Modelling the habitat of the endangered Carpentarian Grasswren (Amytornis dorotheae): The importance of spatio-temporal habitat availability in a fire prone landscape

Species distribution modelling (SDM), a tool increasingly adopted to quantify geographic range size, often predicts species’ distributions as static. However, habitat availability may exhibit spatial and temporal variation when dynamic processes, such as fire, determine suitability. Static SDM approaches may not satisfactorily represent this dynamic process. We investigated the potential use of SDM to quantify dynamic habitat availability by applying the MaxEnt SDM technique to model the habitat of the Carpentarian Grasswren (Amytornis dorotheae), an endangered Australian passerine dependent on long unburnt vegetation in a fire prone system. By adjusting a typical SDM approach to incorporate the dynamic nature of fire, we modelled the spatio-temporal variation of suitable habitat over 12 years and compared it to a static modelling approach. Incorporating fire as a dynamic process increased the importance of the fire variable to models (from 35% permutation importance) and improved model performance, as evaluated by the AUC using cross-validation. Our dynamic model revealed sizeable temporal variation in the area and spatial arrangement of suitable habitat that was not apparent in the static model. This result may partly solve the mystery of why the species occurs as widely separated populations despite the presence of seemingly suitable intervening habitat. In areas where the species is no longer found, habitat availability was less consistent due to frequent fire, and fire refugia was more limited and isolated, when compared to sites with recent records. These results demonstrate that, when compared to a static approach, a dynamic SDM approach can lead to improved understanding of dynamic ecological processes, and their impact on a species