ABSTRACT
More than 2.5 million hectares have been impacted by coal mining in the Appalachian region of the United States. Revegetation to forested cover is a desirable post-mining land use but is often impractical given the challenges of reforestation on abandoned coal mines. Considering a prospective pollination crisis and the potential value of habitat restoration for pollinators, prairie restoration on mine lands offers a practical restoration option. We tested the effect of native prairie restoration in comparison to traditional reclamation with non-native cool-season grassland on pollinator richness, diversity, and community structure at three mined sites in southeastern Ohio. Rather than treatment level effects, we found correlations between overall floral and pollinator richness and beta diversity, as well as varied pollinator diversity and distinct community composition by month. Therefore, judicious design of prairie restoration seed mixes could increase floral diversity and provide across-season forage for pollinators in post-mining landscapes. Our findings also suggest, by the presence of several specialist/uncommon pollinators, that prairie restorations on coal mines can provide habitat for at-risk pollinators.
Restoration Recap
Prairie mixes can successfully establish on reclaimed mine land, providing native plant conservation value. While they will not necessarily increase pollinator diversity and abundance, they may provide habitat for specialist pollinator species.
Pollinator abundance and diversity is directly related to floral abundance and diversity, so seed mixes incorporating diverse floral morphology and color with continuous bloom throughout the growing season will best support pollinators.
This article was prepared by a U.S. government employee as part of the employee’s official duties and is in the public domain in the United States.
Conversion of land from natural systems to human-dominated land cover types has resulted in the endangerment of species and ecosystems. In particular, pollinating insects appear to be at risk from land conversion and land-use intensification, which destroy and fragment forage and nesting resources (Vanbergen 2013). Both wild and domesticated pollinators have experienced recent declines (Potts et al. 2010). For example, several Bombus spp. (bumble bees) in the United States have lost 23–87% of their geographic ranges (Cameron et al. 2011), and 76% of 1,437 bee species native to North America and Hawaii are likely imperiled or declining (Kopec and Burd 2017). Anthropogenic disturbance can impact bee declines; abundance and richness significantly decline in areas with severe habitat destruction (Winfree et al. 2009). However, land use may also change pollinator composition, and specialist pollinators in particular are negatively impacted by native species losses (Harrison et al. 2018). Restoration designed to restore pollinator communities, therefore, needs to consider species assemblages, specialist species, and overall abundance and richness.
Ecosystem loss threatens dependent species and ecosystem services. In the United States, up to 98% of the historical extent of tallgrass prairie has been lost (Comer et al. 2018). This dramatic reduction in prairie extent includes losses of ecosystem services such as maintenance of genetic diversity, carbon and nitrogen sequestration exceeding that of forests (Seastedt and Knapp 1993), soil conservation (Daily 1997), and loss of wildlife and pollinator habitat.
Mined Lands and Prairie Restoration
Surface coal mining in the United States disturbed ∼2.5 million ha from 1930–1977 (Skousen and Zipper 2014). Although rates of land conversion from the practice seem to be slowing (Pericak et al. 2018), it is estimated that ≥10,000 ha are mined each year in Appalachia (Burger 2011, Zipper et al. 2011) and ∼290,000 ha were newly mined between 1985–2015 in central Appalachia (Pericak et al. 2018). Prior to 1977, surface mined lands were typically either abandoned or reforested in the Appalachian region. Following the passage of the Surface Mining Control and Reclamation Act (SMCRA) of 1977, reclamation and recontouring of the soil were required to eliminate high walls and reduce soil erosion, but this practice led to difficulty in successful reforestation due to compact soil. Instead, landowners met short-term revegetation goals with quicker growing non-native grasses and forbs such as Dactylis glomerata (orchard grass), Phleum pratense (timothy), Lotus corniculatus (bird’s-foot trefoil), and Trifolium repens (white clover) (Holl and Cairns 1994).
