91 / 2019-07-31 19:51:43

Resistance to Toughness Catastrophe Energy of Slopes along Mountain Highway Based on Random Forest Assessment Model--A Case Study of Maoxian Highway

Highway Line, Slope Catastrophe, Stochastic Forest Model, Evaluation Factor, Geospatial Data

水库地质灾害防治与生态环境保护 > 地质灾害监测预警

On July 17, 2014, a landslide occurred at K774+600m of 213 National Highway in Maoxian County, Aba Prefecture, Sichuan Province, killing more than 30 people and burying the road. The main reason for the frequent occurrence of slope disasters along mountainous roads is that it has insufficient ability to resist deformation and destruction under the influence of natural or human factors. At present, there are few studies on slope toughness and evaluation methods of slope catastrophe resistance. This study is based on GIS and digital simulation to study the toughness of slopes along the main roads in Mao County.
By studying the mechanism of slope toughness, 13 factors such as elevation, slope gradient, slope direction, slope position, micro-geomorphology, curvature, forward and backward slope, normalized vegetation index, lithology, distance from water system, distance from fault, distance from road and average rainfall for many years were selected as evaluation factors. According to the geospatial division on the 30-meter scale, the geospatial information database of slope catastrophe resistance evaluation is constructed together with the historical slope disaster along the highway, and the database is randomly divided into training data set and verification data set according to 7:3. In the training data set, spatial grids with no historical disasters and historical disasters was randomly selected to construct a training sample with 1200 pieces of data. In order to reduce the accidental error, five groups of data were selected. The random forest method is used to train and model the sample data, and the obtained model is used to predict and analyze the training data set, the verification data set and the whole research area. Finally, the accuracy of the model is verified by statistical methods and ROC curves.
The results show that the disastrous resistance of the slope along the highway can be divided into extremely low, low, medium, high and extremely high, and its area accounts for 4.3%, 8.9%, 12.2%, 28.6% and 46.1% respectively, while the corresponding historical disasters account for 78.8%, 9.6%, 5.1%, 6.2% and 0.2% respectively. the AUC values of training data set, validation data set and the whole research area are 1.000, 0.884 and 0.956%, respectively. At the same time, the recent typical slope disasters are all in the low disaster resistance area. The stochastic forest model has good reliability for evaluating the slope catastrophic resistance along mountain highway.