Content uploaded by Elaine Evans
Author content
All content in this area was uploaded by Elaine Evans on Aug 15, 2019
Content may be subject to copyright.
A preview of the PDF is not available
Monitoring and Habitat Assessment of Declining Bumble Bees in Roadsides in the Twin Cities Metro Area of Minnesota
Abstract and Figures
Several bumble bee species have declined dramatically, including the endangered rusty-patched bumble bee, Bombus affinis. Roadsides offer a unique opportunity to increase habitat for these declining species. The objectives of this study are to: (1) characterize the bumble bee community and floral availability within roadsides in the Minneapolis and Saint Paul, Minnesota, metro area, (2) estimate detection probabilities and occupancy for bumble bees using occupancy modeling, (3) determine the effort needed to detect rusty-patched bumble bees, and (4) examine the relationship of the bumble bee community to the surrounding landscape. We use rapid and broad-scale sampling at randomly selected locations. Despite overall low floral abundance, many bumble bee species, including rare and declining species, use roadsides. Occupancy models predict rusty-patched bumble bees occupy 4% of sites, with a 30% chance of detection if it is at the site. We recommend performing nine surveys in a single season to be 95% sure that B. affinis is detected if it is there. Bumble bee abundances and species numbers increase with more wooded area and floral cover. Crops are negatively associated with bee abundance, species numbers, and the presence of rare bumble bees. Our management recommendations for roadsides to support rare and declining bumble bees are: (1) incorporate additional bumble bee forage, (2) when weed control requires elimination of flowering plants, replace with bumble bee forage, (3) use our estimates for occupancy and abundance as a baseline to assess conservation efforts for bumble bees within roadsides in the metropolitan area of Minneapolis and Saint Paul.
Figures - uploaded by Elaine Evans
Author content
All figure content in this area was uploaded by Elaine Evans
Content may be subject to copyright.
... Cumulative detection probability allows land managers to design monitoring schemes that allocate effort efficiently based on the species detection probability (MacKenzie and Royle, 2005). The value of cumulative detection probability for improving imperiled bee monitoring is just starting to be explored (Evans et al., 2019;Graves et al., 2020;Otto et al., 2023). ...
... Lastly, observer skill and experience levels vary and could impact detection (McNeil et al., 2019;Otto et al., 2023). Although researchers have begun to acknowledge the importance of accounting for imperfect detection in bumble bee monitoring (Evans et al., 2019;Graves et al., 2020;Janousek et al., 2023;Otto et al., 2023), more work is needed to understand the biotic and abiotic factors that influence the detection of rare or declining bumble bee species. ...
... Conversely, a monitoring program specifically targeting the declining B. fervidus could optimize efficiency by sampling later in the season when this species reaches its peak detection probability in our study system (three surveys needed to reach 75 % cumulative detection probability on August 7 versus seven surveys needed on July 8). This example also emphasizes the importance of covariates on the detection process when compared with previously reported detection estimates generated from fitting single species occupancy models without covariates to this dataset (Evans et al., 2019). When detection covariates were excluded, we estimated B. fervidus detection probability was a static 0.26 under all survey conditions (Evans et al., 2019), which would suggest that five surveys are needed to reach 75 % cumulative detection probability on any date within our survey period, potentially misguiding sampling design. ...
... The number of surveys required to achieve a high detection threshold rose considerably in August and September due to a significant drop in detection probability (Fig. 4). Our research is supported by Evans et al. (2019) who estimated a 0.30 probability of detecting B. affinis during 10-min surveys along randomly selected roadside transects in eastern Minnesota from mid-June through late-August. The authors concluded that nine surveys would be required to achieve a 95% detection threshold. ...
... While our study revealed higher detection estimates, it is difficult to directly compare these values because of differences in sampling design, model structure, and site selection. For example, we included weather and other covariates in our analysis while Evans et al. (2019) only considered a model that assumed constant detection probability (null model). Furthermore, we selected sites of high habitat quality with prior history of B. affinis detections whereas Evans et al. (2019) conducted a random selection of sites along roadsides with no prior knowledge of B. affinis detection. ...
... Thus, our study provides a more holistic evaluation of how B. affinis detection probability is expected to change when sampling varies across date ranges, weather conditions, and habitat quality, and the detection estimates we generated have direct relevance to the B. affinis monitoring guidance (USFWS 2019). We note that both Evans et al. (2019) and our study suggest multiple repeat visits may be needed to accurately discern occupancy of B. affinis. ...