Mined Lands as an Opportunity for Pollinator Habitat Restoration
Ecological restoration positively impacts pollinators in vegetative restoration of hedgerows (Morandin and Kremen 2013), landfills (Tarrant et al. 2012), riparian forests (Williams 2011), and roadside grasslands (Hopwood 2008) as well as plant-pollinator networks in restored hay meadows (Forup and Memmott 2005) and invasive shrub dominated mountaintops (Kaiser-Bunbury et al. 2017). Some research has shown that prairie restorations, even if not focused on pollinators, can benefit pollinators (Geroff et al. 2014, Harmon-Threatt and Hendrix 2015, Robson 2008, Tonietto et al. 2017). Research on reclaimed surface coal mines has shown that prairie restorations can benefit pollinators if designed for them (Cusser and Goodell 2014, Novotny and Goodell 2020). However, there are many commercially available prairie seed mixes, some of which are not specifically designed for pollinators. Reviews of these seed mixes exist (Zinnen et al. 2022), but there is need to look at how pollinators respond to mixes not designed to benefit pollinators, and compare these mixes with non-prairie mixes. Tall grass prairie mixes in particular are often species poor (Zinnen et al. 2022), similar to the mix used in this study. While the purpose of this initial study was to evaluate the feasibility of native species establishment in reclaimed abandoned mined lands, the experiment suggests that increasing the quantity and diversity of native species seeded enhances habitat value. As previously reported, both prairie grasses and forbs have been shown to successfully establish on reclaimed mined land (Drake 1980, Heckman et al. 1996, Swab et al. 2017, Scott et al. 2021), with levels of biomass comparable to that produced by non-native cool-season grasses (Rodgers and Anderson 1989). Given the large extent of abandoned surface coal mines yet to be revegetated in Appalachia (120,000 ha in Pennsylvania alone [OSMRE 2018]), restoration presents a unique opportunity to conserve native prairie species in the context of novel ecosystems—formerly forested mined lands which, after reclamation, can more effectively support grassland instead. Seed mixes can be effectively designed to balance both economy and species diversity, making it possible to achieve native prairie mixes similar in cost to current standard non-native reclamation seed mixes (Swab et al. 2017, Scott et al. 2021).
The goal of this study was to evaluate the diversity and community structure of pollinating insects on mined land reclaimed with a seed mix containing mostly native prairie plants in comparison to sites planted with traditional nonnative reclamation seed mixes. Our specific objectives were to determine 1) if reclamation with traditional cool-season seed mixtures compared with native prairie mixtures affects pollinator richness and diversity; 2) how variables such as time of year and floral unit abundance, richness and beta diversity affect pollinator beta diversity, abundance, and composition; and 3) if the composition of pollinator communities differs between traditional non-native cool-season grasslands and native prairie plantings, and whether there are any specialist pollinators supported by the natives.
Methods
Study Sites
Three sites in southeastern Ohio slated for reclamation by the Ohio Division of Mineral Resource Management (DMRM) were revegetated between December, 2014 and July, 2015 (Figure 1). Sites had been mined prior to 1977 and abandoned (left without reclamation) until 2014. The DMRM reclaimed these sites utilizing reclamation standards established by the Surface Mining Control and Reclamation Act of 1977. At each site, the best available topsoil, subsoil, or spoil material was spread across the site to an average depth of 0.3 m, and then the sites were seeded. Three plots of 0.4 ha were established at each site. Plots were adjacent to each other in two sites. In the third (Rose Valley), due to the shape of the reclamation site, plots were spaced ∼200 m apart. At all sites, forest surrounded the reclamation plots, with the forest edge equidistant of all plots (∼200 m, varying by site). Two plots were revegetated with an experimental native prairie seed mix created for this purpose comprised of both non-native species typically used in reclamation, as well as native species. All seed mixes were purchased from Ernst Conservation Seeds in Meadville, Pennsylvania. The native seed mix consisted of the following native species and their proportions: Panicum virgatum (switchgrass, 0.18), Sorghastrum nutans (Indian-grass, 0.14), Chamaecrista fasciculata (partridge pea, 0.18), Coreopsis tinctoria (golden tickseed, 0.01), Rudbeckia triloba (brown-eyed Susan, 0.01), Helianthus maximiliani (Maximilian sunflower, 0.01), Monarda fistulosa (wild bergamot, 0.01) and Asclepias syriaca (common milkweed, 0.01). To ensure that the reclaimed soil would be revegetated, the following proven non-native species were included: L. corniculatus (0.1), Lolium perenne (perennial ryegrass, 0.18), and D. glomerata (0.18). The plots were seeded at rates of 16.81 kg/ha−1 (Native Light) and 33.63 kg/ha−1 (Native Heavy) rates. The third plot at each site was seeded with a non-native mix frequently used in the reclamation of mined lands in Ohio (Traditional), including the three nonnative species above augmented with P. pratense, Trifolium pratense (red clover), and Lolium multiflorum (Italian ryegrass) at a seeding rate of 33.63 kg/ha−1. Reclamation was considered successful on all sites and in all treatments according to industry standards, which are focused on the percentage of vegetative cover. No differences in vegetative cover, species diversity, or richness were observed between the Native Light and Native Heavy seed mixes five years after reclamation (Swab et al. 2017). All three sites, despite minor differences in soil conditions and seeding dates, had successful establishment of vegetation in all three treatment types (Swab et al. 2017). In addition to species planted, there were a number of volunteer species across all plot types.
