The managed Apis mellifera honey bee is one of the most valuable among beneficial insect species. It contributes to the pollination of agricultural crops in nearly all corners of the globe. For more than a decade, elevated colony losses in parts of North America and Europe have cast a spotlight on honey bees, and focused efforts by agricultural producers and the scientific community to tackle such losses. The unique nature of honey bees requires that multiple mitigation measures be compounded.  A honey bee colony, with its free-flying social members, is intimately connected to the surrounding environment – as its foragers search for flower rewards like nectar and pollen far beyond the barriers (e.g., fences) that typically confine livestock and other managed species. However, a honey bee colony also relies on its beekeeper to provide subsistence, especially during periods of dearth, parasitic infections or extreme weather events.

After years of research, there is growing consensus that multiple factors negatively influence honey bees, both in the United States and abroad. For example, recent lab and field experiments have demonstrated the negative consequences of parasitic infections and poor nutrition on honey bee physiology and behavior,  while “Citizen science” initiatives by beekeepers also point to the role of parasites, as well as that of weather conditions. To date, however, very few studies have taken a multi-factorial, broad-scale approach to the identification of risk factors – mainly due to limited availability of appropriate data.

In a recent Scientific Report publication, we employed publicly available data from the USDA long-term honey bee colony status and stressors survey, in conjunction with USDA weather data and PRISM data on land-use, to help identify important drivers of colony loss across the whole territory of the continental United States. Our study, which is the first to provide nation-wide analyses, confirmed important roles for the parasitic mite Varroa destructor, pesticides, and extreme weather conditions.

Because we employed multiple public data sources, we needed an effective way to work with different spatio-temporal resolutions and align the data. For example, weather data from PRISM were collected daily on a 4 by 4 kilometer grid, whereas the USDA colony loss data were collected quarterly and at the state level. To address this, we aggregated higher resolution data using not just their averages, but multiple indexes capturing their distribution and behavior.  This enabled us to reduce the loss of information due to aggregation, and to use more informative predictors when modeling honey bee colony loss. We also employed novel statistical techniques produced by our group to simultaneously remove contaminations from the data (outliers) and select the most important stressors.

Ultimately, our study highlighted the spatio-temporal complexities of honey bee colony loss, and its association to different stressors, across the whole United States. Our results may also serve to focus future efforts by beekeepers, agricultural producers and scientists – as they suggest that varroa mite management, a reduction in pesticide exposure, and preparation for extreme weather events may mitigate honey bee colony loss.