Prescribed fire increases the number of ground-nesting bee nests in tallgrass prairie remnants
Abstract
- Prescribed burning is a common management technique in tallgrass prairie remnants, but there have been few empirical studies that directly examine burning impacts on the nesting preferences and habitat of ground-nesting bees.
- We used emergence traps in remnant tallgrass prairies in western Minnesota, USA to determine whether ground-nesting bees prefer to nest in burned or unburned prairies. We estimated the total number of nests made by actively nesting bees in burned and unburned patches by assessing each specimen for wing and mandible wear, sex, and sociality. We also measured characteristics that may influence bee nesting preferences including bare ground, thatch depth, vegetative cover, and the floral community.
- We found more nests of actively nesting ground-nesting bees in burned patches than unburned patches, but no differences in effective number of species of ground-nesting bees or community composition. Burned patches had higher amounts of percent bare ground and a thinner thatch layer, but no differences in percent vegetative cover, floral abundance, flowering plant species richness, effective number of species of flowers or community composition.
- Our results suggest that ground-nesting bees prefer to nest in burned patches of remnant tallgrass prairies and highlight opportunities for future research to better understand bee nesting ecology in response to prairie management.
INTRODUCTION
The tallgrass prairie ecosystem exists because of both naturally occurring wildfires and prescribed burns, which Indigenous tribes of the Upper Midwest have expertly managed for millennia (Christianson & Krisoff, 2019; Collins & Wallace, 1990; Kimmerer & Lake, 2001; McClain et al., 2021; Stewart, 2002). Typically, prescribed fire consists of intentionally setting a slow-burning, low-intensity fire in a section of prairie (Packard & Mutel, 2005). These fires can mediate changes in plant and animal interactions, such as spurring the growth of grasses for grazing animals, reducing encroachment of woody vegetation into the prairie, and increasing flowering patch quality for foraging pollinators (Collins & Calabrese, 2012; Mason et al., 2021; Mola & Williams, 2018; O'Connor et al., 2020; Pausas & Parr, 2018). European settler colonisation and land dispossession from Indigenous peoples in this region intentionally suppressed cultural burning practices, and it is only within the last few decades that prescribed burning has been recognised by settler land managers as an important tool to promote biodiversity (Hoffman et al., 2022; Lake et al., 2017; McClain et al., 2021). Prescribed fire is often used in concert with other management techniques such as grazing or haying in attempts to emulate historic disturbance processes (Harmon-Threatt & Chin, 2016; Helzer, 2009). Recently, there has been an added emphasis for prairie grassland managers to understand if their management actions support declining pollinator populations, especially in the Upper Midwest (Minnesota Department of Natural Resources, 2018; The White House, 2015).
Many studies across a diversity of ecosystems have found positive effects of prescribed fire on bee abundance and diversity (Brown et al., 2017; Carbone et al., 2019; Mason et al., 2021; Smith DiCarlo et al., 2019; Ulyshen et al., 2021; Williams et al., 2010, but see Griffin et al., 2021 and Leone et al., 2022). The underlying mechanisms for bees' positive responses to fire are generally considered to be altered flowering phenology and floral density, with fire leading to higher floral densities (Mola & Williams, 2018; Ponisio, 2020).
Fire is predicted to support ground-nesting bee habitat through multiple mechanisms (Grundel et al., 2010; Potts & Willmer, 1997; Wray & Elle, 2015). Ground-nesting bees nest deep below the soil surface and thus may avoid the detrimental impacts of burning compared to bees that nest in stems (Bruninga-Socolar et al., 2022; Cane & Neff, 2011). Specifically, prescribed fire exposes bare ground, increases soil surface temperatures and reduces moisture content, factors that are known to influence ground-nesting bee nesting sites (Anderson, 1965; Briggs & Knapp, 1995; Cane, 1991; Ehrenreich & Aikman, 1963; Larson et al., 2020; Wagle & Gowda, 2018; Weissel et al., 2006). These changes to the soil surface last throughout the growing season following a burn, leaving areas of exposed soil between the stems of plants throughout the summer as opposed to thick thatch layers of dried, decaying grasses in unburned patches. However, there may also be ground-nesting bee species that prefer to nest in unburned patches, where decaying litter may provide refuge and protection of their nests (Antoine & Forrest, 2021, Osgood, 1972, Packer & Knerer, 1986).
