Habitat modeling as a predictive tool for analyzing spatial shifts in Antarctic benthic communities due to global climate change

Abstract

Ensemble habitat modeling is a tool in the multivariate analysis of arbitrary species or community distribution which combines models of best fit to an optimized model (ensemble model, EM). To simulate spatial variation of communities and predict the impact of climate change, it is essential to identify the distribution-controlling factors. Macroalgae biomass production in polar regions is determined by environmental factors such as irradiance, which are modified under climate change impact. In coastal fjords of King George Island/Isla 25 de Mayo, Antarctica, suspended particulate matter (SPM) from glacial melting causes shading of algal communities during summer. Ten different species distribution models (SDMs) were applied to predict macroalgae distribution based on their statistical relationships with environmental variables. The suitability of the SDMs was assessed by two different evaluation methods. Those SDMs based on a multitude of decision trees such as Random Forest and Classification Tree Analysis reached the highest predictive ability followed by generalized boosted models and maximum-entropy approaches. We achieved excellent results for the current status EM (true scale statistics 0.833 and relative operating characteristics 0.975). The environmental variables hard substrate and SPM were identified as the best predictors explaining more than 60 % of the modelled distribution. Additional variables distance to glacier, total organic carbon, bathymetry and slope increased the explanatory power proved by cross-validation. Presumably, the SPM load of the meltwater streams on the Potter Peninsula will continue to increase at least linearly. We therefore coupled the EM with changing SPM conditions representing enhanced or reduced melt water input. Increasing SPM by 25% decreased predicted macroalgal coverage by approximately 38%. The ensemble species distribution modelling helps to identify the important factors controlling spatial distribution and can be used to link causes to effects in (Antarctic) coastal community change.