An alpine plant shows no decrease in genetic diversity associated with rapid post‐glacial range expansion Authors: Mackenzie Urquhart‐Cronish, Colette S. Berg, Dylan Moxley, Olivia J. Rahn, Lila Fishman, Amy L. Angert Published: 2025 ABSTRACT Abstract Premise: While range expansion is hypothesized to be a mechanism for species persistence under climate change, many eco‐evolutionary models describe demographic and genetic processes during expansion that may increase genetic drift, decrease genetic variation, and ultimately decrease relative fitness at the leading edge. These predictions assume dispersal occurs from the low‐density leading edge during colonization, common in post‐glacial expansion at the continental scale (~20,000 years ago), but relatively understudied on contemporary timescales, like alpine glacier recession since the end of the Little Ice Age (~150 years ago). Methods: We use the native alpine plant Erythranthe (Mimulus) lewisii to quantify neutral genetic diversity (single nucleotide polymorphisms) and infer signatures of genetic drift across two contemporary instances of range expansion on alpine glacier forelands in Garibaldi Provincial Park, BC, Canada, by testing for the presence of signatures of increased genetic differentiation and decreased genetic variation toward the contemporary range edge relative to the historical range core over space and time. Results: We find weak support for the prediction of increasing genetic differentiation toward the range edge, and no support for decreasing genetic diversity toward the range edge. This suggests dispersal occurring primarily from the leading edge is not characterizing colonization, with the implication that potential relative reductions in range‐edge fitness due to range expansion as predicted by theory are likely not applicable in nature at this spatiotemporal scale. Conclusions: Together, our results suggest that demographic dynamics of colonization following alpine glacier retreat do not result in the loss of genetic diversity over space and time. INTRODUCTION Under changing climatic conditions, many species are shifting their distributions polewards and upslope to track their climatic niches (Chen et al., 2011; Parmesan and Yohe, 2003). While range expansion is hypothesized to be a mechanism for species persistence under climate change, many eco‐evolutionary models describe demographic and genetic processes during range expansion that may increase genetic differentiation, decrease genetic variation, and increase the frequency of deleterious mutations at the leading edge (Klopfstein et al., 2006; Travis et al., 2007), undermining individual fitness and potentially population persistence (Peischl et al., 2013, 2015). If range expansion is primarily driven by dispersal from the low‐density leading edge, genetic diversity of small founding populations is expected to decline relative to the original, larger founder population. Following a population bottleneck, the relative influence of selection can be overwhelmed by genetic drift, and deleterious mutations can reach high frequencies. This can result in reduced fitness at the range edge relative to the range core, defined as the “expansion load” (Peischl et al., 2013, 2015). This prediction relies on the fundamental assumption that during range expansion, dispersal is rare and occurs primarily from the low‐density leading edge, resulting in a range expansion characterized by serial founder events, an increasing influence of genetic drift relative to selection, and a resulting increased frequency of deleterious mutations accumulating at the range edge as a result of colonization dynamics. Before making assumptions about potential relative reductions in fitness of range‐edge populations in nature, it is important to quantify whether real range expansions conform to the assumption that dispersal is primarily from the low‐density leading edge; this can be tested by looking for signatures of genetic drift along pathways of expansion. Empirical phylogeographic studies have demonstrated how patterns of genetic diversity change with post‐glacial range expansion (Hewitt, 2000). A meta‐analysis found that historical range dynamics related to post‐glacial colonization consistently show higher genetic diversity in older, refugial populations relative to younger, recolonizing populations (73% of studies support this claim) and are often associated with a latitudinal gradient in genetic diversity (Pironon et al., 2017). These patterns of decreasing genetic diversity, related to dispersal occurring primarily from the low‐density leading edge, and increasing genetic differentiation, related to patterns of isolation‐by‐distance, after post‐glacial range expansions, are consistent with colonization characterized by serial founder events and sequential population bottlenecks occurring at the leading edge (Excoffier and Ray, 2008). Most studies of post‐glacial range expansion have examined large‐scale patterns, and there is a bias toward studying post‐glacial range expansion on long spatial and temporal scales (i.e., across latitude and Quaternary interglacial cycles) (Hewitt, 2000; Davis and Shaw, 2001; Tiedemann et al., 2004; Kuchta and Tan, 2005; Lee‐Yaw et al., 2008). Post‐glacial range expansion also occurs on contemporary spatiotemporal scales, such as following alpine glacier retreat since the end of the Little Ice Age (i.e., across elevation, over ~150 years) (Naftz et al., 1996). Our understanding of whether dispersal also occurs primarily from the leading edge during range expansion at these smaller scales is relatively underdeveloped (Raffl et al., 2008). As such, it is currently unclear whether these instances of post‐glacial range expansion are defined by the rate of