The few studies that examined responses of ground-nesting bees to prescribed burning in tallgrass prairies found higher abundances or proportions of these bees foraging or flying within patches that were burned, suggesting that they may also be nesting in burned patches (Bruninga-Socolar et al., 2022; Decker & Harmon-Threatt, 2019; Smith DiCarlo et al., 2019). These studies then compared their findings to perceived measures of quality nesting habitat, like the amount of bare ground created by burning (Bruninga-Socolar et al., 2022; Campbell et al., 2018; Decker & Harmon-Threatt, 2019; Miller, 2021; Moylett et al., 2020; Potts & Willmer, 1997; Smith DiCarlo et al., 2019). However, a potential weakness of these studies is that their results could be driven by bees flying into the burned patches of prairie but nesting elsewhere because the methods do not locate nests.
In densely vegetated habitats like tallgrass prairies, little is known about how management techniques, like burning, directly impact nesting habitat availability for ground-nesting bee communities. Bee nesting is challenging to study because it is difficult to detect and monitor their nests, but is crucial to understand because over 80% of all bee species nest underground (Antoine & Forrest, 2021; Decker & Harmon-Threatt, 2019; Harmon-Threatt, 2020; Orr et al., 2022). To directly document bee nesting, researchers often place ‘emergence traps’ on the ground, which are small lightweight tents with a collection bottle at the top. However, a bee caught in an emergence trap may or may not indicate the presence of a current, active nest. For example, some collected bees may be emerging from nests created from the previous year such that the presence of their nest may or may not coincide with when a disturbance or management action took place (Pane & Harmon-Threatt, 2017; Portman, Brokaw & Cariveau, 2022; Sardiñas & Kremen, 2014). It is therefore critical to carefully determine within-year nest-site preferences by ground-nesting bees based on recent prairie management actions, like prescribed burning. Furthermore, the number of estimated nests from emergence traps can be inflated if they are based on estimates derived from specimen abundance alone. This is because males, multiple workers from the same social species, non-ground nesting bees and incidentally collected bees can all be collected by emergence traps but do not necessarily represent the presence of individual nests. Rigorous estimates of nesting are needed to make accurate, data-informed predictions about habitat quality for ground-nesting bees and must be addressed to advance our efforts in tallgrass prairie management and pollinator conservation.
The objective of this study was to determine whether ground-nesting bees prefer to nest in burned patches of tallgrass prairies compared to unburned patches. We used a paired design in prairies that were patch burned, meaning that a predetermined area (a “patch”) of a site was burned and other patches within the same prairie site were left unburned. We placed emergence traps in burned and unburned patches of the prairies to find and quantify actively nesting bees. Furthermore, we assessed bee nesting preferences using a novel methodology wherein we differentiate actively nesting bees from non-nesting bees to estimate the number of active nests. We hypothesized that compared to unburned patches, burned patches would (1) support more active nests and higher Effective Number of Species, (2) support a different composition of ground-nesting bee species and (3) have distinct vegetative and microhabitat differences that are important for bee nesting.
METHODS
Study system
We conducted this study at four patch-burned remnant tallgrass prairies in western Minnesota in 2019 from June 13 until August 22 (Figure 1a, b). Sites ranged in size from approximately 35 to 450 hectares and are managed with regular burns at varying intervals, with two sites managed with grazing in addition to prescribed burns. The surrounding habitat in this region is primarily characterised by industrialised corn and soybean production and cattle grazing. Sites were divided into adjacent burned and unburned patches, where the burned patches we sampled were burned in the spring of 2019 (Figure 1c). All unburned patches sampled in this study had not been burned for at least 2 years. No sites were grazed the year of the study. The presence of burned and unburned patches within each site allowed for direct within-site comparisons of bee nesting responses (see Table S1 for a full description of site management history).
We conducted this study in remnant prairies because remnants are areas that were deemed undesirable for farming by European settlers because of their topography, hydrology or use for railroad development, thus, they were never ploughed or cultivated. Studying ground-nesting bees in remnant prairie reduced any potential confounding effects or legacies of farm management practices that may impact local floral community composition, soil properties and wild bee communities (Barak et al., 2017; Lane et al., 2022; Whisler et al., 2016). Each site was similarly managed, or co-managed by the Minnesota Department of Natural Resources (MN DNR), the US Fish and Wildlife Service (USFWS) and The Nature Conservancy (TNC). To ensure that we sampled sites likely to host abundant and diverse wild bee communities, we sampled sites that were highly ranked by the MN DNR Statewide Biodiversity Significance standards or by the USFWS Morris Wetland Management District Prioritisation Model (Guidelines for Assigning Biodiversity Significance Ranks to Minnesota Biology Survey Sites, 2009, Rohweder, 2017).
