Asymmetric responses by bees and aculeate wasps to dune stabilisation across the southern Canadian prairies

Inland sand dunes in the Great Plains of North America provide refuge for a variety of psammophilous (i.e., sand‐loving) organisms but are threatened by vegetative stabilisation. In a cross‐sectional study, we analysed the effects of dune stabilisation on wild bees and stinging wasps (Hymenoptera: Aculeata)—which comprise some of the most diverse assemblages of dune‐dwelling arthropods—at 13 sites across the southern Canadian prairies. We did not detect changes in overall taxonomic richness with increased stabilisation, with comparable numbers of bee and aculeate wasp species/morphospecies observed among dunes at different stages of stabilisation. However, abundances of species of wasps identified as dune specialists decreased with increased stabilisation. Bees and aculeate wasps responded nonlinearly to the percent of plant cover on dunes dominated by grasses (Poales: Poaceae), whereas abundance and richness were generally high on dunes where the dominant plant was dune scurfpea (Fabales: Fabaceae). Thus, the extent to which these plants cover a particular dune can be indicative of the numbers of individuals and types of aculeate species it can sustain. Overall, differences in aculeate community composition between dune sites were better explained by proximity of sites to one another than stabilisation level, reflecting the large spatial scale at which the dunes were surveyed (i.e., within a 50,350 km2 area). Bees and aculeate wasps, especially open sand specialist species, serve as a promising tool for monitoring the effects of dune stabilisation. This study provides a crucial reference point and methodology for future assessments of dune‐inhabiting aculeates in these unique inland dune systems.


INTRODUCTION
Active sand dunes in the Great Plains of North America represent a unique habitat that sustains diverse communities of psammophiles-organisms that require, prefer or thrive in sandy areas (Acorn, 2011;Hugenholtz et al., 2010). Commonly known as the prairie sandhills, these dunes trace their origins to the Quaternary glacial epochs, when grinding ice (namely the Laurentide Ice Sheet, which retreated approximately 15,000 years ago) and weathering reduced bedrock to gravel, sand and silt (Forman et al., 2001;Hugenholtz et al., 2010).
Unlike coastal and desert dunes, the prairie sandhills are in otherwise largely vegetated inland areas and require a certain level of wind erosion to generate loose sand to be reshaped into dunes and thereby remain active (Hugenholtz et al., 2010;Hugenholtz & Wolfe, 2005).
Throughout much of the northern Great Plains, dune stabilisationthe fixation of sand in place and eventual conversion to grassland by erosion-inhibiting vegetation-has increased dramatically over the past century and threatens the continued existence of various dune-dependent species (Hugenholtz et al., 2010;Hugenholtz & Wolfe, 2005;Wolfe & Hugenholtz, 2009). Declining sand dune activity in the northern Great Plains in Canada since European settlement has been attributed to a multitude of causes, including decreased wind erosion due to climatic factors, less frequent wildfires and reduced biological control of plants, by herbivory, trampling and wallowing (Forman et al., 2001;Hugenholtz et al., 2010). The remaining active sand dunes are islands of biodiversity inhabited by various specialised organisms not found elsewhere in the surrounding environment, such that they are patchily distributed and thus exhibit a metapopulation structure (Hugenholtz et al., 2010).
Sand specialists restricted within Canada to these unique inland dune systems include species of plants, vertebrates and terrestrial arthropods that have been assessed as rare, threatened or endangered by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC; reviewed in part by Hugenholtz et al., 2010). However, little information is available on the status of most dune-dependent arthropods (Hugenholtz et al., 2010), with notable exceptions-tiger beetles of the genus Cicindela Linnaeus (Coleoptera: Carabidae) and dunedependent moths (Lepidoptera: Noctuidae) of the genera Copablepharon Harvey and Schinia Hübner (COSEWIC, 2012(COSEWIC, , 2014(COSEWIC, , 2016(COSEWIC, , 2018(COSEWIC, , 2022. Wild bees and aculeate wasps (Hymenoptera: Aculeata) are visibly abundant and diverse insects inhabiting active dunes in the Canadian Prairies (Acorn, 2011;Finnamore & Buckle, 1999). Sand provides a nesting substrate for both ground-nesting bees and aculeate wasps (Cane, 1991;Evans & O'Neil, 2007;Finnamore & Buckle, 1999;Michener, 2007), which may require its distinct texture and/or moisture retention properties or prefer them to those of other prairie soils.
