How fire type and season influences patchiness and severity across woodland and forest in the southwest of Australia

Fire is a dominant disturbance process in terrestrial systems and is predicted to increase in extent, frequency, and impact with climate change. In most fire-prone systems globally, planned fire (prescribed burning) is used as a management tool to reduce risk to human life and property. Although quantifying burn impact (patchiness, longevity of risk reduction) is essential to informing management it is infrequently quantified. Here we examine the patchiness and spatial heterogeneity of wildfires and planned fires across seasons and their impacts on fire-prone woodlands and dry sclerophyll forests of southwestern Australia. We use Google Earth Engine (GEE) to analyse spatial heterogeneity of planned vs unplanned fire as a way of optimising fire management. Season of fire (planned: spring and autumn, wildfire: summer) as well as antecedent drought, weather conditions during fire, fire history, and topography were used as predictors. Fire severity was defined using normalized burn ratio (NBR), which are derived from Landsat data on GEE. Random Forest were used to select the main features affecting fire severity, and a model predicting fire patchiness was built based on the selected features. The result indicted that the most important features affecting burnt severity are season, fire type and elevation. The model shows 80% accuracy for predicting the burnt severity. By quantifying patchiness of prescribed burning and wildfire relative to disturbance history and spatiotemporal covariates we provide critical insights to fire managers to optimise resources as well as demonstrate GEE's capacity to scale up previously unavailable knowledge.