These prairie remnants and the biodiversity within them exist because of the centuries of caretaking by the Dakota and Anishinaabe peoples (Westerman & White, 2012). All study sites were located on land stolen from Indigenous tribal nations and are now managed by state and federal agencies, and conservation organisations (Case, 2018; Waziyatawin., 2009, Westerman & White, 2012). While we worked directly with MNDNR, USFWS and TNC to obtain permits and signed agreements to sample bees from these sites for our research, their permitting processes did not involve tribal consent. To our knowledge, there is currently no formal process for university researchers to work with tribes or to obtain consent prior to conducting research on privately owned or public lands managed by state or federal agencies, although more recent policies obligate Minnesota state agencies to obtain consent and input from tribes (Minnesota Statute 10.65 Government-to-Government Relationship with Tribal Governments, 2021).
Bee sampling
To collect nesting bees, we used soil emergence traps, which are small tents with an open bottom and collection jar at the top (Figure 2). As bees emerged from nests in the area under the traps, they flew upward and were killed in the collection jar filled with propylene glycol at the top of the tent for the duration of each sample round, which lasted 6–10 days. There were eight sample rounds at each site.
Bees were sampled using two types of emergence traps (Figure 2). We used commercially available emergence traps covering 0.36 m2 (Figure 2a, Bugdorm Soil Emergence Trap, MegaView Science, Taiwan, 60 cm × 60 cm). We also designed and constructed custom emergence traps covering 1.3 m2 (Figure 2b, 114.3 cm × 114.3 cm). Emergence traps were set at sunset throughout the summer to ensure that foraging bees had returned to their nests prior to trap placement.
We placed 15 custom and six commercial traps in both burned and unburned patches at each of the four sites for every sampling round for a total of 42 traps per site, with a total of 168 traps set across sites simultaneously (Figure 1d). Seven traps (five custom and two commercial) were placed in three parallel haphazardly located transects per treatment. There were five meters between traps and between transects. There was a total of six transects per site. The transects were located at least 10–15 m away from the edges of the sites. The transects in the burned versus unburned patches ranged from between 5 to 100 m apart to be within flying distance of most wild bee species (Gathmann & Tscharntke, 2002; Wright et al., 2015), although at one site, traps were placed about 600 m apart due to access issues into the field site between the burned and unburned patch.
Traps were secured to the surface of the soil using metal garden staples along each side of the bottoms of the traps to prevent bees from escaping. Before placing the traps at dusk, the vegetation was cut to about 0.25–0.5 m tall using pruning shears to ensure that bees were able to easily find the collection jar after emergence. Vegetation was cut at dusk because certain bee species are known to use landmarks to find their nests that cutting during the day could disrupt (Danforth et al., 2019).
All bees collected in the traps were pinned and are stored at the UMN Bee Laboratory. Bees were identified to species by Dr. Zachary Portman using keys and revisions, and bee life history traits like nesting substrates and sociality were assigned based on the literature review (Arduser, 2015, Coelho, 2004, Gibbs, 2010, 2011, Gibbs et al., 2017, Johnson, 2018, Oram, 2018, Portman, Lane, et al., 2020; Portman, Arduser, et al., 2022; Rehan & Sheffield, 2011, Wright et al., 2020, Vickruck, 2010). We considered all communal bees to be solitary as we considered them to be individual nests that share a single entrance. For a full list of identification resources and life history traits, refer to Table S2. While the bycatch of this study was not quantitatively assessed, it largely consisted of hymenopteran (mostly wasps), dipteran, hemipteran and orthopteran species.
Vegetation and microhabitat sampling
The open, functional inflorescences of any flowering forbs were counted and identified to species in twenty 1 m2 quadrats in each burned and unburned patch at the beginning of each sample round to determine floral abundance and flowering plant species richness. In cases with highly abundant plant species with clusters of inflorescences that make it difficult to count individual flowers, an average number of flowers per cluster was calculated to determine total number of flowers in a quadrat. Flowers in the Asteraceae family in which a single flower head consists of many individual flowers were counted as one flower. Quadrats were placed in two parallel transects of 10 quadrats separated by approximately four meters and were adjacent to the emergence trap transects.
Following a sample round, microhabitat variables were recorded underneath a random subset of four traps in each transect. Beneath the sampled traps, the percent bare ground, percent vegetation cover, percent thatch, thatch thickness, slope and aspect were recorded by visual estimation by one observer, as these are all factors that may relate to nesting by ground-nesting bees (Potts et al., 2005; Sardiñas & Kremen, 2014). Importantly, the amount of bare ground refers both to open, bare soil present under a trap and any exposed soil has seen at the base of or underneath vegetation. Slope and aspect were categorised based on the presence of a slope versus flat ground and aspects known to be favourable to bee nesting (which exclude northerly aspects).