Additionally, bees, including floral specialists, as well as certain wasps visit flowers of various psammophilous plants growing on or at the edges of active sand dunes (the latter mainly for nectar) (Cane et al., 1996;Danforth et al., 2019;Parker & Griswold, 1982). The dune-dependent mining bee Andrena haynesi Viereck and Cockerell (Andrenidae), which is oligolectic on Helianthus L. (Asterales: Asteraceae), is an example of a prairie sandhill species that relies on sand as a nesting substrate and dune-restricted pollen sources (Parker & Griswold, 1982). Other rare species may similarly depend on dunes for both nesting and floral resources. Moreover, various aculeate wasps patrol open sand while searching for mates (as do some bees), hunting and/or scavenging (Evans & O'Neil, 2007).
Inland sandy areas promote bee and wasp diversity and provide critical refuge for various species in northern temperate regions (Banaszak & Twerd, 2018;Exeler et al., 2009;Heneberg et al., 2013;Seitz et al., 2019). Aculeate wasp species richness was almost 1.7 times greater in sites with aeolian (i.e., wind-generated) sandy soil than those with chernozem (black, humus-rich soil) in southeastern Alberta (Finnamore & Buckle, 1999). This region, known as the Middle Sand Hills, has seen a precipitous (94%) decrease in the area of open sand between the late 1940s and mid-2000s (see fig. 5 in Hugenholtz et al., 2010). Sandhills in southwestern Manitoba experienced comparable losses between the late 1920s and mid-2000s, whereas losses via stabilisation were lower but varied considerably (from 21% to 53%) among dune fields in southern Saskatchewan over the same period (Hugenholtz et al., 2010;Hugenholtz & Wolfe, 2005). Despite the significance of sand to aculeates, the effects of declining sand dune activity in the northern Great Plains on both the resident bee and aculeate wasp faunas have not been methodically investigated.
The objectives of the present study are threefold: (1) to present a list of bee and aculeate wasp species collected in sandhill habitats across the southern Canadian prairies, (2) to understand how aculeate communities respond to dune-to-grassland succession and (3) to identify species at risk of becoming extirpated from areas where dune activity continues to decline.

Study sites
In a cross-sectional study, 13 dunes at various stages of stabilisation were surveyed from southeastern Alberta to southwestern Manitoba (2 in Alberta, 2 in Manitoba and 9 in Saskatchewan) in the spring and summer of 2019. A minimum distance of 670 m and maximum distance of 776 km separated any two sites. Each site was surveyed five times for a total of 65 collection events, with a minimum of 7, maximum of 29 and average of 17 days between same-site visits. Site details and the complete sampling schedule are indicated in Table S1.
All sites were sampled in pairs (i.e., two were visited each collecting day) except for one (at Douglas Provincial Park, SK), which was sampled alone on the days it was visited.

Bee and wasp sampling protocol
Bees and wasps were sampled lethally by pan trapping and timed targeted netting, with the latter serving as a supplemental method to detect species that might be missed by the former (see Portman et al., 2020 for a discussion of the shortcomings associated with using only passive sampling methods). Ants are the only aculeate group excluded from this study, since their small size and terrestrial/fossorial nature made it impractical to sample them in targeted collections and since their numbers in pan traps were strongly influenced by the proximity of individual traps to ant nests. Pan traps (disposable Dollar Tree plastic bowls with a diameter of 17.5 cm and height of 3.8 cm) were prepared in three colours: blue (coated twice with Rust-Oleum ® 253715 Brilliant Blue Painter's Touch ® 2Â Ultra Cover ® Gloss Spray), yellow (coated twice with Rust-Oleum ® N1942830 Yellow Specialty Fluorescent Spray) and white (unpainted). At each site per visit, 30 pans were placed and filled with a trapping solution (water with a small amount of detergent) in alternating colours (blue, white and yellow) along an X-shaped transect, with the pans separated by about 3 m and the two perpendicular linear transects crossing between the seventh and eighth pans. The traps were generally set between 8:00 and 9:00 AM and emptied and retrieved between 3:00 and 4:00 PM, with paired sites visited in the same order at the start and end of the sampling session. In the interim (generally between noon and 3:00 PM), bees and wasps were collected at each site (roughly within the radius of the pan-trapping transects) using a hand-held insect net in timed 45-min sessions.
