Landslide hazard evaluation based on BPNN-SHAP model: A case study of the Yili River Basin, Xinjiang Province
-
摘要:
为进一步提高滑坡危险性预测模型精度、增强模型可解释性，本文以新疆伊犁河流域为研究区，选取8个影响滑坡发生的危险性因子，在反向传播神经网络（BPNN）基础上，借鉴博弈论思想，构建一种可解释BP神经网络模型（BPNN-SHAP），解决神经网络滑坡危险性评价的“黑箱”问题。将数据集分为70%训练集和30%测试集，采用5折交叉验证提高模型稳定性，对比深度神经网络（DNN）、随机森林（RF）和逻辑回归（LR）3个模型的评价精度，并探讨BPNN-SHAP预测结果的可解释性，完成区域滑坡危险性评价。研究结果表明：相较于其他模型，BPNN-SHAP模型的5个精度评价指标均为最高，分别是：准确率（A）=0.904、精准度（P）=0.911、召回率（R）=0.919、F1分数（F1Score）=0.915、曲线下面积（SAUC）=0.901；研究区滑坡极高、高危险区分别占比11.96%、15.53%，其中新源县和巩留县极高、高危险区占比最高，分别为51.1%、45.6%；滑坡主控因子为高程、坡度、降雨量和峰值地面加速度（PGA），定量揭示高程在
1500 ～2000 m、坡度大于14°、年降雨量在260～310 mm、PGA大于0.23 g的区域对滑坡发生起促进作用，表明该区域滑坡可能为高程和坡度主控的降雨型、地震型滑坡。本研究方法可为滑坡危险性评价提供新的技术参考，为伊犁河流域防灾减灾韧性建设提供理论支撑。Abstract:To further improve the accuracy of landslide hazard prediction models and enhance their interpretability, this study selected 8 influencing factors of landslide occurrence, taking the Yili River Basin, Xinjiang province as an example. An interpretable BPNN-SHAP model, based on the back propagation neural network (BPNN) model and the game theory with the aim of addressing the 'black box' issue, was constructed. Firstly, the dataset was divided into 70% training set and 30% test set, and 5-fold cross-validation was used to enhance the robustness of the BPNN-SHAP model. Then, the evaluation accuracy of this model was compared with three other models: Deep Neural Network (DNN), Random Forest (RF), and Logistic Regression (LR). Finally, regional landslide hazard assessment was completed, and the interpretability of BPNN-SHAP was also discussed. The results showed that the BPNN-SHAP model achieved the highest statistical values in the following metrics: Accuracy (A)=0.904, Precision (P)=0.911, Recall (R)=0.919, F1Score=0.915, and SAUC=0.905. The very high and high danger areas for landslides in the study region accounted for 11.96% and 15.53%, respectively. Among these regions, Xinyuan and Nileke County occupy the highest proportions, at approximately 51.1% and 45.6%, respectively. The primary controlling factors for landslides were elevation, slope, rainfall, and peak ground acceleration (PGA). Specifically, areas with an elevation of
1500 m to2000 m, slopes greater than 14°, annual rainfall between 260 mm and 310 mm, and PGA greater than 0.23 g are prone to landslides, indicating that the predominant types of landslides are rainfall-induced and earthquake-induced. Our research method is expected to provide a new technical reference for landslide hazard assessment and theoretical support for disaster prevention, mitigation, and resilience construction in the Yili River Basin. -
-
表 1 影响因子数据来源
Table 1. Sources of data on influence factors
影响因子 数据来源 土地利用类型 https://livingatlas.arcgis.com/landcover/ PGA https://zenodo.org/ 高程、坡度、距河流距离 https://www.gscloud.cn/ 年均降雨量 https://climate.copernicus.eu/climate-reanalysis 距断层距离、工程地质岩组 新疆维吾尔自治区自然资源档案馆（ http://zrzyt.xinjiang.gov.cn ）
下载: 导出CSV
表 2 模型参数取值表
Table 2. Model parameter values
参数名 优化器 批处理大小 学习率 激活函数 单次训练迭代次数 取值 Adam 8 0.001 隐藏层(ReLU)，输出层(Sigmoid) 50
下载: 导出CSV
表 3 模型性能对比
Table 3. Comparison of model performance
模型 A P R F1Score SAUC LR 0.852 0.858 0.869 0.863 0.809 RF 0.874 0.877 0.877 0.877 0.847 DNN 0.891 0.902 0.914 0.908 0.888 BPNN（本文） 0.904 0.911 0.919 0.915 0.901
下载: 导出CSV
表 4 各分区危险性灾害分布表
Table 4. Distribution of landslide hazard areas
危险性分区 分区滑坡数量 分区滑坡数量占比 分区面积占比 极低危险区 24 2.01% 26.67% 低危险区 76 6.34% 28.89% 中危险区 125 10.43% 16.95% 高危险区 461 38.48% 15.53% 极高危险区 512 42.74% 11.96%
下载: 导出CSV
-
[1] 艾力哈木·艾克拉木,2022. 伊犁河流域平原区地下水水质特征及其形成机理研究[D]. 新疆:新疆农业大学.
