基于BP神经网络的九寨沟地区地震滑坡危险性预测研究  

EARTHQUAKE-TRIGGERED LANDSLIDE SUSCEPTIBILITY PREDICTION IN JIUZHAIGOU BASED ON BP NEURAL NETWORK

作　　者：张迎宾[1] 徐佩依 林剑锋 伍新南 柳静 相晨琳 何云勇 杨昌凤 许冲 ZHANG Yingbin;XU Peiyi;LIN Jianfeng;WU Xinnan;LIU Jing;XIANG Chenlin;HEYunyong;YANG Changfeng;XU Chong(School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,China;Sichuan Highway Planning,Survey,Design and Research Institute Co.,Ltd.,Chengdu 610041,China;National Institute of Natural Hazards,Ministry of Emergency Management of China,Beijing 100085,China)

机构地区：[1]西南交通大学土木工程学院,成都610031 [2]四川省公路规划勘察设计研究院有限公司,成都610041 [3]应急管理部国家自然灾害防治研究院,北京100085

出　　处：《工程地质学报》2024年第1期133-145,共13页Journal of Engineering Geology

基　　金：国家自然科学基金项目(资助号:41977213);科技部、中国科学院第二次青藏高原科考项目(资助号:2019QZKK0906);四川省科技厅(资助号:2020YFH0017,2021YFS0321);四川省交通运输科技项目(资助号:2021-A-03);黑龙滩高烈度红层地区边坡工程动力灾变机理及防护关键技术研究(资助号:R110121H01092)。

摘　　要：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(PGA)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。The BP neural network is widely employed in regional landslide susceptibility prediction due to its excellent nonlinear fitting ability and generalization capability.This paper establishes a landslide susceptibility assessment model based on the BP neural network and applies it to Jiuzhaigou,Sichuan Province.The study focuses on 4834 historical landslides caused by the M S7.0 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,including 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 landslides.Results indicate that the seismogenic fault,lithology,and slope have a significant influence on the distribution 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 fitting ability in predicting regional landslide susceptibility.The evaluation results provide a reference for future earthquake-triggered landslide disaster prediction and earthquake prevention and mitigation efforts.

关 键 词：九寨沟地区 BP神经网络 地震滑坡 危险性评价 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] P642.22[自动化与计算机技术—控制科学与工程] P315.9[天文地球—工程地质学]