论文部分内容阅读
以西藏地区的冰湖为研究对象,基于冰湖的野外调查、遥感影像与历史文献资料,选取32个冰湖作为样本进行了危险性分析。根据研究对象的区域特性,选取冰湖坝顶宽度、湖水位距坝顶高度与湖坝高度之比、冰舌前端距冰湖距离、冰舌段坡度、冰湖面积和补给冰川面积6个预测指标作为变量,通过对样本进行逻辑回归分析,提出了西藏地区冰湖溃决的危险性概率预测模型。基于分类阈值为50%的逻辑回归模型交叉验证结果表明,模型对已溃决冰湖样本预测的准确率为82%,对未溃决冰湖样本预测的准确率为95%,对所有冰湖样本预测的准确率为91%。按冰湖溃决的概率范围,将冰湖溃决的危险性划分为4个等级,概率值<0.30的为低等级,0.30~0.50的为中等级,0.50~0.80的为高等级,>0.80的为极高等级。
Taking glacial lakes in Tibet as the research object, 32 ice lakes were selected as the samples for risk analysis based on the field investigation, remote sensing images and historical documents of glacial lakes. According to the regional characteristics of the study object, six projections of the width of the ice dam, the ratio of the height of the lake to the height of the dam, the distance from the front of the ice tongue to the ice lake, the gradient of the ice tongue, the area of glacial lake and the area of recharge glacier are selected Indicators as variables, through the logistic regression analysis of the sample, put forward the risk probability prediction model of glacial lake rupture in Tibet. Based on the logistic regression model with a classification threshold of 50%, the cross-validation results show that the model has an accuracy of 82% for the predicted glacial lake samples and 95% for the non-predicted ice lake samples, The accuracy rate is 91%. According to the probability range of ice lake rupture, the danger of ice lake rupture is divided into 4 grades, the probability value <0.30 is the low grade, 0.30 ~ 0.50 is the middle grade, 0.50 ~ 0.80 is the high grade,> 0.80 is Very high grade.