serial founder events originating at the range‐edge or the availability of new habitat to colonize as defined by the rate of ice retreat. Understanding and evaluating the demographic pathways of post‐glacial range expansion in the alpine will inform the relevance of applying current theory (Peischl et al., 2013) and latitudinal results (Willi et al., 2018; Perrier et al., 2020; Zeitler et al., 2023) to this understudied spatiotemporal scale. A handful of previous studies have investigated patterns of intraspecific genetic diversity and genetic differentiation following alpine glacier retreat across the entire glacier foreland, defined as the area between the historical glacial maxima (i.e., the location of the oldest terminal moraine) and the contemporary glacial extent (i.e., the current location of the glacier) (Raffl et al., 2006, 2008; Pluess, 2011; Powolny et al., 2016). These studies used previous knowledge of glacier recession pathways over space and time to apply a chronosequence approach where a space‐for‐time substitution along a gradient of terrain age at a single site was used to define the pathways of post‐glacial colonization (Walker et al., 2010). These studies sampled along existing chronosequences within previously mapped glacier forelands to assess patterns of genetic differentiation and diversity along the pathway of post‐glacial range expansion in the alpine. Together, they have described patterns that are inconsistent with the demographic pathway that predicts relative reductions in range‐edge population fitness; studies have found increasing (Powolny et al., 2016) or no detected change in genetic diversity within colonizing populations (Raffl et al., 2006, 2008; Pluess, 2011) and mixed support for increased genetic differentiation along the glacier foreland chronosequence zones (yes: Raffl et al., 2006, 2008; Pluess, 2011; Powolny et al., 2016). Among these four case studies, focal species have represented a variety of life histories (e.g., long vs. short lifespan, wind vs. insect pollinated, capable or not capable of clonal growth) and foreland size (1–5 km). However, all studies to date were conducted in the European Alps and used older types of genetic markers (AFLPs and microsatellites) that generated relatively fewer polymorphic loci than modern sequencing methods (i.e., single‐nucleotide polymorphisms or SNPs). To further test the role of dispersal during post‐glacial range expansion on relatively shorter spatiotemporal scales, we sampled Erythranthe (Mimulus) lewisii [(Phrymaceae) (Pursh) G.L.Nesom and N.S.Fraga]—Northern pink monkeyflower—a perennial alpine plant that is an early colonizer of disturbed riparian habitats, along previously mapped alpine glacier recession pathways (Koch et al., 2009) in Garibaldi Provincial Park, BC, Canada. We used modern sequencing methods to generate a dataset with thousands of genetic markers (SNPs), resulting in a test with more statistical power to detect differences in genetic differentiation and diversity than previous studies. In this study, we refer to “isolation‐by‐chronodistance” to examine patterns of genetic differentiation and diversity incorporating space‐for‐time substitution, and “isolation‐by‐topographic‐distance” to refer to spread over space only. We hypothesize that rare dispersal occurring primarily from the low‐density leading edge during range expansion structures genetic diversity during the colonization of alpine glacier forelands, setting the stage for founder events, genetic bottlenecks, and associated increased inbreeding and genetic drift. If dispersal occurs primarily from the low‐density leading edge during range expansion, we predict genetic signatures of isolation‐by‐chronodistance and decreasing genetic diversity toward more recent chronosequence zones. Alternatively, if dispersal during range expansion does not occur primarily from the leading edge, the predicted patterns of increasing isolation‐by‐chronodistance and decreasing genetic diversity would not be detected. This scenario could arise if initial founder events are followed by the eventual build‐up of high densities at the range edge, swamping genetic signatures of previously restricted dispersal, or if dispersal (via seed sources outside of the foreland and/or from within the foreland) or gene‐flow (via pollinator movement) is high across all chronosequence zones. Ultimately, we are interested in quantifying whether there are lasting signatures of genetic drift during range expansion in the alpine because this informs whether the same risk of reduction in relative fitness of range‐edge population described for species undergoing range expansion across long spatiotemporal scales also applies to short spatiotemporal timescales. We discuss how our results fit in with the handful of existing studies to begin to create consensus about the risk of negative genetic consequences (i.e., relative reductions in range‐edge fitness) for species undergoing range expansion across short spatiotemporal timescales. MATERIALS AND METHODS Study system Erythranthe lewisii (Phrymaceae) or Northern pink‐monkeyflower is an herbaceous perennial plant that is native to subalpine and alpine riparian habitats throughout northwestern North America. Aspects of the natural history of E. lewisii (e.g., riparian, early colonizer of disturbed habitat) tie it to the ecology of the glacier foreland, making it an ideal focal species to study post‐glacial range expansion. It is primarily bumblebee‐pollinated, but possesses hermaphroditic flowers and is capable of self‐fertilization and can reproduce clonally via rhizomes. Seed dispersal in monkeyflowers is thought to primarily occur via water (Waser et al., 1982), but dispersal upslope along the glacier foreland could