Abstract Deep learning and computer vision hold enormous potential for automated monitoring of biodiversity, including pollinators and other insects. Efficient, scalable monitoring of insect pollinators is crucial given pollinators' role in supporting biodiversity and agricultural productivity amidst declining pollinator populations. However, several practical challenges limit the broad adoption of automated monitoring techniques. Existing approaches often depend on passive image capture (e.g. timelapse) that can generate impractically large datasets (especially for large‐scale monitoring) and rely on application‐specific models that may not generalize well to novel contexts. This creates barriers to broader use in ecological applications where limited training data exist and underscores the need for flexible, reliable automated monitoring systems. Here we introduce AutoPollS (Autonomous Pollinator Sampler), an open‐source system for automated, image‐based monitoring of flower‐visiting insects. AutoPollS uses multi‐camera monitoring using lightweight deep learning models in the field (to reduce data storage and processing) followed by higher performance detection and classification models offline. AutoPollS uses a modular camera system that can be adapted to different plant–pollinator communities. We validate AutoPollS in two agroecosystems (sunflowers and apple orchards) and a multi‐species alpine meadow. In all three systems, AutoPollS provided robust in‐field monitoring using broadly trained models fine‐tuned with only limited application‐specific data, including high detection and species classification accuracy (>95%) in honey bees (Apis spp.) and bumble bees (Bombus spp.). However, models did show reduced performance in real‐world applications compared to training datasets—likely associated with image quality and curation—highlighting the challenges of monitoring in real‐world ecological conditions. Overall, our results demonstrate the utility of our approach in diverse ecological applications, including non‐lethal monitoring of endangered species and characterizing pollinator responses to dynamic environmental conditions. Finally, we highlight the potential of this approach to bridge practical applications and model development by facilitating broader adoption. This could support generation of larger and more diverse datasets to improve model performance.

Organic agriculture is considered a more biodiversity-friendly alternative to conventional agriculture, the latter being a major driver of biodiversity loss. Perennial woody crops, like apple, provide nutritious food for humans and stable habitats for biodiversity. Yet, a synthesis of the effects of organic management on pollinators and pollination has not been done in this crop. Here, we synthesized data from seven studies to assess how organic versus non-organic management in apple cultivation influences bee communities (honeybee abundance, wild bee abundance, and species richness), and pollination (fruit set, fruit weight, and seed set). Organic management increased wild bee richness but had no effect on honeybee or wild bee abundance. Fruit set was unaffected by management or bee metrics. Fruit weight was lower under organic management and decreased with increasing honeybee abundance. Seed set was higher under organic management, and the effects of both wild bee abundance and species richness on seed set were management dependent, as we concluded a positive relationship between wild bees and seed set in non-organic orchards, whereas no effect was detected in organic orchards. We conclude that organic management supports greater bee richness in apple orchards, and effects on pollination vary between the considered pollination metrics. These findings challenge the expectation that organic management has a general positive effect on pollinators and their pollination services. We suggest avoiding excessively high honeybee densities as they may negatively affect pollination. Our study underscores the need for more research on more specific management actions that benefit pollinators, pollination and yield in order to develop management strategies that simultaneously enhance ecological and agronomic outcomes.

Abstract Introduction Human land use intensification and the resulting habitat loss are primary drivers of insect pollinator declines. Habitat restoration offers a promising approach to counteract these declines, yet landscape‐level evaluations of bee responses to restoration and management remain limited. We conducted a 2‐year, landscape‐scale study in Wisconsin, United States, to assess how different intensities of tallgrass prairie restoration and management affect bumble bees ( Bombus spp.). Objectives This study aimed to determine whether (1) bumble bee abundance and diversity increase with active restoration, and (2) outcomes differ between low‐ (seeded only) and moderate‐intensity (seeded and managed with prescribed fire) interventions. Methods Using catch‐and‐release surveys, we measured bumble bee abundance and diversity at 32 sites representing a gradient in restoration intervention: no‐intervention (passive recovery), low‐intervention, and moderate‐intervention. Results Bumble bee abundance and diversity were higher at active restoration sites (low‐ and moderate‐intervention) than at no‐intervention sites. Although both abundance and diversity tended to be greater at moderate than low‐intervention intensities, these differences were not statistically significant. Bumble bee community composition also differed across intervention intensity, driven by shifts in dominant species (e.g. Bombus impatiens and B. griseocollis ). Rarer taxa, including endangered and vulnerable species, occurred only at active restoration sites, with the largest populations at moderate‐intervention sites. Across all sites, bumble bee responses were strongly and positively associated with floral abundance, but not with semi‐natural habitat in the surrounding landscape. Conclusion Our findings demonstrate that active grassland restoration can effectively increase bumble bee abundance and diversity, supporting its value as a conservation practice for pollinators.

