攀枝花市矿产资源丰富,采矿历史悠久,开采活动强度大,地质结构复杂,生态环境脆弱,矿山地质灾害频发.基于2019年的三江(怒江,澜沧江,金沙江)北段矿产开发环境遥感监测调查资料,在分析矿山地质灾害发育规律与影响因素关系的基础上,选取坡...攀枝花市矿产资源丰富,采矿历史悠久,开采活动强度大,地质结构复杂,生态环境脆弱,矿山地质灾害频发.基于2019年的三江(怒江,澜沧江,金沙江)北段矿产开发环境遥感监测调查资料,在分析矿山地质灾害发育规律与影响因素关系的基础上,选取坡度、工程地质岩组、距断层的距离、降水量、植被覆盖度、距河流的距离以及距开采活动面的距离7个评价因子,借助ArcGIS软件平台,采用确定性系数(Certainty Factor,CF)模型、信息量模型以及CF与信息量耦合模型开展攀枝花市矿山地质灾害易发性评价研究.结果表明,工程地质岩组、植被覆盖度、距开采活动面的距离是影响矿山地质灾害分布的控制因子;经过ROC(Receiver Operating Characteristic Curve,ROC)曲线检验,CF与信息量耦合模型的AUC(Area Under the Curve,AUC)值高达0.909,表明耦合模型比单一模型的评价精度高,其精度为耦合模型>信息量模型>确定性系数模型;耦合模型的易发性分区为极高易发区(5.24%)、高易发区(11.67%)、中易发区(41.66%)和低易发区(41.43%),其中极高和高易发区主要分布在开采活动强度较大的煤矿、铁矿和花岗岩矿矿山,即主要分布在仁和区、东区、盐边县和米易县.展开更多
This study explores a comparative study of three susceptibility assessment models based on remote sensing(RS) and geographic information system(GIS). The Lenggu region(China) was selected as a case study. At first, a ...This study explores a comparative study of three susceptibility assessment models based on remote sensing(RS) and geographic information system(GIS). The Lenggu region(China) was selected as a case study. At first, a landslide inventory map was compiled using data from existing geology reports, satellite imagery, and coupling with field observations. Subsequently, three models were built to map the landslide susceptibility using analytical hierarchy process(AHP), fuzzy logic(FL) and certainty factors(CF). The resulting models were validated and compared using areas under the curve(AUC). The AUC plot estimation results indicated that the three models are promising methods for landslide susceptibility mapping. Among the three methods, CF model has highest prediction accuracy than the other two models. Similarly, the outcome of this study reveals that streams, faults, slope and elevation are the main conditioning factors of landslides. Especially, the erosion of streams plays a key role of the landslide occurrence. These landslide susceptibility maps, to some extent, reflect spatial distribution characteristics of landslides in alpine-canyon region of southwest China, and can be used for land planning and hazard risk assessment.展开更多
文摘构建准确的滑坡预测模型和确定环境因子的贡献程度,对滑坡易发性评价具有重要意义。在以往研究中,最大熵物种分布(maximum entropy model,MaxEnt)模型因其对样本量要求低、预测精度高和可避免模型过度拟合等优点,被广泛运用在生态学领域。以沅陵县为研究区,基于342处滑坡灾害点数据和9个环境变量,分别采用确定性系数(certainty factor,CF)模型、逻辑回归(Logistic)模型和MaxEnt模型对沅陵县进行滑坡易发性分区预测。同时采用刀切法(Jackknife)检验环境因子对预测结果的贡献程度,确定滑坡地质灾害的主要影响因素。结果表明:确定性系数模型、逻辑回归模型和MaxEnt模型的受试者特征曲线(receiver operating characteristic,ROC)下面积(area under the curve,AUC)值分别为0.827、0.803、0.911,3种模型的预测精度均较高,且MaxEnt模型精度最高,表现较好;河流是影响研究区滑坡灾害发生贡献程度最高的环境因子;滑坡灾害主要发育在以河流为中心向外延伸100 m范围内,集中分布在沅江、深溪和兰溪附近。研究能为沅陵县地质灾害易发性评价提供一种新的方法。
文摘攀枝花市矿产资源丰富,采矿历史悠久,开采活动强度大,地质结构复杂,生态环境脆弱,矿山地质灾害频发.基于2019年的三江(怒江,澜沧江,金沙江)北段矿产开发环境遥感监测调查资料,在分析矿山地质灾害发育规律与影响因素关系的基础上,选取坡度、工程地质岩组、距断层的距离、降水量、植被覆盖度、距河流的距离以及距开采活动面的距离7个评价因子,借助ArcGIS软件平台,采用确定性系数(Certainty Factor,CF)模型、信息量模型以及CF与信息量耦合模型开展攀枝花市矿山地质灾害易发性评价研究.结果表明,工程地质岩组、植被覆盖度、距开采活动面的距离是影响矿山地质灾害分布的控制因子;经过ROC(Receiver Operating Characteristic Curve,ROC)曲线检验,CF与信息量耦合模型的AUC(Area Under the Curve,AUC)值高达0.909,表明耦合模型比单一模型的评价精度高,其精度为耦合模型>信息量模型>确定性系数模型;耦合模型的易发性分区为极高易发区(5.24%)、高易发区(11.67%)、中易发区(41.66%)和低易发区(41.43%),其中极高和高易发区主要分布在开采活动强度较大的煤矿、铁矿和花岗岩矿矿山,即主要分布在仁和区、东区、盐边县和米易县.
基金Supported by the National Natural Science Foundation of China(41602354)the Chongqing Research Program of Basic Research and Frontier Technology(2017jcyjAX0300)the Fundamental Research Funds for the Central Universities(XDJK2016B027)
文摘This study explores a comparative study of three susceptibility assessment models based on remote sensing(RS) and geographic information system(GIS). The Lenggu region(China) was selected as a case study. At first, a landslide inventory map was compiled using data from existing geology reports, satellite imagery, and coupling with field observations. Subsequently, three models were built to map the landslide susceptibility using analytical hierarchy process(AHP), fuzzy logic(FL) and certainty factors(CF). The resulting models were validated and compared using areas under the curve(AUC). The AUC plot estimation results indicated that the three models are promising methods for landslide susceptibility mapping. Among the three methods, CF model has highest prediction accuracy than the other two models. Similarly, the outcome of this study reveals that streams, faults, slope and elevation are the main conditioning factors of landslides. Especially, the erosion of streams plays a key role of the landslide occurrence. These landslide susceptibility maps, to some extent, reflect spatial distribution characteristics of landslides in alpine-canyon region of southwest China, and can be used for land planning and hazard risk assessment.
