We examined the scale-dependent relationship between spatial fire likelihood or burn probability (BP) and some key environmental controls in the southern Sierra Nevada, California, USA. Continuous BP estimates were generated using a fire simulation model. The correspondence between BP (dependent variable) and elevation, ignition density, fuels and aspect was evaluated at incrementally increasing spatial scales to assess the importance of these explanatory variables in explaining BP. Results indicate the statistical relationship between BP and explanatory variables fluctuates across spatial scales, as does the influence of explanatory variables. However, because of high covariance among these variables, it was necessary to control for their shared contribution in order to extract their 'unique' contribution to BP. At the finest scale, fuels and elevation exerted the most influence on BP, whereas at broader scales, fuels and aspect were most influential. Results also showed that the influence of some variables tended to mask the true effect of seemingly less important variables. For example, the relationship between ignition density and BP was negative until we controlled for elevation, which led to a more meaningful relationship where BP increased with ignition density. This study demonstrates the value of a multi-scale approach for identifying and characterising mechanistic controls on BP that can often be blurred by strong but correlative relationships. © IAWF 2011. Reproduced from the International Journal of Wildland Fire (Sean A. Parks, et al, 2011) with the kind permission of CSIRO Publishing on behalf of the International Association of Wildland Fire.