The environmental and social problems that industrial agriculture creates and relies upon, but also suffers from and responds to, are massive compared to, for instance, conservation interventions on the margins of fields. We need to transform food production that create opportunities for current and future farmers to evolve and grow into farming that is genuinely regenerative, that is, that build soil and foster diversity of crops, landscapes, and people. Metrics of success for agriculture should indicate soil accretion, clean water, habitat for biodiversity, and an abundance and diversity of thriving and vital farmers and rural communities. Farming that is profitable to the farmer, regenerates their ability to produce for society, and adds value to society’s overall ecosystem, community, and individual health and wellbeing portfolio. Agrifood system transformation is not possible within the productivist paradigm that incentivizes farmers to produce as much as they possibly can from the land and livestock. We must reward farmers for providing diverse portfolios of outcomes that are profitable to the farm, while providing for us all today, and building capacity for future generations to do the same.

Understanding the capacity of mobile organisms such as insects to utilize resources across different patches in a landscape can reveal strategies for their conservation. Past research suggests that higher levels of non-crop habitat or landcover diversity in agricultural regions typically benefit generalist predators who can fortify their diets with prey from multiple adjacent habitats. For some taxa such as lady beetles (Coccinellidae), dietary diversity is associated with improved fitness, but foraging patterns in real landscapes are hard to measure. We used a DNA metabarcoding approach to explore how the presence and taxonomic richness of arthropod prey in lady beetle diets varied by local habitat (crop vs. non-crop) and landscape complexity (non-crop habitat and landcover diversity in a 250 m radius). We collected over 500 individual lady beetles from a range of landscapes in 2019 and 2021 in southern Wisconsin (USA), performed whole-body DNA extractions, amplified arthropod DNA using primers optimized for insectivore diets, and used Illumina sequencing to characterize the taxonomic composition and diversity of prey. We found 50 unique prey taxa in lady beetle guts from eight arthropod orders (mostly flies, true bugs, and thrips). Lady beetles in landscapes with a greater proportion of crops were slightly more likely to have prey in their gut, and community-level prey richness was strongly positively correlated with surrounding landcover diversity. This effect was dampened slightly in high-crop landscapes, likely due to a smaller prey species pool available to predators. Our results enhance knowledge of lady beetle trophic ecology and demonstrate that supplementation of diets through increased habitat diversity may be an important mechanism for the success of mobile generalists in complex landscapes.

A central challenge in predicting biological responses to climate change is bridging the mismatch between coarse changes in climate and the fine-scale environments organisms experience.1,2 Behavior plays a crucial, understudied role in bridging these scales. For small ectotherms like pollinating insects, behavioral responses to microclimate variation can buffer or amplify thermal exposure, generating temperature differences that meet or exceed mean climate warming.3,4,5,6,7 Realized climate impacts on pollinators and pollination depend strongly on behavioral responses to thermal environments, response variation across taxa, and the dynamic microclimatic heterogeneity organisms experience. Yet, our understanding of microclimatic dynamics at insect-relevant scales-or how pollinators exploit them-remains limited. We combined a novel, automated, deep-learning-based pollinator monitoring system8 with physical operative temperature models to quantify pollinator responses to dynamic microclimate gradients at fine spatial (meters) and temporal (minutes) scales. In experimental shade microclimate manipulations, insect pollinators dynamically tracked thermal microclimates. In naturalistic plant communities, microclimate heterogeneity varied over time, driven by short-term shifts in temperature, solar radiation, and diurnal cycles, and increased seasonally, coinciding with vegetative growth. Pollinator taxa showed divergent responses to microclimatic changes, indicating potential thermal buffering and amplification, suggestive of differential thermal tolerance and life history constraints. Data-driven modeling suggests taxon-specific responses to microclimate could drive divergent thermal risk under identical environmental conditions, altering the spatial structure of plant-pollinator communities, and impacting pollination efficacy. Together, we show that climate impacts are shaped by pollinator behavior and microclimate dynamics, suggesting that managing for microclimatic variation may be impactful for pollinators and pollination.