Sampling took place on warm and sunny or mostly clear days. The

Specimen identification and vouchering
Bees (with some assistance; see Acknowledgements) and aculeate wasps were identified to species with reference to a large body of taxonomic literature (see Supporting Information: Appendix A; Table S2), vouchered material, including available type specimens and authoritatively identified non-type specimens or images thereof, and a library of DNA barcodes for sequenced specimens. A detailed account of the identification methods used, with information on specimen and sequence vouchering, is presented in Supporting Information:

Estimates of floral composition and percent plant cover
Plant cover was estimated at each site in five 1 m by 1 m quadrats on the last day of sampling (in August), when vegetative cover was at a maximum. The quadrats were centred where the four outermost pan traps along the X-shaped transect were placed as well as the point at which the two perpendicular linear transects crossed. Digital images (see Supporting Information: Appendix B) were taken of each quadrat (with the pans temporarily moved and placed outside the frame) in the morning, immediately after the pans were set out. Floral composition was determined to the lowest taxonomic level possible using various literature sources (see Supporting Information: Appendix A), and the percent of plant cover was estimated to the nearest 5% in, and subsequently averaged across, the five quadrats at each site.

Variable selection
In our study, we distinguished between bees and aculeate wasps because they differ markedly in their feeding habits as larvae. Bees use pollen-nectar provisions, whereas wasps consume insects and spiders (Goulet & Huber, 1993;Michener, 2007). Thus, despite some overlap, their respective roles in ecosystem functioning are primarily as pollinators and arthropod predators. Bees are a monophyletic group (Anthophila) nested within apoid wasps (Branstetter et al., 2017). We also distinguished between open sand specialists and non-specialists as described in Supporting Information: Appendix A (species with insufficient information were included among the latter).
We investigated the effects of stabilisation in qualitative and quantitative terms. The qualitative measure was stage of succession

Statistical model testing
Statistical analyses were performed in RStudio version 2022.07.0+548 (RStudio Team, 2022). The effects of dune stabilisation on target communities of bees, aculeate wasps, sand specialists and non-specialists were examined using non-parametric permutational analyses of variance (PERMANOVA; with 10,000 permutations) using the function 'adonis2' in the R package vegan (Oksanen et al., 2022). In each analysis, a Bray-Curtis dissimilarity matrix of bee and/or aculeate wasp species composition was computed from species capture abundance data. The main and all possible interaction effects of (aculeate or ecological) group, stabilisation level and dominant plant type were investigated, all of which were treated as fixed factors. Permutations were constrained within sites using the strata argument. Homogeneity of dispersion was assessed for all significant main and interaction effects using the functions 'betadisper' and 'permutest' (with 10,000 permutations) in vegan. Distance matrices were visualised through two-dimensional principal coordinates analysis.
The effects of dominant plant type and percent plant cover on bees and aculeate wasps were investigated with generalised additive models (GAMs), calculated using the function 'gam' in the R package mgcv (Wood, 2011). The use of GAMs made it possible to model non-linear interactions with percent plant cover, with the smooths defined using the s() term. Changes in overall abundance and species richness were analysed using a negative binomial log link function.
