Growth, drought response, and climate‐associated genomic structure in whitebark pine in the Sierra Nevada of California

Abstract Whitebark pine (Pinus albicaulis Engelm.) has experienced rapid population declines and is listed as threatened under the Endangered Species Act in the United States. Whitebark pine in the Sierra Nevada of California represents the southernmost end of the species' distribution and, like other portions of its range, faces threats from an introduced pathogen, native bark beetles, and a rapidly warming climate. Beyond these chronic stressors, there is also concern about how this species will respond to acute stressors, such as drought. We present patterns of stem growth from 766 large (average diameter at breast height >25 cm), disease‐free whitebark pine across the Sierra Nevada before and during a recent period of drought. We contextualize growth patterns using population genomic diversity and structure from a subset of 327 trees. Sampled whitebark pine generally had positive to neutral stem growth trends from 1970 to 2011, which was positively correlated with minimum temperature and precipitation. Indices of stem growth during drought years (2012 to 2015) relative to a predrought interval were mostly positive to neutral at our sampled sites. Individual tree growth response phenotypes appeared to be linked to genotypic variation in climate‐associated loci, suggesting that some genotypes can take better advantage of local climatic conditions than others. We speculate that reduced snowpack during the 2012 to 2015 drought years may have lengthened the growing season while retaining sufficient moisture to maintain growth at most study sites. Growth responses may differ under future warming, however, particularly if drought severity increases and modifies interactions with pests and pathogens.