Integrated pest management (IPM) has been criticized for its inability to move agriculture beyond "curative" approaches for crop management. Today, IPM is primarily a set of tactics aimed at controlling pests, still relying on pesticides and pesticide management as a main tool. Yet, from its inception, the intent of IPM was to redesign agricultural systems, through planning and reliance on ecological processes, such that pest problems were largely avoided in the first place. The pathways and processes for achieving widespread cropping system redesign, however, were never fully articulated in the IPM concept. Frameworks from the field of sustainability transitions, such as the multi-level perspective (MLP), can shed light on the dynamics and drivers of change in socio-technological systems. In this framing, the ecological intensification and diversification of farms and landscapes that can create pest-suppressive environments can be conceptualized as novel "technological" innovation niches that challenge the dominant socio-technological regime. The conventional agri-food system, however, is dominated by "lock-ins" that have coevolved over time among farmers, supply chain participants, businesses, research institutions, policy, and public institutions to be resilient to change. Enabling IPM to transform agroecosystems requires overcoming a range of interconnected hurdles. When seen through the lens of the MLP, successful IPM will require not only new technologies and management practices but also social and political innovations that support collective action, social learning and transdisciplinary partnerships. Novel decision support tools for managing and predicting the dynamics of agroecosystems at farm and landscape scales can be instrumental for planning and implementing diversified production systems. Ultimately, institutional supports are needed to foster innovations currently obstructed by political pressure and to reconfigure power dynamics that maintain the status quo. Finally, conceptualizing agriculture and pest management as one component of a broader landscape that affects human well-being reveals points of leverage and synergy to facilitate a much needed transition away from the pesticide-intensive approaches that still dominate today.

Roughly 40% of global agri-food production is lost to pests during an era when productivity gains are essential to humanity. Restoring farmland biodiversity for conservation biological control offers potential to secure win-win outcomes for yield and the environment. However, achieving this is hindered by gaps in our un-derstanding of agrobiodiversity, including a lack of data on the occurrence, identity, and interactions of farm-dwelling (plant, animal, microbial) biota. Limited interdisciplinary collaboration and weak policy frameworks exacerbate these is-sues. Comprehensive data capture using standardized metrics, universal proto-cols, farmer-scientist cooperation, and next-generation tools could consolidate the evidence base on which to reform farming practice. This will involve ecologists stepping outside their comfort zones to promote behavioral change and make ecological intensification a reality.

Agroecosystem modeling tools can provide insights into cover crop performance under varying environmental and management combinations. This study aims to (1) simulate winter cereal rye cover crops in Agro-IBIS, a process-based terrestrial ecosystem model and (2) evaluate Agro-IBIS performance in predicting aboveground biomass (AGB) of winter cereal rye cover crops. To achieve this, the winter wheat plant functional type (PFT) in Agro-IBIS was adapted to represent winter cereal rye as a cool-season winter annual grass cover crop. We adjusted the specific leaf area (SLA), maximum Rubisco activity at 15 °C (Vc,max), growing degree days (GDD) base temperature, GDD upper threshold, and planting and termination dates as indicated by observed data. Model performance was evaluated using observed data from continuous maize and maize-soybean rotation systems in southern Wisconsin. The model effectively represented interannual variability of winter cereal rye cover crop AGB that was measured in southern Wisconsin in continuous maize and maize-soybean rotation systems. This demonstrated the efficacy of Agro-IBIS in representing establishment success, cold-hardening, spring green-up, and AGB accumulation of winter cereal rye cover crops in conventional annual grain cropping systems. Environmental drivers like growing season length, accumulated GDDs, precipitation amount, and solar radiation were key drivers of cover crop AGB production, which is generally represented by Agro-IBIS. This suggests the model would be an accurate tool to use when investigating the impact of climate change or increased weather variability on the success of cover crops across the Midwest and beyond.

Pasture yield is crucial to the economic viability of grass-based livestock enterprises, yet the difficulty in predicting yields under various environmental and management conditions prevents effective planning. We used USDA-SSURGO data to create a random forest model that predicts pasture yield potential based on soil properties for the state of Wisconsin (USA). This model is highly accurate (RMSE = 0.11 tons/acre, or 4% of the average yield), predicting pasture yields in Wisconsin grasslands to range from 1.0 to5.3 tons/acre, with an average yield of 2.6 tons/acre. We then integrated this model with guidelines from a USDA-NRCS grazing planning tool to adjust pasture yield potential for different levels of grazing intensity. The adjustments were multiplied to the random forest model output and ranged from 0.65 for continuously grazed pasture to 1.2 for pastures rotated more than once per day. The model is available to use within an online decision support tool through an R-shiny interface and can be easily replicated for other states in the Midwest US. The tool is easy to use and can support farmer analysis of the costs and benefits of grass-based agriculture.