The U.S. Fish and Wildlife Service developed national guidelines to track species recovery of the endangered rusty patched bumble bee [Bombus affinis Cresson (Hymenoptera: Apidae)] and to investigate changes in species occupancy across space and time. As with other native bee monitoring efforts, managers have specifically acknowledged the need to address species detection uncertainty and determine the sampling effort required to infer species absence within sites. We used single-season, single-species occupancy models fit to field data collected in four states to estimate imperfect detection of B. affinis and to determine the survey effort required to achieve high confidence of species detection. Our analysis revealed a precipitous, seasonal, decline in B. affinis detection probability throughout the July through September sampling window in 2021. We estimated that six, 30-min surveys conducted in early July are required to achieve a 95% cumulative detection probability, whereas >10 surveys would be required in early August to achieve the same level of confidence. Our analysis also showed B. affinis was less likely to be detected during hot and humid days and at patches of reduced habitat quality. Bombus affinis was frequently observed on Monarda fistulosa (Lamiales: Lamiaceae), followed by [Pycnanthemum virginianum Rob. and Fernald (Lamiales: Lamiaceae)], Eutrochium maculatum Lamont (Asterales: Asteraceae), and Veronicastrum virginicum Farw. (Lamiales: Plantaginaceae). Although our research is focused on B. affinis, it is relevant for monitoring other bumble bees of conservation concern, such as B. occidentalis Greene (Hymenoptera: Apidae) and B. terricola Kirby (Hymenoptera: Apidae) for which monitoring efforts have been recently initiated and occupancy is a variable of conservation interest.
... As these models are relatively new, they have not been used in many previous bumblebee ecology studies, perhaps also because they require repeat visits to survey sites which necessarily means sampling fewer different sites (under a fixed amount of survey effort). However, even when employed, previous work has not accounted for the potential influence of habitat variables on detection probability, either not including any parameters on detection (Evans et al., 2019) or only including time of year, time of day, or survey effort as predictors of detection (Cole et al., 2019;Loffland et al., 2017). ...
... These results highlight an important but often overlooked component of observational studies of bees, where the influence of habitat characteristics on whether or not you see a species of interest (and not whether or not it occurs there) are largely unknown. Previous work employing occupancy modeling for observational studies of bumble bee ecology did not incorporate habitat predictors on detection probability(Cole et al., 2019;Evans et al., 2019;Loffland et al., 2017). Although an occupancy framework imparts additional logistical constraints in requiring revisits of sites that directly trades off with the number of sites sampled, our results highlight the importance of differentiating between likelihood of occupancy versus likelihood of detection. ...
Bumble bees are important pollinators in temperate forested regions where fire is a driving force for habitat change, and thus understanding how these insects respond to fire is critical. Previous work has shown bees are often positively affected by the postfire environment, with burned sites supporting greater bee abundance and diversity, and increased floral resources. The extent to which fire impacts variation in bumblebee site occupancy is not well-understood, especially in higher latitude regions with dense, primarily coniferous forests. Occupancy models are powerful tools for biodiversity analyses, as they separately estimate occupancy probability (likelihood that a species is present at a particular location) and detection probability (likelihood of observing a species when it is present). Using these models, we tested whether bumblebee site occupancy is higher in burned locations as a result of the increase in canopy openness, floral species richness, and floral abundance. We quantified the impact of fire, and associated habitat changes, on bumblebee species' occupancy in an area with high wildfire frequency in British Columbia, Canada. The burn status of a site was the only significant predictor for determining bumblebee occurrence (with burned sites having higher occupancy); floral resource availability and canopy openness only impacted detection probability (roughly, sample bias). These findings highlight the importance of controlling for the influence of habitat on species detection in pollinator studies and suggest that fire in this system changes the habitat for bumble bees in positive ways that extend beyond our measurements of differences in floral resources and canopy cover.
Petition to list the variable cuckoo bumble bee (Bombus variabilis) to the Endangered Species Act.
