Fire spread prediction models for surface fuels in subtropical forests of southern China

Abstract Subtropical forest fires are characterized by relatively small fire areas and high frequency of occurrence, with surface fires being the primary mode of forest fires spread. There is limited research on simulating forest fire spread in subtropical regions, which hinders the development and application of appropriate fire spread models. In this study, we assess the suitability and accuracy of the Rothermel model and a Random Forest model built with experimental data for predicting the rate of spread (ROS) across different types of surface fine fuel in subtropical forests. We consider fine fuels from seven typical forest types in the subtropical region of China. A total of 288 indoor experiments were conducted to simulate the fire spread process under no-wind conditions, varying the fuel moisture content at four levels (5%, 10%, 15%, and 20%) and slope angle at four inclinations (0°, 10°, 20°, and 30°). The ROS values obtained from these experiments were used to compare and analyze the predictive accuracy of the Rothermel model, a modified Rothermel model, and the Random Forest model to determine the optimal predictive model. Our findings show: (i) The prediction of the ROS under conditions of high slope angle and low fuel moisture content is not satisfactory when directly using the Rothermel model, with a coefficient of determination (R2) of 0.795, mean absolute error (MAE) of 0.204 m·min−1, and mean relative error (MRE) of 37.7%); (ii) Both the modified Rothermel model (R2: 0.902, MAE: 0.098 m·min−1, MRE: 20.2%) and the Random Forest model (R2: 0.902, MAE: 0.074 m·min−1, MRE: 13.7%) demonstrate good predictive performance with similar accuracy; (iii) Given, its physical principles and therefore potentially increased transportability, we consider the modified Rothermel model to be the most suitable of the examined models for predicting the ROS in typical forest types of southern Jiangxi Province, China, across slopes ranging from 0° to 30°. Our research provides valuable guidance for the management and suppression of subtropical forest fires.