西藏林区森林火险等级中短期预报方法

Short and Mid-term Prediction Method of Forest Fire Weather Danger Classification Over Tibetan Forest

利用2009/2010、2010/2011和2011/2012西藏林区防火期（11月—翌年4月）气象观测资料和T639数值预报资料,基于人工神经网络BP算法,建立了西藏林区森林火险等级1~7天预报模型,历史拟合率超过85%;通过对2012/2013防火期间的森林火险等级试报检验结果表明,前3天的平均绝对误差不超过0.5级,7天的平均绝对误差不超过0.6级;与直接利用数值预报模式气象要素预报结果相比,有效地纠正了数值模式要素预报的系统偏差,表明模型预报效果良好。该模型的建立提高了对西藏高原森林火险等级的预报准确性,为森林火险防御和消防调度提供了参考。

Using the 2009/2010,2010/2011 and 2011/2012 Tibet forest fire prevention period（from November to next April） meteorological observations and T639 numerical forecast data,based on artificial neural network BP algorithm（BPM）,the forest fire danger rating forest model in 7 days in Tibet was established.The model history matching rate was more than 85%,through the 2012/2013 fire prevention period fire danger rating test report results showed that：the average absolute error in 3 days was less than 0.5,and the average absolute error in 7 days was less than 0.6;comparing with using numerical weather prediction model,the BPM effectively correct prediction of system deviation factor.It indicated that：the BPM had a good prediction results.The model improved on the Tibetan forest fire danger rating prediction to improve forest fire prevention.