAVO	1,910	99	0.57	0.65	3.32
		Lassen NF	LNF	1,675	7	0.54	0.50	2.53
	princeps	Crater Lake NP	CRLA	1,728	122	0.47	0.53	2.82
Northwest Great Basin (NWGB)	schisticeps	Hart Mountain Range	HMAR	1,631	44	0.35	0.38	2.05
		Warner Mountains	WAMT	1,757	12	0.40	0.53	2.54
Northwest Nevada (NWNV)	schisticeps	Crook’s Carter	CRCA	1,705	4	0.36	0.43
		Crook’s Carter/Vya	CRCA_VYAa	1,710	10	0.47	0.42	2.29
		Grassy Canyon	GRCA	1,793	12	0.58	0.55	2.66
		Hays Canyon Range	HCRA	1,747	15	0.47	0.52	2.68
		Macy Flat	MAFL	1,776	2	0.57	0.30
		Massacre Rim	MARI	1,800	2	0.33	0.50
		Nut Mountain	NUMO	1,825	2	0.62	0.44
		Nut Mountain/Grassy Canyon	NUMO_GRCAa	1,793	14	0.59	0.54	2.52
		Sheldon Wildlife Refuge	SHWR	1,855	54	0.49	0.55	2.67
		South Sheldon	SOSH	1,844	7	0.42	0.45	2.37
		Vya	VYA	1,715	6	0.51	0.39
Southern Great Basin (SGB)	schisticeps	Mount Jefferson	MTJE	2,282	10	0.36	0.36	2.05
		Sweetwater	SWWA	2,230	4	0.74	0.67
		White Mountains	WTMT	2,261	21	0.49	0.54	2.71
Sierra Nevada (SN)	schisticeps	Echo Lake	ECHO	2,287	6	0.56	0.53
		Sequoia and Kings Canyon NP	SEKI	2,756	31	0.50	0.67	3.43
		Yosemite NP	YOSE	2,915	78	0.67	0.71	3.60
Northern Rocky Mountains (NRM)	princeps	Salmon-Challis NF	CHNF	2,617	9	0.60	0.69	3.55
		Grand Teton NP	GRTE	2,432	194	0.63	0.70	3.60
		Rocky Mountain NP (North)	ROMO_N	3,183	66	0.64	0.73	3.74
		Yellowstone NP	YELL	2,299	26	0.68	0.77	4.00
Snake River Plane	princeps	Craters of the Moon NP	CRMO	1,733	60	0.50	0.59	3.02
Southern Rocky Mountains	saxatilis	Great Sand Dunes NP	GRSA	2,801	54	0.64	0.76	3.79
		Rocky Mountain NP (South)	ROMO_S	3,070	157	0.61	0.75	3.91

^aSites were combined after STRUCTURE analyses for subsequent genetic diversity analyses. NP = National Park; NM = National Monument; NF = National Forest.

Fig. 1.

Map of sites where fecal samples of Ochotona princeps schisticeps were collected for genetic analysis between 2010 and 2019 in parts of the NWGB, CR, and SGB. Samples in NWNV and SN were collected in summer of 2019 and 2021. Subregions (NWGB, NWNV, SN, SGB, and LABE) are indicated by circles. Refer to Fig. 3 for NWNV site codes.

Sample collection.

We collected fecal samples for genetic analyses with a combination of targeted and opportunistic sampling methods. Sites within the NWNV region were selected using information on pika occupancy from surveys conducted by NDOW between 2014 and 2021. In the other 3 regions, targeted sampling involved exhaustive searches of potential suitable pika habitat identified using remotely sensed data (Jeffress et al. 2013). Opportunistic sampling involved searching talus areas while moving between targeted sampling localities. We aimed to obtain a minimum of 10 to 20 samples per site. Within each site, we collected from multiple patches >50 m apart to reduce the chance of collecting duplicate samples from the same individual. To minimize the risk of contamination between individuals we collected samples from a single defecation event, i.e., found in a single, discrete pile, all pellets similar in appearance, and not in contact with older fecal pellets. Samples were collected using forceps sterilized using a 10% bleach solution followed by flame sterilization. GPS coordinates of each sample were recorded with up to ~15 m accuracy.

