Performance of operational fire spread models in California

Background Wildfire simulators allow estimating fire spread and behaviour in complex environments, supporting planning analysis of incidents real time. However, uncertainty derived from input data quality model inherent inaccuracies may undermine the utility such predictions.Aims We assessed performance models for initial attack used California through rate (ROS) 1853 wildfires.Methods retrieved observed growth FireGuard (FG) database, ran an automatic simulation with Analyst Enterprise accuracy simulations by comparing predicted ROS well-known error bias metrics, analysing main factors influencing accuracy.Key results The errors biases were reasonable performed automatically. identified environmental variables that predictions, especially timber areas where some fuel underestimated ROS.Conclusions models’ is line studies developed other regions are accurate enough to be time assess fires.Implications This work allows users better understand operational environments opens new research lines further improve current models.