Combining community-level spatial modelling and expert knowledge to inform climate adaptation in temperate grassy eucalypt woodlands and related grasslands

Many studies predict effects of future climate scenarios on species distributions, but few predict impacts on landscapes or ecological communities, the scales most relevant to conservation management. We combined expert knowledge with community-level spatial modelling (using artificial neural networks, ANN, and generalised dissimilarity modelling, GDM) to inform climate adaptation management in widespread but highly threatened temperate grassy ecosystems (TGE) of Australian agricultural landscapes. GDM predicted high levels of 'biotically-scaled environmental stress' (scaled in terms of potential change in species composition of communities) for plants, reptiles and snails within the TGE under medium, and especially high, 2070 climate scenarios. Predicted stress was lower for birds, mammals and frogs, possibly owing to generally wider species distributions, but these models do not account for changing habitat characteristics. ANN predicted environments within the current TGE biome will become increasingly favourable for formations such as chenopod shrublands, L. forests and Vent. forests by 2070, although classification error for eucalypt woodland in current climates was high. Expert knowledge and GDM suggest these predictions may be mediated by attributes such as environmental heterogeneity that confer resilience, but GDM confirms that widespread degradation has greatly compromised the capacity of TGE to adapt to change. Based on model predictions and expert knowledge we discuss five potential climate change outcomes for TGE: decreasing fire frequency, structural change, altered functional composition, exotic invasion, and cascading changes in ecological interactions. Although significant ecological change in TGE is likely, it is feasible to ameliorate non-climatic limits to adaptation and promote reassembly by native rather than exotic species. Current conservation efforts already target similar goals, and reinforcing and adjusting these approaches offer the highest priority, lowest risk climate adaptation options. We conclude that despite high uncertainties, combining community-level modelling with expert knowledge can guide climate adaptation management.