| INTRODUC TI ON
Whitebark pine (Pinus albicaulis Engelm., Pinaceae subgenus Strobus) is a long-lived tree found in high-elevation forests of western North America. Because of the multiple roles that whitebark pine plays in subalpine ecosystems, it has been described as a keystone and foundation species. Over much of its range whitebark pine is declining due to a suite of stressors including a non-native, invasive pathogen (Cronartium ribicola J.C. Fisch., the causal agent of white pine blister rust), outbreaks of native pests (mountain pine beetle, Dendroctonus ponderosae Hopkins), altered fire regimes, and a rapidly changing climate (Keane et al., 2012;Schwandt et al., 2010;Tomback & Achuff, 2010). As a result, whitebark pine has been listed as an en- The Sierra Nevada of California is the southernmost range extent of whitebark pine, where it experiences warmer, drier conditions relative to other portions of its distribution (Arno & Hoff, 1989).
Whitebark pine in the Sierra Nevada is genetically differentiated from other regions, which may reflect adaptations to climatic differences and reduced gene flow (Richardson et al., 2002;Syring et al., 2016). Additionally, whitebark pine in the Sierra Nevada ecoregion of California and Nevada (https://www.epa.gov/eco-resea rch/ ecore gions -north -america, accessed 3/9/2022) currently appears to be less impacted by biotic and abiotic stressors relative to other portions of its range. A recent synthesis of whitebark pine status from a US national forest inventory database showed slightly over half of the standing whitebark pine ≥12.5 cm DBH (stem diameter at breast height [1.37 m]) were dead; however, within the Sierra Nevada alone only <10% of stems were dead (Goeking & Izlar, 2018). The lower mortality in the Sierra Nevada likely reflects the lower incidence of stressors. While white pine blister rust has been present in the Sierra Nevada at least since the 1960s (Kinloch, 2003), the proportion of infected trees across the Sierra Nevada is highly variable. Maloney et al. (2012) found that approximately 35% of whitebark pine were infected within the relatively mesic Tahoe Basin, which contrasts with ≤1% of infected whitebark pine in the cooler, drier sites of the central and southern Sierra Nevada (Dudney et al., 2020;Nesmith et al., 2019). There are currently few published reports of mountain pine beetle outbreaks in whitebark pine in the Sierra Nevada (Meyer et al., 2016;Millar et al., 2012). Historical fire return intervals in subalpine forests of the Sierra Nevada are estimated to be greater than a century (Safford & Van de Water, 2014), so fire exclusion over the past century is unlikely to have greatly affected whitebark pine in this ecoregion.
The persistence of whitebark pine may be impacted by ongoing changes in climate (Chang et al., 2014;Keane et al., 2017;McKenney et al., 2007). Subalpine forests are generally thought to be energy limited; as a result, warming has been associated with increasing tree growth, stand densification, and treeline advance Harsch et al., 2009;Salzer et al., 2009;Trant et al., 2020).
In subalpine forests, precipitation often arrives as snow, which can be negatively related to tree growth that year (e.g., Dolanc, Westfall, et al. 2013; but see Graumlich, 1991). However, warming has also led to greater proportions of precipitation arriving as rain, reducing depth, and duration of annual snowpack (Fyfe et al., 2017;Marlier et al., 2017;Mote et al., 2016;Pederson et al., 2011). This could potentially benefit high-elevation forests in the near-term, but may lead to severe moisture deficits as warming intensifies under expected future climates (Ullrich et al., 2018). Continued warming may also portend a future with increasing incidence of fire, pest and pathogen activity, and intensified competition Dudney et al., 2021;Gergel et al., 2017;Harpold, 2016;Millar et al., 2004;Schwartz et al., 2015).
Beyond projected changes in average environmental conditions, extreme events, including droughts, will also likely affect whitebark pine. From 2012 to 2016 California experienced a severe drought, and 2014 to 2015 represented some of the warmest, driest years in the instrumental record (Diffenbaugh et al., 2015;Griffin & Anchukaitis, 2014;Ullrich et al., 2018;Williams et al., 2015). Snow course data in California showed that 2015 had record low April 1 snow water equivalents, particularly at lower elevations (Mote et al., 2016). Range-wide estimates of snowpack in the Sierra Nevada suggested 2012 to 2015 had low cumulative snowfall, including substantial reductions in snowpack at high elevations in 2015 (Margulis et al., 2016). These conditions caused deep soil drying at low elevations (Goulden & Bales, 2019), leading to vegetation moisture stress and widespread tree mortality (Asner et al., 2016;Young et al., 2017). Tree mortality during and immediately following the drought was concentrated in lower elevation forests and was often driven by interactions with bark beetle activity (Fettig et al., 2019;Stephenson et al., 2019). Pest and pathogen activity is often less severe at high elevations in the Sierra Nevada (Das et al., 2016;Dudney et al., 2020), but it is unclear whether bark beetle activity is changing in these subalpine forests.
Tree growth provides an indication of how individuals respond to drought and changing environmental conditions. Secondary growth, measured by stem diameter, is responsive to environmental conditions (Speer, 2010). Moisture deficit may induce trees to reallocate resources away from noncritical functions, such as growth (Waring, 1987). As a result, short-term reductions in stem growth may be a symptom of stress for individual trees and can be predictive of mortality (Cailleret et al., 2019;Dudney et al., 2021). Given the broad latitudinal range of whitebark pine in the Sierra Nevada, trees are unlikely to respond uniformly to changing environmental conditions. We expect tree growth to respond to broad latitudinal gradients in climate, while varying with microclimatic differences in elevation, aspect, and soils. Moreover, local or population-level genetic variation may impact whitebark pine adaptation to climatic conditions (Depardieu et al., 2020;Trujillo-Moya et al., 2018). Patterns of genetic variation within a species are shaped by both historical and contemporary vicariant and adaptive processes, including selection, seed and pollen dispersal patterns, range expansions and contractions and range fragmentation due to geographic features | 3 of 20 van MANTGEM et al. (Slatkin, 1987;Wright, 1978). Previous range-wide studies of genetic variation in whitebark pine have shown that the Sierra Nevada is genetically distinct from the rest of the species range (Richardson et al., 2002;Syring et al., 2016). Therefore, gene exchange with the rest of the species distribution is likely limited. While genetic structure of whitebark pine across the entire Sierra Nevada has not been well-characterized, studies of limited areas within this range reported relatively high levels of genetic similarity among stands (Jorgensen & Hamrick, 1997;Lind et al., 2017;Richardson et al., 2002;Rogers et al., 1999).
In this study, we examined longer-term patterns in whitebark pine stem growth and interannual climate and estimated stem growth during the recent meteorological drought in the Sierra Nevada.
Further, we contextualize growth responses in relation to population genomic structure (genome-wide SNP analysis). We expected stem growth responses would provide insight into the important climate drivers over longer time periods, while short-term patterns in radial growth should provide insight into the short-term growth responses during meteorological drought. Finally, genetic data describe the degree of local population adaptation to a particular site and may provide a genetic perspective of drought vulnerability.
We used growth and genetic data to answer the following questions: (1) how did recent whitebark pine growth respond to variability in interannual climate in the Sierra Nevada? (2) What was the growth response of whitebark pine during the recent meteorological drought years and how did it vary across the Sierra Nevada?
and (3) Were growth responses during the drought years linked to climate-associated genetic variation? By combining measurements of growth and genomic structure over a common set of individual trees and sites, we aim to provide a framework to characterize the environmental and intrinsic genetic drivers of variation in growth.
This study potentially provides insight into responses and adaptation of whitebark pine to possible future climatic conditions in its southernmost ecoregion.