We collected samples throughout NWNV, the Warner Mountains (WAMT), and Sequoia Kings Canyon National Park (SEKI) between May and August 2021. NDOW provided samples from NWNV and the Sweetwater Range (SWWA) that were collected between 2015 and 2020. Samples from Mount Jefferson, Nevada (MTJE); the White Mountains, California (WTMT); and Echo Lake, California (ECHO) were collected in the summer of 2019. We combined this new data set with previously genotyped samples collected in Yosemite National Park (YOSE); Lava Beds National Monument (LABE); and Lassen Volcanic National Park (LAVO), as well as sites outside O. p. schisticeps range between June and September 2010 to 2014 (Castillo et al. 2016).

DNA extraction and genotyping.

We extracted DNA from fecal samples collected after 2015 using the Qiagen DNeasy Blood and Tissue sampling Kit (Qiagen Inc., Valencia, California) following the manufacturer’s extraction protocol with modifications for fecal samples: fecal pellets were processed whole and placed in a rotisserie during the digestion step. Samples were extracted in a UV laminar flow hood in an isolated room within the lab space. We genotyped individuals at 22 microsatellite loci in 4 multiplex polymerase chain reactions (PCR) using a Qiagen Multiplex PCR kit (Qiagen) following (Castillo et al. 2016). We visualized PCR products using an ABI 3730 capillary sequencer, then scored genotypes using GENEMAPPER V4.1 (Applied Biosystems). We amplified each sample a minimum of 3 times for each multiplex PCR. Each allele was considered confirmed if it was typed at least twice in independent amplifications. If an allele was seen only once, we repeated PCR for those individuals to construct consensus genotypes. We identified samples with more than 2 confirmed microsatellite peaks at any locus as contaminated and removed them from further analysis. All individuals with >50% missing data were removed from further analysis. We calculated the probability of identity among full siblings (P_(ID)sib) for each locus and combined loci using the R package “PopGenUtils” (Tourvas 2021). We screened for duplicate individuals and removed all but 1 genotype for each set, identified as those that differ by fewer than 2 loci. We then tested for linkage disequilibrium and significant deviations from expected Hardy–Weinberg genotype frequencies using the “adegenet” package (Jombart 2008) in R.

Population genetic analysis.

We evaluated population genetic structure, genetic diversity, and gene flow between and within all sites and regions within our data set. We used the program STRUCTURE (Pritchard et al. 2000) to determine the optimal number of genetic clusters and assign individuals to genetic clusters based on allele frequencies and deviations from expectations of Hardy–Weinberg equilibrium. We performed 3 separate analyses with 10 independent runs for each K (assumed number of genetic clusters), with admixture models and correlated allele frequencies and incorporating the sampling sites as prior information (LOCPRIOR). Each run included a burn-in of 50,000 Markov chain Monte Carlo steps followed by an additional 50,000 iterations while K ranged from 1 to 22 for the full O. p. schisticeps data set, 1 to 11 for NWGB, and 1 to 9 for NWNV.

We determined the best-supported K number of clusters using STRUCTURE HARVESTER (Earl and VonHoldt 2012) according to the delta K method (Evanno et al. 2005). Individual membership assignments estimated in STRUCTURE were summarized in the program CLUMPP (Jakobsson and Rosenberg 2007) with either the greedy or large K greedy algorithm when K > 10. Additionally, we performed a discriminant analysis of principal components (DAPC) on the NWNV data set using the R package “adegenet” (Jombart 2008). This approach differs from STRUCTURE in that clusters are assigned based on minimizing genetic differentiation within groups and maximizing differentiation between groups, rather than Hardy–Weinberg assumptions.

We quantified genetic distance as F_ST and chord distance (Dch) using “hierfstat” (Goudet 2005) in R. Next, we assessed isolation by distance with Mantel tests of correlation between individual genetic distance (Dps) and geographic distance (log(m)) matrices using “ecodist” (Goslee and Urban 2015) in R.

