Transferable, predictive models of benthic communities informs marine spatial planning in a remote and data‐poor region

Abstract

Abstract Systematic conservation planning requires spatial information on biodiversity. Such information is often unavailable, forcing spatial planning to rely on assumed relationships between species and environmental features. This problem is particularly acute in large, remote marine protected areas that are proliferating rapidly. Here, we use models to predict whether (a) macrobenthic biodiversity across four taxa (gorgonians, soft corals, hard corals, and sponges) with different life histories are congruent within seascape features through regional space; and (b) models generated in an intensively‐sampled area in one region can predict the occurrence of habitat‐forming macrobenthos in neighboring ones. All four taxa studied showed similar habitat preferences, but high variability in distributions among and within features suggesting factors other than simple geomorphology influence these regional biodiversity patterns. Nonetheless, models derived from one region accurately predicted the presence and absence of the same taxa hundreds of kilometers away. This transferability of models of species occurrences has the potential to deliver improved reserve design in data‐deficient regions.

Keywords

• boosted regression trees
• conservation planning
• marine protected areas
• predictive models
• spatial planning
• transferability