Deep learning for identifying bee species from images of wings and pinned specimens.

One of the most challenging aspects of bee ecology and conservation is species-level identification, which is costly, time consuming, and requires taxonomic expertise. Recent advances in the application of deep learning and computer vision have shown promise for identifying large bumble bee (Bombus) species. However, most bees, such as sweat bees in the genus Lasioglossum, are much smaller and can be difficult, even for trained taxonomists, to identify. For this reason, the great majority of bees are poorly represented in the crowdsourced image datasets often used to train computer vision models. But even larger bees, such as bumble bees from the B. vagans complex, can be difficult to separate morphologically. Using images of specimens from our research collections, we assessed how deep learning classification models perform on these more challenging taxa, qualitatively comparing models trained on images of whole pinned specimens or on images of bee forewings. The pinned specimen and wing image datasets represent 20 and 18 species from 6 and 4 genera, respectively, and were used to train the EfficientNetV2L convolutional neural network. Mean test precision was 94.9% and 98.1% for pinned and wing images respectively. Results show that computer vision holds great promise for classifying smaller, more difficult to identify bees that are poorly represented in crowdsourced datasets. Images from research and museum collections will be valuable for expanding classification models to include additional species, which will be essential for large scale conservation monitoring efforts.

Landscape-scale floral resource discontinuity decreases bumble bee occurrence and alters community composition.

Agricultural practices and intensification during the past two centuries have dramatically altered the abundance and temporal continuity of floral resources that support pollinating insects such as bumble bees. Long-term trends among bumble bees within agricultural regions suggest that intensive agricultural conditions have created inhospitable conditions for some species, while other species have maintained their relative abundances despite landscape-level changes in flower availability. Bumble bee responses to spatiotemporal resource heterogeneity have been explored at the colony and behavioral level, but we have yet to understand whether these conditions drive community structure and ultimately explain the diverging patterns in long-term species trends. To explore the relationship between landscape-level floral resource continuity and the likelihood of bumble bee occurrence, we mapped the relative spatial and temporal availability of floral resources within an intensive agricultural region in the US Upper Midwest and related this resource availability with bumble bee species relative abundance. Across the bee community, we found that relative bumble bee occurrence increases in landscapes containing more abundant and more temporally continuous floral resources. Declining species, such as Bombus terricola, exhibited the strongest, positive responses to resource abundance and continuity whereas common, stable species, such as Bombus impatiens, showed no statistical relationship to either. Together with existing experimental evidence, this work suggests that efforts to increase spatiotemporal flower availability, along with overall flower abundance at landscape scales may have positive effects on bumble bee communities in the US Upper Midwest.

Disease-related population declines in bats demonstrate non-exchangeability in generalist predators.

The extent to which persisting species may fill the functional role of extirpated or declining species has profound implications for the structure of biological communities and ecosystem functioning. In North America, arthropodivorous bats are threatened on a continent-wide scale by the spread of white-nose syndrome (WNS), a disease caused by the fungus Pseudogymnoascus destructans. We tested whether bat species that display lower mortality from this disease can partially fill the functional role of other bat species experiencing population declines. Specifically, we performed high-throughput amplicon sequencing of guano from two generalist predators: the little brown bat (Myotis lucifugus) and big brown bat (Eptesicus fuscus). We then compared changes in prey consumption before versus after population declines related to WNS. Dietary niches contracted for both species after large and abrupt declines in little brown bats and smaller declines in big brown bats, but interspecific dietary overlap did not change. Furthermore, the incidence and taxonomic richness of agricultural pest taxa detected in diet samples decreased following bat population declines. Our results suggest that persisting generalist predators do not necessarily expand their dietary niches following population declines in other predators, providing further evidence that the functional roles of different generalist predators are ecologically distinct.

Numbers matter: how irruptive bark beetles initiate transition to self-sustaining behavior during landscape-altering outbreaks.