Estimating the number of active nests
A major advancement of this research is that we estimated the number of nests in burned and unburned patches that were actively being used by ground-nesting bees the year of the study (hereafter ‘active nests’), instead of inferring based on bee abundance or the presence or absence of bees within a trap. This analysis allowed us to avoid over- or under- estimating nest densities within our treatments (Portman, Brokaw, & Cariveau, 2022).
To determine the estimated number of active nests, we scored all specimens based on their species, sex, overwintering strategy and the amount of wing and mandible wear, which would accrue from foraging and digging nests. Two independent observers (Julia Brokaw and Zachary M. Portman) determined if a bee was ‘worn’ or ‘unworn’ from the presence of frayed or ragged wing edges and from the dullness or worn-down tips of the mandibles.
Active nests of solitary species were characterised by the presence of a worn female in a trap. For social species, we assigned all individuals of a given species caught in a single trap as originating from one nest to avoid inflating the number of nests. Traps were placed starting in mid-June to avoid catching overwintering adult bees emerging from hibernacula. Because the burn treatment was in the spring of 2019, we also excluded bees that were likely emerging from nests provisioned the previous year, which were characterised by the presence of single or multiple unworn solitary ground-nesting individuals of the same species that overwinter in the natal cell, all caught in the same trap. Cleptoparasites, stem-nesting bees, cavity-nesting species and males were excluded from the analysis because they do not excavate their own nests underground (see Portman, Brokaw, & Cariveau, 2022) for a full description of this methodology).
Statistical analysis
Impacts of prescribed burns on actively nesting ground-nesting bee abundance and diversity
To determine if there were differences in the number of active nests between burned and unburned patches (or ‘burn treatment’), we used the total estimated number of active nests per site per sample round as a response variable. Burn treatment was a fixed effect. Site was a random intercept. We used a generalised linear mixed effects model with a negative binomial distribution from the ‘lme4’ package in R version 3.6.2 (Bates et al., 2014; R Core Team, 2021).
To account for differences in the relative abundance of each ground-nesting bee species between burned and unburned patches, we calculated the Effective Number of Species (ENS) of actively nesting bees between burned and unburned patches for each site and sample round using the ‘vegan’ package. We used the ENS because it gives a more meaningful comparison of community differences than diversity metrics because the units of ENS are ‘units of species’ opposed to an index and ENS controls for the nonlinearity of diversity indices (Dauby & Hardy, 2012; Oksanen et al., 2013). The ENS was calculated by taking the exponent of the Shannon Diversity index for each site, sample round and treatment. The bee community matrix used for this analysis was the total estimated number of active nests of each species per site and sample round. The ENS between burned and unburned patches was compared using a linear mixed effects model with a Gaussian distribution with site as a random effect.
Impacts of prescribed burns on actively nesting ground-nesting bee community composition
We also conducted a PERMANOVA test with a Bray Curtis dissimilarity index using the adonis2() function in the ‘vegan’ package to determine if bee community composition varied between burned and unburned patches (Oksanen et al., 2013). Permutations were constrained using the setBlock() function to account for the paired design of the study. The bee species matrix used for this analysis summed the estimated number of active nests of different bee species found in each treatment across sites.
Impact of prescribed burning on the microhabitat, vegetation and floral community
Microhabitat variables were assessed for collinearity. Percent thatch cover was highly correlated with percent bare ground (r = −0.88) and was excluded from this analysis. We favoured bare ground for comparison, as bare ground is more commonly used in studies of ground-nesting bee preferences than thatch cover (Antoine & Forrest, 2021). Bare ground was also correlated with thatch depth (r = −0.68), but thatch depth was not excluded because it describes an important microhabitat feature and that may relate to bee nesting preferences.
To test for effects of prescribed burning on the microhabitat and vegetative data we collected, we used a separate linear mixed-effects model for each response variable with burn treatment as a fixed effect and site as a random effect. The response variables were means of percent bare ground, thatch depth and percent vegetative cover by site, sample round and treatment. Two of the variables required transformation prior to analysis: percent bare ground was arcsine square root transformed and average thatch depth was log-transformed.
We also compared whether our traps were similarly distributed with respect to slope and aspect between treatments using a chi-square test, as slope and aspect may influence bee preferences for nesting (Lybrand et al., 2020; Potts & Willmer, 1997; R Core Team, 2021).