To analyse trends in abundance, the numbers of sampled specimens were summed across taxa by aculeate (bee or wasp) and ecological (sand specialist or non-specialist) groups per site. Since there were few sand specialist species, the numbers of species/morphospecies were summed only by aculeate group. Effects on abundance were analysed with aculeate group, ecological group and dominant plant type treated as fixed factors, for which the main and all possible interaction effects were investigated, and smooths were produced for each combination of aculeate group, ecological group and dominant plant type. Effects on species richness were analysed with aculeate group and dominant plant type treated as fixed factors, for which the main and interaction effects were investigated. Smooths were produced separately for bees and wasps in one model and two site categories characterised by different dominant plant types in another. A basis dimension of k = 5 was specified for both the abundance and richness models. In all models, thin plate regression splines were used to represent smooth terms. The function gam. check in mgcv was used to confirm that the models converged and that the residuals were randomly distributed. Mixed-effects models implementing the spatial autocorrelation structures 'corEXP' and 'corSpher' were tested (using the function 'gamm') to account for possible effects resulting from the proximity of sites to one another considering their geographic coordinates. However, since no significant differences in fit were detected between the abundance and richness models with and without correlation structures (ΔAIC = 2 in all cases), simpler models (i.e., GAMs without correlation structures) were employed instead.

Estimation and visualisation of species richness and beta diversity
For comparisons of species richness among dune stabilisation categories, individual-based rarefaction curves were constructed (to standardise for differences in sample size). Species richness was T A B L E 1 PERMANOVA summary results for variation in aculeate community composition (based on Bray-Curtis dissimilarity) in response to three categorical variables-(aculeate or ecological) group, stabilisation level and dominant plant type (main and all possible interaction effects)analysed separately for subsets of the same data partitioned in two ways: by aculeate (bee or wasp) and ecological (sand specialist or nonspecialist) groups. For significant partial effects (p adonis2 < 0.05), permutation tests for multivariate homogeneity of dispersion were performed, with p βdisper >0.05 indicating no significant differences in dispersion among groups. Note: Interaction effects are indicated by a colon. *p < 0.05. **p < 0.01. ***p < 0.001. rarefied in EstimateS 9.1.0 (Colwell, 2013), resampling the data set 100 times without replacement.
To investigate changes in aculeate community composition between the sampled dune sites, Bray-Curtis dissimilarity indices were computed from species abundance data using the function 'vegdist' in vegan. These were subjected to hierarchical clustering analysis (using the function 'hclust' in Base R) to construct a dendrogram. An accompanying projected site map was constructed using the R packages raster (Hijmans, 2022), rgdal (Bivand et al., 2022) and sf (Pebesma, 2018), with spatial data for global administrative boundaries (source: https://gadm.org/) acquired using the function 'get- F I G U R E 1 GAM thin plate regression splines fitting abundance (summed across taxa by aculeate and ecological groups per site) to percent plant cover, with smooths produced separately for each combination of aculeate group, ecological group and dominant plant type. In all plots, the y-axis is on the scale of the log link function based on a negative binomial distribution. Grey-shaded bands above and below the curves represent 95% confidence intervals. In summary statistics, edf = estimated degrees of freedom for the model terms; Ref.df = degrees of freedom for reference distributions; p = approximate significance of smooth terms. Summary statistics for parametric terms are presented in Table 2. sampled specimens but more than a quarter (28.3%) of the species/ morphospecies detected. The 14 species identified as open sand specialists comprised nearly a fifth (18.0%) of all sampled specimens but only 3.7% of the species detected.
T A B L E 3 GAM summary results for two models of variation in species richness (summed across taxa by aculeate group per site) in response to two categorical variables-aculeate group and dominant plant type (main and interaction effects)-and percent plant cover, modelled by smooth functions, with smooths produced separately for bees and wasps (Model 1) and sites dominated by dune scurfpea and sites dominated by grasses (Model 2). Summary statistics for smooth terms are presented in Figure 2. F I G U R E 2 GAM thin plate regression splines fitting species richness (summed across taxa by aculeate group per site) to percent plant cover, with smooths produced separately for (a) bees and wasps (Model 1) and (b) sites dominated by dune scurfpea and sites dominated by grasses (Model 2). In all plots, the y-axis is on the scale of the log link function based on a negative binomial distribution. Grey-shaded bands above and below the curves represent 95% confidence intervals. In summary statistics, edf = estimated degrees of freedom for the model terms; Ref. df = degrees of freedom for reference distributions; p = approximate significance of smooth terms. Summary statistics for parametric terms are presented in Table 3.