Irruptive forest insects such as bark beetles undergo intermittent outbreaks that cause landscape-scale tree mortality. Despite their enormous economic and ecological impacts, we still have only limited understanding of the dynamics by which populations transition from normally stable endemic to irruptive densities. We investigated density-dependent changes in mountain pine beetle reliance on stressed hosts, host selection, spatial configuration of attacks, and the interaction of host selection and spatial configuration by performing a complete census of lodgepole pine across six stands and 6 years. In addition, we compared the dynamics of mountain pine beetle with those of other bark beetles. We found that as population size increased, reliance on stressed trees decreased and new attacks shifted to larger trees with thicker phloem and higher growth rates that can support higher offspring production. Moreover, the spatial configuration of beetle-attacked trees shifted from random to spatially aggregated. Further, we found evidence that beetle utilization of larger trees was related to aggregation behavior as the size of tree attacked was positively correlated at 10-25 m, within the effective distance of pheromone-mediated signaling. In contrast, non-irruptive bark beetle species did not exhibit such density-dependent spatial aggregation at the stand scale or switches in host selection behavior. These results identify how density-dependent linkages between spatial configuration and host utilization can converge to drive population transitions from endemic to irruptive phases. Specifically, a combination of stand-level spatial aggregation, behavioral shifts, and higher quality of attainable hosts defines a critical threshold beyond which continual population growth becomes self-driving.

Climate-induced outbreaks in high-elevation pines are driven primarily by immigration of bark beetles from historical hosts.

Warming temperatures are allowing native insect herbivores to expand into regions that previously exceeded their thermal tolerance, encounter new host species, and pose significant threats to native communities. However, the dynamics of these expansions remain poorly understood, particularly in the extent to which outbreaks remain reliant on emigration from historical hosts or are driven by local reproduction within novel hosts in the expanded range. We tested these non-mutually exclusive hypotheses using spatially explicit data on mountain pine beetle (Dendroctonus ponderosae), which historically undergoes intermittent outbreaks in low-elevation lodgepole pine (Pinus contorta), but is now causing severe mortality in a high-elevation endangered species, whitebark pine (Pinus albicaulis). We compiled data from 2000 to 2019 across British Columbia, Canada, at 1-km2 resolution, and analyzed spatiotemporal patterns of beetle infestations, lodgepole pine distributions, expansion into habitats dominated by whitebark pine, and the likelihood of future outbreaks in all pine communities under simulated conditions. Overall, we found strong support for the hypothesis of emigration from the historical host species continuing to be a major driver of outbreaks in the more recently accessed host. First, beetle population pressure was consistently the best predictor of infestation severity in both lodgepole and whitebark pine, and appeared to be mostly unidirectional from lodgepole to whitebark pine. Second, infestations in lodgepole pine were of a longer duration than those in whitebark pine, which appeared too brief to sustain transitions from endemic to eruptive dynamics. Furthermore, resource depletion appears to drive emigration from lodgepole pine, whereas in whitebark pine drought appears to favor establishment of immigrants although bioclimatic factors and stand structure preclude self-sustaining outbreaks. Finally, we project that most pine in British Columbia will be at risk in the event of a new major outbreak. We describe implications for conserving and protecting whitebark pine and to other climate-driven range expansions.

Historical decrease in agricultural landscape diversity is associated with shifts in bumble bee species occurrence.

Agricultural intensification is a key suspect among putative drivers of recent insect declines, but an explicit link between historical change in agricultural land cover and insect occurrence is lacking. Determining whether agriculture impacts beneficial insects (e.g. pollinators), is crucial to enhancing agricultural sustainability. Here, we combine large spatiotemporal sets of historical bumble bee and agricultural records to show that increasing cropland extent and decreasing crop richness were associated with declines in over 50% of bumble bee species in the agriculturally intensive Midwest, USA. Critically, we found that high crop diversity was associated with a higher occurrence of many species pre-1950 even in agriculturally dominated areas, but that current agricultural landscapes are devoid of high crop diversity. Our findings suggest that insect conservation and agricultural production may be compatible, with increasing on-farm and landscape-level crop diversity predicted to have positive effects on bumble bees.

Assessing the potential for deep learning and computer vision to identify bumble bee species from images.