Several species of bumblebees have recently experienced range contractions and possible extinctions. While threats to bees are numerous, few analyses have attempted to understand the relative importance of multiple stressors. Such analyses are critical for prioritizing conservation strategies. Here, we describe a landscape analysis of factors predicted to cause bumblebee declines in the USA. We quantified 24 habitat, land-use and pesticide usage variables across 284 sampling locations, assessing which variables predicted pathogen prevalence and range contractions via machine learning model selection techniques. We found that greater usage of the fungicide chlorothalonil was the best predictor of pathogen (Nosema bombi) prevalence in four declining species of bumblebees. Nosema bombi has previously been found in greater prevalence in some declining US bumblebee species compared to stable species. Greater usage of total fungicides was the strongest predictor of range contractions in declining species, with bumblebees in the northern USA experiencing greater likelihood of loss from previously occupied areas. These results extend several recent laboratory and semifield studies that have found surprising links between fungicide exposure and bee health. Specifically, our data suggest landscape-scale connections between fungicide usage, pathogen prevalence and declines of threatened and endangered bumblebees. © 2017 The Author(s) Published by the Royal Society. All rights reserved.
Appropriate large-scale citizen-science data present important new opportunities for biodiversity modelling, due in part to the wide spatial coverage of information. Recently proposed occupancy modelling approaches naturally incorporate random effects in order to account for annual variation in the composition of sites surveyed. In turn this leads to Bayesian analysis and model fitting, which are typically extremely time consuming. Motivated by presence-only records of occurrence from the UK Butterflies for the New Millennium data base, we present an alternative approach, in which site variation is described in a standard way through logistic regression on relevant environmental covariates. This allows efficient occupancy model-fitting using classical inference, which is easily achieved using standard computers. This is especially important when models need to be fitted each year, typically for many different species, as with British butterflies for example. Using both real and simulated data we demonstrate that the two approaches, with and without random effects, can result in similar conclusions regarding trends. There are many advantages to classical model-fitting, including the ability to compare a range of alternative models, identify appropriate covariates and assess model fit, using standard tools of maximum likelihood. In addition, modelling in terms of covariates provides opportunities for understanding the ecological processes that are in operation. We show that there is even greater potential; the classical approach allows us to construct regional indices simply, which indicate how changes in occupancy typically vary over a species’ range. In addition we are also able to construct dynamic occupancy maps, which provide a novel, modern tool for examining temporal changes in species distribution. These new developments may be applied to a wide range of taxa, and are valuable at a time of climate change. They also have the potential to motivate citizen scientists.
Occupancy modelling has received increasing attention as a tool for differentiating between true absence and non-detection in biodiversity data. This is thought to be particularly useful when a species of interest is spread out over a large area and sampling is constrained. We used occupancy modelling to estimate the probability of three phylogenetically independent pairs of native—introduced species [Megachile campanulae (Robertson)—Megachile rotundata (Fab.), Megachile pugnata Say—Megachile centuncularis (L.), Osmia pumila Cresson—Osmia caerulescens (L.)] (Apoidea: Megachilidae) being present when repeated sampling did not always find them. Our study occurred along a gradient of urbanization and used nest boxes (bee hotels) set up over three consecutive years. Occupancy modelling discovered different patterns to those obtained by species detection and abundance-based data alone. For example, it predicted that the species that was ranked 4th in terms of detection actually had the greatest occupancy among all six species. The native M. pugnata had decreased occupancy with increasing building footprint and a similar but not significant pattern was found for the native O. pumila. Two introduced bees (M. rotundata and M. centuncularis), and one native (M. campanulae) had modelled occupancy values that increased with increasing urbanization. Occupancy probability differed among urban green space types for three of six bee species, with values for two native species (M. campanulae and O. pumila) being highest in home gardens and that for the exotic O. caerulescens being highest in community gardens. The combination of occupancy modelling with analysis of habitat variables as an augmentation to detection and abundance-based sampling is suggested to be the best way to ensure that urban habitat management results in the desired outcomes.
Wild bee declines have been ascribed in part to neonicotinoid insecticides. While short-term laboratory studies on commercially bred species (principally honeybees and bumblebees) have identified sub-lethal effects, there is no strong evidence linking these insecticides to losses of the majority of wild bee species. We relate 18 years of UK national wild bee distribution data for 62 species to amounts of neonicotinoid use in oilseed rape. Using a multi-species dynamic Bayesian occupancy analysis, we find evidence of increased population extinction rates in response to neonicotinoid seed treatment use on oilseed rape. Species foraging on oilseed rape benefit from the cover of this crop, but were on average three times more negatively affected by exposure to neonicotinoids than non-crop foragers. Our results suggest that sub-lethal effects of neonicotinoids could scale up to cause losses of bee biodiversity. Restrictions on neonicotinoid use may reduce population declines.
