Computational models of wildfires are necessary for operational prediction and risk assessment. These models require accurate spatial fuel data and remote sensing techniques have ability to provide high spatial resolution raster data for landscapes. We modelled a series of fires to understand and quantify the impact of the spatial resolution of fuel data on the behaviour of fire predictive model. Airborne laser scanning data was used to derive canopy height models and percentage cover grids at spatial resolutions ranging from 2 m to 50 m for Mallee heath fire spread model. The shape, unburnt area within the fire extent and extent of fire areas were compared over time. These model outputs were strongly affected by the spatial resolution of input data when the length scale of the fuel data is smaller than connectivity length scale of the fuel. At higher spatial resolutions breaks in the fuel were well resolved often resulting in a significant reduction in the predicted size of the fire. Our findings provide information for practitioners for wildfire modelling where local features may be important, such as operational predictions incorporating fire and fuel breaks, and risk modelling of peri-urban edges or assessment of potential fuel reduction mitigations.