potentially occur via wind and mammals (Vickery et al., 1986) and seed dispersing from source populations located on moraine slopes above/outside the glacier foreland. Here, we focus on E. lewisii in Garibaldi Provincial Park (BC, Canada) within two alpine glacier forelands: Helm and Garibaldi/Lava (Figure 1A–C). In Garibaldi Provincial Park, E. lewisii plants are found growing along slow‐moving alpine streams, rivers, and seeps. Their growing season begins with snow melt (mid‐July) and continues until snowfall (early September). Study sites in Garibaldi Provincial Park, BC, and mapped glacier recession pathways with sampling locations. (A) Location of glacier forelands, with the Little Ice Age (LIA) maximum mapped (Orange—Helm; Light blue—Garibaldi/Lava), sampled in this study. (B and C) Glacier recessions pathways mapped from Koch et al. (2009) and sampling locations for the (B) Helm Glacier and (C) Garibaldi/Lava Glacier forelands. Sampling locations are numbered by foreland (1–10 for the Helm foreland; 11–23 for the Garibaldi/Lava foreland) and year. Field collections along glacier forelands Fieldwork took place in Garibaldi Provincial Park (located in southwestern BC) during August and September 2020 (Letter of Authorization: 98700‐20). Here, we sampled E. lewisii along two distinct alpine glacier forelands located (Figure 1A) ~18 km apart, accounting for topography. The glacier forelands we selected have known records of the timing of glacier retreat (Koch et al., 2009) (Appendix S1), where Koch et al. (2009) characterized glacier recession pathways since the end of the Little Ice Age (~150 years ago) using a combination of historical images, dendrochronology, and physical evidence of terminal and lateral moraines. We designed our field collections based on these mapped chronosequence zones (Helm: 1928–2003, ~3 km; Garibaldi/Lava: 1850–1977, ~5 km) (Appendix S1), including sites within and outside of the alpine glacier foreland. Within the glacier foreland, we selected sites where we found E. lewisii growing (i.e., in suitable habitat) closest to the border of the mapped chronosequence zones (Figure 1B, C) (Table 1). Many of the adjacent sites were separated by gaps in habitat, and we kept a minimum distance of 200 m between sites. Per site, we sampled leaf tissue from ~10–20 plants with at least ~1 m distance between sampled plants to avoid sampling clonal genets and stored field‐collected leaf tissue in manila coin envelopes placed inside bags filled with non‐toxic indicator desiccant (Dry and Dry, B0725LNZ24, Brea, California, USA). Table 1Locations of sampling sites and timing of glacier recession history within forelands with per‐site genetic and nucleotide diversity data.Glacier forelandSite IDForeland zoneLatitudeLongitudeElevation (m) He Ho π Tajima's D Helm1 <185049.9644−123.031917700.0960.0770.0990.2482 <185049.9695−123.022317360.0980.0760.1010.4343192849.9677−123.017717380.0760.0580.0780.2324194949.9636−123.010317500.1000.0730.1030.3915 <185049.9840−123.005516630.1040.0950.1080.4436192849.9801−123.004217000.0950.0760.0990.5107192849.9743−123.000917850.0980.0820.1010.3938194949.9686−123.002417660.0940.0620.0970.3079197749.9720−122.999017930.1040.0760.1070.30810200349.9700−122.992318600.1010.0790.1040.372Garibaldi/Lava11 <185049.8174−122.975315080.0920.0670.0940.36812 <185049.7988−122.992814330.2130.1680.220.97813185049.8062−122.986312600.1030.0760.1080.38414185049.8194−122.979715230.1040.0690.1070.39615192849.8144−122.983513580.1230.0890.127−0.80116194949.8243−122.979915390.0990.0750.1020.31617196949.8270−122.978315480.1070.0790.1100.51518197749.8290−122.975316300.1110.0900.1150.38819 <185049.8228−122.944715120.1070.0860.1100.31420192849.8088−122.956513270.1070.0840.1100.35221194949.8175−122.953514110.1060.0820.1090.38122196949.8227−122.953814580.1220.0830.1260.08923197749.8246−122.953614900.1100.0760.1170.301John Wiley & Sons, Ltd. Locations of sampling sites and timing of glacier recession history within forelands with per‐site genetic and nucleotide diversity data. ddRADseq genotyping We extracted DNA from field‐collected leaf tissue using a CTAB‐chloroform protocol modified from Xin and Chen (2012). Briefly, we added 100 μg/mL Proteinase K (NEB P8107S, Ipswich, Massachusetts, USA) to the CTAB Lysis buffer to reduce enzymatic oxidation and increased centrifugation speeds to 6000 rcf following the DNA‐CTAB complex precipitation. For DNA purification, we suspended 5 μL of MagAttract beads (Qiagen 1026901, Germantown, Maryland, USA) in 55 μL of DNA plus TE Buffer (10 mM TRIS, 0.1 mM EDTA, pH 8.0) and performed multiple washes. We transferred 54 μL of the sample to a clean PCR plate, and sample purity was assessed using 2 μL of DNA elution on a Nanodrop 2000 (Thermo Scientific, Wilmington, Delaware, USA), and DNA concentration was quantified with the Qubit 2.0 Broad Spectrum kit (Invitogen, Waltham, Massachusetts, USA) using 2 μL of DNA elution. The full‐length protocol can be found in Moxley et al. (2025). We then used a double‐digest restriction‐site associated DNA sequencing (ddRADseq) protocol to generate genome‐wide sequence clusters (tags) for 357 individuals across 23 collection sites, following the BestRAD library preparation protocol and methods from Kolis et al. (2022), using restriction enzymes PstI and BfaI (New England Biolabs, Ipswich, Massachusetts, USA; both methylation insensitive) to digest DNA samples. After enzymatic digestion, sets of 48 individual DNA samples were labeled with unique in‐line barcoded oligos, and each set was pooled into a single tube. Pools were barcoded using NEBNext indexing oligos with a degenerate barcode, PCR amplified, and size selected (300–600 bp fragments) using a Bluepippin (Sage Science, Beverly, Massachusetts, USA) agarose