Discussion

With catastrophic fire events gaining global concern, realistic fire danger assessment tools are needed more than ever. Here we present a framework to predict fuel hazard at a fine spatial resolution that is directly useful for operational fire management. It is based on mechanism-informed random forest models that make use of field-based observations of fuel hazard, gridded soil and topographic attributes, long-term climate trends, as well as plant responses to changing environment. Our modelling approach provides an important step towards a mechanism-informed fire risk assessment system. The predictions could also be used in fire behaviour models as well as to evaluate other vegetation model projections.