Near-real-time updating of ROS adjustment factors based on geostationary satellite observation data

This paper proposes a new method to improve the prediction accuracy of wildfire spread simulators by updating the Rate of Spread (ROS) adjustment factor in near-real-time using wildfire detection data from geostationary satellites. While geostationary satellites can collect wildfire detection data in near-real-time with minimal constraints from external environmental conditions, the advantage of this capability requires updating the ROS with each new data collected. To address this, the proposed algorithm describes the time required for wildfire to spread over the distance equivalent to one pixel in geostationary satellite observation data as a linear model of various factors. Subsequently, ROS adjustment factors are determined by least square estimation of the model parameters based on the spread duration obtained from multiple pixels. Using each newly collected observation dataset, the algorithm updates the ROS adjustment factors and incorporates them into FARSITE for a continuous updating of the wildfire spread prediction results. This paper demonstrates the efficiency and applicability of the proposed algorithm by a numerical example of the 2020 Creek Fire in California, USA. The results confirm that the proposed algorithm significantly enhances wildfire spread prediction accuracy by updating the ROS adjustment factors. The proposed algorithm is expected to effectively support wildfire response decision-making in situations where high-resolution data collection is challenging because of bad weather conditions such as strong winds, thereby reducing the overall impact of wildfires regardless of weather conditions.