gel. Libraries were prepared using NEBNext Ultra II library preparation kits for Illumina (New England Biolabs, Ipswich, Massachusetts, USA). The libraries were paired‐end sequenced (150 bp) on 12 lane of Illumina NovaSeq SP sequencer (University of Oregon Genomics and Cell Characterization Core Facility, Eugene, Oregon, USA). Illumina reads were demultiplexed, and true PCR duplicates (same restriction site, plus same degenerate barcode) were removed using a custom Python script. Adapters were trimmed and low‐quality reads were removed using STACKS function process_shortreads (Catchen et al., 2011, 2013). Reads were mapped to the E. erubescens version2 reference genome (http://mimubase.org/FTP/Genomes/LF10g_v2.0/) using BWA MEM with default parameters and read group information included for each sample (Li and Durbin, 2009) and indexed with SAMtools (Li et al., 2009). SNPs were called using GATK (McKenna et al., 2010) for aligned sequence data. After alignment and initial filtering (quality >30, maximum missing data =0.2), we retained 102,835 sites. After filtering for sites with mean depth (≥10× and ≤30×), and mean depth per individual ≥7×, we retained 64,316 variable sites for 332 individuals (Helm N = 147 and Garibaldi/Lava N = 185) representing all 23 collection sites (8–17 samples per site). For analyses of genetic diversity, we used genomic data where we did not filter for minor allele frequencies or linkage disequilibrium. For analyses of population structure, we filtered minor allele frequency (maf) >0.1 with bcftools (Danecek et al., 2021) and linkage disequilibrium (100 kb window) with PLINK (Purcell et al., 2007) and retained a total of 6358 variable sites. Analyses of population genetic structure, differentiation, and diversity Population genetic structure and genetic differentiation To analyze population genetic structure between and within glacier forelands, we generated a data set for principal component analysis using PLINK (Purcell et al., 2007) to summarize the overall genetic similarity of all samples. We ran the program ADMIXTURE (Alexander et al., 2009) with all samples for K=1−23 (maximum number of sites in sampling between forelands) and on vcf's subset to include only samples within a foreland (Helm: K=1−10 and Garibaldi/Lava: K=1−13 to generate between and within foreland individual‐level ancestry assignments. For principal component analysis (PCA), we calculated the 95% confidence levels of a priori foreland assignment using the stat_elipse() function from the ggplot2 package, assuming a multivariate t‐distribution since the data were not normally distributed. We calculated pairwise genetic distances (Weir and Cockerham's FST) between forelands and all sampled sites using functions from the R package heirfstat (Goudet and Jombart, 2022) within the R package graph4lg (Savary et al., 2021). We generated two measures of geographic distance: (1) “isolation‐by‐chronodistance” represents the space‐for‐time substitution of the chronosequence approach, where we calculated pairwise differences in year of glacier recession zone to represent the pathway of range expansion, and (2) “isolation‐by‐topographic‐distance” represents topographic distances between sampling sites (without the spatiotemporal context of range expansion). To calculate topographic distances, we accessed topographic data via the open source website OpenTopography and used the R package elevatr (Hollister et al., 2023) to generate a topographic raster layer of the area of our study sites. We then used the R package topoDistance (Wang, 2020) to calculate a pairwise matrix of topographic distances between all sampled sites. To test for patterns of isolation‐by‐chronodistance and isolation‐by‐topographic‐distance within distinct forelands, we used the R packages ade4 (Dray and Dufour, 2007) and vegan (Oksanen et al., 2022) to run separate Mantel tests per site, with 9999 permutations. We tested the correlation between our two distance matrices—isolation‐by‐chronodistance and isolation‐by‐topographic‐distance—within forelands and within Helm, found a weak‐moderate correlation that is significant (Mantel statistic r:0.38,p=0.02,r:0.38,p=0.02) and within Garibaldi‐Lava, a weak correlation that is non‐significant (Mantel statistic r:0.16,p=0.1). Genetic diversity We used the R package graph4lg to calculate the following measures of genetic diversity within each distinct site: mean observed heterozygosity (Ho), mean expected heterozygosity (He), and mean allelic richness. Genetic diversity statistics were calculated within individuals and then averaged across all individuals per population. We tested the relationship between chronosequence zone, foreland identity, and genetic diversity by running a linear model using the R package lm() and a Type III ANOVA to test the individual relationship and interactions between time (chronosequence zone) and genetic diversity across forelands. To investigate nucleotide diversity, we used an “all sites” vcf (including variant and invariant sites) to avoid biasing our estimates of nucleotide diversity (Korunes and Samuk, 2021) generated with GATK (McKenna et al., 2010) and filtered for depth and missing data in the same way as above (retained 96,795 sites). We then used the software pixy (Korunes and Samuk, 2021) to calculate average nucleotide diversity within (π) populations over a 1 Mb window per chromosome. We then averaged the per‐region values to generate a per‐individual value. Tajima's D is a method to detect shifts in the site frequency spectrum away from neutral expectations by comparing the number of segregating sites and average pairwise base pair differences. We calculated average values of Tajima's D using vcftools (Danecek et al., 2011) and the “all