Genetic diversity.

We evaluated the relationship between allelic richness and environmental and climatic variables for 15 populations spanning the range of O. princeps throughout the northwestern United States, including 4 of the 5 recognized subspecies. Climatic and environmental variables considered include: minimum, mean, and maximum daily temperature; mean annual precipitation; NDVI; elevation; minimum and maximum daily vapor pressure deficit (VPD; the difference between the amount of moisture in the air and the amount of moisture that the air can hold when it is saturated); daily global shortwave solar radiation; and atmospheric transmittance (cloudiness). We obtained climate data from PRISM Climate Group (http://prism.oregonstate.edu) and NDVI and elevation data from the United States Geological Survey Earth Resources Observation and Science Center (USGS EROS). PRISM data represent the 30-year normal from 1991 to 2020, while NDVI data were averaged over the years 2001 to 2020. For each raster data set, we calculated the mean values for a 20-km radius surrounding the centroid of our sample localities in R.

We calculated allelic richness for populations with 7 or more individuals using the “hierfstat” (Goudet et al. 2021) package in R, resulting in rarefied estimates for a sample size of 7. Paired t-tests with Bonferroni corrections were run for each pair of populations to test for a significant difference in mean allelic richness. We evaluated the relationships between all climatic and environmental variables and allelic richness using generalized linear models and model selection using AICc with the “MuMIn” package (Barton 2020) in R. We first evaluated each variable independently using simple linear regression. Where appropriate, we included log-transformed and quadratic terms in our independent variables for simple linear regression. All variables determined to have a significant relationship with allelic richness were included in our multivariate regression models. Model selection was performed on all variable combinations, excluding pairs of variables with a correlation coefficient >0.75. We repeated these analyses for 2 sets of populations: O. p. schisticeps populations only; and all populations with a sample size >7 genotyped individuals. For the latter analyses, we used the average of the NWNV sites for both our independent and dependent variables.

Results

Sample collection, DNA extraction, and genotyping.

We collected 52 new samples in NWNV/WAMT and 75 samples in SEKI. Including the previously collected samples, we extracted and genotyped 376 samples for this study. Additionally, we used 1,072 samples that had been genotyped previously. After removing duplicate individuals and samples that failed to amplify consistently (>50% missing data), our final data set contained 82 individuals combined with previously genotyped samples for a total of 1,154 samples (Table 1). Most of our sample loss was due to inconsistent amplification. We removed 7 loci that failed to amplify consistently, for a total of 15 loci. The cumulative probability of identity among full siblings (P_(ID)sib) was $7.14\times {10}^{-8}$ for the full species range data set, $1.29\times {10}^{-6}$ for O. p. schisticeps, and $5.04\times {10}^{-5}$ for NWNV. A P_(ID)sib < 0.001 was achieved with approximately 6, 7, and 10 loci for the full, O. p. schisticeps, and NWNV data sets, respectively, indicating that 15 loci were more than sufficient to identify individuals.

Population structure.

From our STRUCTURE analysis of the O. p. schisticeps data set, we determined K = 2 was the optimum number of clusters using the delta K method (Supplementary Data SD1). Individuals either clustered with the NWGB or the SN, with limited admixture (Supplementary Data SD2). Notably, WAMT included individuals assigned to either Great Basin or SN clusters as well as individuals with 50/50 assignment to both clusters (Supplementary Data SD2). The SGB populations, WTMT and MTJE, clustered with the SN populations (Supplementary Data SD2). In the data set of NWGB samples only, K = 3 was the best supported, with K = 9 also supported (Supplementary Data SD3). In the K = 3 model, all localities showed some admixture, with HMAR being the most distinct (Supplementary Data SD4). In the K = 9 model, WAMT and HCRA were each assigned to distinct clusters that shared little admixture with other populations (Supplementary Data SD5). A closer examination of fine-scale structure within only the NWNV data set showed that K = 3 was the best supported, with K = 6 also showing some support (Supplementary Data SD6). In the K = 3 model, NUMO, GRCA, CRCA, and VYA again cluster together (Fig. 2A). However, in the K = 6 model, NUMO + GRCA and HCRA show little admixture while the remaining populations were more admixed (Fig. 2B; Fig. 3). Due to the lower sample size, proximity of these populations, and near identical cluster assignments in all 3 analyses we combined NUMO + GRCA and CRCA + VYA for all subsequent analyses.