Locations of surface coal mines reclaimed with an experimental prairie seed mix in 2014–2015 described by county: Middleton Run (Jackson County), Joyce Hill (Tuscarawas County), and Rose Valley (Belmont County). Abandoned mine land (AML) and other mine land (OML) represent surface coal mines in Ohio, USA classified as either abandoned or active/inactive/released, respectively, by the Ohio Department of Natural Resources’ Division of Mineral Resources Management (updated through 2018). Pollinator surveys were conducted at the three sites in summer 2019.
General Sampling Design
Each site was sampled for pollinators once per month from June–September 2019 on warm (21–32°C) days with no rainfall and ≤ 4.0 m/s winds. During these sampling events, within designated areas (described below) within each seed mix, we collected insects observed visiting flowers for later identification, and recorded the species of flower visited. Within each month, one site was sampled per day and subsequent monthly sampling at each site was approximately four weeks apart. Each day, sampling was divided into three rounds: early morning (09:00–10:30), late morning (10:30–12:00), and afternoon (12:30–14:00) to encompass the daily activity patterns of pollinators. Each plot was sampled once during each round, randomly ordered, but with at least 30 minutes between sampling at each plot. Within each plot, three 25-m transects were established from a central point, choosing the direction with a random number generator each time. A different transect in each plot was surveyed each round, for a total of three per plot. Thus, each sampling was in an area undisturbed by previous sampling. Each time a plot was sampled, a random number generator was used to place two 1-m2 quadrats along each transect. If a quadrat without open fresh flowers was randomly selected, the next closest quadrat with flowers was surveyed instead. If no quadrats along the transect had flowers, no pollinators were surveyed, and floral resources were recorded as zero. The proportion of quadrats along the transect with flowers was recorded.
Flower-visiting Insect and Floral Resource Sampling
Within each quadrat, for four minutes all pollinating insects observed probing floral reproductive parts were collected by net for identification. Once all sampling rounds were complete, a six-minute collection occurred across the entire plot to ensure sampling of less abundant and/or underrepresented flowers. Our priority was to assess pollinator richness during the six-minute collections. Insects were identified to lowest taxonomic level possible in the lab using microscopes and with consultation of expert Karen Goodell.
Floral resource surveys immediately followed four-minute pollinator surveys. In each quadrat, the number of floral units of each species were recorded. Floral units were separated by a pollinator’s ability to fly or walk between individual flowers (Cusser and Goodell 2013). For example, hundreds of individual flowers of Solidago canadensis (Canada goldenrod) could be considered one unit if a pollinator could walk across them without flying, or split into separate units if there were unwalkable gaps between flowers. This placed the available floral resources in the context of the pollinators, who are likely to walk from flower to flower if they are able, but might move to another plant entirely once they take off (personal observation). Number and identity of flower species in bloom within the entire plot was recorded.