sites” vcf with a 1 Mb window size, and averaged values per site and then per population. Tajima's D is a useful statistic to make inferences about whether deviations from neutral evolutionary expectations are occurring, due to selection and/or demography. When values are strongly positive, this suggests an excess of intermediate‐frequency variation in observed samples, which can be driven demographically by a recent population bottleneck or as a result of balancing selection. If values are strongly negative, this suggests a deficit of observed genetic variation/an excess of singletons, and this can be driven demographically by recent population growth or as a result of positive selection/selective sweeps (Hahn, 2019). We were interested in making inferences about demography (recent founder events) and tested the relationship between chronosequence zone and Tajima's D. We ran a linear model using the R package lm() and a Type III ANOVA to test the relationship between space/time (interaction between numerical chronosequence zones and categorical forelands) and deviations from neutral evolution. RESULTS Distinct population structure detected between and within forelands In the principal component analysis, PC1 explained 15.6% and PC2 explained 10.1% of the variation in population genetic diversity (Figure 2A). Individuals from each foreland clustered in distinct groups along PC1, as demonstrated by ellipses depicting the 95% confidence levels of foreland identity. We also found evidence of weak genetic differentiation between forelands (FST=0.035). While PC1 separated individuals between foreland sites, PC2 appeared to summarize genetic structure present within sites, specifically identifying genetic dissimilarity between a single site within the Helm foreland (site 3, Appendix S2A) and elevated genetic differentiation relative to the rest of the sampled sites within the foreland (Appendix S3). Population genetic structure suggests two distinct genetic groups with greater similarity among individuals sampled within a given foreland. (A) PCA plot with data points colored by sampling location (foreland) with ellipses representing the 95% confidence intervals around foreland identities (Orange = Helm, Blue = Garibaldi/Lava; <20 km distance between forelands). (B) ADMIXTURE analysis results for K=2 plotting proportion assigned ancestry. X‐axis labels the site (Figure 1B, C) and the glacier recession zone; sites 1–10 are Helm and 11–23 are Garibaldi/Lava. Colors match the legend in panel A. Note, the color of individuals in panel A corresponds to sampling sites (foreland identity) and not the ancestral assignments in panel B. Based on our sampling design (two forelands) and the visual representation of genetic structure defined by the PCA (two to three distinct clusters), we plotted the ADMIXTURE results for K=2 (Figure 2B) and K=3 (Appendix S2B) ancestral clusters. We also examined the CV error plot from the ADMIXTURE analysis; however, clustering algorithms can overfit data easily, and caution is recommended when interpreting these inferred values of K (Pritchard et al., 2000) (in our case, K=7,6,& 8). For these reasons, we based our ancestral cluster assignment on the visual structure in PC space (Figure 2A). Most individuals within sites were assigned mixed ancestry, but with the majority of ancestry assignment belonging to the respective foreland within which each individual was sampled. Site 3 in the Helm foreland, which had some individuals forming a distinct cluster along PC2 (site 3, Figure S2A), was assigned to a distinct ancestry cluster when K=3 (Appendix S2B). Since we detected differentiation between glacier forelands, we conducted further investigations of within‐foreland population genetic structure related to the expansion pathways along glacier chronosequences. In the Helm foreland, we visualized the within‐foreland population structure in PC space and did not detect any clear temporal patterns among the genetic similarity of samples (Figure 3A). We visualized ancestry assignments within the foreland for K=2,3,& 4 ancestral clusters (Figure 3B–D), with K=4 being the lowest CV error value from the ADMIXTURE analysis. As the ancestral cluster number increased, the proportion of individuals assigned mixed ancestry increased, with Site 3 maintaining a unique genetic structure regardless of the value of K. Overall, no clear spatiotemporal patterns across chronosequence zones emerge from visualizing population structure in the Helm foreland. Within‐foreland population genetic structure is not clearly structured by glacier chronosequence. Within‐foreland genetic structure: Helm (A, B, C, D) in orange and Garibaldi/Lava (E, F, G, H) in blue. (A) PCA of Helm samples colored by year of chronosequence zone. (B–D) ADMIXTURE analysis results for K=2,3,& 4 plotting proportion assigned ancestry. (E) PCA of Garibaldi/Lava samples colored by year of chronosequence zone. (F–H) ADMIXTURE analysis results for K=2,3,& 4 plotting proportion assigned ancestry. X‐axis labels the site (Figure 1B, C) and the glacier recession zone; sites 1–10 are Helm and 11–23 are Garibaldi/Lava. Note, the color of individuals in panels A and E corresponds to chronosequence zones within the foreland and not the ancestral assignments in the ADMIXTURE plots. We repeated the same analyses in the Garibaldi/Lava foreland, finding a similar lack of temporal structure within‐foreland (Figure 3E) and no clear spatiotemporal patterns across chronosequence zones (Figure 3F–H). Within the Garibaldi/Lava foreland, Site 12 maintained a generally unique genetic structure regardless of the value of K (Figure 3F–H). Mixed evidence for isolation‐by‐chronodistance within forelands To assess whether dispersal was limiting the rate