Fig. 2.

STRUCTURE bar plots where (A) K = 3 and (B) K = 6 for populations of Ochotona princeps schisticeps in NWNV. Each bar represents an individual while clusters are defined by color. Sampling locations within NWNV are indicated below the chart with black lines indicating the boundary between sites.

Fig. 3.

Map of all sites sampled for Ochotona princeps schisticeps in NWNV (bounding box in inset). The map shows a pie plot of NWNV STRUCTURE results of K = 6 showing relative proportion of cluster assignment for each site (n = 9): Crook’s Carter (CRCA), Grassy Canyon (GRCA), Hay’s Canyon (HCRA), Macy Flat (MAFL), Massacre Rim (MARI), Nut Mountain (NUMO), Sheldon Wildlife Refuge (SHWR), and South Sheldon (SOSH).

The DAPC analysis of the NWNV data set differentiated 4 or 5 genetic clusters. K = 4 was chosen as the optimal number of clusters for all sites within this data set due to higher assignment probability for individuals, which differs from the results of the STRUCTURE analysis slightly with more admixture overall, but HCRA again is the most genetically distinct (Supplementary Data SD7).

Population differentiation.

Pairwise F_ST ranged from 0.053 (CRCA_VYA, MARI) to 0.506 (MTJE, SOSH; MTJE, LNF; Supplementary Data SD8). Pairwise Dch ranged from 0.118 (HMAR, SHWR) to 0.852 (MAFL, ECHO; Supplementary Data SD8). Overall, comparisons with MTJE had the highest F_ST and Dch values. Mantel r values for genetic similarity among individuals (Dps) versus geographic distance (log(m)) ranged from −0.344 (HCRA, P = 0.004) to 0.085 (NUMO + GRCA, P = 0.748; Supplementary Data SD9). There was a general negative trend between genetic similarity and geographic distance between individuals, with all sites showing statistically significant relationships except MTJE, CRCA–VYA, NUMO–GRCA, and SOSH (Supplementary Data SD9). Additionally, Mantel r values were less negative in SEKI and YOSE compared to NWNV populations (Supplementary Data SD9).

Genetic diversity.

Mean allelic richness within O. p. schisticeps ranged from 2.05 (HMAR and MTJE) to 3.60 (YOSE; Table 1). Great Basin populations exhibited the lowest mean allelic richness while SN populations exhibited the highest (Fig. 4). Analysis of variance confirmed that there was a significant difference in allelic richness among populations (P < 0.0001), with YOSE and SEKI having significantly higher allelic richness than HMAR, MTJE, and SOSH (Supplementary Data SD10). Allelic richness across O. p. schisticeps was best predicted by precipitation, with allelic richness positively correlated with increased mean annual precipitation (Supplementary Data SD11 and SD12; Fig. 4). Great Basin populations experience both the lowest allelic richness and mean annual precipitation (Fig. 4). VPD, temperature, NDVI, and elevation were also significantly correlated with allelic richness (Supplementary Data SD11). Multivariate regression likewise included precipitation as the best-supported model, with multivariate models including precipitation + temperature (tmin, tmean, or tmax) or precipitation + elevation within 2 ΔAICc of the model containing only precipitation (Supplementary Data SD11). All other model combinations were beyond the 2 ΔAICc threshold.

Fig. 4.

Linear regression of mean allelic richness versus mean annual precipitation for Ochotona princeps schisticeps by region. Colors represent regions for each sampling locality (points).

