Abstract: [Objective] The Rothermel model is the most widely used semi-physical prediction model at present. The aim of this study was to investigate the applicability of the Rothermel forest fire spread rate prediction model and the other two kinds of spread rate prediction models based on the Rothermel model to eight typical surface dead fine fuels in southern China to provide theoretical support and guidance for the forest fire spread rate prediction.[Method] Eight typical surface dead fine fuels in southern China were chosen as the research material, according to the actual conditions of the research material in the field, the fuel bed with different fuel moisture contents, fuel load and height was constructed in the wind tunnel laboratory of Maoer Mountain Experiment Forest Station of Northeast Forestry University. The 36 burning tests were carried out for each fuels under flat and windless condition, totaling 288 burning tests. The spread rate of each fuel type in different proportions was recorded. The most suitable prediction models for the forest fire spread rates of eight typical fuels in southern China were obtained by comparing the Rothermel model, re-estimating the Rothermel model parameters and the self-built model after improving the form of Rothermel model.[Result] 1) Under the flat and windless conditions, the maximum spread rate of eight typical fuels in southern China was 0.55 m·min^-1. The average spread rate from big to small was as follows:Pinus armandi, Pinus yunnanensis, Phyllostachys edulis, Cryptomeria fortune, Cunninghamia lanceolate, Pinus massoniana, Quercus acutissima, Cyclobalanopsis glauca. 2) The Rothermel model was used to predict the spread rate of forest fire, the prediction error was large, the mean absolute error was 0.18 m·min^-1, and the average relative error was 70.0%. 3) The prediction accuracy of re-estimating the Rothermel model parameters and self-built model was significantly higher than that of Rothermel model. The average absolute error of the re-estimating Rothermel model was 0.04 m·min^-1, the average relative error was less than 18%; the average error of the self-built model was 0.037 m·min^-1, the average relative error was 16.45%. 4) The prediction error was not significantly different between the re-estimating the Rothermel model parameters and the self-built models. The correlation coefficient between the predicted value and the measured value of the self-built model was 0.71-0.90, and the average value was 0.80.[Conclusion] For the eight typical surface dead fine fuel in southern China, the error of re-estimating the Rothemel model parameters and the self-built model was not significant, but the format of the self-built model was more simple and easy to use. The self-built model can be used to predict the surface fire spread rate of typical fuels in southern China.

Key words: Rothermel model, spread rate, fire behave, fuel bed height, moisture content