Recent and future declines of a historically widespread pollinator linked to climate, land cover, and pesticides.

The acute decline in global biodiversity includes not only the loss of rare species, but also the rapid collapse of common species across many different taxa. The loss of pollinating insects is of particular concern because of the ecological and economic values these species provide. The western bumble bee (Bombus occidentalis) was once common in western North America, but this species has become increasingly rare through much of its range. To understand potential mechanisms driving these declines, we used Bayesian occupancy models to investigate the effects of climate and land cover from 1998 to 2020, pesticide use from 2008 to 2014, and projected expected occupancy under three future scenarios. Using 14,457 surveys across 2.8 million km2 in the western United States, we found strong negative relationships between increasing temperature and drought on occupancy and identified neonicotinoids as the pesticides of greatest negative influence across our study region. The mean predicted occupancy declined by 57% from 1998 to 2020, ranging from 15 to 83% declines across 16 ecoregions. Even under the most optimistic scenario, we found continued declines in nearly half of the ecoregions by the 2050s and mean declines of 93% under the most severe scenario across all ecoregions. This assessment underscores the tenuous future of B. occidentalis and demonstrates the scale of stressors likely contributing to rapid loss of related pollinator species throughout the globe. Scaled-up, international species-monitoring schemes and improved integration of data from formal surveys and community science will substantively improve the understanding of stressors and bumble bee population trends.

Neonicotinoid Pesticides Cause Mass Fatalities of Native Bumble Bees: A Case Study From Wilsonville, Oregon, United States.

In June of 2013 an application of dinotefuran on an ornamental planting of European linden trees (Tilia cordata Mill. [Malvales: Malvalceae]) in a shopping mall parking lot in Wilsonville, Oregon provoked the largest documented pesticide kill of bumble bees in North America. Based on geographic information systems and population genetic analysis, we estimate that between 45,830 and 107,470 bumble bees originating from between 289 and 596 colonies were killed during this event. Dinotefuran is a neonicotinoid that is highly effective in exterminating and/or harming target pest insects and non-target beneficial insects. Analysis to detect the concentration of pesticides in flowers that received foliar application revealed that the minimum reported dinotefuran concentration of a sampled T. cordata flower was 7.4 ppm, or in excess of 737% above the LC50 of the beneficial pollinator, the honey bee (Apis mellifera Linnaeus, 1758 [Hymenoptera: Apidae]). Furthermore, sampled Vosnesensky bumble bees (Bombus vosnesenskii Radoskowski, 1862 [Hymenoptera: Apidae]) were found to have an average dinotefuran concentration of 0.92 ppm at the time of death, which exceeds the maximum LC50 of A. mellifera (0.884 ppm). Our study underscores the lethal impact of the neonicotinoid pesticide dinotefuran on pollinating insect populations in a suburban environment. To our knowledge, the documentation and impact of pesticide kills on wild populations of beneficial insects has not been widely reported in the scientific literature. It is likely that the vast majority of mass pesticide kills of beneficial insects across other environments go unnoticed and unreported.

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.