Data Analysis
The counts of pollinators and floral units, respectively, were summed by taxa across quadrats and six-minute walks for each plot by sampling month combination (3 quadrats × 3 transects × 4 months, n = 36). All statistics were performed using R statistical programming language (R Core Team 2021). Pollinator diversity and floral unit diversity were calculated using the Shannon-Weiner diversity index; evenness was calculated using Pielou’s evenness. Diversity indices were calculated using the ‘vegan’ package (Oksanen et al. 2019). Linear mixed effect models (LMMs) with restricted maximum likelihood were fit to assess the independent effects (fixed) of seed mixes (treatment), sampling month (month), floral unit abundance, richness, and evenness on response variables: pollinator richness and pollinator diversity. Pollinator evenness response was not evaluated as it could not be transformed to meet model assumptions. Model fitting was performed using lmer function of the package ‘lme4’ (Bates et al. 2015). Random intercepts of site, plot, and interaction of site by plot were included, in all univariate models (lmer syntax: y∼ x + (1|Plot) + (1|Site) + (1|Plot:Site). This random effect structure was maintained even when variances of zero were detected in order to account for plots nested within sites and repeated measures on plots. After fitting models, a check for multicollinearity of continuous variables was performed using the vif function in the car package (Fox and Weisberg 2019). To maintain a low (< 5) variance inflation factor, Shannon’s diversity of plants was not included in models because plant richness and evenness are the components of Shannon’s diversity. Model assumptions of homoskedasticity, normality of residuals, and approximately normal distribution of random effects were visually verified. The small sample size corrected Akaike Information Criterion (AICc) was used to determine relative quality of models evaluated separately for each response variable. Starting with a global model, the dredge function in the ‘MuMIn’ package (Barton 2019) was used to select the combination of fixed effects that always included treatments (the only manipulated predictor) that minimized AICc. The selected model and the full model were refit as maximum likelihood models and compared using a chi-squared test. P-values were calculated using the ‘lmerTest’ package (Kuznetsova et al. 2017), with type 3 sums of squares and Kenward-Roger denominator degree of freedom corrections. Significance thresholds for univariate predictors were adjusted with a Bonferroni correction (i.e., the original alpha [0.05] was divided by the number of fixed predictors [6] for a corrected alpha of 0.0083). Marginal (accounts for fixed effects only; R2m) and conditional (accounts for fixed and random effects; R2c) pseudo-R2 values were calculated using the r.squaredGLMM function in the ‘MuMIn’ package (Barton 2019). Additional models with the same structure were used to test if floral richness and floral evenness differed among treatments. When categorical effects were significant, contrasts were evaluated with Šidák-corrected estimated marginal means using the ‘emmeans’ package (Lenth 2019).
Multivariate community statistics were performed with the ‘vegan’ package. Gamma diversity was calculated as the total number of species encountered within each seeding treatment using the specnumber function. Permutational multivariate analysis of variance (PERMANOVA) was performed with the adonis2 function using marginal P-values. PERMANOVA was used to identify variables that were associated with differences in community composition. The PERMANOVA analysis used treatment, month, floral unit abundance, richness, and evenness as predictors and a plot by site interaction block. A principal coordinate ordination was created to display pollinator community composition. Because PERMANOVA assumes homogenous dispersion, significant categorical predictors were subjected to a permutational dispersion test using the betadisper function with a block of plot by site interaction. Continuous variables that were significant in the PERMANOVA were subjected to vector fitting on the PCoA ordination. Pairwise PERMANOVA was performed for all significant categorical predictors using the pairwise.adonis function (Martinez Arbizu 2020). All variables that influenced pollinator community composition according to PERMANOVA were then included in a constrained PCoA ordination. Significance of the constrained ordination was tested using a permutation test. To test the correlation of pollinator and plant beta diversity, a mantel test of Bray-Curtis ecological distances of pollinator and flowering plant communities was performed using the Pearson method.
Indicator species analysis was performed for significant categorical PERMANOVA predictors. Indicator species analysis was calculated as the product of frequency and abundance of species to a group (De Cáceres et al. 2012), using the multipatt function of the ‘indicspecies’ package (De Caceres and Legendre 2009). Indicator species values were reported for all continuous combinations of month.
Results
Across all sampling times and locations, 501 individuals from 58 pollinator species were encountered in the Native Heavy treatment, 48 species in Native Light, and 53 in Traditional. The top four taxa encountered, each with 40–67 individuals collected, were Apis mellifera (western honey bee), Toxomerus marginatus (margined calligrapher), Chauliognathus pensylvanius (goldenrod soldier beetle), and Lasioglossum spp. (sweat bees). A list of all taxa with individual counts by site can be seen in Supplementary Table S1. The most parsimonious model of pollinator richness including the imposed seeding treatment also had floral richness as a predictor (month, floral abundance, and floral evenness dropped from model). The selected model was not significantly different from the full model (χ211 = 10.55, p= 0.482). Pollinator richness responded positively to floral richness (Figure 2A). Pollinator richness was not significantly affected by treatment (F2, 3.99 = 0.19, p = 0.835; Figure 2 B). The most parsimonious model of pollinator Shannon’s diversity including the imposed seeding treatment also had floral richness as a predictor (month, floral abundance, and floral evenness dropped from model). The selected model was not significantly different from the full model (χ211 = 9.19, p = 0.604). The pollinator richness model predicts between 5 and 12 pollinator species over the range of 1 to 11 flowering plant species. Pollinator Shannon’s diversity did not significantly increase with floral richness (F1, 29 = 5.33, p = 0.028). Pollinator Shannon’s diversity was not significantly affected by treatment (F2, 3.99 = 0.19, p = 0.989; Figure 2 C). Floral richness (F2, 4 = 0.32, p = 0.745) did not differ among treatments. The statistical model comparing floral evenness between treatments did not converge, likely because of little variation among fixed effects.