Finally, we assessed the differences in average number of flowers (floral abundance), flowering species richness (raw number of flower species) and floral ENS per site and sample round. The average number of flowers per site and sample round was log-transformed and analysed using a linear mixed-effects model with a Gaussian distribution with site as a random effect and the number of flower species was analysed using a linear mixed-effects model with a Poisson distribution with site as a random effect. To calculate the ENS of the floral community, we created a site-by-floral species matrix and the total number of flowers per species was summed across quadrats per site and sample round. The ENS of the floral community was compared between burned and unburned patches using a linear mixed effects model with site as random effect. We used permutational multivariate analysis of variance (PERMANOVA) with a Bray Curtis dissimilarity index to determine if the floral communities between burned and unburned patches were significantly dissimilar. We used the setBlock() function to constrain permutations and to account for the paired study design. To create the site-by-floral species matrix for the PERMANOVA, the total number of flowers was summed across quadrats per site (Oksanen et al., 2013).
All analyses were conducted in R statistical programing assisted by using the ‘tidyverse’ package (R Core Team, 2021 version 3.6.2, Wickham et al., 2019). Model fit and residuals were assessed using the ‘DHARMa’ package (Hartig & Hartig, 2017).
RESULTS
We collected a total of 507 bee specimens in emergence traps. Of the 507 bees, 456 specimens were ground-nesting species and 330 of those were actively nesting. Using our protocol for estimating the number of active nests represented by our collected specimens, we estimated a total of 205 active nests across the season in both burned and unburned patches combined (Table S3). We analysed active nests instead of the total number of specimens of actively nesting ground-nesting bees to account for incidentally collected bees or the presence of many social species present in traps that likely only represent only one nest. In burned patches, there were a total of 131 active nests and in unburned patches there were 74 active nests. There were 30 species of 10 genera of actively nesting bees across the study.
The overall rate of detected active nests in the emergence traps across the study was 0.15 active nests per trap. In burned patches, the trap rate was 0.19 active nests per trap and in unburned patches it was 0.11 active nests per trap.
Impacts of prescribed burns on bee nesting and community composition
There were significantly more active nests of ground-nesting bees in the burned patches than unburned patches of remnant prairies (Figure 3a., Table 1a, Table S4, estimate = −0.77, ±0.26 SE, z = −2.91, p = 0.003). On average, we found 4.09 ± 0.75 SE active nests per site and sample round in burned patches and 2.31 ± 0.47 SE active nests in unburned patches.
| (a) Total estimated active bee nests | Estimate | z | p |
|---|---|---|---|
| Intercept | 1.36 ± 0.32 | 4.22 | <0.0001*** |
| Treatment unburned | −0.77 ± 0.26 | −2.91 | 0.003** |
| (b) ENS of bees | Estimate | t | p |
|---|---|---|---|
| Intercept | 2.03 ± 0.57 | 3.54 | 0.017* |
| Treatment unburned | −0.55 ± 0.38 | −1.43 | 0.157 |
We found 25 species of actively nesting bees from 9 genera in the burned patches and 16 species of 9 genera of actively nesting bees in the unburned patches. Between the 2 treatments, 11 species were found nesting in both the burned and unburned patches (Table S5). The average ENS in burned patches was 2.29 ± 0.34 SE and 1.55 ± 0.28 SE in unburned patches (Figure 3c). The ENS of the nesting bee community was not significantly different between burned and unburned patches (Table 1b, estimate = −0.55 SE = 0.38, d.f. = 59.28, t = −1.43, p = 0.157).
PERMANOVA did not show significant differences in the composition of the actively nesting ground-nesting bee communities between burned and unburned patches (p = 0.25, d.f. = 1, sum of squares = 0.30, R2 = 0.14, F = 0.98).
Impact of prescribed burning on microhabitat, vegetation and floral community
Percent bare ground in burned patches was significantly higher than in unburned patches (Figure 4a, Table 2a, estimate = −0.41, SE = 0.03, d.f. = 59, t = −15.24, p = <0.0001). In burned patches, there was an average of 22.98% ± 2.33 SE bare ground per site and sample round and in unburned patches there was an average of 1.86% ± 0.68 SE bare ground per site and sample round. To control for the potential confounding effects of the association of higher amounts of bare ground in the burned patches and the impacts on bee nesting, the residuals from the model of bee nesting and burn-treatment were plotted against the average percent bare ground for each site and sample round. Through visual examination, we determined that there was no relationship between average percent cover of bare ground and model residuals.