Ailhamu A,2022. Characteristics of groundwater quality and its formation mechanism in the plain area of Yili River Basin [D]. Xinjiang:Xinjiang Agricultural University (in Chinese with English abstract).
[2] 陈绪钰,李明辉,王德伟,等,2019. 基于GIS和信息量法的四川峨眉山市地质灾害易发性定量评价[J]. 沉积与特提斯地质, 39(4):100 − 112.
Chen X Y,Li M H,Wang D W,et al.,2019. Quantitative evaluation of geological hazard vulnerability in Emeishan City,Sichuan Province based on GIS and information method[J]. Sedimentary Geology and Tethyan Geology, 39(4):100 − 112 (in Chinese with English abstract).
[3] Collini E,et al.,2022. Predicting and understanding landslide events with explainable AI[J]. IEEE Access,10:31175 − 31189. doi: 10.1109/ACCESS.2022.3158328
[4] Dai F C,Lee C F,2002. Landslide characteristics and slope instability modeling using GIS,Lantau Island,Hong Kong[J]. Geomorphology,42(3-4):213 − 228. doi: 10.1016/S0169-555X(01)00087-3
[5] 范贺娟,2023. 天山野果林区大小莫合流域山体滑坡灾害生态风险评价[D]. 新疆:新疆师范大学.
Fan H J,2023. Ecological risk assessment of landslide disaster in the Big and small Mohe river area of wild fruit forest in Tianshan Mountains [D]. Xinjiang:Xinjiang Normal University (in Chinese with English abstract).
[6] 傅贵,2021. 伊犁某典型黄土区滑坡易发性评价研究[D]. 安徽:安徽理工大学.
Fu G,2021. Evaluation of landslide susceptibility in a typical loess area of Yili [D]. Anhui:Anhui University of Science and Technology (in Chinese with English abstract).
[7] 弓小平,王正刚,马宏兵,等,2018. 新疆伊犁谷地地质灾害成因及评价研究[M]. 北京:地质出版社.
Gong X P,Wang Z G,Ma H B,et al.,2018. Study on causes and evaluation of geological hazards in Yili Valley,Xinjiang [M]. Beijing:Geological Publishing House (in Chinese with English abstract).
[8] 胡杨,张紫昭,林世河,2023. 基于证据权与逻辑回归耦合的新疆伊犁河谷地区滑坡易发性评价[J]. 工程地质学报,31(4):1350 − 1363.
Hu Y,Zhang Z Z,Lin S H,2023. Evaluation of landslide susceptibility in the Yili Valley region of Xinjiang based on the coupling of right-of-evidence and logistic regression[J]. Journal of Engineering Geology,31(4):1350 − 1363 (in Chinese with English abstract).
[9] Jibson R W,2011. Methods for assessing the stability of slopes during earthquakes—A retrospective[J]. Engineering Geology,122(1-2):43 − 50. doi: 10.1016/j.enggeo.2010.09.017
[10] 李浩,2024. 基于多模融合和数据异常检测的滑坡预测和预警[D]. 江苏:中国矿业大学.
Li H,2024. Landslide prediction and early warning based on multimode fusion and data anomaly detection [D]. Jiangsu:China University of Mining and Technology (in Chinese with English abstract).
[11] 李俊峰,张小琼,马滔,等,2023. 基于XGBoost和SHAP的可解释性滑坡位移预测模型[J/OL]. 工程地质学报:1 − 16.
Li J F,Zhang S Q,Ma T,et al.,2023. Interpretable landslide displacement prediction model based on XGBoost and SHAP[J/OL]. Journal of Engineering Geology:1 − 16 (in Chinese with English abstract).
[12] 李帅,陈建波,姚远,等,2021. 基于GIS的地震滑坡危险性分析研究——以伊犁地区为例[J]. 内陆地震, 35(1):38 − 47.
Li S,Chen J B,Yao Y,et al.,2021. Research on GIS-based seismic landslide hazard analysis-Taking Yili area as an example[J]. Inland Earthquake, 35(1):38 − 47 (in Chinese with English abstract).
[13] 李扬,2023. 基于随机森林模型解释的越野路面识别算法研究[D]. 吉林:吉林大学.
Li Y,2023. Research on off-road pavement recognition algorithm based on random forest model interpretation[D]. Jilin:Jilin University (in Chinese with English abstract).