ABSTRACTLand use change threatens global biodiversity and compromises ecosystem functions, including pollination and food production. Reduced taxonomic α‐diversity is often reported under land use change, yet the impacts could be different at larger spatial scales (i.e., γ‐diversity), either due to reduced β‐diversity amplifying diversity loss or increased β‐diversity dampening diversity loss. Additionally, studies often focus on taxonomic diversity, while other important biodiversity components, including phylogenetic diversity, can exhibit differential responses. Here, we evaluated how agricultural and urban land use alters the taxonomic and phylogenetic α‐, β‐, and γ‐diversity of an important pollinator taxon—bees. Using a multicontinental dataset of 3117 bee assemblages from 157 studies, we found that taxonomic α‐diversity was reduced by 16%–18% in both agricultural and urban habitats relative to natural habitats. Phylogenetic α‐diversity was decreased by 11%–12% in agricultural and urban habitats. Compared with natural habitats, taxonomic and phylogenetic β‐diversity increased by 11% and 6% in urban habitats, respectively, but exhibited no systematic change in agricultural habitats. We detected a 22% decline in taxonomic γ‐diversity and a 17% decline in phylogenetic γ‐diversity in agricultural habitats, but γ‐diversity of urban habitats was not significantly different from natural habitats. These findings highlight the threat of agricultural expansions to large‐scale bee diversity due to systematic γ‐diversity decline. In addition, while both urbanization and agriculture lead to consistent declines in α‐diversity, their impacts on β‐ or γ‐diversity vary, highlighting the need to study the effects of land use change at multiple scales.

Pesticide use and habitat loss are major anthropogenic drivers of bee decline, raising global concerns about impaired crop pollination. However, the relative importance of these stressors and their combined impact on bee assemblages comprising species with different traits, such as body size or nesting strategy, remains unknown. Here we addressed these key knowledge gaps in a global quantitative synthesis analysing bee assemblage data from 681 crop fields across three continents. We found that both local pesticide hazards and decreasing proportions of semi-natural habitats in surrounding landscapes negatively affected wild bee abundance and species richness in crop fields, while pesticides additionally reduced functional and phylogenetic diversity. Semi-natural habitat availability did not buffer against these negative pesticide effects, nor did we identify any specific traits rending bees more vulnerable to one of the two drivers. Our findings highlight the pressing need to reduce non-target effects of pesticide use and emphasize that conservation and restoration of semi-natural habitats successfully promote wild bees, but are insufficient strategies to mitigate pesticide-driven losses of wild bee pollinators from crop fields.

AimMany broad-scale ecological inventory and monitoring efforts collect multi-species (or otherwise multivariate) data under unstructured study designs. Unstructured designs are vulnerable to preferential sampling, where residual covariance between locations selected for sampling and the response variable of interest may render predictions strongly biased.InnovationWe extend previous work to address preferential sampling in spatial single-species distribution models to a multivariate context. Using spatially structured latent variables to approximate residual covariance between species occurrence probabilities and sampling inclusion probabilities, we present ways to account for sampling that may be preferential to varying degrees across multiple species, where (analogously) multiple datastreams might be preferential to varying degrees for a single species, or both. We use simulation to explore our proposed model and present an application that delineates the distributions of 13 bumble bee species across Wisconsin, USA and evaluates evidence for preferential sampling within 3 citizen science datastreams.Main ConclusionsSimulation results suggest that our proposed model improves out-of-sample predictions of species occurrence or richness when the sampling design is preferential and residual covariance between sampling and species occurrence exhibits spatial structure compatible with model assumptions, reducing bias in predictions of species occurrence or richness. Empirically, volunteers appeared to sample preferentially with respect to bumble bee distributions, being more likely to sample in locations where the federally listed Bombus affinis was more likely to occur. Our approach enables practitioners a means to triage preferential sampling within increasingly popular multi-species or integrated distribution models and can be modified slightly to deal with a variety of other response variables.