| (a) Average percent bare ground | Estimate | t | p |
|---|---|---|---|
| Intercept | 0.48 ± 0.06 | 8.47 | 0.002** |
| Treatment unburned | −0.41 ± 0.03 | −15.24 | <0.0001*** |
| (b) Average thatch depth | Estimate | t | p |
|---|---|---|---|
| Intercept | 0.78 ± 0.05 | 16.66 | <0.0001*** |
| Treatment unburned | 1.23 ± 0.07 | 18.53 | <0.0001*** |
| (c) Average vegetative cover | Estimate | t | p |
|---|---|---|---|
| Intercept | 0.73 ± 0.03 | 27.09 | <0.0001*** |
| Treatment unburned | −0.02 ± 0.02 | −0.97 | 0.336 |
Thatch depth was significantly lower in burned patches than unburned patches (Figure 4b, Table 2b, estimate = 1.23, SE = 0.07, d.f. = 62, t = 18.53, p = <0.0001). The average amount of thatch per site and sample round was 1.26 cm ± 0.11 SE and 6.74 cm ± 0.38 SE in burned and unburned patches, respectively. Average percent cover of vegetation did not differ significantly with burn treatment (Figure 4c, Table 2c, estimate = −0.02, SE = 0.02, d.f. = 59, t = −0.97, p = 0.336). In burned patches, the average was 73.01% ± 1.75 SE and in unburned patches the average was 70.72% ± 1.84 SE per site and sample round. Slope and aspect did not differ with burn treatment (χ21 = 2.28, d.f. = 1, p = 0.13).
Across the study, there were 69 species of flowering forbs with 51 species in burned patches and 48 species in unburned patches. Thirty species overlapped between burn treatment. Within each site and sample round, there were 20.63 ± 6.25 SE flowers counted per quadrat in burned patches and 15.82 ± 5.52 SE flowers counted per quadrat in unburned patches. The average ENS of the flowering plant community was 2.07 ± 0.20 SE and 2.31 ± 0.25 SE in burned and unburned patches, respectively.
We did not find significant differences in the number of flowers (Figure 5a, Table 3a, estimate = −0.21, SE = 0.32, d.f. = 56.97, t = −0.65, p = 0.519), the flower species richness (Figure 5b. Table 3b, estimate = −0.02, SE = 0.13, z = −0.13, P = 0.897), nor the ENS (Figure 5c, Table 3c, estimate = 0.24, SE = 0.32, d.f. = 60, t = 0.74, p = 0.464) between burned and unburned patches per site and sample round. The floral communities between burned and unburned patches were not significantly dissimilar across sites (d.f. = 1, sum of squares = 0.32, R2 = 0.11, F = 0.74, p = 0.625).
| (a) Average number of flowers | Estimate | t | p |
|---|---|---|---|
| Intercept | 2.12 ± 0.32 | 6.68 | 0.0009*** |
| Treatment unburned | −0.21 ± 0.32 | −0.65 | 0.519 |
| (b) Number of flower species | Estimate | z | p |
|---|---|---|---|
| Intercept | 1.36 ± 0.09 | 14.98 | <0.0001*** |
| Treatment unburned | −0.02 ± 0.13 | −0.13 | 0.897 |
| (c) ENS of flowers | Estimate | t | p |
|---|---|---|---|
| Intercept | 2.07 ± 0.23 | 9.14 | <0.0001*** |
| Treatment unburned | 0.24 ± 0.32 | 0.74 | 0.464 |
The species most frequently observed in quadrats across the study period were Amorpha canescens in the burned patches and Lotus corniculata in the unburned patches (see Table S6 for lists of flower species in each treatment).
DISCUSSION
Within the same year as a prescribed burn, we found that the burned patches of remnant tallgrass prairies had a higher number of estimated active nests than nearby unburned patches. However, contrary to our expectations, we found that there was no difference in effective number of species of bees or composition between burned and unburned patches. We also found distinctive differences in the microhabitat resulting from the burn that may have influenced bee nesting choices, notably that burned patches had higher amounts of bare ground beneath the vegetation and a thinner layer of thatch than unburned patches. This suggests that ground-nesting bees preferred to nest in burned patches of prairies and that burning provides high- quality nesting habitat.