[14] Lian L,Yu M Z,Xiu J L,2023. RMDGCN:Prediction of RNA methylation and disease associations based on graph convolutional network with attention mechanism[J]. Plos Computational Biology,19(12):e1011677 − e1011677. doi: 10.1371/journal.pcbi.1011677
[15] 梁世川,乔华,吕东,等,2023. 伊犁谷地地质灾害分布特征及主控因素分析[J]. 干旱区地理,46(6):880 − 888.
Liang S C,Qiao H,Lu D,et al.,2023. Distribution characteristics of geologic hazards in the Yili Valley and analysis of the main controlling factors[J]. Arid Zone Geography,46(6):880 − 888 (in Chinese with English abstract).
[16] 刘冰,2012. 公布遇难者名单是对生命的尊重[N]. 新疆:新疆日报(汉).
Liu B,2012. Publishing the List of Victims is a Respect for Life[N]. Xinjiang:Xinjiang Daily(Chinese) (in Chinese with English abstract).
[17] 刘任鸿,李明辉,邓英尔,等,2021. 基于GIS的华蓥市地质灾害易发性评价[J]. 沉积与特提斯地质, 41(1):129 − 136.
Liu R H,Li M H,Deng Y R,et al.,2021. GIS-based assessment of geological hazard susceptibility in Huaying City[J]. Sedimentary Geology and Tethyan Geology, 41(1):129 − 136 (in Chinese with English abstract).
[18] Lundberg S M,Erion G,Chen H,et al.,2020. From local explanations to global understanding with explainable AI for trees[J]. Nature Machine Intelligence,2(1):56 − 67. doi: 10.1038/s42256-019-0138-9
[19] Mitchell R,Frank E,Holmes G,2022. GPUTreeShap:Massively parallel exact calculation of SHAP scores for tree ensembles[J]. PeerJ Computer Science,8:e880.
[20] Molnar C,2020. Interpretable machine learning [M]. Raleigh:Independently Published.
[21] Ribeiro M,Singh S,Guestrin C,2016. Why should I trust you? Explaining the predictions of any classifier [C]. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics:Demonstrations,97 − 101.
[22] 孙星,2023. 断层影响下中深孔采场围岩不连续变形特征及稳定性评价[J]. 中国矿业,32(6):113 − 122. doi: 10.12075/j.issn.1004-4051.20230278
Sun X,2023. Characteristics of discontinuous deformation and stability evaluation of the peripheral rock of medium-deep hole quarry under the influence of fault[J]. China Mining Industry,32(6):113 − 122 (in Chinese with English abstract). doi: 10.12075/j.issn.1004-4051.20230278
[23] Tao K,Wang L,Qian X,2016. Multi-factor constrained analysis method for geological hazard risk[J]. International Journal of Engineering and Technology,8(3):198. doi: 10.7763/IJET.2016.V8.884
[24] 陶妍,角媛梅,丁银平,等,2023. 降雨诱发型滑坡的降雨阈值及机理研究进展与展望[J]. 云南师范大学学报:自然科学版,43(4):71 − 78.
Tao Y,Jiao Y M,Ding Y P,et al.,2023. Progress and prospects of rainfall threshold and mechanism of rainfall-induced landslides[J]. Journal of Yunnan Normal University (Natural Science Edition),43(4):71 − 78 (in Chinese with English abstract).
[25] 铁永波,葛华,高延超,等,2022. 二十世纪以来西南地区地质灾害研究历程与展望[J]. 沉积与特提斯地质,42(4):653 − 665.
Tie Y B,Ge H,Gao Y C,et al.,2022. The research progress and prospect of geological hazards in Southwest China since the 20th Century[J]. Sedimentary Geology and Tethyan Geology,42(4):653 − 665 (in Chinese with English abstract).
[26] Tien B D,Tuan T A,Klempe H,et al.,2016. Spatial prediction models for shallow landslide hazards:A comparative assessment of the efficacy of support vector machines,artificial neural networks,kernel logistic regression,and logistic model tree[J]. Landslides,13:361 − 378. doi: 10.1007/s10346-015-0557-6
[27] 王存智,张炜,李晨冬,等,2022. 基于GIS和层次分析法的沙溪流域滑坡地质灾害易发性评价[J]. 中国地质调查,9(5):51 − 60.
Wang C Z,Zhang W,Li C D,et al.,2022. Evaluation of landslide geohazard susceptibility in Shaxi watershed based on GIS and hierarchical analysis[J]. Chinese Geological Survey,9(5):51 − 60 (in Chinese with English abstract).
[28] 王家柱,高延超,铁永波,等,2023. 基于斜坡单元的山区城镇滑坡灾害易发性评价:以康定为例[J]. 沉积与特提斯地质,43(3):640 − 650.