| Whitebark pine in the Sierra Nevada
Whitebark pine in the Sierra Nevada of California is found from north of the Lake Tahoe basin to south of Sequoia National Park (approximate latitudinal range = 36°-39°; Figure 1). The Sierra Nevada is characterized by a Mediterranean climate with warm, dry summers and cold, wet winters. Whitebark pine grows in the subalpine zone of the Sierra Nevada, where temperatures are low, growing seasons are short (2 to 3 months), and most annual precipitation falls as snow. In the Sierra Nevada, whitebark pine occurs between 2100 and 3700 m (https://www.fs.fed.us/rm/highe levat ionwh itepi nes/ About/ dist.htm#calif ornia, accessed 3/10/2022), growing as vertical stems at low elevations and as a more shrub-like growth form (krummholz) near treeline (Millar et al., 2004(Millar et al., , 2020 Whitebark pine has a mutualistic relationship with a seeddispersing bird, Clark's nutcracker (Nucifraga columbiana), which promotes a clumped pattern of individual trees and influences population genetic structure from tree to regional scales. Clark's nutcracker creates caches of 1 to 15 seeds (Tomback, 1978(Tomback, , 1982, which may develop into multiple, often genetically distinct, basally fused stems (Linhart & Tomback, 1985;Rogers et al., 1999). Whitebark F I G U R E 1 Locations of 27 study sites across the whitebark range in the Sierra Nevada. Yellow area shows the current estimated distribution of whitebark pine (PIAL) in our study area (https://www.fs.usda.gov/detai l/r5/landm anage ment/ gis/?cid=fsepr d697598, accessed March 19, 2021). The locations of outgroup populations (sites 28-31) used in genetic analysis are shown in the inset map, with Sierra Nevada sites inside the box. Two outgroup populations at Crater Lake National Park were in close proximity and appear as a single symbol. pine seeds do not have wings and are dependent on the Clark's nutcracker for long-distance dispersal, with dispersal distances of up to 32.6 km recorded from radiotagged birds along the eastern Cascade mountains (Lorenz et al., 2011). The pollen, however, is wind dispersed (Fryer, 2002) potentially allowing a high degree of genetic mixing across longer distances (Richardson et al., 2002).
Even in cases where overall genomic similarity and gene flow are high, variation in allele frequency in certain genes has been noted at local scales. For example, Lind et al. (2017) reported significant differences at loci associated with soil water availability across trees sampled in the Tahoe Basin, suggesting that local adaptation occurs in this species at relatively fine spatial scales despite overall high genomic similarity and outbreeding opportunities provided by the seed caching behavior of Clark's nutcracker and pollen dispersal. Across the species range, Syring et al. (2016) reported a subset of genes with very high differentiation across geographic regions and strong associations with latitude, suggesting that these genes were tracking climatic differences.

