Evaluating models for predicting microclimates across sparsely vegetated and topographically diverse ecosystems

Aim Microclimate information is often crucial for understanding ecological patterns and processes, including under climate change, but typically absent from biogeographic studies owing to difficulties in obtaining microclimate data. Recent advances modelling, however, suggest that conditions can now be predicted anywhere at any time using hybrid physically empirically based models. Here, we test these methods across a sparsely vegetated topographically diverse sub-Antarctic island ecosystem (Macquarie Island). Innovation predictions were generated height of 4 cm above the surface on 100 × m elevation grid snow-free season (Oct–Mar), with models driven by either reanalysis data (CRA) or CRA augmented meteorological observations island's automatic weather station (AWS+CRA). These compared simple lapse rate model (LR), where an elevational adjustment was applied hourly temperature measurements AWS. Prediction errors tended lower AWS+CRA-driven models, particularly when CRA-driven The AWS+CRA LR had similar prediction averaged Tmin Tmean, Tmax much smaller former. within-site correlation between observed daily Tmean average >0.8 all months >0.7 predictions, consistently predictions. Main conclusions ecologically relevant spatial temporal scales possible provide added value over extremes situ observations. will help add dimension aid delivery change-resilient conservation planning change-exposed ecosystems.