Species distribution modeling as an approach to studying the processes of landscape domestication in central southern Mexico

Context Landscape domestication (LD) has been an important mechanism for subsistence in traditional rural societies. In the last decade, important theoretical advances have been accomplished with valuable archaeological, historical, and ecological approaches for understanding LD processes, while integrated spatial approximations are still scanty. Objectives To assess the usefulness of species distribution modeling for addressing LD processes associated with human influence on plant community distribution, using as case study the palm-stands of Brahea dulcis in central-southern Mexico. Methods We used MaxEnt for building distribution models, including social factors as triggers of palm-stand distribution, and environmental variables as covariates. Model performance, predicted surface and variable importance were statistically evaluated. The best distribution model was chosen, and it was further assed by visual inspection and ground-truth validation to contrasts predicted versus observed palm-stand distribution. Results Social factors contributed the most to the model fit, proving to be strongly associated with palm-stand presence. Visual inspection of the predicted spatial distribution resulted coherent with observed palm-stand distribution. Ground-truth proved to reach a certainty of circa 80% between predicted and observed areas with palm-stands. Conclusions Ethnic identity, distance to roads, and land tenure were the strongest explanatory variables to accurately predict the observed palm-stand spatial distribution. This revealed the high importance of social factors for palm-stand presence and support the assertion that palm-stand are the result of long-standing human actions, rather than purely ecological and physical attributes. The identification of driving social factors allowed to inquire into the underlying LD processes.