| Study sites and tree selection
In the summers of 2018 and 2019, we collected tree cores from 27 sites across the range of whitebark pine in the Sierra Nevada ( Figure 1). We considered areas that were on average <2 km from randomly located U.S. Forest Service and National Park Service whitebark pine monitoring plots to ensure accessibility and whitebark pine presence in the area (Nesmith et al., 2019;Slaton et al., 2019).
Within these areas, we identified three to five polygons that were primarily south-facing aspects (135°-225°) with slopes <25° using a digital elevation model (Gesch et al., 2002). We stratified these polygons based on contrasting snow accumulation and temperatures from 2012 to 2015 (at least ±1 SD from the mean value across the range of whitebark pine in the Sierra Nevada) estimated from the Basin Characterization Model (BCM; Flint et al., 2021;see Methods, Climate). We determined final sampling sites based on crew safety and the presence of multiple non-krummholz whitebark pines. We avoided krummholz due to their complex and potentially differing growth responses to environmental conditions (Millar et al., 2020).
Starting from a random point within each site, we selected trees along a transect perpendicular to the slope to maintain consistent aspect and elevation along the transect. In sites where non-krummholz whitebark pine was sparse or where many trees failed to meet selection criteria, most or all individuals within a site were sampled, so trees cannot be considered to be randomly selected. Transect length varied by whitebark pine density, with the transect length extending >1 km where whitebark pine was sparse. Whitebark pine can be extremely difficult to differentiate in the field from other cooccurring five-needled pines, such as western white pine and limber pine. We minimized the potential for species misidentification by sampling within areas of known whitebark pine presence, focused on large trees that have field identification characteristics such as male and female cones, and used extremely experienced field crews.
To further confirm species identification, we tested a subset of samples using a molecular assay that distinguishes whitebark from limber pine based on differences in chloroplast genes (see Alongi et al., 2019 for methods). All samples were confirmed as whitebark pine.

| Field collections
We collected 800 tree cores. After removing duplicate and unmeasurable cores (see below) along with one core from a tree with symptoms of white pine blister rust, we had 766 cores used in analysis. At each site, we had an average of 28 individual whitebark pines with usable tree core data (range = 17 to 30 trees). Sampled trees were required to be growing upright, free of significant damage, disease, and beetle activity. Stem diameters for sampled trees ranged from 6.3 to 89.0 cm DBH (average = 25.2 cm), and tree heights ranged from 2.3 to 22.0 m (average height = 7.3 m). There were five trees with missing DBH measurements and two trees with missing height measurements. One sampled tree had symptoms of white pine blister rust and was removed from analysis of growth but was retained for the genetics analyses. If stems were clumped, we cored only one arbitrarily selected stem per clump. Needle samples were collected from cored trees for genetic analysis (NPS Permit #: YOSE-2019-SCI-0061). We were working largely in wilderness areas, and the removal of cone-bearing branches for herbarium specimens would have been impractical. notably low or high growth). Second, series within each sampling location were statistically cross-dated following standard dendrochronology procedures using the dplR package for R (Bunn, 2008). Third, internally cross-dated series were statistically cross-dated against a recently developed regional chronology for whitebark pine (Millar et al., 2012).

| Population genomic data collection
Needle tissue for DNA analysis was collected from each sampled tree. Additional samples were provided from opportunistically collected needles at Crater Lake National Park, Lassen Volcanic National Park (Klamath Inventory and Monitoring Network, National Park Service), and the Jarbidge Mountains in northeastern Nevada (University of Nevada, Reno), to act as outgroup sites for genetic analysis. For each tree, five-needle fascicles were placed into coin envelopes with two 5-gram silica packets. Upon return from the field site, silica packets were removed, and envelopes were transferred to a 45°C oven to accelerate desiccation and aid in DNA preservation (Bainard et al., 2010;Chase & Hills, 1991;Doyle & Dickson, 1987) and stored in silica gel desiccant prior to extraction.
Genomic DNA was extracted from dried needles using the E-Z Note: N series, number of tree cores; N rings, total number of tree rings; Mean series length, mean number of rings per series; Start year, earliest calendar year captured at the site; Mean AR1, mean first-order autocorrelation coefficient and 1 standard deviation; Mean interseries correlation, mean pairwise Spearman correlation coefficient between series AR1 = first-order autocorrelation and 1 standard deviation. The final line of the table shows the summed number of series and rings, the overall mean series length, the minimum start year, and overall average AR1, and overall average interseries correlation across all cores.