Univariate responses of pollinator richness and Shannon’s diversity. (A) Relationship of pollinator richness and flowering plant richness. (B) Non-significant response of pollinator richness to seeding treatment. (C) Non-significant response of pollinator Shannon’s diversity to seeding treatment. In boxplots, whiskers represent the most extreme values observed, the box represents the interquartile range, the line represents the median, and points represent the estimated marginal means with standard errors.
Sampling month best explained pollinator community structure (PERMANOVA: p = 0.001; Figure 3A). All months significantly differed in pairwise centroid location comparison (pairwise PERMANOVA: p < 0.05). In other words, each sampling month had a distinct pollinator community composition. Month did not significantly vary in dispersion (PERMDISP: p = 0.931), suggesting the significant PERMANOVA result represents true differences in centroid location. Floral evenness also had a significant PERMANOVA result (p = 0.044) but had a non-significant vector fit (p = 0.375). The non-significant vector fit suggests that the PERMANOVA result is an artifact of differences in dispersion along the floral evenness gradient. Month (p = 0.959), floral richness (p = 0.294), floral abundance (p = 0.866), and the interaction of treatment and month (p = 0.993) were all unrelated to pollinator community composition according to PERMANOVA. A PCoA constrained by month was highly significant (p = 0.001; Figure 3 B). Pollinator beta diversity (Bray-Curtis ecological distances between all plots) was positively correlated with plant beta diversity (Mantel: r = 0.24, p = 0.001).
Multivariate pollinator community responses. (A) Unconstrained principal coordinates (PCoA) ordination with ellipses displaying months. (B) Unconstrained PCoA ordination with ellipses displaying seeding treatment. (C) Constrained PCoA ordination with ellipses displaying month.
June pollinator community was indicated by T. marginatus (IV = 0.80, p = 0.005), Chauliognathus marginatus (margined leatherwing) (IV = 0.67, p = 0.010), and Sphaerophoria spp. (hoverflies) (IV = 0.67, p = 0.005). July pollinator community was indicated by Bombus bimaculatus (two-spot bumblebee) (IV = 0.70, p = 0.005) and Ceratina dupla (doubled ceratina) (IV = 0.66, p= 0.015). June–July pollinator community was indicated by Bombus griseocollis (brown-belted bumblebee) (IV = 0.62, p = 0.020). August–September pollinator community was indicated by C. pensylvanicus (IV = 0.85, p = 0.005). June–August pollinator community was indicated by the genus Lasioglossum (IV = 0.72, p = 0.020).
Discussion
Expanding on previous research at this site, this study shows that while native prairie plants can establish and succeed on abandoned surface coal mines (Swab et al. 2017, Scott et al. 2021), this may not increase pollinator richness and diversity. However, natives can provide habitat for specialist pollinator species which would otherwise not exist in these novel ecosystems.
While some studies show grassland and prairie restoration can increase pollinator richness and diversity (Ries et al. 2001, Hopwood 2008, Tarrant et al. 2012), in this case the lack of difference in pollinator richness and diversity between treatments was likely related to the lack of difference in vegetation richness and diversity between native and traditional plots. However, in contrast to vegetation, which had different composition in native and traditional sites (Scott et al. 2021), the composition of the pollinator community also did not differ between treatments. It is important to note that our experimental native seed mix was comprised of only ten plant species, most of which were not insect pollinated. The limited suite of species was intended to decrease costs and incentivize use; it contained two papilionaceous legumes, two radially symmetrical asters, one bilateral mint, and one asclepiadaceous milkweed. During the time of the surveys, 24 flowering plant species were found in total at the three sites, with each site having 13 or 14 species (See Supplementary Table S2). Given this limited seed mix and the contribution to all treatments by volunteer species, floral diversity (measured here), an important factor for floral visiting insects, did not differ between treatments. However, like other studies (Potts et al. 2003, Ebeling et al. 2008, Fründ et al. 2010), pollinator richness was positively correlated with floral richness. Pollinator community composition expectedly varied by time of year, but alpha diversity had little variation explained by month (dropped from both pollinator richness and pollinator alpha diversity models). Pollinator beta diversity exhibited a weak, but significant, positive correlation with floral beta diversity. Though the vegetative communities were different between the prairie and traditional treatments, floral bloom diversity did not differ between treatments, and therefore pollinators were unaffected by treatment as a whole. Only four of the planted species were native flowering species, and thus floral diversity overall was largely unaffected despite differences in the entire vegetative community. Most of the blooms available at any one point in time were volunteer species, and equally likely to be in native or traditional plots, rather than species from the seed mixes.
