Wildfire spread prediction using geostationary satellite observation data and directional ROS adjustment factor.

This paper introduces a novel data-driven approach to enhance the prediction accuracy of regional-scale wildfire spread simulators using geostationary satellite wildfire detection data. First, we propose a new algorithm that maximizes accuracy by innovatively utilizing the conventional uniform Rate of Spread (ROS) adjustment factor. The ROS adjustment factor is estimated by a genetic algorithm based on near-real-time observations from geostationary satellites and the FARSITE wildfire simulator. The accuracy of wildfire spread prediction is improved by applying the estimated ROS adjustment factor within FARSITE. In addition, this paper proposes a Directional ROS adjustment factor approach, which applies different ROS adjustment factors to a given fuel model depending on their location, thereby allowing for a more accurate representation of a wildfire's perimeter in FARSITE-based simulations. The proposed methodology is demonstrated through the 2020 Creek Fire in California, USA. The results demonstrate that the proposed algorithm significantly enhances accuracy. Moreover, incorporating the directional ROS adjustment factor leads to more precise predictions than the conventional uniform ROS adjustment factors. The proposed methodology offers significant advantages in developing targeted wildfire response strategies, particularly under conditions where external environmental conditions make high-resolution data collection challenging. This approach enhances the accuracy and reliability of response planning, enabling more effective allocation of resources and potentially reducing the overall impact of wildfires.