Effective monitoring of native bee populations requires accurate estimates of population size and relative abundance among habitats. Current bee survey methods, such as netting or pan trapping, may be adequate for a variety of study objectives but are limited by a failure to account for imperfect detection. Biases due to imperfect detection could result in inaccurate abundance estimates or erroneous insights about the response of bees to different environments. To gauge the potential biases of currently employed survey methods, we compared abundance estimates of bumblebees (Bombus spp.) derived from hierarchical distance sampling models (HDS) to bumblebee counts collected from fixed‐area net surveys (“net counts”) and fixed‐width transect counts (“transect counts”) at 47 early‐successional forest patches in Pennsylvania. Our HDS models indicated that detection probabilities of Bombus spp. were imperfect and varied with survey‐ and site‐covariates. Despite being conspicuous, Bombus spp. were not reliably detected beyond 5 m. Habitat associations of Bombus spp. density were similar across methods, but the strength of association with shrub cover differed between HDS and net counts. Additionally, net counts suggested sites with more grass hosted higher Bombus spp. densities whereas HDS suggested that grass cover was associated with higher detection probability but not Bombus spp. density. Density estimates generated from net counts and transect counts were 80%–89% lower than estimates generated from distance sampling. Our findings suggest that distance modelling provides a reliable method to assess Bombus spp. density and habitat associations, while accounting for imperfect detection caused by distance from observer, vegetation structure, and survey covariates. However, detection/non‐detection data collected via point‐counts, line‐transects and distance sampling for Bombus spp. are unlikely to yield species‐specific density estimates unless individuals can be identified by sight, without capture. Our results will be useful for informing the design of monitoring programs for Bombus spp. and other pollinators.
Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence, Second Edition, provides a synthesis of model-based approaches for analyzing presence-absence data, allowing for imperfect detection. Beginning from the relatively simple case of estimating the proportion of area or sampling units occupied at the time of surveying, the authors describe a wide variety of extensions that have been developed since the early 2000s. This provides an improved insight about species and community ecology, including, detection heterogeneity; correlated detections; spatial autocorrelation; multiple states or classes of occupancy; changes in occupancy over time; species co-occurrence; community-level modeling, and more. Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence, Second Edition has been greatly expanded and detail is provided regarding the estimation methods and examples of their application are given. Important study design recommendations are also covered to give a well rounded view of modeling.
1.Occupancy models are widely used to describe the distribution of rare and cryptic species— those that occur on only a small portion of the landscape and cannot be detected reliably during a single survey. However, model estimates of occupancy (ψ) and detection probabilities (p) are often least accurate under these circumstances.
2.Available sampling designs for occupancy surveys include standard design, wherein each of S sites is visited K times, and removal design, wherein S sites are visited K times each or until the species of interest is detected. We propose a new conditional design, wherein each of S sites is visited one time, and sites where the species of interest is encountered during the first survey are visited an additional (K−1) times to better estimate detection probability. We used large sample properties of maximum-likelihood estimators and Markov chain Monte Carlo simulations to characterize our proposed conditional design and compare it to standard and removal designs across a wide range of true occupancy and detection probabilities (ψ, p = 0.1 to 0.9 by 0.1 increments), maximum visits (K) and total sampling effort (E, the number of surveys accrued across all sites).
3.The conditional design provided more accurate estimates (lower standard or root mean squared error) of occupancy than standard or removal designs in our calculations and simulations when species were rare (ψ≤0.3) as well as more accurate estimates of detection probability over most combinations of ψ and p. These low-occupancy improvements are achieved by expending a greater proportion of effort at occupied sites, improving estimates of p and thus ψ. When species are common (ψ≥0.5) the removal design generally provided the most accurate occupancy estimates, whereas the standard design performed best when ψ was intermediate and during MCMC simulations when p and K were low.
4.We recommend the conditional design for surveys of rare species and pilot studies. For multi-species surveys that include mixtures of rare and common species, a hybrid standard-conditional design with 2-3 replicates at all sites and additional replicates at sites where rare species are detected improves occupancy estimates of rare species.
