广西师范大学学报（自然科学版） ›› 2026, Vol. 44 ›› Issue (6): 54-68.doi: 10.16088/j.issn.1001-6600.2026033102

• 物理与电子工程 • 上一篇    下一篇

联合物理信息与深度学习的滑坡易发性评价方法

黄泽华1,2,3, 李泽龙2, 周绍园2, 邱芹军4, 龚秉生5, 梁栋6, 马凯1,2*   

  1. 1.水电工程智能视觉监测湖北省重点实验室(三峡大学), 湖北 宜昌 443002;
    2.三峡大学 计算机与信息学院, 湖北 宜昌 443002;
    3.长江水利委员会网信中心, 湖北 武汉 430010;
    4.中国地质大学(武汉) 计算机与信息学院, 湖北 武汉 430010;
    5.长江水利委员会水文局长江中游水文水资源勘测局, 湖北 武汉 430010;
    6.水利部长江勘测技术研究所, 湖北 武汉 430010
  • 收稿日期:2026-03-31 修回日期:2026-05-28 出版日期:2026-11-05 发布日期:2026-11-10
  • 通讯作者: 马凯(1980—),男,河南南阳人,三峡大学教授,博士。E-mail: makai@ctgu.edu.cn
  • 基金资助:
    国家自然科学基金(42302337, 42301492)

A physics-informed deep learning approach for landslide susceptibility assessment

Huang Zehua1,2,3, Li Zelong2, Zhou Shaoyuan2, Qiu Qinjun4, Gong Bingsheng5, Liang Dong6, Ma Kai1,2*   

  1. 1. Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering, China Three Gorges University, Yichang Hubei 443002, China;
    2. College of Computer and Information Technology, China Three Gorges University, Yichang Hubei 443002, China;
    3. Information Research Centre, Changjiang Water Resources Commission, Wuhan Hubei 430010, China;
    4. School of Computer Science, China University of Geosciences (Wuhan), Wuhan Hubei 430074, China;
    5. Hydrological and Water Resources Survey Bureau of the Middle Reaches of the Yangtze River, CJWRC, Wuhan Hubei 430000, China;
    6. Changjiang Research Institute of Geotechnique and Survey, MRW, Wuhan Hubei 430011, China
  • Received:2026-03-31 Revised:2026-05-28 Online:2026-11-05 Published:2026-11-10

摘要： 滑坡易发性评价是滑坡风险预警的重要手段。针对评价过程中对环境因子物理作用和内在关联信息提取应用能力不足、难以有效融合多元因子精确预测的问题,本文提出一种联合物理信息与深度学习的滑坡易发性评价方法。首先,根据环境因子在滑坡中的物理作用将其分为自然条件和驱动因素;然后,基于深度学习方法设计自然条件评价模块和驱动因素评价模块,分别评估潜在滑坡风险和外界驱动强度;最后,结合因子顺序、因果关系融合评估结果,精确预测滑坡概率。结果表明:新方法的AUC(0.898 4)、总体精度(0.829 7)、准确率(0.795 8)和召回率(0.863 6)显著优于7种常用方法;方法内自注意力机制能够准确识别滑坡的关键驱动因素,为区域滑坡易发性评价提供精细指导。

关键词: 滑坡易发性评价, 深度学习, Transformer, 自注意力机制, 物理信息

Abstract: Landslide susceptibility assessment is an important tool for landslide risk early warning. To address the issue of insufficient capability in extracting and utilizing the physical effects and intrinsic correlations of environmental factors during the assessment process, which makes it difficult to effectively integrate multiple factors for accurate prediction, this study proposes a landslide susceptibility assessment method that combines physics-informed information with deep learning. First, environmental factors are divided into natural conditions and driving factors based on their physical roles in landslides. Then, based on deep learning methods, a natural condition assessment module and a driving factor assessment module are designed to evaluate potential landslide risks and external driving intensity, respectively. Finally, the assessment results are integrated by considering the sequence and causal relationships of the factors to accurately predict landslide probability. The results show that the new method significantly outperforms seven commonly used methods in terms of AUC (0.898 4), overall accuracy (0.829 7), precision (0.795 8), and recall(0.863 6). The self-attention mechanism within the method can accurately identify the key driving factors of landslides, providing precise guidance for disaster prevention and mitigation efforts.