Pollinators are undergoing a global decline. Although vital to pollinator conservation and ecological research, species-level identification is expensive, time consuming, and requires specialized taxonomic training. However, deep learning and computer vision are providing ways to open this methodological bottleneck through automated identification from images. Focusing on bumble bees, we compare four convolutional neural network classification models to evaluate prediction speed, accuracy, and the potential of this technology for automated bee identification. We gathered over 89,000 images of bumble bees, representing 36 species in North America, to train the ResNet, Wide ResNet, InceptionV3, and MnasNet models. Among these models, InceptionV3 presented a good balance of accuracy (91.6%) and average speed (3.34 ms). Species-level error rates were generally smaller for species represented by more training images. However, error rates also depended on the level of morphological variability among individuals within a species and similarity to other species. Continued development of this technology for automatic species identification and monitoring has the potential to be transformative for the fields of ecology and conservation. To this end, we present BeeMachine, a web application that allows anyone to use our classification model to identify bumble bees in their own images.

Predator preferences shape the diets of arthropodivorous bats more than quantitative local prey abundance.

Although most predators are generalists, the majority of studies on the association between prey availability and prey consumption have focused on specialist predators. To investigate the role of highly generalist predators in a complex food web, we measured the relationships between prey consumption and prey availability in two common arthropodivorous bats. Specifically, we used high-throughput amplicon sequencing coupled with a known mock community to characterize seasonal changes in little brown and big brown bat diets. We then linked spatiotemporal variation in prey consumption with quantitative prey availability estimated from intensive prey community sampling. We found that although quantitative prey availability fluctuated substantially over space and time, the most commonly consumed prey items were consistently detected in bat diets independently of their respective abundance. Positive relationships between prey abundance and probability of consumption were found only among prey groups that were less frequently detected in bat diets. While the probability of prey consumption was largely unrelated to abundance, the community structure of prey detected in bat diets was influenced by the local or regional abundance of prey. Observed patterns suggest that while little brown and big brown bats maintain preferences for particular prey independently of quantitative prey availability, total dietary composition may reflect some degree of opportunistic foraging. Overall, our findings suggest that generalist predators can display strong prey preferences that persist despite quantitative changes in prey availability.

A global synthesis reveals biodiversity-mediated benefits for crop production.

Human land use threatens global biodiversity and compromises multiple ecosystem functions critical to food production. Whether crop yield-related ecosystem services can be maintained by a few dominant species or rely on high richness remains unclear. Using a global database from 89 studies (with 1475 locations), we partition the relative importance of species richness, abundance, and dominance for pollination; biological pest control; and final yields in the context of ongoing land-use change. Pollinator and enemy richness directly supported ecosystem services in addition to and independent of abundance and dominance. Up to 50% of the negative effects of landscape simplification on ecosystem services was due to richness losses of service-providing organisms, with negative consequences for crop yields. Maintaining the biodiversity of ecosystem service providers is therefore vital to sustain the flow of key agroecosystem benefits to society.

Grassland harvesting alters ant community trophic structure: An isotopic study in tallgrass prairies.

Disturbances have long been recognized as important forces for structuring natural communities but their effects on trophic structure are not well understood, particularly in terrestrial systems. This is in part because quantifying trophic linkages is a challenge, especially for small organisms with cryptic feeding behaviors such as insects, and often relies on conducting labor-intensive feeding trials or extensive observations in the field. In this study, we used stable isotopes of carbon and nitrogen to examine how disturbance (annual biomass harvesting) in tallgrass prairies affected the trophic position, trophic range, and niche space of ants, a widespread grassland consumer. We hypothesized that biomass harvest would remove important food and nesting resources of insects thus affecting ant feeding relationships and trophic structure. We found shifts in the feeding relationships inferred by isotopic signatures with harvest. In particular, these shifts suggest that ants within harvest sites utilized resources at lower trophic levels (possibly plant-based resources or herbivores), expanded trophic breadth, and occupied different niche spaces. Shifts in resource use following harvest could be due to harvest-mediated changes in both the plant and arthropod communities that might affect the strength of competition or alter plant nitrogen availability. Because shifts in resource use alter the flow of nutrients across the food web, disturbance effects on ants could have ecosystem-level consequences through nutrient cycling.

Harvesting effects on wild bee communities in bioenergy grasslands depend on nesting guild.

