Assessing the impact of past climate and land use change on bird occupancy dynamics in North America

Evidence is accumulating that the ranges of many species change in response to environmental change. How different global change drivers contribute to colonisation and local extinction across species is, however, poorly understood.

Here, we leveraged dynamic occupancy models to simulate metapopulation dynamics and attribute observed changes in occupancy to climate and land use drivers. We fitted the models to 25 years (1995-2019) of occurrence data for 159 bird species from the North American Breeding Bird Survey using a Bayesian approach. In the dynamic occupancy models, we related initial occupancy, site-level colonisation and extinction to annually resolved climate and land use variables. Simulated occupancy dynamics were evaluated using spatial and temporal block cross-validation. For 80 species, the models showed fair predictive performance, and we subsequently only considered these species for impact attribution. We assessed the relative importance of climate and land use change for the simulated occupancy dynamics based on counterfactual scenarios with detrended climate and constant land use with the base year 1995, the first year of our observation time series, and controlled for differences in prediction error between models.

Our results suggest that climate change has affected occupancy dynamics of North American birds more strongly over the last three decades than land use change. For 75 % of the assessed species, climate change has had a negative impact on their overall occupancy trend, according to our simulations. For many of these species, we found that climate change is more important to reproduce the observed occupancy dynamics than for species for which we found a positive impact of climate change on the overall trend. While land use change turned out to be less important than climate change for the observed occupancy dynamics, its effect on the overall trend is negative for almost all of the assessed species.

Our data-driven statistical framework allows robust attribution of multiple global change drivers on transient bird occupancy dynamics. This can provide valuable information for species management under environmental change and represents an important step towards operationalising detection and attribution in biodiversity science.