TA B L E 1
Overall summary statistics of tree-ring data by sampling site.
little variation in annual site-level average annual sBAI, with pseudo r 2 = .03.

| Stem growth during drought years
Site-level BAI drought response indices were generally positive ( Figure 4)

| Genomic data quality and diversity
We estimated population genomic diversity and structure across our overall missing data, and a minimum minor allele frequency of 0.02.
More stringent filtering using a minimum minor allele frequency of 0.05 and/or removing individuals with more than 30% missing data slightly reduced the number of SNPs (<1%) and reduced the number of individuals (20%), but had little to no effect on population structure or genetic diversity estimates (Table S1). Therefore, we used the full dataset for all final analyses.
Using the full dataset, we found minimal population structure across the Sierra Nevada and among outgroup sites. Sierra Nevada sites clustered separately from outgroup sites, further confirming that the Sierra Nevada represents a unique genetic cluster of whitebark pine ( Figure 5). There was a predominant latitudinal geographical trend among Sierra Nevada sites along the primary coordinate  Table 3).

| Genotype-climate associations
We sought to determine whether climate could help iden- were four significant RDA axes (p < .001) and outlier analysis identified 1270 putatively adaptive SNPs that loaded ±3 SD from the mean loading on one or more axes. In total, these outlier SNPs exhibited higher genetic diversity ( Individual trees showed variable loadings within and among sites with respect to the RDA axes, with some sites more tightly clustered and others displaying greater variation and overlap. This suggests some sites may contain a greater adaptive range associated with certain climate factors, but a much tighter range in others (Figure 7).
Individual BAI drought response indices were correlated with RDA3 (r s = .12, p = .05) and RDA4 (r s = .13, p = .03). RDA3 appeared to differentiate high and low AET, with individuals in sites 3, 6, and 21 exhibiting the highest genotype loading and individuals from sites 14 and 19 lowest (Figure 7b). In addition, site 6 contained the highest variance in RDA3 scores. RDA4 differentiated lower PPT (sites 10, 21) from higher PPT (site 9; Figure 7b).

| DISCUSS ION
The objectives of our study were to describe how growth of whitebark pine in the Sierra Nevada responded to variability in climate (including meteorological drought) and to better understand patterns of genetic variation associated with this response.
Whitebark pine at our sites generally had positive to neutral recent stem growth trends between 1970 and 2011. Stem growth, as measured by sBAI, appeared to be weakly positively correlated with water year annual minimum temperature and precipitation.
Counterintuitively, growth indices during the meteorological drought years for BAI were generally positive to neutral. This implies that on average whitebark pine was not strongly moisture limited during the meteorological drought years (2012 to 2015). Individual growth response phenotypes appeared to be linked to genotypic variation in climate-associated loci, indicating that drought growth response was not only controlled by site-level environmental differences, but that some genotypes can take better advantage of local climatic conditions than others, as suggested in Millar et al. (2012).
We speculate that average stem growth, as measured by BAI drought response index, may have increased during the drought years due to an extended growing season. Whitebark pine in the Sierra Nevada often occupies energy-limited environments, characterized by low minimum temperatures and often high snowpack.

F I G U R E 5 (a) Principal component analysis of whitebark pine sampling locations including Sierra Nevada sites and outgroup sites
from Crater Lake National Park, Lassen National Park, and Eastern Nevada. Each point represents an individual, each color indicates site identity, and color gradient corresponds to site latitude (north to south; dark to light). Inset shows relative values of eigenvectors. (b) FastSTRUCTURE barplots where each bar represents an individual and each color represents identity to one of K genetic clusters. Sites are separated by vertical black bars and ordered by latitude (north to south; left to right).
Warm air temperatures during the drought years may have been related to warm soil temperatures, potentially providing additional days where soil water was unfrozen and available to support photosynthesis. It may also be possible that snowpack losses during the drought years lengthened the growing season while still retaining sufficient moisture to permit growth. At Crater Lake National Park in Oregon, whitebark pine radial growth was negatively associated with instrumental records of snowpack (Jules et al., 2016). The mismatch between meteorological and agricultural drought (vegetation responses to moisture stress, such as reduced growth) is consistent with a number of other studies that specifically measured responses across environmental gradients. Lebourgeois et al. (2010) found that Abies alba and Picea abies were not strongly negatively affected by extreme drought in high-elevation stands in France. Linares and Tíscar (2010) found that older trees growing at higher elevations maintained almost steady basal area increment during drought events, and Marqués et al. (2016) found that Pinus sylvestris was not strongly drought sensitive at higher elevation sites. These studies, as well as ours, suggest that high-elevation sites that are more energy limited than water limited may not experience water limitation during periods of low precipitation (i.e., meteorological drought).
Additional analyses using ring-width data from whitebark pine at our sites suggest that most of our sampled sites were primarily energy limited (rather than water limited), supporting the idea that drought conditions could lead to increased stem growth (Dudney et al., In press). Future research that captures snow-free days and soil moisture would provide insight into likely mechanisms driving growth patterns of whitebark pine in the Sierra Nevada.

