首页|基于BP神经网络的九寨沟地区地震滑坡危险性预测研究

基于BP神经网络的九寨沟地区地震滑坡危险性预测研究

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
BP神经网络因具有良好的精度和拟合能力,被广泛地运用在区域性滑坡危险性预测中.本文建立了基于BP神经网络的地震滑坡危险性评价模型并应用于四川九寨沟地区,以2017年8月8日的九寨沟Ms7.0地震引发的4834个历史滑坡为例,将其随机划分为70％的训练样本集用于九寨沟地区地震滑坡危险性预测,以及30％的验证样本集对预测结果的精度进行评估.选取高程、坡度、坡向、平行发震断层距离、垂直发震断层距离、震中距离、距道路距离、地面峰值加速度(PG4)以及岩性共9个影响因子,分析发震断层对地震滑坡的控制作用,并总结九寨沟地区地震滑坡空间分布规律特征,其中发震断层、岩性和坡度对九寨沟地区地震滑坡分布产生重要影响.利用模型得到九寨沟地震滑坡危险性预测图,结果显示73.19％的滑坡位于极高和高危险区域,与实际地震滑坡分布基本相符.通过30％的验证样本集来绘制预测成功率曲线,结果表明模型预测成功率(AUC值)为0.90,证实了 BP神经网络在九寨沟地区地震滑坡危险性预测中具有良好的精度和拟合能力,评价结果为后续地震滑坡灾害预测和防震减灾工作提供了科学的参考.
EARTHQUAKE-TRIGGERED LANDSLIDE SUSCEPTIBILITY PREDICTION IN JIUZHAIGOU BASED ON BP NEURAL NETWORK
The BP neural network is widely employed in regional landslide susceptibility prediction due to its ex-cellent nonlinear fitting ability and generalization capability.This paper establishes a landslide susceptibility assess-ment model based on the BP neural network and applies it to Jiuzhaigou,Sichuan Province.The study focuses on 4834 historical landslides caused by the Ms7.0 earthquake in Jiuzhaigou in August 2017.Seventy percent of them are randomly divided into a training sample for landslide susceptibility prediction in Jiuzhaigou,while the remaining 30％form a validation sample set to evaluate the accuracy of the predicted results.Nine influencing factors,inclu-ding elevation,slope,aspect,distance to parallel seismogenic fault,distance to vertical seismogenic fault,distance to the epicenter,distance to the road,peak ground acceleration(PGA),and lithology,were selected to discuss the control effect of seismogenic faults on earthquake-triggered landslides and conduct the correlation analysis of these influencing factors.The study then summarizes the spatial distribution characteristics of earthquake-triggered land-slides.Results indicate that the seismogenic fault,lithology,and slope have a significant influence on the distribu-tion of earthquake-triggered landslides in Jiuzhaigou.The study obtains the prediction map of earthquake-triggered landslide susceptibility in Jiuzhaigou through the model.The results reveal that 73.3％of the landslides are located in the extremely high and high susceptibility areas,which is consistent with the actual distribution of earthquake-triggered landslides.Using the 30％validation sample set to predict the success rate curve,the results show that the model's prediction success rate(AUC)is 0.90,proving that the BP neural network exhibits good accuracy and fit-ting ability in predicting regional landslide susceptibility.The evaluation results provide a reference for future earth-quake-triggered landslide disaster prediction and earthquake prevention and mitigation efforts.

Jiuzhaigou areaBP neural networkEarthquake-triggered landslideHazard assessment

张迎宾、徐佩依、林剑锋、伍新南、柳静、相晨琳、何云勇、杨昌凤、许冲

展开 >

西南交通大学土木工程学院,成都 610031,中国

四川省公路规划勘察设计研究院有限公司,成都 610041,中国

应急管理部国家自然灾害防治研究院,北京 100085,中国

九寨沟地区 BP神经网络 地震滑坡 危险性评价

国家自然科学基金项目科技部、中国科学院第二次青藏高原科考项目四川省科技厅四川省科技厅四川省交通运输科技项目&&

419772132019QZKK09062020YFH00172021YFS03212021-A-03R110121H01092

2024

10.13544/j.cnki.jeg.2022-0013

工程地质学报
中国科学院地质与地球物理研究所

工程地质学报

CSTPCD北大核心
影响因子：1.215
ISSN：1004-9665
年,卷(期)：2024.32(1)
  • 61