Lessons learned using species' distribution models for conservation planning in the Golden Gate Biosphere reserve.

Conservation practitioners responsible for maintaining biodiversity and ecosystem services within protected areas require information about how dominant plant species may reassemble under rapid global change. Although species' distribution models (SDMs) alone do not account for multiple threats or species population dynamics, they can provide robust assessments of where species may persist or disperse to in the future, especially if carefully constructed and thoroughly evaluated. We used ensemble SDMs to evaluate how climate change may alter suitable habitat for six dominant plant species representing different life forms within the Golden Gate Biosphere Network (GGBN), located along the central to northern coast of California. We trained the models on presence-absence data and 23 environmental predictors, including climate, topography, and soils, using six algorithms. We projected habitat suitability to late-century climate conditions using three GCMs under RCP 8.5 and summarized areas of agreement, expansion, and refugia. Model results can be summarized into several lessons learned, many of which are consistent with previous research. The first is that projected habitat may either expand or contract, and the direction of change varies by individual species, even within the same genus. Many projected changes reflected species' relationships with key climate variables relative to their future projected trends. Disagreement across scenarios was largely driven by uncertainty in projected precipitation changes, while non-climatic variables, particularly soils, were also important in mediating projected habitat change. Contrary to common assumptions, projected habitat shifts were not always upslope or poleward. Finally, although species face multiple threats from other global changes, SDMs can provide a valuable baseline for conservation decisions within small reserves, particularly through identification of refugia and comparison of model scenarios. However, habitat changes within protected areas may not reflect dynamics elsewhere across species' ranges, underscoring the need for multi-scale conservation planning.