基于Logistic回归的泥石流灾害预警模型

Logistic Regression for Early Warning of Debris Flow Hazards

为提前预警泥石流灾害,基于Logistic回归模型,定量分析引起泥石流的雨量评判指标的影响因子(降雨强度和累计雨量),形成了一种能预测泥石流是否发生的灾害预警方法。西南地区2002—2010年已有93次泥石流灾害监测数据模型估计结果表明:各个降雨参数和总模型的模拟效果良好,且降雨参数能有效地解释泥石流发生概率变化的过程;相比累计雨量参数而言,降雨强度对泥石流发生概率的影响更为显著;回归模型对未来时空模拟效果良好,预测精度能达到85%以上。该方法可为我国西南地区的泥石流灾害预警提供一定的启发。

To obtain an effective warning of debris flow disasters a disaster warning was designed that predicts occurrence of mudslides,based on the logistic regression model,and the quantitative factors influencing the rainfall evaluation indicators of rainfall intensity and accumulated rainfall that cause mudslides analysis.Analysis of the data of the 93 debris flows in southwest China between 2002 to2010 indicated the simulation results of each rainfall parameter and the total model were good,and the rainfall parameters could explain the process of the probability change of debris flow effectively.Compared with the accumulated rainfall parameters,rainfall intensity has more significant influence on the probability of debris flow occurrence.The regression model has a good effect on the future space-time simulation with the prediction accuracy reaching 85% and above.This method can provide scientific and technical support for early warning of debris flow hazards in southwest China.