When including all sampled subspecies and regions, mean allelic richness ranged from 2.05 (HMAR and MTJE) to 4.00 (YELL; Table 1). Allelic richness had a significant negative correlation with temperature and VPD, a significant positive correlation with NDVI and elevation, and a significant quadratic relationship with precipitation (Supplementary Data SD13 and SD15). The best-supported predictor variable was temperature (order of support: tmean, tmax, and tmin; Fig. 5). Multivariate model selection supported NDVI + elevation as the top model, while models containing only temperature were within 2 ΔAICc (Supplementary Data SD14). Allelic richness was positively correlated with both NDVI and elevation (Supplementary Data SD15).

Fig. 5.

Linear regression of mean annual temperature (°C) and allelic richness of all sampled populations of Ochotona princeps. Populations are indicated above their corresponding points on the graph. Points are colored by region. Additionally, shapes indicate the subspecies of each point on the graph.

Discussion

We examined population structure and genetic diversity of O. p. schisticeps throughout most of its range, including Oregon, Nevada, and California. We assessed the influence of primary productivity and climate-related variables on genetic diversity within O. p. schisticeps, as well as at the species level to better understand both local determinants and broader, range-wide trends. In support of our hypothesis that genetic diversity of this subspecies would be shaped by environmental variation, particularly at the extremes, we found that genetic diversity was best predicted by mean annual temperature and mean annual precipitation. Genetic diversity decreased with increased mean annual temperature and increased with increased mean annual precipitation. At the subspecies level, likely due to the general aridity of this system as compared to other lineages sampled, we did not see declines in genetic diversity at the highest precipitation levels. Populations in the Great Basin exhibited low levels of genetic diversity and population connectivity with some populations exhibiting high levels of isolation relative to populations in the SN. Great Basin populations experience both the hottest temperatures and the lowest mean annual precipitation of any of the regions in our data set. These results suggest that populations in this region of the NWGB are especially vulnerable to climate-driven range contraction. In recent regional surveys, fewer than 20% of talus sites remained occupied (Jeffress et al. 2017). A hotter and drier climate can limit the amount of suitable habitat which, in turn, limits gene flow between these already fragmented metapopulations (Castillo et al. 2016). Loss of population connectivity can decrease genetic diversity, increase the risk of inbreeding depression, and subsequently increase the risk of extirpation.

Population structure.

All analyses identified the NWNV populations as genetically distinct from all other populations of O. p. schisticeps. This is consistent with the degree of geographic isolation characteristic of these populations. Notably, at K = 2—the most strongly supported number of clusters identified by Program STRUCTURE—individuals in WAMT appeared strongly admixed between NWNV and SN populations, with a 50/50 assignment to both SN and NWNV clusters. While this could suggest that contemporary gene flow between NWNV and SN has occurred, it is most likely not recent. We reach this conclusion for 2 reasons: (1) our sample size from WAMT is minimal with only 12 individuals sampled from this region; and (2) at higher values of K, WAMT is a more distinct population cluster compared to other populations in the NWGB (Supplementary Data SD4 and SD5). Nevertheless, this result warrants further investigation into the level of connectivity between WAMT and NWNV, particularly with populations at the southern extent of WAMT and HCRA. We also observed moderate to high population structure among the 9 subpopulations sampled within NWNV. The landscape in this region is highly fragmented, with talus forming distinct patches separated by non-habitat sometimes greater than the known pika dispersal capacity of 2 to 5 km (Smith and Weston 1990; Castillo et al. 2016). Connectivity between these patches likely continues to be reduced as conditions become hotter and drier (Castillo et al. 2016). Our results indicate that many of the subpopulations within NWNV are not well connected, particularly HCRA. HCRA is the most geographically isolated from the other populations in our analysis (31.4 km from NUMO + GRCA and 49.5 km from CRCA + VYA). Additionally, the potential habitat surrounding this population is largely unoccupied (Jane Van Gunst, personal observation). Therefore, our observations of increased isolation are consistent with field observations and occupancy surveys and may indicate a population at particularly high risk. In contrast, NUMO + GRCA and CRCA + VYA each appear to have pikas freely dispersing between localities, thus acting as interconnected metapopulations.