Conversion of annual crops to native perennial grasslands for bioenergy production may help conserve wild bees by enhancing nest and food resources. However, bee response to the disturbance of biomass harvesting may depend on their nesting location, thus their vulnerability to nest destruction, and the response of the forb community on which they forage. Moreover, because bees have long foraging ranges, effects of local harvesting may depend on the amount of natural habitat in the surrounding landscape. We performed a large-scale one- and two-year experiment in Michigan and Wisconsin, USA, respectively, to examine how grassland harvesting, landscape context, and study year affect the forb community, above- and belowground-nesting bee species richness, community composition, trap nest emergence, and visitation rate. In Wisconsin, harvesting increased forb richness, cover, and evenness compared to unharvested control sites. Harvesting negatively affected aboveground-nesting bee richness and emergence from trap nests, possibly because of nest destruction during the previous harvest. By contrast, harvesting positively affected belowground-nesting bee richness, possibly because of the greater food resource availability and reduced thatch allowing greater access to nesting sites in the soil. Harvesting also affected bee community composition, reflecting the increase in belowground-nesting species at harvested sites. Despite harvesting effects on forb and bee communities, there was no effect on flower visitation rate, indicating little effect on pollination function. We did not find a harvest by landscape context interaction, which, in combination with the negative harvesting effect on trap nest emergence, suggests that harvesting can affect local population growth rather than simply affecting forager aggregation in different resource environments. For bees, there was no harvest by study year interaction, indicating a consistent response over a short timescale. Similarly, in Michigan, belowground-nesting species also responded positively to harvesting, which was more pronounced in sandier soils that are preferred for nesting. However, other components of the Michigan bee and forb communities were not significantly affected by biomass harvesting. Overall, our study demonstrates that harvesting grasslands can positively affect the ~80% of bee species that nest belowground by enhancing nest and/or forage resources, but that conserving aboveground nesters may require leaving some area unharvested.

An improved method for utilizing high-throughput amplicon sequencing to determine the diets of insectivorous animals.

DNA analysis of predator faeces using high-throughput amplicon sequencing (HTS) enhances our understanding of predator-prey interactions. However, conclusions drawn from this technique are constrained by biases that occur in multiple steps of the HTS workflow. To better characterize insectivorous animal diets, we used DNA from a diverse set of arthropods to assess PCR biases of commonly used and novel primer pairs for the mitochondrial gene, cytochrome oxidase C subunit 1 (COI). We compared diversity recovered from HTS of bat guano samples using a commonly used primer pair "ZBJ" to results using the novel primer pair "ANML." To parameterize our bioinformatics pipeline, we created an arthropod mock community consisting of single-copy (cloned) COI sequences. To examine biases associated with both PCR and HTS, mock community members were combined in equimolar amounts both pre- and post-PCR. We validated our system using guano from bats fed known diets and using composite samples of morphologically identified insects collected in pitfall traps. In PCR tests, the ANML primer pair amplified 58 of 59 arthropod taxa (98%), whereas ZBJ amplified 24-40 of 59 taxa (41%-68%). Furthermore, in an HTS comparison of field-collected samples, the ANML primers detected nearly fourfold more arthropod taxa than the ZBJ primers. The additional arthropods detected include medically and economically relevant insect groups such as mosquitoes. Results revealed biases at both the PCR and sequencing levels, demonstrating the pitfalls associated with using HTS read numbers as proxies for abundance. The use of an arthropod mock community allowed for improved bioinformatics pipeline parameterization.

Collaboration Matters: Honey Bee Health as a Transdisciplinary Model for Understanding Real-World Complexity.

We develop a transdisciplinary deliberative model that moves beyond traditional scientific collaborations to include nonscientists in designing complexity-oriented research. We use the case of declining honey bee health as an exemplar of complex real-world problems requiring cross-disciplinary intervention. Honey bees are important pollinators of the fruits and vegetables we eat. In recent years, these insects have been dying at alarming rates. To prompt the reorientation of research toward the complex reality in which bees face multiple challenges, we came together as a group, including beekeepers, farmers, and scientists. Over a 2-year period, we deliberated about how to study the problem of honey bee deaths and conducted field experiments with bee colonies. We show trust and authority to be crucial factors shaping such collaborative research, and we offer a model for structuring collaboration that brings scientists and nonscientists together with the key objects and places of their shared concerns across time.

Crop pests and predators exhibit inconsistent responses to surrounding landscape composition.

