江西南昌典型林分地表死可燃物含水率预测——方法优选与FWI适用性分析

Prediction on moisture contents of typical forest-land surface dead fuels in Nanchang city, Jinagxi province: method optimization and FWI applicability

亚热带针阔混交林是我国南方重要的植被类型，森林火灾发生次数多。对该植被类型的典型可燃物含水率将有助于提高火险预报的准确性。以位于该区域内的南昌市茶园山林场5种典型林分地表死可燃物为研究对象，通过对其含水率的连续观测，分析了死可燃物含水率与气象因子的关系。采用气象要素、FWI因子和两者的结果，区分不同含水率范围，分别建立的各可燃物3种预测模型。结果表明：研究地区防火期内可燃物含水率从11．8％到276．6％，只有1／4的时间低于35％，具备发生森林火灾的条件，其余的时间燃烧性都很低，平均火险不高。但最小含水率已低到10％左右，具备发生大火的潜在可能性。3种预测方法中，FWI模型误差最大，气象要素回归模型和混合模型误差相似，考虑到简单方便，该地区的含水率预测可以采用气象要素回归法。含水率小于35％时，模型MAE2．25％～4．67％，平均3．73％；MRE11．14％～23．73％，平均18．63％。含水率〉35％时，模型MAE6．34％～18．33％，平均8．63‰MRE10．49％～18．16％，平均14．43％。在全范围含水率时，模型MAE7．46％～16．43％，平均10．51％；MRE18．78％～23．64％，平均20．99％。FWI指标与研究地区可燃物含水率关系密切，也可以用于含水率预测和火险预报，如果考虑全国统一的预报模型，对于该地区，FWI系统是适用的，但为提高预测的准确性，应进行进一步的修正。

Subtropical mixed broadleaf-conifer forest is an important vegetation type in south China, where forest fires occurred frequently. Study on the moisture content of typical forest-land surface dead fuels in the region is of benefit to increase the accuracy of fire danger forecast. By taking the dead forest land surface fuels in 5 typical stands in Chayuanshan Forest Farm, Nanchang city, Jiaxi province as the studied objects, through continuous observation on moisture of five dead fuels, the relations between the dead fuels and weather variables were investigated. Based on the weather factors, FWI factor and the results of both, the moisture content ranges of different dead fuels were distinguished, and three predictive models for five dead fuels were respectively established. The results show that the fuel moisture in the region in the study period ranged from 11.8% to 276.6%, one fourth of time they were less than 35% which is ready to burn and in the rest time the burn probability of the fuels had lower combustibility, which suggests low fire hazard in the region. Even so, since the minimum moisture was below 10%, it still had the potential to burn largely. For the three types of models established, the FWI model had the largest error, higher than the vapor exchange models and the mixed model, where there was no significant difference between the rest two models. Considering simple computation requirements, the vapor exchange models are the best choice for predicting fuel moisture in the region, for all different moisture ranges. The MAE were 2.25%-4.67% with a average of 3.73%, the MRE were from 11.14% to 23.73% with average of 18.63%. The FWI indexes were closely correlated with fuel moisture in the region, which indicates that they were suitable for the region and for creating a nationally unique prediction model, But systematically modification should be conducted rather than the simple one done here for higher accuracy.