| Recent stem growth
We found increasing to neutral sBAI trends from 1970 to 2011.
Similar trends have been reported elsewhere for trees near or at treeline in the Sierra Nevada (Graumlich, 1991;Millar et al., 2004Millar et al., , 2020Salzer et al., 2009). Whitebark sBAI was positively correlated with annual water year temperature, with a smaller, positive effect of annual water year precipitation, in broad agreement with previous studies (Dolanc, Westfall, et al. 2013;Millar et al., 2012). While we found annual climatic predictors were statistically significant, these variables explained very little of the variation in site-level average stem growth. There was high variation among individuals, perhaps reflecting intrinsic differences in tree genotypes (see below), tree size (see below), and microclimate (Bunn et al., 2011). Similar to F I G U R E 6 FastSTRUCTURE (Raj et al., 2014) barplots for K values between 2 and 5, where each bar represents an individual and each color represents identity to one of K genetic clusters. Sites are separated by vertical black bars and ordered by latitude (north to south; left to right).
our results, Millar et al. (2012Millar et al. ( , 2020 reported that annual growth of whitebark pine was positively correlated with precipitation at centennial scales (with varying relationships with temperature).
Potentially important differences with these earlier studies and our results include length of measurement (e.g., Millar et al., 2020 used >100 years of tree-ring records), site differences (eastern versus primarily western slope of the Sierra Nevada), and methodological differences (e.g., Millar et al., 2020 used  forms, trees affected by pests and pathogens, and trees that did not survive the recent drought. That is, we cannot draw population-level inferences from our samples because we did not exhaustively sample the population (Brienen et al., 2012). Future work may want to consider variability across tree conditions, growth forms, and live and dead trees to better describe population-level growth responses to changing environmental conditions.
We found a negative relationship between sBAI and tree diameter. We believe this is not an artifact of sample design or data processing. Sample design, tree demography, and how growth is defined can strongly influence growth trends in tree-ring data (Brienen et al., 2012(Brienen et al., , 2017Nehrbass-Ahles et al., 2014;Trouillier et al., 2020). At each site, non-krummholz trees were sampled across a wide range of stem sizes and we focused on recent growth, so it is unlikely our results reflect a strong size or age bias. By normalizing each annual basal area increment by the standard deviation of the basal area increments within each tree, sBAI represents the relative direction and magnitude of the growth trend for each tree, independent of stem size. It is unclear whether the relationship between growth trends and stem size is intrinsic to tree size or age or reflects other variables covarying with size or age. Even the largest individual trees we sampled were short-statured (maximum height = 22 m), so hydraulic limitations found in tall conifers are unlikely. Trees were often in open-grown conditions, so changes in stand density were unlikely strongly biasing growth responses. One possibility is that larger older trees have established on sites with different conditions F I G U R E 7 Redundancy analysis (a) axes 1 and 2, and (b) 3 and 4, where each point represents an individual multilocus genotype, color indicates site identity, and color gradient corresponds to site latitude (north to south; dark to light). Ellipses are 95% confidence intervals by site with label offset from ellipse center for legibility. Arrow direction and magnitude indicate how each significant climate variable loads onto the axes. Proximity of a genotype point to a climate vector indicates the strength of the association.
due to fine-scale topographic mediation of snow depth and duration (Zald et al., 2012), and differential microsite conditions can result in different growth trends and growth-climate relationships (Oberhuber & Kofler, 2000). The importance and causes of size and age-related growth trends in whitebark pine in the Sierra Nevada warrant further research but are beyond the scope of this study.

