Background <p>Prescribed burning is an important management tool for mitigating wildfire risk and maintaining ecological values in fire-prone landscapes. Burnability, defined as the probability that a location will burn, varies across space and time at fine scales. Understanding these patterns and how they respond to forecast weather could help practitioners identify suitable burn windows and anticipate burn outcomes. We aimed to develop a spatially explicit logistic regression model to predict burnability and tested whether incorporating dynamic variables (microclimate and soil moisture) improved predictions relative to models with only static variables (e.g., topography). Using a spatial dataset of historic prescribed burns in eucalypt forests, we related point-scale burn outcomes (burnt vs. unburnt) to gridded topographic, weather, and moisture variables as well as the fuel management zone as a proxy for burn objectives.</p> Results <p>Dynamic predictors improved model performance. In the dynamic model, in-forest vapor pressure deficit and soil moisture supplemented topographic position index and fuel management zone to predict spatial patterns in burnability at a 30-m resolution, and how it varies day-by-day. Overall predictive performance was moderate (cross-validated AUC = 0.68), indicating limited ability to fully discriminant between burned and unburned areas.</p> Conclusions <p>Burnability models provide an objective, pre-ignition assessment of likely burn coverage. This can support practitioners in identifying suitable windows for prescribed burning. Incorporating dynamic microclimatic variables improves predictive performance and enables models to respond to changing weather conditions. However, predictive accuracy remains constrained by the inability to fully represent on-ground management decisions.</p>