Key words: landslide susceptibility assessment, deep learning, transformer, self-attention mechanism, physical information

中图分类号:  P694

[1] Chen B, Li Z H, Song C, et al. Unveiling the long-term cascading effects of the 2018 Baige landslide and subsequent outburst flood with satellite radar observations[J]. Remote Sensing of Environment, 2026, 334: 115231. DOI:10.1016/j.rse.2026.115231.
[2] 中华人民共和国国家统计局.2025中国统计年鉴[M]. 北京: 中国统计出版社, 2025.
[3] Ng C W W, Pan T L, Ma P F. A novel deep learning model for landslide mapping using cross-resolution change technique[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2025, 229: 254-269. DOI:10.1016/j.isprsjprs.2025.08.004.
[4] Kang Y, Lu Z, Zhao C Y, et al. An extended vector inclination method for inferring detailed slip surfaces beneath landslides from SAR and optical satellite remote sensing image[J]. Remote Sensing of Environment, 2026, 334: 115225. DOI:10.1016/j.rse.2025.115225.
[5] Reichenbach P, Rossi M, Malamud B D, et al. A review of statistically-based landslide susceptibility models[J]. Earth-Science Reviews, 2018, 180: 60-91. DOI:10.1016/j.earscirev.2018.03.001.
[6] Myronidis D, Papageorgiou C, Theophanous S. Landslide susceptibility mapping based on landslide history and analytic hierarchy process (AHP)[J]. Natural Hazards, 2016, 81(1): 245-263. DOI:10.1007/s11069-015-2075-1.
[7] Ling X, Ming D P, Zhang W H, et al. White-boxed landslide susceptibility mapping using unsupervised causal model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 4508613. DOI:10.1109/TGRS.2025.3605274.
[8] Lv L, Chen T, Liu G, et al. A comparative study of model interpretability considering the decision differentiation of landslide susceptibility models[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 4401218. DOI:10.1109/TGRS.2024.3522165.
[9] Arabameri A, Saha S, Roy J, et al. Landslide susceptibility evaluation and management using different machine learning methods in the gallicash river watershed, Iran[J]. Remote Sensing, 2020, 12(3): 475. DOI:10.3390/rs12030475.
[10]Amatya P, Kirschbaum D, Stanley T. Use of very high-resolution optical data for landslide mapping and susceptibility analysis along the karnali highway, Nepal[J]. Remote Sensing, 2019, 11(19): 2284. DOI:10.3390/rs11192284.
[11]Huang Y, Zhao L. Review on landslide susceptibility mapping using support vector machines[J]. Catena, 2018, 165: 520-529. DOI:10.1016/j.catena.2018.03.003.
[12]Liu Z X, Zhang Z J, Zhang W C, et al. InSAR-based landslide susceptibility analysis in karst erosion landforms using a non-landslide sampling strategy[J]. International Journal of Applied Earth Observation and Geoinformation, 2025, 145: 104997. DOI:10.1016/j.jag.2025.104997.
