Influencing factors and prediction models of surface fire spreading rate of typical coniferous forest in karst ecosystem of Guizhou Province, southwestern China

Objective The karst ecosystem is relatively vulnerable to forest fire due to interlaced agriculture and forestry and the fire source, which poses serious impact on the restoration of karst ecosystem. Thus, investigating the spreading rate of forest fire under no-wind and zero-slope conditions was of great significance for the restoration of vegetation and forest fire prevention in this region.

Method This study focuses on the surface combustibles under three typical coniferous forests within the karst ecosystem. Under indoor no-wind and zero-wind conditions, coniferous beds layers similar to those found in the wild were constructed, and 188 ignition experiments were conducted with different gradient combinations of moisture content, thickness, and loads factors as variables.

Result (1) Under the same conditions, there was no significant difference in forest fire spreading rate between Pinus yunnanensis and Pinus armandii, but both rates were significantly higher than that of Pinus massoniana (P < 0.001). (2) The moisture content of fuelbed could significantly retard the spreading rate of all fuelbed, while the height of fuelbed had a significant promoting effect, and the loading had promoting effect on the spreading rate of the needle bed of Pinus yunnanensis and Pinus armandii. (3) Fitting with additive model, the mean absolute error of spreading rate of fuelbed of Pinus massoniana, Pinus yunnanensis and Pinus armandii was 0.013, 0.029 and 0.020 m/min, respectively, and the mean absolute error of multiplier model of forest fire spreading rate of fuelbed of Pinus massoniana, Pinus yunnanensis and Pinus armandii was 0.014, 0.023, 0.018 m/min, respectively. Besides, both prediction models could explain over 75% of the dynamic of forest fire spreading rate. (4) The mean absolute error and mean relative error of the extrapolation error of additive model were 0.108 m/min and 42.50%, respectively. The extrapolation effect of multiplicative model was better than that of the additive model, with the mean absolute error and relative error of 0.086 m/min and 28.20%, respectively.

Conclusion This study reveals the influence of fulebed characteristics on the forest fire spreading rate in karst ecosystem. The spreading rate of three coniferous beds is high under the experiment condition, and it is possible to spread and cause disasters. Based on prediction accuracy and extrapolation effect, it is suggested to select multiple model to simulate forest fire spreading rate. These findings provide valuable understandings and insights for forest fire prevention in karst ecosystem.