F I G U R E 3 Rarefied species richness of aculeates (standardised for differences in the numbers of individuals collected) in sand dunes at three different stages of stabilisation. The number of sites sampled per stabilisation category is indicated in parentheses. Effects of stabilisation on bee and wasp species assemblages PERMANOVA revealed significant differences in species capture abundances between bees and wasps and open sand specialists and non-specialists (Table 1). The main effects of stabilisation level and dominant plant type were also significant (Table 1; see Figure S1 for a visual representation). The only significant interaction was between stabilisation level and dominant plant type, and this was the case for bees, wasps, sand specialists and non-specialists. The strongest contributor to variation in bee and aculeate wasp capture abundances was ecological group (r 2 = 55% and 36%, respectively), with each species assigned to only one of the two (open sand specialist or non-specialist) categories. Similarly, with each species F I G U R E 4 (a) Map showing the sampling localities (black circles) in the three Prairie Provinces (note that paired sites are not distinguishable at the scale shown). (b) Hierarchical clustering dendrogram of Bray-Curtis distances illustrating aculeate community compositional turnover (in terms of species identity and abundances) between sand dunes at different stages of stabilisation across the southern Canadian prairies. Site details are indicated in Table S1. categorised as either a bee or wasp, the strongest contributor to variation in the capture abundances of sand specialists and nonspecialists was aculeate group (r 2 = 48% and 42%, respectively). In the models, which tested the effects separately on bees, wasps, sand specialists and non-specialists, the second-strongest contributor to variation in species capture abundance was stabilisation level (r 2 = 5%, 9%, 8% and 6%, respectively).
GAM of overall abundance (summed across taxa by aculeate and ecological groups per site) did not reveal significant differences between bees and aculeate wasps captured; however, non-specialists greatly outnumbered the sand specialists across our samples (Table 2; Figure S2). Additionally, more bees and aculeate wasps were collected in sites where dune scurfpea was identified as the dominant plant than in sites dominated by grasses. Of the parametric terms, the only significant interaction was between aculeate group and ecological group. The percent of plant cover was a significant smooth term for wasp abundance in sites dominated by grasses, with numbers peaking around 50% estimated vegetative cover, and this affected both sand specialists and non-specialists ( Figure 1). Only sand specialist wasps, which were far more abundant in our samples than sand specialist bees, declined with increased plant cover in sites dominated by dune scurfpea. No significant relationship between overall abundance and percent plant cover was detected for any of the remaining possible combinations of aculeate group, ecological group and dominant plant type.
GAM of species richness (summed across taxa by aculeate group per site) did not detect significant differences between bees and aculeate wasps captured or sites dominated by dune scurfpea and grasses (Table 3). However, species richness varied significantly and nonlinearly with increased plant cover in sites dominated by grasses, with the observed relationship resembling those between the overall abundances of sand specialist and non-specialist wasps and percent plant cover in those sites (Figure 2). Accounting for differences in the numbers of individuals sampled, comparable levels of aculeate species richness were observed among sites assigned to the three different stabilisation categories (Figure 3). However, whereas the ratio of bee species to individuals was highest in active dunes ( Figure S3), the ratio of wasp species to individuals was highest in stabilised dunes ( Figure S4).

Comparisons of beta diversity among dune sites
Hierarchical clustering of Bray-Curtis dissimilarities revealed that compositional turnover was generally lowest between dunes closest to one another (separated by ≤3.1 km), regardless of stabilisation level (Figure 4). Specifically, aculeate assemblages were most similar between five of the six paired neighbouring sites sampled simultaneously (i.e., on the same days) throughout the course of the collecting season. The greatest dissimilarity observed was between the 3 easternmost sampling sites (in south-central Saskatchewan and southwestern Manitoba) and the 10 westernmost ones (i.e., all the others, in southwestern Saskatchewan and southeastern Alberta).

