【摘要】 基于Logistic回归模型分析黑龙江黑河地区林火发生与气象因子间的关系，确定影响林火发生主要气象因子并建立林火发生概率预测模型。以1981—2015年黑河地区日历史林火数据和日气象数据为基础，随机划分80%的建模数据和20%的校验数据。使用R软件，通过Logistic逐步回归建立5组中间模型并筛选出显著气象因子(P<0.05)。运用Logistic回归模型建立全样本林火发生概率预测模型，分析主要气象因子。结果表明:日最大风速、日最高气温、日平均相对湿度、日平均水汽压和日降水量是影响林火发生的主要显著气象因子(P<0.05)，其中日最大风速和日最高气温与林火发生呈正相关，日平均相对湿度、日平均水汽压和日降水量与林火发生呈负相关;基于Logistic回归的各样本模型预测准确率为82.83%～86.12%，受试者工作特征(ROC)曲线的线下面积(AUC)值在0.884～0.920之间，模型具有较高的预测准确率和拟合度，适用于该地区基于气象因子的林火发生概率预测。更多还原

【Abstract】 Based on the Logistic regression model, this study analyzed the relationship between the occurrence of forest fires and meteorological factors, determined the main meteorological factors, and established the prediction model for Heihe region in Heilongjiang Province. Historical forest fire and meteorological data from 1981 to 2015 in Heihe region were randomly divided into 80% building data and 20% validation data. R software was used to establish five sub-models and screen out significant meteorological factors by Logistic stepwise regression(P<0.05). Moreover, a forest fire prediction model was established using the Logistic regression model to analyze the main meteorological factors. The results showed that daily maximum wind speed, daily maximum temperature, Daily average water vapor pressure, daily average relative humidity, and daily precipitation are the significant meteorological factors affecting the occurrence of forest fires(P<0.05). The former two meteorological factors are positively correlated with forest fire occurrence, whereas the latter three factors are negatively correlated with such occurrence. The prediction accuracy of the Logistic regression model is 82.83%～86.12%. Area under the curve ranges from 0.884 to 0.920. The Logistic regression model has good prediction accuracy and high goodness of fit, which is suitable for forest fire occurrence probability prediction research based on meteorological factors in Heihe region.更多还原