Comparison of individual genetic similarity within sites (Dps) versus geographic distance (log m) using Mantel tests showed that isolation by distance varies considerably among sites, suggesting isolation by resistance or barrier in addition to distance, as reported previously for pikas (Castillo et al 2016). Pikas are philopatric by nature but typically will disperse to their nearest patch (Peacock 1997; Peacock and Smith 1997). Dispersal is limited by the spatial configuration of habitat patches as well as the conditions between patches. Temperature alone could act as a physiological barrier to pika dispersal (Smith 1974). Topographic complexity such as cliffs or ravines and bodies of water can also act as barriers to pika dispersal (Castillo et al. 2016). In NWNV, several large alkaline lakes in the valleys between mountain ranges likely present dispersal barriers. Much of the landscape in NWNV is managed by the Bureau of Land Management in a multiuse framework where livestock grazing is a dominant use. Livestock grazing near pika-occupied sites may negatively alter foraging behavior and forage availability of pikas (Millar 2011). While these sites occur at the lowest elevations where local climate may make extirpation more likely (Beever et al. 2003), land-use change can exacerbate climatic changes (Rowe 2007) altering the pace and location of extirpation processes. Furthermore, in the Great Basin, there has been an expansion of non-native grasses, i.e., Cheatgrass (Bromus tectorum), that has significantly altered fire regimes (Morris and Rowe 2014) which may alter vegetation and dispersal patterns.

Genetic diversity and its relationship with climate and primary productivity.

Populations in both the SGB and the northern Great Basin exhibited lower mean allelic richness when compared to the SN and southern CR; SGB populations were more genetically similar to the SN than to the NWGB populations. Continuous habitat area is an important predictor of pika persistence (Beever et al. 2003; Stewart et al. 2015; Schwalm et al. 2016). Great Basin populations are naturally more fragmented than populations in the SN where the talus is more continuously distributed. In such systems, a decrease in gene flow between populations will increase population structure, leading to a decrease in genetic diversity as a result of genetic drift and inbreeding (Frankham 2005). While we did not assess habitat area due to the lack of fine-scale habitat data for many of our sites, the best predictor of allelic richness for O. p. schisticeps was precipitation, such that allelic richness increased with increasing mean annual precipitation. Likewise, there was a negative relationship with both VPD and maximum annual temperature. VPD reflects how dry (high VPD) or humid (low VPD) the air is and contributes to environmental water balance through its role in evapotranspiration, particularly in plants. The SN and southern CR populations exhibit both the highest allelic richness values and highest mean annual precipitation within O. p. schisticeps. The exception is Lassen National Forest (LNF), which has relatively high mean annual precipitation but low allelic richness—likely due to the higher degree of habitat fragmentation compared to the SN sites. In contrast, Great Basin populations all experience similarly low mean annual precipitation and high temperatures, regardless of latitude and elevation. Our results are consistent with previous studies indicating that precipitation plays an important role in pika distribution and persistence (Erb et al. 2011; Wilkening et al. 2011). In addition to precipitation, water balance was shown to be an important predictor of pika occupancy in the Great Basin (Beever et al. 2016). These results suggest a more complex relationship between temperature, precipitation, and water balance with respect to habitat quality.

In our analysis of genetic diversity across 4 subspecies of the American pika, we found additional support for our hypothesis that genetic diversity was lower at environmental extremes. Mean annual temperature was the best predictor of allelic richness across the species range, as allelic richness decreased with increasing mean annual temperature. While precipitation was not among the top models at the species range level it was quadratically related to precipitation, supporting our prediction that at the species level, genetic diversity would be highest at moderate precipitation. The top multivariate model did not contain temperature or precipitation, but rather elevation + NDVI. As temperature decreases predictably with increasing elevation and NDVI is affected by precipitation and temperature, this further supports the hypothesis that extreme precipitation may result in persistent snowpack, reducing foraging and dispersal opportunity. In the Great Basin, primary productivity increases with precipitation and that ecoregion is one of the more sensitive in the western United States to fluctuations in precipitation (Maurer et al. 2020). Therefore, differences in our O. p. schisticeps and species-level analyses are not surprising because variation in climate-related variables is significantly greater at the species level. However, temperature was consistently among the top-ranked models. Postglacial pika extinctions were driven by increased temperatures and decreased moisture (Grayson 1993) and these factors still appear to drive trends in genetic diversity and population persistence for pikas.

