Aim: Human activity has reshaped ecological communities for thousands of years. While these activities have typically led to habitat loss, some species have successfully exploited human environments. However, the effects of long-term human land-use on the distributions of such species are poorly understood. Here, we investigated how land-use change over the last 12,000 years has altered the distribution of suitable habitat for a widespread human-commensal bird species, the Pacific swallow (Hirundo tahitica). 

Location: Southeast Asia, Melanesia

Methods: We built species distribution models using community science records and evaluated approaches to controlling for spatial biases that arise from using unstructured survey data to model species associated with anthropogenic environments. We then assessed the distribution of suitable habitat for Pacific swallows under three alternative scenarios of land-use change that incorporated only climate variables, only human land-use variables, and both variable sets. Lastly, we hindcasted alternative models at 1,000-year intervals over the last 12,000 years to evaluate how habitat availability has changed due to long-term human activity.

Results: Models that included climate and human land-use variables were the best fit to occurrence records. Standard methods for controlling for spatial bias performed poorly compared to fully sampling the environmental background, highlighting unique considerations for modeling human-associated species. Hindcasting alternative land-use scenarios showed that suitable human-altered habitat arose within the Pacific swallow range at least 4000 years ago and caused significant expansions of suitable habitat over the last 2000 years. 

Main conclusions: Human land-use over the last several thousand years has likely provided Pacific swallows with substantial new habitat, which may have led to population size expansions. Incorporating long-term human land-use into species distribution models offers insights into when associations with human environments may have arisen and generates testable predictions for how populations respond and adapt to human land-use change over millennial timescales.