Testing the ability of distribution models to map rare plant communities in Waterton Lakes National Park

Abstract

Effective biodiversity conservation requires understanding the distributions of both individual species and the communities they form. However, few studies have assessed methods for modelling the distributions of ecological communities. I compared two modelling approaches for predicting seven rare plant communities in Waterton Lakes National Park, Alberta, Canada: community distribution models (CDMs), which treat communities as units, and stacked species distribution models (s-SDMs), which model the component species of each community and then assemble these predictions. Consistent with the individualistic view of ecological communities, s-SDMs consistently outperformed CDMs, especially for communities with few occurrence records. Model accuracy was largely unaffected by recent wildfire disturbance, suggesting that these models are just as useful in disturbed landscapes. My findings show that s-SDMs are an effective approach for mapping rare vegetation communities and can help managers locate, monitor, and protect these assemblages.