of range expansion across the foreland, we examined whether genetic differentiation increased with increasing distance from the historical range core to the contemporary range edge using pairwise comparisons of FST and timing of glacier recession (our proxy for recency of range expansion). We found no support for isolation‐by‐chronodistance in the Helm foreland (r=−0.15, p>0.05) (Figure 4A) and a significant relationship in the Garibaldi/Lava foreland (r=0.39,p=0.008) (Figure 4B). We also investigated the relationship between topographic distance, ignoring the influence of range expansion history, and genetic distance, and found no evidence for isolation‐by‐topographic‐distance in either foreland (Appendix S4). Relationships between range expansion history (years) versus genetic differentiation (FST), evidence of restricted gene flow between glacier recession zones, within forelands are varied (A) Helm (B) Garibaldi/Lava. Mantel tests with 9999 permutations are non‐significant for Helm (Mantel statistic r:−0.15,p=0.79) and significant for Garibaldi/Lava (Mantel statistic r:0.39,p=0.0086). Linear regression was used to fit lines to the scatterplots. No clines in genetic diversity detected along pathways of range expansion We examined patterns of genetic diversity following the timing of glacier retreat to detect patterns associated with founder effects (i.e., decreased heterozygosity) due to leading edge colonization, characteristic of limited dispersal during range expansion. We quantified Ho,He, and allelic richness for all sites sampled within forelands and used linear models to test the relationship between the timing of expansion and levels of genetic diversity. Generally, our model results revealed no relationship between timing of expansion and genetic diversity, foreland identity, or their interactions (Table 2); however, we did detect a main effect of foreland site on levels of allelic richness (df=1,F=4.7,p=0.04). Comparisons between He and Ho within sites all showed He>Ho, suggesting non‐random mating, perhaps due to inbreeding, within sites. Tajima's D values did not provide support for demographic signatures of recent population bottlenecks (or recent population growth) related to range expansion following alpine glacier retreat (chronosequence zone year: df=1,F=0.33,p=0.6; foreland: df=1,F=1.21,p=0.3; chronosequence zone year * foreland: df=1,F=0.45,p=0.5) (Figure 5). Table 2Linear model results testing the relationship between chronosequence zone years, foreland identity, and their interactions on values of genetic diversity. Since no interactions were significant, the results of an additive model are reported for tests of individual variables.Response variableVariabledfSum. Sq. F value p‐value He Chronosequence zone year10.00030.510.5Foreland10.0023.40.08Chronosequence zone year * foreland10.00020.370.5 Ho Chronosequence zone year10.0010.140.7Foreland10.034.7 0.04 Chronosequence zone year * foreland10.000020.050.8Allelic richnessChronosequence zone year10.0003−0.70.5Foreland11.1−0.50.6Chronosequence zone year * foreland10.0010.140.7 Note: Bold values indicate statistical significant at p < 0.05.John Wiley & Sons, Ltd. Linear model results testing the relationship between chronosequence zone years, foreland identity, and their interactions on values of genetic diversity. Since no interactions were significant, the results of an additive model are reported for tests of individual variables. Note: Bold values indicate statistical significant at p < 0.05. No evidence of population bottlenecks characterizing range expansion as measured by genetic (A, B, C) and nucleotide (D) diversity across chronosequence zones in each distinct foreland (Helm and Garibaldi/Lava) (A) Mean expected heterozygosity by locus by site (B) Mean observed heterozygosity by locus by site (C) Mean allelic richness by locus by site, and (D) Mean Tajima's D, a metric that represents deviations from neutral evolutionary processes (demographic processes or selection). No relationships between the timing of glacier recession and genetic diversity (A, B, C) or nucleotide diversity were detected. Note, due to differences in timing of glacier recession (Appendices S1, S3), populations from Helm are absent from the <1850 core zone, and the 1850 time period represents habitat outside the LIA glacial maximum. DISCUSSION In this study, we demonstrated that an alpine plant shows inconsistent evidence of isolation‐by‐chronodistance and no evidence of decreased genetic diversity following rapid post‐glacial range expansion. In one of our two replicate forelands, we found that glacier recession zones that were farther apart in time had increased genetic differentiation, suggesting reduced gene flow between populations in older vs. more recently exposed habitats. However, we did not find strong or lasting genetic patterns that would be consistent with dispersal occurring predominantly from the low‐density leading edge during expansion, suggesting either that founder effects during expansion were weak or that any effects on genetic diversity were transient. Overall, our results demonstrate limited evidence for the expected effects of genetic drift during range expansion and, combined with the existing literature, suggest that the predictions of the original theory of expansion load (Peischl et al., 2013) are unlikely to apply to post‐glacial range expansions on short spatiotemporal scales in the alpine. Population structure emerging on small space and time scales We detected evidence of population genetic structure between the two glacier forelands, even with only 18 km of topographic distance separating both sites. Genetic differentiation across this relatively small spatial scale could be due to landscape features that