Our findings align with previous research that used specimens collected from flowers or passive sampling methods to conclude that prescribed burns and wildfires in grassland habitats increase ground-nesting bee abundance and the overall proportion of ground-nesting bees (Bruninga-Socolar et al., 2022; Decker & Harmon-Threatt, 2019; Smith DiCarlo et al., 2019). Unlike our findings, other studies also found higher ground-nesting bee species richness and evenness in burned patches (Decker & Harmon-Threatt, 2019; Miller, 2021; Smith DiCarlo et al., 2019). Our bee community composition data suggest that patterns we observed were not driven by strong species-specific associations in burned or unburned patches, as was true of other studies (Ulyshen et al., 2021). Future research in prairies should examine how this pattern in bee composition may change through time, especially given how prairie management can impact soil properties across years, especially when sites are managed with a combination of burning and grazing (Buckles & Harmon-Threatt, 2019). While Buckles and Harmon-Threatt (2019) found a negative association between grazing and bee nesting, the two sites in our study with a history of grazing had the highest numbers of active nests in both burned and unburned areas. Thus, more work is needed to document whether prescribed burning confers fitness benefits to the ground-nesting bee community and how other management techniques like grazing or haying may influence nest site choices (Leone et al., 2022).
A major contribution of this research was the methodology for estimating the total number of active nests of ground-nesting bees by incorporating sociality and the estimated age of collected bees. This contrasts to other emergence trap studies that analyse total bee abundance or that rely on trap-level presence-absence data (Cope et al., 2019; Kim et al., 2006; Manion, 2020; May, 2015, Rivers et al., 2018; Sardiñas & Kremen, 2014; Sardiñas et al., 2016; Ulyshen et al., 2021). Our methodology is useful for describing patterns of within-year nest site choices by ground-nesting bees that can be applied to other studies using emergence traps (Portman, Brokaw, & Cariveau, 2022). This likely provided more robust results by excluding incidental captures and prevented inflating nesting counts due to the collection of many individuals of social species in traps.
There are a few potential factors that may have influenced the increased nesting in burned patches in our study system. First, burned patches had higher amounts of bare ground than unburned patches. Importantly, ‘bare ground’ in our study included any bare soil between stems of plants underneath leafy vegetative cover in addition to areas of open, bare soil. This distinction is important because the average percent cover of vegetation between burned and unburned patches was not significantly different. Both the amount of bare ground and access to the soil surface from a thinner thatch layer is often noted to be an important nesting resource for ground-nesting bees (Antoine & Forrest, 2021; Cane, 1991; Harmon-Threatt, 2020; Nichols et al., 2020; Olynyk et al., 2021; Sardiñas & Kremen, 2014; Stöckli, 2021; Wuellner, 1999). In addition, prairie thatch decreases soil temperature and retains moisture (Briggs & Knapp, 1995; Collins & Calabrese, 2012; Ehrenreich & Aikman, 1963), which are known to influence bee nesting choices because many species prefer well-drained soils exposed to the sun (Danforth et al., 2019; Potts & Willmer, 1997; Weissel et al., 2006; Wuellner, 1999). While we found significantly higher numbers of actively nesting bees in burned patches, we still found many ground-nesting bees nesting in the thick thatch of the unburned patches at two study sites. This requires further study to better understand specific drivers of nest site choices by ground-nesting bees in prairie habitats. In addition, it is important to leave patches unburned as refugia for the persistence of other insect and animal groups who may be more sensitive to disturbance caused by fire, especially stem-nesting bees (Bruninga-Socolar et al., 2022; Kral et al., 2017; Swengel & Swengel, 2007).
Previous studies examining burning management impacts on bees infer that the benefits of burning on the bee community are mediated by indirect effects of the local floral resources in addition to other factors like surrounding landscape composition (Buckles & Harmon-Threatt, 2019; Griffin et al., 2021; Miller, 2021; Mola & Williams, 2018; Smith DiCarlo et al., 2019). It is well documented that spring burning can increase floral diversity, abundance and plant reproduction of certain species in remnant tallgrass prairies depending on the timing of burn and other factors (Copeland et al., 2002; Ehrenreich & Aikman, 1963; Howe, 1994; Lovell et al., 1982; Wagenius et al., 2020). In our study, the higher number of bee nests in burned patches could not be attributed to any observed differences in floral richness, abundance or composition between burned and unburned patches. This may be due to the relatively few numbers of sites sampled compared to most plant-community-focused studies. Regardless, the ground-nesting bees captured in our traps had the opportunity to fly from their nest to any preferred patch of flowers within their foraging range, and within the range of traps placed in burned and unburned patches, because of the paired design within each site. This allows us to infer that their nesting choices were likely independent of the floral community composition that directly surrounded their nesting sites during any given sample round. Nevertheless, the evidence that burning benefits bees by increasing floral abundance and diversity while also creating preferred nesting habitat for ground-nesting species demonstrates the importance of this management tool for ground-nesting bee conservation.