DISCUSSION
Our sampling of 13 dunes across the southern Canadian prairies revealed diverse assemblages of dune-inhabiting bees and aculeate wasps. That two-fifths (40.1%) of the species in our samples were, at their time of capture, previously unpublished from one or more provinces in which they were collected (some of these records have since been published elsewhere; e.g., Gibbs et al., 2023) illustrates the potential importance of surveying aeolian sites within this region for the purpose of species discovery and documentation. It is obvious that many of our discoveries reflect a lack of extensive surveying across the larger region historically and/or available expertise to identify previously vouchered specimens. Still, it is notable that one of the five most abundant species of bees in our samples-Lasioglossum onuferkoi Gardner and Gibbs (Halictidae), represented by 360 specimens, was undescribed at the time of capture, despite the subgenus to which it belongs having recently been revised for Canadian species (Gibbs, 2010). Its belated discovery has been attributed to the species' restriction to rather isolated sandy sites in a relatively small area within Canada (Gardner & Gibbs, 2022).
Few species (14) were open sand specialists, but they accounted for about a quarter (24.2%) of the individuals from active sand dunes.
On stabilised dunes, sand specialists decreased to 6.5% of individuals, and these may have come from nearby active dunes. Stabilised dunes were adjacent to or within 100 m of active dunes/blowouts, and wild bees and aculeate wasps are known to forage at much greater distances from their nesting sites and disperse over large distances (Gathmann & Tscharntke, 2002;Nalepa et al., 2013). Additionally, as none of the sampled dunes was completely vegetated, we cannot rule out the possibility that open sand specialists were using small areas of exposed sand on stabilised dunes.
The sand specialists in this study exhibit different biogeographic patterns and varying degrees of affinity to large, open dunes. The reasons for their association with large areas of open sand are not entirely clear but presumably varied. For example, whereas the cleptoparasitic bee Nomada fervida Smith (Apidae) is regarded as a dune specialist (Droege et al., 2010;Gibbs et al., 2023), its suspected host bee-Agapostemon splendens (Lepeletier) (Halictidae)-is not, despite nesting in areas with deep sand (Portman et al., 2022;Roberts, 1969).
We therefore speculate that A. splendens (if indeed the host) only occurs in sufficient numbers to sustain populations of N. fervida on active sand dunes or that the latter has physiological or biogeographical constraints not shared by the host. Crabro denningi Bohart (Crabronidae) and L. onuferkoi have the smallest ranges of all the specialists observed, apparently being restricted to larger dune systems in the Canadian Prairies and, for C. denningi, also the adjacent United States. Among the open sand specialists, these two are of greatest conservation concern because of their small ranges and habitat specificity. L. onuferkoi presumably occurs in neighbouring Montana but so far is known only from the specimens collected through our surveys in Alberta and Saskatchewan and another 17 specimens collected by A.T. Finnamore and D. Pollock in the 1990s in the Canadian Forces Base near Suffield, Alberta (Gardner & Gibbs, 2022). Other species also have their Canadian centre of distribution in the Prairies but range more widely. Among the wasps, Podalonia argentifrons (Cresson) (Sphecidae), Philanthus psyche Dunning Labrador and Newfoundland, whereas P. argentifrons and E. auranus merely reach the southern edge of the boreal zone where they can be found in jack pine dune habitat. The latter species is remarkable for being the only eumenine wasp that evolved a tarsal rake that aids in the excavation of nest burrows (Buck et al., 2008), a morphological adaptation that is widespread in other families of ground-nesting aculeate wasps. Interestingly, this species is less associated with sand areas in the southern part of its range, which extends through the western half of the United States into Mexico.