| Adaptive and neutral genomic structure
Despite low genetic differentiation overall, we detected significant associations between a subset of gene loci and climate. Our results indicate genetic associations related to differences in annual precipitation, minimum temperature, climate water deficit, and actual evapotranspiration. We observed higher heterozygosity and pairwise F ST among study sites in this subset of loci indicating that this variation is more strongly structured across the sampling range. Genetic diversity was highest in northern sites and lowest in southern sites, a pattern that is consistent with a past range expansion (reviewed in Excoffier et al., 2009). A north-to-south diversity cline within the Sierra Nevada was also apparent in allozyme loci (Jorgensen & Hamrick, 1997). In the Sierra Nevada, highelevation whitebark pine range expansion may have occurred after the last major glacial retreat, ~12,000 to 13,000 years BP (Clark & Gillespie, 1997). Strong isolation by distance, low F ST , and weak structuring across a known sampling gap suggest that whitebark pine likely consists of a single genetic population across the Sierra Nevada. Relatively high genetic exchange across the Sierra Nevada likely facilitates a wide range of genetic variation and local adaptation to different environmental conditions (Tigano & Friesen, 2016).
In particular, simulations suggest that when local adaptation is driven by many alleles of small effect size, good levels of standing variation in local populations are needed when gene flow is high (Yeaman, 2015). Maintaining gene flow may be an important consideration to help buffer against declines due to future climate change. Gene flow will likely be maintained through continued longdistance seed dispersal by Clark's nutcracker and wind-dispersed pollen, potentially assisting the dissemination of adaptive genotypes (Richardson et al., 2002;Rogers et al., 1999).
Gene-environment association studies represent an import-  (Crepeau et al., 2017), and loblolly pine (Neale et al., 2014) could provide a starting point (reviewed in Neale et al., 2017). Additionally, provenance tests (e.g., Chhin et al., 2018) or common garden studies (e.g., Warwell & Shaw, 2017) could be useful in determining the relative impacts of genetic and environmental factors underlying growth trends and variance, which are confounded in our approach. Provenance tests would also help to refine a set of screening loci associated with drought-resilient genotypes for future restoration (sensu Axelsson et al., 2020). These studies could be particularly important in documenting seedling and early growth which may be vital stages in which mortality is high and selection is strong (Bower & Aitken, 2006). Our current dataset is limited in that it targeted only larger-stemmed trees (≥6.3 cm DBH) and did not include krummholz. Expanded sampling and comparisons of seedling diversity to mature trees in stands could provide further information in diversity and adaptive trends over growth types and generations (Aravanopoulos, 2016).

| Implications for conserving whitebark pine
The combination of population stability in some long-term monitoring locations (Nesmith et al., 2019), neutral to positive growth trends in most sampling sites, and high genetic connectivity may be signs that whitebark pine was relatively healthy in specific regions of the Sierra Nevada. This contrasts with other parts of the species' range, where whitebark pine is in rapid decline (Goeking & Izlar, 2018;Keane et al., 2012). Increasing temperatures may be related to faster growth for Sierran whitebark pine, although future changes to growing conditions or more extreme drought may eventually reduce growth (Ullrich et al., 2018). Warming trends are also expanding the range of white pine blister rust into whitebark pine in the southern Sierra Nevada (Dudney et al., 2021). Long-term projections of population stability will be unreliable without including likely shifts in pest and pathogen responses to forecasted warming.
Conserving Sierra Nevada whitebark pine is particularly important given that it represents a genetically distinct, southernmost population of a widely distributed species (Richardson et al., 2002).
We expect whitebark pine in the Sierra Nevada to occupy sites that are generally warmer and drier relative to the majority of its range, potentially providing source material for recovery efforts under warming climates. Screening for rust resistance genotypes is ongoing in the Sierra Nevada and throughout the much of the range of whitebark pine (reviewed in King et al., 2010, Sniezko & Koch, 2017. Ultimately, disease and drought-resistant genotypes could be preferentially planted in strategic areas to facilitate conservation and restoration (e.g., Landguth et al., 2017). Such strategies have been generally proposed for forest management in California (Young et al., 2020). However, much of the whitebark pine populations in the Sierra Nevada are in wilderness areas with more restricted opportunities for planting seedlings. Direct planting of seeds is an alternative action that may avoid some of these difficulties (Pansing & Tomback, 2019). Our results suggest that efforts to facilitate the spread of adaptive genotypes could include selection of individuals from a broad geographic region (e.g., Mahalovich & Hipkins, 2011).