Abstract:
|
Longer fire seasons due to anthropogenic climate change have increased the frequency and intensity of wildland fires in many fire prone areas. These fires lead to catastrophic loss with respect to property, human life, and ecological diversity. Although there are numerous deterministic models for fire spread, both operational and in research environments, the operational models are limited in their ability to model realistic fire spread. Here, we consider a Bayesian cellular automata model for fire spread that considers a multi-state categorical response with latent spatio-temporal dynamics that accommodate advective spread. Importantly, this approach allows for the inclusion of important spatial and temporal covariates (e.g., elevation, fuel load, land use, background wind, etc.) and provides a coherent uncertainty quantification to the prediction of the fire spread through time. This model is demonstrated on controlled burn fire observations.
|