The idea that noncrop habitat enhances pest control and represents a win-win opportunity to conserve biodiversity and bolster yields has emerged as an agroecological paradigm. However, while noncrop habitat in landscapes surrounding farms sometimes benefits pest predators, natural enemy responses remain heterogeneous across studies and effects on pests are inconclusive. The observed heterogeneity in species responses to noncrop habitat may be biological in origin or could result from variation in how habitat and biocontrol are measured. Here, we use a pest-control database encompassing 132 studies and 6,759 sites worldwide to model natural enemy and pest abundances, predation rates, and crop damage as a function of landscape composition. Our results showed that although landscape composition explained significant variation within studies, pest and enemy abundances, predation rates, crop damage, and yields each exhibited different responses across studies, sometimes increasing and sometimes decreasing in landscapes with more noncrop habitat but overall showing no consistent trend. Thus, models that used landscape-composition variables to predict pest-control dynamics demonstrated little potential to explain variation across studies, though prediction did improve when comparing studies with similar crop and landscape features. Overall, our work shows that surrounding noncrop habitat does not consistently improve pest management, meaning habitat conservation may bolster production in some systems and depress yields in others. Future efforts to develop tools that inform farmers when habitat conservation truly represents a win-win would benefit from increased understanding of how landscape effects are modulated by local farm management and the biology of pests and their enemies.

Do managed bees have negative effects on wild bees?: A systematic review of the literature.

Managed bees are critical for crop pollination worldwide. As the demand for pollinator-dependent crops increases, so does the use of managed bees. Concern has arisen that managed bees may have unintended negative impacts on native wild bees, which are important pollinators in both agricultural and natural ecosystems. The goal of this study was to synthesize the literature documenting the effects of managed honey bees and bumble bees on wild bees in three areas: (1) competition for floral and nesting resources, (2) indirect effects via changes in plant communities, including the spread of exotic plants and decline of native plants, and (3) transmission of pathogens. The majority of reviewed studies reported negative effects of managed bees, but trends differed across topical areas. Of studies examining competition, results were highly variable with 53% reporting negative effects on wild bees, while 28% reported no effects and 19% reported mixed effects (varying with the bee species or variables examined). Equal numbers of studies examining plant communities reported positive (36%) and negative (36%) effects, with the remainder reporting no or mixed effects. Finally, the majority of studies on pathogen transmission (70%) reported potential negative effects of managed bees on wild bees. However, most studies across all topical areas documented the potential for impact (e.g. reporting the occurrence of competition or pathogens), but did not measure direct effects on wild bee fitness, abundance, or diversity. Furthermore, we found that results varied depending on whether managed bees were in their native or non-native range; managed bees within their native range had lesser competitive effects, but potentially greater effects on wild bees via pathogen transmission. We conclude that while this field has expanded considerably in recent decades, additional research measuring direct, long-term, and population-level effects of managed bees is needed to understand their potential impact on wild bees.

Gut microbiomes of mobile predators vary with landscape context and species identity.

Landscape context affects predator-prey interactions and predator diet composition, yet little is known about landscape effects on insect gut microbiomes, a determinant of physiology and condition. Here, we combine laboratory and field experiments to examine the effects of landscape context on the gut bacterial community and body condition of predatory insects. Under laboratory conditions, we found that prey diversity increased bacterial richness in insect guts. In the field, we studied the performance and gut microbiota of six predatory insect species along a landscape complexity gradient in two local habitat types (soybean fields vs. prairie). Insects from soy fields had richer gut bacteria and lower fat content than those from prairies, suggesting better feeding conditions in prairies. Species origin mediated landscape context effects, suggesting differences in foraging of exotic and native predators on a landscape scale. Overall, our study highlights complex interactions among gut microbiota, predator identity, and landscape context.

Understanding Barriers to Participation in Cost-Share Programs For Pollinator Conservation by Wisconsin (USA) Cranberry Growers.

The expansion of modern agriculture has led to the loss and fragmentation of natural habitat, resulting in a global decline in biodiversity, including bees. In many countries, farmers can participate in cost-share programs to create natural habitat on their farms for the conservation of beneficial insects, such as bees. Despite their dependence on bee pollinators and the demonstrated commitment to environmental stewardship, participation in such programs by Wisconsin cranberry growers has been low. The objective of this study was to understand the barriers that prevent participation by Wisconsin cranberry growers in cost-share programs for on-farm conservation of native bees. We conducted a survey of cranberry growers (n = 250) regarding farming practices, pollinators, and conservation. Although only 10% of growers were aware of federal pollinator cost-share programs, one third of them were managing habitat for pollinators without federal aid. Once informed of the programs, 50% of growers expressed interest in participating. Fifty-seven percent of growers manage habitat for other wildlife, although none receive cost-share funding to do so. Participation in cost-share programs could benefit from outreach activities that promote the programs, a reduction of bureaucratic hurdles to participate, and technical support for growers on how to manage habitat for wild bees.

