Integrated modelling of insect population dynamics at two temporal scales

Rising minimum temperatures contribute to 50 years of occupancy decline among cold-adapted Arctic and boreal butterflies in North America

Global climate change has been identified as a potential driver of observed insect declines, yet in many regions, there are critical data gaps that make it difficult to assess how communities are responding to climate change. Poleward regions are of particular interest because warming is most rapid while biodiversity data are most sparse. Building on recent advances in occupancy modeling of presence-only data, we reconstructed 50 years (1970-2019) of butterfly occupancy trends in response to rising minimum temperatures in one of the most under-sampled regions of North America. Among 90 modeled species, we found that cold-adapted species are far more often in decline compared with their warm-adapted, more southernly distributed counterparts. Furthermore, in a post hoc analysis using species' traits, we find that species' range-wide average annual temperature is the only consistent predictor of occupancy changes. Species with warmer ranges were most likely to be increasing in occupancy. This trend results in the majority of butterflies increasing in occupancy probability over the last 50 years. Our results provide the first look at macroscale butterfly biodiversity shifts in high-latitude North America. These results highlight the potential of leveraging the wealth of presence-only data, the most abundant source of biodiversity data, for inferring changes in species distributions.

rGAI: An R package for fitting the generalized abundance index to seasonal count data

The generalized abundance index (GAI) provides a useful tool for estimating relative population sizes and trends of seasonal invertebrates from species' count data and offers potential for inferring which external factors may influence phenology and demography through parametric descriptions of seasonal variation. We provide an R package that extends previous software with the ability to include covariates when fitting parametric GAI models, where seasonal variation is described by either a mixture of Normal distributions or a stopover model which provides estimates of life span. The package also generalizes the models to allow any number of broods/generations in the target population within a defined season. The option to perform bootstrapping, either parametrically or nonparametrically, is also provided. The new package allows models to be far more flexible when describing seasonal variation, which may be dependent on site‐specific environmental factors or consist of many broods/generations which may overlap, as demonstrated by two case studies. Our open‐source software, available at https://github.com/calliste‐fagard‐jenkin/rGAI, makes these extensions widely and freely available, allowing the complexity of GAI models used by ecologists and applied statisticians to increase accordingly.