Wang J Z,Gao Y C,Tie Y B,et al.,2023. Landslide susceptibility assessment based on slope units of mountainous cities and towns:A case study of Kangding city[J]. Sedimentary Geology and Tethyan Geology,43(3):640 − 650 (in Chinese with English abstract).
[29] 王鑫盈,马超,吕立群,等,2024. 黄土高原不同土地利用类型区浅层滑坡侵蚀特征分析——以蔡家川滑坡事件为例[J/OL]. 干旱区研究:1 − 10.
Wang X Y,Ma C,Lü L Q,et al.,2024. Characterization of shallow landslide erosion in different land-use types of the Loess Plateau--A case study of the Caijiachuan landslide[J/OL]. Arid Zone Research:1 − 10 (in Chinese with English abstract).
[30] 王颖慧,丁建丽,李晓航,等,2022. 伊犁河流域土地利用/覆被变化对生态系统服务价值的影响——基于强度分析模型[J]. 生态学报,42(8):3106 − 3118.
Wang Y H,Ding J L,Li X H,et al.,2022. Impacts of land use/cover change on ecosystem service values in the Yili River Basin - based on an intensity analysis model[J]. Journal of Ecology,42(8):3106 − 3118 (in Chinese with English abstract).
[31] 王娅美,张紫昭,张艳阳,等,2023. 基于多种组合模型的新疆巩留县滑坡危险性评价研究[J]. 工程地质学报,31(4):1375 − 1393.
Wang Y M,Zhang Z Z,Zhang Y Y,et al.,2023. Study on landslide hazard evaluation in Gongliu County,Xinjiang based on multiple combination models[J]. Journal of Engineering Geology,31(4):1375 − 1393 (in Chinese with English abstract).
[32] 肖婷,2021. 三峡库区万州区及重点库岸段滑坡灾害风险评价[D]. 武汉:中国地质大学(武汉).
Xiao T,2021. Landslide Disaster Risk Assessment of Wanzhou District and Key Bank Sections in Three Gorges Reservoir Area[D]. Wuhan:China University of Geosciences(Wuhan) (in Chinese with English abstract).
[33] 杨杏丽,2021. 分类学习算法的性能度量指标综述[J]. 计算机科学,48(8):209 − 219. doi: 10.11896/jsjkx.200900216
Yang X L,2021. A review of performance metrics for classification learning algorithms[J]. Computer Science,48(8):209 − 219 (in Chinese with English abstract). doi: 10.11896/jsjkx.200900216
[34] 曾韬睿,王林峰,张俞,等,2024. 基于CatBoost-SHAP模型的滑坡易发性建模及可解释性[J]. 中国地质灾害与防治学报,35(1):37 − 50.
Zeng T R,Wang L F,Zhang Y,et al.,2024. Landslide susceptibility modeling and interpretability based on CatBoost-SHAP model[J]. China Journal of Geological Hazards and Prevention,35(1):37 − 50 (in Chinese with English abstract).
[35] 张迎宾,徐佩依,林剑锋,等,2024. 基于BP神经网络的九寨沟地区地震滑坡危险性预测研究[J]. 工程地质学报,32(1):133 − 145.
Zhang Y B,Xu P Y,Lin J F,et al.,2024. Seismic landslide hazard prediction in Jiuzhaigou area based on BP neural network[J]. Journal of Engineering Geology,32(1):133 − 145 (in Chinese with English abstract).
[36] 周昌,黄顺,2023. 新疆伊犁黄土工程地质特征及致灾机理研究综述[J]. 工程地质学报,31(4):1247 − 1260.
Zhou C,Huang S,2023. A review of engineering geological characteristics and disaster mechanism of loess in Yili,Xinjiang[J]. Journal of Engineering Geology,31(4):1247 − 1260 (in Chinese with English abstract).
[37] 周超,甘露露,王悦,等,2023. 综合非滑坡样本选取指数与异质集成机器学习的区域滑坡易发性建模[J]. 地球信息科学学报, 25(8):1570 − 1585.
Zhou C,Gan L L,Wang Y,et al.,2023. Regional landslide susceptibility modeling based on non-landslide sample selection index and heterogeneous integrated machine learning[J]. Journal of Geoinformation Science, 25(8):1570 − 1585 (in Chinese with English abstract).
[38] 周新植,2023. 滑坡易发性机器学习优化模型及可解释性研究[D]. 重庆:重庆大学.
Zhou X Z,2023. Landslide susceptibility machine learning optimization model and interpretability study[D]. Chongqing:Chongqing University (in Chinese with English abstract).
-
图(11)
表(4)
计量
- 文章访问数: 398
- PDF下载数: 195
- 施引文献: 0