Increased duration of aquatic resource pulse alters community and ecosystem responses in a subarctic plant community.

Allochthonous resource movement across ecosystem boundaries creates episodic linkages between ecosystems. The sensitivity of the community to external resources of varying duration can alter the baseline upon which future pulses of allochthony can act. We explored the terrestrial ecosystem response to pulsed inputs of lake-derived resources with a manipulative experiment in a subarctic heathland where we assessed plant community and nutrient availability responses to additions of midge carcasses (Diptera: Chironomidae). Insect carcasses were added as either a one-time pulse or a 4-yr press to simulate differing durations of allochthony, which is common in the area. We found that midge pulses significantly elevated soil inorganic [N] in the first year (7× over background levels) but were significantly diminished (1.5×) by the second year after the initial pulse. The press treatment continued to elevate total soil inorganic [N] to 13× over background levels by the fourth year of midge additions, but then declined to 3.6× background in year 5 when experimental midge additions had ceased. In contrast to the soil inorganic N response, plant biomass was similar in pulse-addition and control plots over the course of the experiment. However, by the second year of the study plant biomass in press-addition plots were significantly higher than controls (>50%), and continued to increase over the 4 yr of the press treatment. Midge addition stimulated dominance of graminoids and thatch litter in plots that had previously been primarily heathland vegetation, a response that persisted 4 yr post-midge addition. Our findings suggest that soil and plant community responses to persistent insect carcass deposition (e.g., press) into heathland vegetation has the potential to carry forward in a way that modifies the baseline ecosystem conditions upon which additional allochthony may act.

A global synthesis of the effects of diversified farming systems on arthropod diversity within fields and across agricultural landscapes.

Agricultural intensification is a leading cause of global biodiversity loss, which can reduce the provisioning of ecosystem services in managed ecosystems. Organic farming and plant diversification are farm management schemes that may mitigate potential ecological harm by increasing species richness and boosting related ecosystem services to agroecosystems. What remains unclear is the extent to which farm management schemes affect biodiversity components other than species richness, and whether impacts differ across spatial scales and landscape contexts. Using a global metadataset, we quantified the effects of organic farming and plant diversification on abundance, local diversity (communities within fields), and regional diversity (communities across fields) of arthropod pollinators, predators, herbivores, and detritivores. Both organic farming and higher in-field plant diversity enhanced arthropod abundance, particularly for rare taxa. This resulted in increased richness but decreased evenness. While these responses were stronger at local relative to regional scales, richness and abundance increased at both scales, and richness on farms embedded in complex relative to simple landscapes. Overall, both organic farming and in-field plant diversification exerted the strongest effects on pollinators and predators, suggesting these management schemes can facilitate ecosystem service providers without augmenting herbivore (pest) populations. Our results suggest that organic farming and plant diversification promote diverse arthropod metacommunities that may provide temporal and spatial stability of ecosystem service provisioning. Conserving diverse plant and arthropod communities in farming systems therefore requires sustainable practices that operate both within fields and across landscapes.

Ecoinformatics (Big Data) for Agricultural Entomology: Pitfalls, Progress, and Promise.

Ecoinformatics, as defined in this review, is the use of preexisting data sets to address questions in ecology. We provide the first review of ecoinformatics methods in agricultural entomology. Ecoinformatics methods have been used to address the full range of questions studied by agricultural entomologists, enabled by the special opportunities associated with data sets, nearly all of which have been observational, that are larger and more diverse and that embrace larger spatial and temporal scales than most experimental studies do. We argue that ecoinformatics research methods and traditional, experimental research methods have strengths and weaknesses that are largely complementary. We address the important interpretational challenges associated with observational data sets, highlight common pitfalls, and propose some best practices for researchers using these methods. Ecoinformatics methods hold great promise as a vehicle for capitalizing on the explosion of data emanating from farmers, researchers, and the public, as novel sampling and sensing techniques are developed and digital data sharing becomes more widespread.