Our results highlight the utility of emergence traps to directly document where bees are likely actively nesting (as opposed to foraging), which can help inform conservation efforts to better protect their nesting habitat (Antoine & Forrest, 2021; Harmon-Threatt, 2020; Portman, Bruninga-Socolar, et al., 2020; Requier & Leonhardt, 2020; Tepedino & Portman, 2021). In addition, our methodology of setting traps in the evenings, using high numbers of both commercial and custom-designed traps deployed simultaneously across sites, and frequently rearranging traps throughout the season allowed for a comparably high sample size of ground-nesting bees from emergence traps in this region (Buckles & Harmon-Threatt, 2019; Manion, 2020; Pane & Harmon-Threatt, 2017). However, emergence traps should be used with caution because they catch a considerable amount of non-bee bycatch that is rarely analysed (Spears & Ramirez, 2015).
Patch burning in remnant tallgrass prairies can support pollinator conservation by increasing the availability of nesting habitat for ground-nesting bee species. This is especially important because violent settler colonial projects are actively destroying remnant prairies, threatening bee habitat and contributing to their declines (Alstad et al., 2016; IPBES, 2016; Lark et al., 2019; Whyte, 2016). In recent decades, settler scientists and land managers have just begun to recognise the importance of prescribed burning for biodiversity conservation and the need to protect remnant prairie ecosystems. However, Indigenous and traditional knowledge of prairies and cultural burning practices have always been crucial for prairie health (Adlam et al., 2021; Christianson & Krisoff, 2019; Hoffman et al., 2022; Marks-Block & Tripp, 2021, Native Community Restores Prairie Through Controlled Burns, 2022; Nowacki & Abrams, 2008). Thus, future efforts to use prairie management to support bee conservation must also address legacies of colonial land dispossession by supporting movements for Indigenous sovereignty and land rematriation (Christianson & Krisoff, 2019; Hoffman et al., 2022, Waziyatawin, 2009).
AUTHOR CONTRIBUTIONS
Julia Brokaw: Conceptualization; investigation; funding acquisition; writing – original draft; methodology; writing – review and editing; visualization; project administration; data curation; supervision; formal analysis. Zachary M. Portman: Conceptualization; writing – original draft; methodology; writing – review and editing; supervision. Bethanne Bruninga-Socolar: Writing – review and editing; formal analysis. Daniel Cariveau: Writing – review and editing; formal analysis; methodology; funding acquisition; conceptualization.
ACKNOWLEDGEMENTS
The project would not have been possible without the support from H. Morales Uribe, Y. Metreaud, G. Reuter and M. Mackert, for their help in designing the custom-designed emergence traps. Many thanks to M. Vohs, S. Gedlinske, M. Shanahan, A. Waananen, L. Gedlinske, C. Dolph, M. Carr-Markel, Victor Ramírez-Juárez, A. Tona, C. Herron-Sweet, K. Friedrich, G. Pardee, R. Morris, L. Fulton, M. Dutta, D. Harder, T. Saba, E. Evans, I. Bur and M. Goblirsch, for support building traps and/or supporting the fieldwork or specimen processing, to M. Marek-Spartz, M. Shanahan, V. Wauters, I. Lane, J. Beck, A. Brokaw, M. Spivak, E. Snell-Rood, D. Larkin for their input on analysis and manuscript revisions that greatly improved the article, and to J. Gardner for support with bee identification. We also would like to thank the reviewers for their thoughtful contributions that greatly improved the manuscript. We gratefully acknowledge the US Fish and Wildlife Service, The Nature Conservancy, the Minnesota Department of Natural Resources who manage the sites where we conducted this study and supported. The authors would like to thank the following organisations for their financial support: the Environmental and Natural Resources Trust Fund (M.L. 2017, Chp. 96, Sec. 2, Subd. 03n) to DPC, the Small Grants Program from Prairie Biotic Research Inc, the Student Restoration Research Grant from the Society for Ecological Restoration Midwest Great Lakes Chapter, The Wallance and Mary Lee Dayton Fund of the Bell Museum Natural History Award, and the National Science Foundation Graduate Research Fellowship Program (Grant No. 1839286).
POSITIONALITY STATEMENT
All authors are white, settler scientists conducting research in Minnesota tallgrass prairies and thus our perspectives on prairie stewardship and management are limited. We are committed to actions that support Indigenous sovereignty and are open to comments and critiques that arise from this article.
CONFLICT OF INTEREST STATEMENT
The authors do not have any to declare.
Open Research
DATA AVAILABILITY STATEMENT
The data and code to reproduce the analysis will be made free and publicly available online after publication and all data used for analysis are also found in the supplementary materials. https://doi.org/10.13020/5sjk-8b69.