The expansion and intensification of agriculture in the last century has reduced floral resources for wild insect pollinators, contributing to their decline and potentially lowering pollination services for crop production. Flowering cover crops that can overwinter in harsh climates, such as winter camelina (Camelina sativa [L.] Crantz), can provide key resources for spring-emerging insects and fit into forage cropping systems in the Upper Midwest region of the United States where corn silage production is an important source of dairy forage. However, the amount of floral cover and length of time that cover is available, as well as the practicality of integrating camelina in annual forage cropping systems, depends on fall planting time and cover crop mix. We performed a plot-scale, randomized block experiment to measure spring floral cover and insect flower visitation of winter camelina and uncultivated flowers. This experiment occurred across 2 years in 3 cover crop mixes: 1) camelina monoculture, 2) camelina, triticale, hairy vetch mix, and 3) camelina, cereal rye, hairy vetch mix. In the second year, 3 camelina monocultures were planted at three different fall planting times, which constituted additional treatments: 4) early, 5) mid, and 6) late plantings. We included an unseeded fallow treatment, i.e., no cover crop, for comparison. Cover crop mixes and camelina in monoculture planted simultaneously (all mid-fall plantings) provided equal amounts of floral cover and supported comparable insect visitation rates the following spring. For monocultures, camelina planted earlier in the fall had the highest spring floral cover, and the latest planting time provided virtually no flower cover. Despite the large proportion of total floral cover attributed to dandelion in many plots, flower visitation exclusively increased with increasing camelina floral cover. However, there was an upper asymptote at which adding more camelina did not further increase visitation. Our study demonstrates that winter camelina can provide resources for insects in early spring when planted in a monoculture or mix the previous September as compared to winter fallow, though the period of flowering is short. We also show that although dandelion may provide floral cover early in the growing season, camelina may be more attractive to insects and provides an important spring floral resource.

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.

Loss of natural habitat due to increases in agricultural extent raises the question of whether human-dominated landscape types can support biodiversity, particularly for declining insect pollinators. Compared to more rural agricultural landscapes, urban areas may confer benefits for bumble bee populations by providing stable and diverse floral resources. However, disentangling the effects of local- and landscape-scale characteristics on bumble bee populations in human-modified landscapes is challenging. Here, we assessed bumble bee occupancy using a repeated transect sampling design conducted during the summers of 2019 and 2020 within the metropolitan area of Madison, WI, and the surrounding agricultural landscape. We fit hierarchical occupancy models to estimate the detection (p) and occupancy (psi) probabilities relative to local habitat quality (floral abundance and floral richness) and landscape (agricultural-urban gradient) features for eight bumble bee species. We hypothesised that bumble bees were more likely to occupy urban areas, serving as refugia, relative to agricultural sites. We found that the detection probability of all bumble bee species was seasonal and influenced by maximum floral abundance at survey sites, independent of the surrounding land cover type. After accounting for species-specific detection probabilities, the effect of urbanisation on bee occupancy was weak, and no species were less likely to occupy urban than rural agricultural areas. Our findings suggest that bumble bee occupancy is associated with a 'honeypot effect' where local resource availability, in the form of higher floral abundance, is most important in limiting the occupancy of bumble bees across urban and agricultural landscapes. We found that the detection probability of all bumble bee species was seasonal and influenced by maximum floral abundance at survey sites, independent of the surrounding land cover type. After accounting for species-specific detection probabilities, the effect of urbanisation on bee occupancy was weak, and no species were less likely to occupy urban than rural agricultural areas. Bumble bee occupancy is associated with a 'honeypot effect' where local resource availability is most important in limiting the occupancy of bumble bees across urban and agricultural landscapes. image

In North America, white-nose syndrome (WNS) has caused precipitous declines in hibernating bat populations, raising the question of whether the rapid loss of arthropodivorous bats may affect the abundance of their prey. During the summers of 2015-2018 (1 year after the arrival of WNS in Wisconsin, USA), we performed intensive arthropod black-light trapping, ultrasonic acoustic monitoring, and emergence counts at 10 little brown (Myotis lucifugus) and big brown (Eptesicus fuscus) bat maternity roosts with paired control sites. For little brown bats, which are severely affected by WNS, roost counts declined by 95% over the four-year period, compared to a 38% decline in big brown bat roost counts. Total arthropod abundance decreased by 49%, although decreases among common little brown bat prey were less severe. Our natural predator exclusion experiment supports existing evidence that bats can have measurable trophic impacts on arthropod communities, primarily via top-down effects on common prey.