Abstract: In forest fires, extreme fire behaviors such as spotting and fire whirl frequently occur, which can ignite unburned areas ahead of the fire front and accelerate surface fire spread. This work selected a forest fire in Changjiang County, Hainan, on June 7, 2020, which occurred on flat terrain, as the research object. By altering the number of fire points and the rate of spread adjustment factor, the accuracy of fire spread simulation results using the FARSITE system was analyzed. The results showed that with the increase of the adjustment factor, the prediction accuracy first increases and then decreases. When the adjustment factor is 1, the intersection area under single ignition source conditions accounts for 5% of the actual burned area. Using four initial fire points and changing their positions, the intersection area proportions increased by 10% and 18%, respectively. When the optimal adjustment factor is reached, the fire spread simulation accuracy is the highest, and the intersection area proportion in the single fire source simulation increases to 72% with an accuracy of 0.70. Reasonably setting the locations and number of fire points could improve the accuracy of fire spread prediction, and reduce the impact of extreme fire behaviors or inaccuracies in model data on prediction results, but may also lead to overprediction of the burned area.

Key words: FARSITE, fire spread, number of fire points, rate of spread adjustment factor (RAF), extreme fire behavior