Our results, in conjunction with the recent evidence of population extirpation within the Great Basin and ongoing monitoring efforts, suggest that the future is unclear for American pikas in NWNV. Many of these populations are experiencing high site turnover (Jane Van Gunst, personal observation). Additionally, pika detection in NWNV was significantly more difficult than in more ideal pika habitat such as Sequoia—Kings Canyon in the southern SN. American pikas, especially the SN lineage, experience a wide range of environmental and geographic extremes. Much of the NWGB is outside of the typical characterization of suitable pika habitat, yet they still exist there. However, existence does not equal future persistence. Our results in conjunction with past studies provide further evidence that climate change can alter connectivity between populations (Epps et al. 2006; Rubidge et al. 2012). Although pikas still exist at the edge of their climatic tolerance in the Great Basin, evidence of population structure and lower levels of genetic diversity in NWNV indicate that these populations are becoming increasingly isolated. Particularly when populations are small, low genetic diversity indicates limited movement between populations and, as more patches become unoccupied, connectivity is further decreased. Maintaining connectivity between populations is key to maintaining genetic diversity and thus having the underlying genetic variation necessary to adapt in a rapidly changing environment. Therefore, in the face of intensifying climate change, conservation managers must prioritize strategies that preserve connectivity, promote movement among populations, and thus enhance evolutionary potential.

Supplementary data

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1. Delta K plot associated with the Ops data set.

Supplementary Data SD2. STRUCTURE bar plot of the Ops data set with K = 2 clusters.

Supplementary Data SD3. Delta K plot associated with the NWGB data set.

Supplementary Data SD4. STRUCTURE bar plot for the NWGB data set with K = 3 clusters.

Supplementary Data SD5. STRUCTURE bar plot for the NWGB data set with K = 9 clusters.

Supplementary Data SD6. Delta K plot associated with the NWNV data set.

Supplementary Data SD7. DAPC scatterplot and bar plot of NWNV populations, K = 4.

Supplementary Data SD8. Pairwise F_st (below diagonal) and Dch (above diagonal) values for all Ochotona princeps schisticeps populations. Pairwise Dch values range from 0.118 (HMAR:SHWR) to 0.852 (MAFL:ECHO). Pairwise F_ST values from 0.053 (CRCA_VYA:MARI) to 0.506 (MTJE:SOSH, MTJE:LNF).

Supplementary Data SD9. Mantel test summary of genetic similarity (Dps) versus geographic distance (m) between individuals in all populations.

Supplementary Data SD10. Allelic richness paired t-test P-values for all populations with n > 7 individuals genotyped at each locus.

Supplementary Data SD11. Summary of simple linear regression results showing top supported variables in order of delta AIC values.

Supplementary Data SD12. Summary of multivariate model selection showing results for delta AIC values <2.

Supplementary Data SD13. Summary of univariate model selection on the full O. princeps data set.

Supplementary Data SD14. Summary of multivariate model selection on the full O. princeps data set showing results for delta AICc values <2.

Supplementary Data SD15. Linear regression of AR versus elevation, NDVI, max vapor pressure deficit, and mean precipitation (mm) across all O. princeps.

Author contributions

ENK: formal analysis, investigation, methodology, visualization, writing—original draft, writing—review & editing. JVG: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, writing—review & editing. MJH: investigation, writing—review & editing. YL: formal analysis, investigation, writing—review & editing. MV: formal analysis, writing—review & editing. RSC: formal analysis, writing—review & editing. CWE: conceptualization, funding acquisition, project administration, supervision, writing—review & editing. JACV: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, visualization, writing—original draft, writing—review & editing.

Funding

Funding was provided by San Jose State University, Nevada Department of Wildlife, and the Animal Welfare Institute.

Conflict of interest

None declared.

Data availability

Climate data are freely available from PRISM Climate Group (http://prism.oregonstate.edu) and NDVI and elevation data from the US Geological Survey Earth Resources Observation and Science Center (USGS EROS).