[13]Liu B, Li D R, Xiao X W, et al. Enhancing Sentinel-2 landslide change detection by integrating multispectral, deformation, and topographic information[J]. International Journal of Applied Earth Observation and Geoinformation, 2026, 146: 105116. DOI:10.1016/j.jag.2026.105116.
[14]Guo W X, Ye J, Liu C B, et al. An approach for predicting landslide susceptibility and evaluating predisposing factors[J]. International Journal of Applied Earth Observation and Geoinformation, 2024, 135: 104217. DOI:10.1016/j.jag.2024.104217.
[15]Jamaluddin I, Chen Y N, Fan K C. Spatial-spectral-temporal deep regression model with convolutional long short-term memory and transformer for the large-area mapping of mangrove canopy height by using sentinel-1 and sentinel-2 data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 1-17. DOI:10.1109/tgrs.2024.3362788.
[16]Jiang S G, Li N Y, Xu M, et al. SQformer: spectral-query transformer for hyperspectral image arbitrary-scale super-resolution[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 5532315. DOI:10.1109/TGRS.2024.3463745.
[17]Park J Y, Saad O M, Oh J W, et al. Transformer-based seismic image enhancement: a novel approach for improved resolution[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 5901210. DOI:10.1109/TGRS.2024.3510863.
[18]叶润青, 付小林, 郭飞, 等. 三峡水库运行期地质灾害变形特征及机制分析[J]. 工程地质学报, 2021, 29(3): 680-692. DOI:10.13544/j.cnki.jeg.2021-0191.
[19]姬建, 崔红志, 佟斌, 等. 基于物理过程不确定性的降雨诱发浅层滑坡易发性快速区划: GIS-FORM技术开发与应用[J]. 岩石力学与工程学报, 2024, 43(4): 838-850. DOI:10.13722/j.cnki.jrme.2023.0724.
[20]黄发明, 叶舟, 姚池, 等. 滑坡易发性预测不确定性: 环境因子不同属性区间划分和不同数据驱动模型的影响[J]. 地球科学, 2020, 45(12): 4535-4549.
[21]杜鹏, 陈宁生, 伍康林, 等. 基于随机森林模型的藏东南地区滑坡易发性评价及主控因素分析[J]. 成都理工大学学报(自然科学版), 2024, 51(2): 328-344. DOI:10.3969/j.issn.1671-9727.2024.02.12.
[22]毛成忠, 肖湛臻, 杨志彪, 等. 地理经纬双向过渡带和复杂地形下气候的过渡性和特异性: 以宜昌市为例[J]. 中低纬山地气象, 2023, 47(1): 31-37. DOI:10.3969/j.issn.1003-6598.2023.01.005.
[23]陈剑, 李晓, 杨志法. 三峡库区滑坡的时空分布特征与成因探讨[J]. 工程地质学报, 2005, 13(3): 305-309. DOI:10.3969/j.issn.1004-9665.2005.03.004.
[24]傅角今. 湖南地理志[M]. 雷树德, 校点. 长沙: 湖南教育出版社, 2008.
[25]尹蓉, 李雨欣, 周明薇, 等. 基于GIS的湖南柑橘种植气候适宜性区划研究[J]. 农业科学, 2025, 15(12): 1456-1462. DOI:10.12677/hjas.2025.1512182.
[26]张磊, 宫兆宁, 王启为, 等. Sentinel-2影像多特征优选的黄河三角洲湿地信息提取[J]. 遥感学报, 2019, 23(2): 313-326.
[27]GOOGLE. Google earth engine[EB/OL].[2026-06-05]. https://code.earthengine.google.com/.
[28]地理空间数据云. 地理空间数据云[EB/OL].[2026-06-05]. https://www.gscloud.cn/search.
[29]熊立华, 郭生练, Kieran M.O Connor. 利用DEM提取地貌指数的方法述评[J]. 水科学进展, 2002, 13(6): 775-780. DOI:10.14042/j.cnki.32.1309.2002.06.021.