We detected significant changes in aculeate abundance with decreasing sand dune activity. However, the effect of vegetative cover was not consistent across sites, aculeate groups or ecological groups (Figures 1 and 2). Our results suggest that vegetative cover may be a poor indicator of overall abundance and species richness, except for certain guilds and/or dunes dominated by grasses. Invasive grass cover was more consequential for the numbers of sandburrowing arthropods than the overall vegetative cover (measured inversely as the amount of open sand) in dunes in central California (Slobodchikoff & Doyen, 1977). We found that wasp abundance and aculeate species richness varied nonlinearly with increased percent plant cover where grasses dominated, with peak levels observed on dunes with 50% estimated vegetative cover. Stabilised dunes were dominated by grasses, but dune scurfpea was often dominant in sites with open sand. Dune scurfpea is among the first species of plants to colonise active inland sand dunes in North America and dominates the pioneer stage but is replaced by other plants in subsequent stages of succession (Chadwick & Dalke, 1965;Hulett et al., 1966). The extent to which the plant itself is responsible for the observed higher levels of aculeate abundance on dunes where it appears to dominate is unclear. Dune scurfpea flowers are visited by bees (see Discoverlife. org for associated bee records compiled by J. Pickering, searchable by plant species), and insects associated with the plant may be an important source of prey for aculeate wasps. However, dune scurfpea may just be an indicator of high-quality habitat for bees and wasps.
Regardless, sampling of additional sites is needed to determine whether the observed patterns hold broadly.
Although circumstantial changes in aculeate abundance and species richness were detected in response to stabilisation, turnover was better explained by site proximity than stabilisation level. This likely reflects the large scale at which our sites were surveyed. Whereas the maximum distance between any (simultaneously sampled) paired sites was a little over 3 km, the minimum distance separating any sites sampled on different days was about 18 km. It is thus unsurprising that despite equal sampling effort, the numbers of individuals collected over the course of the 2019 spring and summer seasons varied greatly among sites, from a low of 431 to a high of 1674 (Table S2). The two sites with the fewest captured individuals were closest to one another, as were the two sites with the most captured individuals, even though in both cases, the paired sites were at different stages of stabilisation. Bee and wasp species assemblages show high beta diversity at the regional/landscape scale, reflecting differences in species pools between widely separated areas (Clough et al., 2007;Murray et al., 2012;Rubene et al., 2015). We speculate that similarities in species assemblages among dunes at similar stages of stabilisation might have been observed at the local level, with sampling of dunes that are all close to one another, as opposed to the regional level, at which our surveys were made. Finnamore and Buckle (1999) found higher turnover of wasp species between active dune blowouts and stabilised dunes than between any sites in the same dune stabilisation category in a study with a much smaller geographic scope focused on the Canadian Forces Base Suffield National Wildlife Area in Alberta.

Summary and implications
We infer that stabilisation changes the structure of dune-inhabiting aculeate communities, with asymmetric responses observed by different bee and wasp guilds. Psammophilous species behaved consistently with the initial assumption and decreased with increased stabilisation. We provide additional examples of species potentially at risk of extirpation from areas where dune activity continues to decline. Elsewhere, sand-specialist Hymenoptera have been classified as regionally extinct or critically endangered (Tropek et al., 2013). Canada's prairie sandhills are mainly used as rangeland for cattle grazing and natural gas extraction, but notable uses also include conservation and ecotourism. Two of our sampled dune systems are located within provincial parks, and Saskatchewan's Great Sandhills are a protected area lying within the Great Sandhills Ecological Reserve. Although cattle are a poor substitute for migratory bison herds, which historically helped maintain dunes (Fox et al., 2012), management interventions to maintain large areas of open sand, including using livestock to trample and graze on sand-stabilising vegetation, may be necessary to conserve the remaining active sand dunes in this region (Hugenholtz et al., 2010). In the mid-to-late 2000s, the Canadian Forces Base Suffield National Wildlife Area initiated efforts to reactivate stabilised dunes in the Middle Sand Hills using various treatments, including manual digging and prescribed fires as well as targeted grazing and trampling by livestock (Hugenholtz et al., 2010). Whether similar initiatives will be implemented to reactivate other recently stabilised dunes in the region remains to be seen. Meanwhile, long-term effects on dune-dependent species should be monitored. To that effect, our study provides a preliminary checklist of bee and aculeate wasp species in prairie sandhill environments, which will be available as a baseline for future surveys.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available under Supporting Information.