Bumble bees (Bombus spp.) are declining across many regions in the Northern Hemisphere, leading to a need for management actions that will protect and enhance their habitats. In the Sierra Nevada of California, USA, montane chaparral is prevalent across the landscape, particularly after forest fires, and may provide important floral resources for pollinators. However post-fire montane chaparral is often targeted for removal during reforestation efforts, to reduce competition with young trees. In 2015 and 2016, we conducted non-lethal bumble bee surveys within 2 areas in the Sierra Nevada that burned in forest fires in 2004. Our goals were to describe bumble bee abundance and species richness in a post-fire landscape, to compare results from chaparral-dominated upland vegetation with results from interspersed patches of riparian vegetation, and to identify characteristics of individual chaparral stands that might make some stands more valuable to bumble bees than others. We captured 2,494 bumble bees of 12 species, and used Bayesian hierarchical modeling to determine that bumble bee abundance was substantially greater in riparian plots (modeled capture rate = 1.10 ± 0.31 [SD] bees/survey in 2015, and 2.96 ± 0.83 bees/survey in 2016) than in upland plots ( = 0.47 ± 0.07 bees/survey in 2015, and 1.27 ± 0.18 bees/survey in 2016), which comprised a mix of chaparral shrubs and associated herbaceous plants. Modeled species richness was also greater in riparian plots, with an average mean richness of 4.1 ± 1.8 bumble bee species in riparian plots versus 2.3 ± 1.3 species in upland plots across the 2 years of the study. Within upland and riparian areas, plots dominated by herbaceous vegetation had greater abundance and species richness. One chaparral shrub species, bearclover (Chamaebatia foliolosa), was foraged on preferentially over all other shrub species and over all but 1 forb taxon, and was associated with increased occupancy probability in the Vosnesensky bumble bee (Bombus vosnesenskii), the most abundant bumble bee species on our study plots. A complex of closely related herbaceous species in the genus Phacelia, commonly associated with upland chaparral in our study area, was the plant taxon most frequently used by bumble bees, and appeared to be particularly important during mid-summer after bearclover flowers became scarce. Our findings suggest that post-fire chaparral communities are generally less intensively used by bumble bees than nearby riparian vegetation but may nevertheless provide important habitat. When chaparral removal is part of post-fire forest regeneration strategies, bumble bees will likely benefit from retention of a mosaic of upland habitat patches dominated by herbaceous vegetation and, in our study area, bearclover, which may provide foraging resources throughout the life cycle of local bumble bee colonies. Because habitat characteristics affected the occupancy of individual bumble bee species differently, managers should consider foraging preferences of target bumble bee species when making land management decisions.
Bumblebees (Bombus spp.) are important pollinators of both crops and wild flowers. Their contribution to this essential ecosystem service has been threatened over recent decades by changes in land use, which have led to declines in their populations. In order to design effective conservation measures it is important to understand the effects of variation in landscape composition and structure on the foraging activities of worker bumblebees. This is because the viability of individual colonies is likely to be affected by the trade-off between the energetic costs of foraging over greater distances and the potential gains from access to additional resources.
We used field surveys, molecular genetics and fine resolution remote sensing to estimate the locations of wild bumblebee nests and to infer foraging distances across a 20 km2 agricultural landscape in southern England. We investigated five species, including the rare B. ruderatus and ecologically similar but widespread B. hortorum. We compared worker foraging distances between species and examined how variation in landscape composition and structure affected foraging distances at the colony level.
Mean worker foraging distances differed significantly between species. Bombus terrestris, B. lapidarius and B. ruderatus exhibited significantly greater mean foraging distances (551 m, 536 m, 501 m, respectively) than B. hortorum and B. pascuorum (336 m, 272 m, respectively). There was wide variation in worker foraging distances between colonies of the same species, which was in turn strongly influenced by the amount and spatial configuration of available foraging habitats. Shorter foraging distances were found for colonies where the local landscape had high coverage and low fragmentation of semi-natural vegetation, including managed agri-environmental field margins. The strength of relationships between different landscape variables and foraging distance varied between species, for example the strongest relationship for B. ruderatus being with floral cover of preferred forage plants.
Our findings suggest that management of landscape composition and configuration has the potential to reduce foraging distances across a range of bumblebee species. There is thus potential for improvements in the design and implementation of landscape management options, such as agri-environment schemes, aimed at providing foraging habitat for bumblebees and enhancing crop pollination services.