separate the two forelands, such as Garibaldi Lake and the Garibaldi névé (snow field and mountains) (Figure 1A), which may prevent the successful dispersal of seed and/or pollen between forelands. Other studies that have found fine‐scale population genetic differentiation across topographically complex areas have attributed these patterns to local geographic features (e.g., mountain ridges, lakes) (Hotaling et al., 2018; Trense et al., 2021). The different forelands are weakly genetically differentiated, with instances of three individuals from Helm and one individual from Garibaldi/Lava that clustered with the opposite foreland in population structure analyses (Figure 2). These specific individuals were all collected in later chronosequence zones (1977 and 2003) near the glaciers' contemporary extent, and we interpret these patterns as potential evidence of recent seed dispersal across the two forelands. We also detected potential evidence of isolation‐by‐environment (as opposed to a process related to range expansion) for a single site within the Helm foreland (Appendix S2A, B). Since site 3 belongs to an earlier chronosequence zone and not the range edge, this differentiation could be related to distinct ecological conditions (e.g., wetter, cooler climate) of the site; however, we did not quantify site‐level conditions. Testing within‐foreland structure, in the Helm and Garibaldi/Lava forelands, we found no clear temporal pattern in PC space (Figure 3A, E) or spatiotemporal patterns in ancestry assignment (Figure 3B–D; F–H), suggesting that any existing population genetic structure cannot be attributed to range expansion across the replicate chronosequences. Evaluating patterns of limited dispersal Genetic differentiation Within forelands, we found support for a pattern of isolation‐by‐chronodistance only in the Garibaldi/Lava foreland. Again, landscape differences at the scale of the individual foreland, such as size, length of glacier recession history, and elevation (Table 1), may potentially underlie differences in the degree of isolation‐by‐chronodistance at this relatively small spatial scale. For example, compared to the Helm foreland, the Garibaldi/Lava foreland is larger, has a longer recession history, and also has a larger elevational range, and these factors may aid in restricting gene flow among chronosequence zones following colonization. However, while the Helm foreland is approximately half the size of the Garibaldi/Lava foreland, previous work investigating patterns of intraspecific genetic variation in an alpine plant (Trifolium pallescens) along glacier recession pathways in Switzerland found significant patterns of isolation‐by‐chronodistance in a foreland as small as 1 km (Raffl et al., 2008). The riparian natural history of E. lewisii, specifically seed dispersal via water (i.e., downstream), could facilitate greater propagule mixing among younger and older chronosequence zones, decreasing overall levels of genetic differentiation, especially in the smaller Helm foreland. Additionally, phenological differences in E. lewisii across the forelands may interact with foreland size in shaping spatial patterns of genetic differentiation. For example, differences across elevation in the timing of snow melt early in the season are expected to stagger phenological overlap within a foreland, influencing early‐season gene‐flow dynamics via pollinator activity (Albrecht et al., 2010). A larger foreland size—such as the Garibaldi/Lava foreland relative to the Helm foreland—could further exacerbate these phenological limits to gene flow and preserve spatial genetic structure created by dispersal occurring primarily from the low‐density leading edge during colonization. Differences in the timing of glacier recession may also matter, as supported by a meta‐analysis that found weak evidence of an association between longer timescales of colonization history and positive signals of isolation‐by‐distance (Crispo and Hendry, 2005). We evaluated evidence for signatures of dispersal occurring predominantly from the low‐density leading edge during expansion by investigating spatiotemporal patterns in population genetic summary statistics. Other methods can be used to infer demographic pathways of range expansion, such as demographic inference using the site‐frequency‐spectrum and Approximate Bayesian Computation modeling (Marchi et al., 2021). We selected summary statistics since we believed the spatiotemporal scale of our study may have limited our use of more complex methods, but future work could assess the validity of these methods in detecting demographic change—including population sizes—during range expansion in alpine glacier forelands (Epps and Keyghobadi, 2015). Genetic diversity Allelic and nucleotide diversity patterns over chronodistance supported the alternative hypothesis that range expansion following alpine glacier retreat was buffered from the effects of genetic drift at the leading edge. The patterns we detected could be due to the absence of colonists being confined to the low‐density leading edge during range expansion, or could arise if predominantly leading edge dispersal during colonization occurs but is transient. In the case of the absence of a predominately leading‐edge dispersal during range expansion, the rate of expansion toward the leading edge may be determined by the rate of ice retreat rather than the rate of propagule movement, such that edge populations are able to reach high population densities, therefore conferring higher genetic diversity, before new habitat opens up via glacier recession for subsequent colonization. Alternatively, genetic signals of a leading‐edge range expansion characterized by genetic drift could be obscured over time