昆明典型地表死可燃物含水率预测模型的研究

Study on moisture content prediction model of dead surface fuels in typical stands, Kunming

云南省是我国重点火险区,准确预测该省可燃物含水率对于提高火险预报准确性十分必要。在2013年防火期,通过对昆明地区8个林分地表凋落物可燃物含水率的连续观测,分析了其动态变化规律和影响因子,并采用气象要素回归法、FWI法和两种方法混合建立了相应的死可燃物含水率预测模型。这些模型采用的预报因子都是现有常规气象站方便观测的气象要素,模型误差在同类研究的控制水平内,可以在该地区的森林火险预报中直接应用。对于〈35%的可燃物含水率的预测,采用混合模型的误差最小,考虑到计算的方便,实际中可以使用气象要素回归模型,平均绝对误差（MAE）2.1%～6.0%,平均3.6%;平均相对误差（MRE）11.4%～32.7%,平均21.3%。如果考虑降雨后的可燃物含水率,即全部范围的可燃物含水率的预测,三种模型没有显著差异,为计算方便,仍建议采用气象要素回归模型,MAE 8.2%～14.2%,平均10.6%;MRE48.7%～91.3%,平均值61.4%。FWI指标与地表死可燃物含水率有关,但不如气象要素那么紧密。

Yunnan Province is a high fire risk region in China. It is necessary to accurately forecast fire danger and improve the accuracy of fire danger forecast. Successive observation of fuel moisture contents of dead surface fuels in 8 stands in Kunming, Yunnan province were conducted in 2013 forest fire prevention periods, the dynamics and affecting factors of fuel moisture content were analyzed. Moisture prediction models were established by using vapor exchange method, FWI method and method with mixed weather variables and FWI indexes. The predictors employed by the models all are the easily obtained weather variables from weather stations, and the models＇ errors in the same control level within the have accuracy within the similar studies, and hence the models can be directly used application in forest fire danger forecast in the area. For prediction of fuel moisture 〈 35%, the vapor exchange models are the best choices, resulting minimal errors; By taking into account easy-to-calculate, in actual application, meteorological element regression models can be used, leading mean absolute error （MAE） 2.1% ~ 6.0%, averaged 3.6%, leading the average relative error （MRE） 11.4% ~ 32.7%, an average of 21.3%. For predicting fuel moisture content after rain, no significant difference existed among the three types of models. Considering easy computation, vapor exchange models are still the best choices with MAE 8.2% ~ 14.2%, average 10.6%, and MRE 48.7% ~ 91.3%, average 61.4%. FWI indexes are correlated with local fuel moisture but not as close as weather variables.