基于Gamma广义线性模型对塔河地区林火面积与主要气象驱动因子关系的分析

Relationship between Forest Burned Area and Main Driving Weather Factors Based on Gamma Model in Tahe Region

为了选取最优模型来判别对林火面积有显著影响的气象因子,应用Gamma广义线性模型和多元线性回归模型在年尺度上对大兴安岭塔河地区林火面积和当年及前一年防火期气象因子的关系进行了分析。应用AIC、BIC和RMSE等统计检验方法,对Gamma模型与传统多元线性回归模型进行拟合度比较。结果显示:Gamma全模型对林火面积与当年及前一年气象因子的拟合均要优于相对应的多元线性回归模型,且模型中自变量因子的显著性水平较高。根据最优模型选择标准,对Gamma全模型自变量中的不显著因子进行逐一删除,并获得最优Gamma模型(模型中的气象因子均与林火显著相关)从而确定影响塔河地区林火面积的主要决策气象因子。

We applied Gamma model （a generalized linear regression model） and multiple linear regression model to analyze the relationship between burned area and weather factors including the current-year and the one-year-before weather factor under year scale to select the best model for identifying the main weather driving factors on forest burned area. We made a comparison between Gamma and multiple linear regression model based on Arc, BIC, and RMSE test. Gamma model was superior over the multiple linear regression model on both ＂current-year＂ and ＂one-year-before＂ data fitting. The significance of parameters was improved in Gamma models compared to the corresponding muhiple linear regression model. We chose the best model according to the Best Model Selection Standard to identify the main weather driving factors on local fire burned area in Tahe region.