due to increased population densities at the range edge or if pollinator‐facilitated gene flow increased levels of genetic diversity across chronosequence zones post‐range expansion. While these different processes (i.e., dispersal not occurring from the low‐density leading edge vs. transient leading‐edge dispersal dynamics) would lead to the same pattern of genetic diversity, attributing pattern to process would require additional work, such as a pedigree study in the field to investigate the role of gene flow post‐expansion (Gaudeul and Till‐Bottraud, 2008). We did detect a main effect of foreland site on levels of allelic richness, with allelic richness being slightly higher on average in the Garibaldi/Lava foreland than at the Helm foreland. This again could be related to differences in foreland size and age, with the Garibaldi/Lava foreland being older and larger, potentially supporting larger effective population sizes, an important correlate with higher allelic richness (Foulley and Ollivier, 2006). Genetic consequences of range expansion following alpine glacier recession Based on our results, we can infer that the negative genetic consequences predicted by the original parameter space of theoretical models of expansion load (Peischl et al., 2013; Peischl and Excoffier, 2015) are unlikely to be applicable on these spatiotemporal scales. This suggests range expansions following alpine glacier retreat are likely to have more in common with more recent ecologically complex models of range expansion, allowing range edge populations to overcome the demographic effects of dispersal occurring via the low‐density leading edge during range expansion (Gilbert et al., 2017; Peischl and Gilbert, 2020; Urquhart‐Cronish et al., 2024). We can extend our findings on genetic diversity to suggest that plant populations undergoing range shifts in alpine ecosystems may be more resilient to ecological change than an influential and recent range expansion model suggests (Peischl et al., 2013; Peischl and Excoffier, 2015). Our research, combined with other studies on patterns of intraspecific genetic diversity in the alpine, shows range‐edge populations do not experience decreased levels of neutral genetic diversity relative to the range core. Populations that are able to maintain levels of standing genetic variation during range expansion may possess greater evolutionary potential in the context of changing ecological conditions (Kardos et al., 2021). More generally, our results lead to the inference that range expansions upslope and across elevation appear to experience different dispersal dynamics than their latitudinal counterparts, so researchers looking to apply theoretical models to distinct spatial scales in nature should be aware of the potential mismatch in predicted demographic pathways of range expansion. CONCLUSIONS This study of post‐glacial range expansion along alpine glacier recession pathways found weak support for the prediction of stronger genetic differentiation and no support for lower genetic diversity along the pathways of expansion, from older to younger chronosequence zones. While our results had greater statistical power to detect differences among genetic diversity markers than any previous studies, they generally agree with previous studies that did not detect genetic diversity decreasing from older to younger chronosequence zones on alpine glacier forelands. We did not find evidence in support of strong or lasting founder effects characterizing colonization during range expansion, with the implication that the accumulation of negative genetic consequences at the range edge is not applicable on these spatiotemporal scales. Together, our results suggest that loss of genetic diversity for range shifting species in the alpine is likely not lowering the evolutionary potential of range‐edge populations (Reed and Frankham, 2003; Kardos et al., 2021), and that ecological interactions (i.e., novel competitors, herbivores, and pollinators) might be more important in driving patterns of species persistence at range edges in the face of climatic change (Speed et al., 2012; Richman et al., 2020; Usinowicz and Levine, 2021). Finally, we emphasize the importance of evaluating the essential underlying theoretical assumptions of model parameter space before extending investigations to conduct empirical tests of more complex eco‐evolutionary theory in nature. BENEFIT‐SHARING STATEMENT Benefits Generated: Benefits from this research accrue from the sharing of our data and results on public databases as described in our Data Availability Statement. AUTHOR CONTRIBUTIONS M.U.C. and A.A. conceived of the study and designed the field sampling. M.U.C. and O.J.R. conducted fieldwork. M.U.C. conducted all DNA extractions with the training and assistance of D.M. M.U.C. and C.B. conducted library prep lab work with L.F. providing space, reagents, and assistance with troubleshooting protocols. M.U.C. ran the genomic analyses with demultiplexing scripts provided by C.B. M.U.C. generated and analyzed all data. M.U.C. wrote the first draft of the manuscript with input and contributions from all authors. Funding from NSERC DG to ALA [RGPIN‐2022‐03113] and M.U.C. (CGS‐D graduate scholarship to M.U.C.), the University of British Columbia (Four‐year Fellowship and Tuition funds to M.U.C.), and research grant funding from the Society for the Study of Evolution (SSE) R.C. Lewontin Early Award to M.U.C. Supporting information Appendix S1. Mapped glacier recession pathways without sampling plotted. Appendix S2. PCA and ADMIXTURE plot for three ancestry clusters. Appendix S3. Heatmap of pairwise FST values for the Helm Glacier foreland. Appendix S4. No evidence of restricted gene flow with topographic distance.