Improving prediction of future habitat distributions under climate change by combining multiple habitat suitability models
Understanding future habitat distributions is important for land managers in preparation for managing under various climate change scenarios but predicting future habitats is not easy due to many unknown factors. Mapping reasonably accurate future habitats under climate change requires using the latest machine learning techniques for model development to tackle uncertainty. One of the limitations of habitat suitability models is that a single model could result in errors in future predictions due to suboptimal fit. In addition, a single response such as abundance may not capture multiple aspects and nuances of tree habitat distribution well enough. A Forest Service scientist developed a technique that combines multiple models to predict average and consensus responses. The technique also combines multiple responses like basal area, tree density, stand age, and height into a synthetic index called “habitat-fitness” that better captures the essence of these multiple responses compared to any single response. This allows assessment of habitats where one has maximum confidence (consensus model) and also areas where the confidence is higher (average model). Land managers have used this approach to assign regions of high confidence of future habitat trends and to improve the prediction capabilities of habitat models under climate change.
Forest Service Partners