Our results support two potential improvements regarding seed mix design. First, rather than relying on commercial seed mixes which may be high in graminoids and low in forbs (Zinnen et al. 2022), seed mixes for prairie restorations should be designed with high floral resource richness as this can improve pollinator diversity. For example, Lane et al. 2020 found more florally rich prairie plantings attract more diverse bee communities irrespective of percent of surrounding agriculture. Second, providing diverse forage throughout the season should be prioritized to maintain diverse pollinator communities, especially on degraded sites such as abandoned coal mines (Havens and Vitt 2016). Pollinating insects exhibit interspecific phenological variation to time their emergence with the availability of host plants and cannot meet their dietary needs in periods of floral scarcity. Therefore, managers should design seed mixes to ensure the continuous availability of flowers (ideally with diverse colors and morphology) during months of insect activity. Which species are best will vary by region, but local soil and water conservation district offices should be able to provide advice or suggestions by region, as well as providing recommendations for local seed suppliers.
While many taxa collected from our sites, regardless of treatment, were diet generalists (Hilty 2017), which can typically cope with habitat fragmentation more easily (Ashworth et al. 2004), several signs indicate that the native prairie plantings are providing unique resources for pollinators that would be rarely encountered outside the native plantings. Andrena helianthi (sunflower miner bee), a bee that specializes on Helianthus spp., and Megachile inimica (hostile leafcutter bee), an Asteraceae specialist, were collected from H. maximiliani only within native treatments. Additionally, we collected one to two workers of Bombus auricomus (black and gold bumble bee), Bombus perplexus (confusing bumble bee), and Bombus vagans (half-black bumble bee), three less common bumble bee species in Ohio (Lanterman et al. 2021), feeding exclusively from M. fistulosa only in native treatments. Maximilian sunflower and wild bergamot, both planted, would otherwise likely be absent on these abandoned coal mine sites. Specialist animals, with their narrower habitat requirements, may have a harder time colonizing restored sites due to the absence of source populations and/or their inability to navigate matrices of nonhabitat (Menz et al. 2011). Nonetheless, the presence of uncommon/rare species in the native seeded treatments shows promise in their capability to provide habitat for species of conservation concern.
Conclusion
There is potential value in restoring degraded vegetation communities for the conservation of pollinators (Hopwood 2008, Tarrant et al. 2012). Here we show that native prairie restoration treatments on reclaimed coal mines did not outright increase floral visitor richness, alpha diversity, or result in a distinct pollinator assemblage composition compared to traditional cool-season grassland four to five years post reclamation. However, we observed a positive correlation between floral and pollinator richness as well as a positive relationship between floral and pollinator beta diversity. We also found specialist pollinator species only within the native treatment. Overall, these results suggest that even the inclusion of only a few native flowering species can promote a more diverse pollinator community with specialist species. To better improve pollinator habitat, we recommend incorporating a more diverse set of flowering species into the reclamation mixture, including species that present color and morphological variability, as well as continuous bloom across the growing season.
Acknowledgments
We thank Dr. Karen Goodell and Dr. Stephen Spear for their assistance in study design as well as Alexys Nolan, Eliza Nolan, and Andrea Malek for help in the field. This project was funded through the Ohio Department of Natural Resources. We would like to acknowledge the following ODNR site managers for their assistance with site knowledge and access: Chad Kinney, Kaabe Shaw, Kevin Brachter, Michael Gosnell, Jeff Calhoun, and Ben McCament.
Footnotes
Supplementary materials are available online at: https://er.uwpress.org
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.