[30]中国科学院南京土壤所. 中国土壤科学数据库[DB/OL].中国科学院南京土壤所[2026-06-05]. http://vdb3.soil.csdb.cn/extend/jsp/index_soiltype.
[31]冉有华. 中国土地覆盖数据集[DS/OL].国家青藏高原科学数据中心[2026-06-05].https://www.tpdc.ac.cn/zh-hans/data/f1aaacad-9f42-474e-8aa4-d37f37d6482f/.
[32]USGS. CHIRPS: rainfall estimates from rain gauge and satellite observations[DS/OL].Climate Hazards Center[2026-06-05]. https://www.chc.ucsb.edu/data/chirps.
[33]OSM. Road data[DS/OL]. OSM[2026-06-05]. https://welcome.openstreetmap.org/.
[34]尚敏, 张惠强, 廖芬, 等. 考虑降雨滞后效应的八字门滑坡位移预测研究[J]. 自然灾害学报, 2022, 31(3): 242-250. DOI:10.13577/j.jnd.2022.0325.
[35]刘传正, 崔原, 陈春利, 等. 辽宁抚顺西露天矿南帮滑坡成因[J]. 地质通报, 2022, 41(5): 713-726. DOI:10.12097/j.issn.1671-2552.2022.05.001.
[36]卢书强, 易庆林, 易武, 等. 库水下降作用下滑坡动态变形机理分析: 以三峡库区白水河滑坡为例[J]. 工程地质学报, 2014, 22(5): 869-875. DOI:10.13544/j.cnki.jeg.2014.05.015.
[37]古腾飞, 段平, 王伟鹏, 等. 滑坡易发性评价方法综述[J]. 自然灾害学报, 2024, 33(6): 1-16. DOI:10.13577/j.jnd.2024.0601.
[38]刘纪平, 梁恩婕, 徐胜华, 等. 顾及样本优化选择的多核支持向量机滑坡灾害易发性分析评价[J]. 测绘学报, 2022, 51(10): 2034-2045.
[39]Fu Y B, Li W M, Fan S S, et al. CAL-net: conditional attention lightweight network for in-orbit landslide detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 4408515. DOI:10.1109/TGRS.2023.3321716.
[40]张玘恺, 凌斯祥, 李晓宁, 等. 九寨沟县滑坡灾害易发性快速评估模型对比研究[J]. 岩石力学与工程学报, 2020, 39(8): 1595-1610. DOI:10.13722/j.cnki.jrme.2020.0029.
[41]李文彬, 范宣梅, 黄发明, 等. 不同环境因子联接和预测模型的滑坡易发性建模不确定性[J]. 地球科学, 2021, 46(10): 3777-3795.
[42]田乃满, 兰恒星, 伍宇明, 等. 人工神经网络和决策树模型在滑坡易发性分析中的性能对比[J]. 地球信息科学学报, 2020, 22(12): 2304-2316.
[43]Kavzoglu T, Teke A, Yilmaz E O. Shared blocks-based ensemble deep learning for shallow landslide susceptibility mapping[J]. Remote Sensing, 2021, 13(23): 4776. DOI:10.3390/rs13234776.
[44]Cai J N, Ming D P, Liu F, et al. An enhanced spatiotemporal prediction method on landslide displacement with LDP-ConvFormer and MT-InSAR observations[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2026, 232: 594-612. DOI:10.1016/j.isprsjprs.2025.12.027.
[45]安雪莲, 密长林, 孙德亮, 等. 基于不同评价单元的三峡库区滑坡易发性对比: 以重庆市云阳县为例[J]. 吉林大学学报(地球科学版), 2024, 54(5): 1629-1644. DOI:10.13278/j.cnki.jjuese.20230071.
[46]Li Y C, Chen T, Lv L, et al. IED-GCN: an internal and external decoupled graph convolutional network for landslide susceptibility assessment[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 4414717. DOI:10.1109/TGRS.2025.3595205.
[47]滕超, 王雷, 刘宝华, 等. 辽宁抚顺西露天矿南帮滑坡应力变化规律及影响因素分析[J]. 中国地质灾害与防治学报, 2018, 29(2): 35-42. DOI:10.16031/j.cnki.issn.1003-8035.2018.02.06.
[1] 何荣泽, 王庆祝, 夏道勋, 周晓军. 基于傅里叶特征映射的物理信息残差网络求解高阶非线性偏微分方程[J]. 广西师范大学学报（自然科学版）, 2026, 44(6): 134-145.
[2] 唐程华, 易见兵, 吴欣, 熊文武, 王敬永. 跨域少样本图像语义分割方法综述[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 1-27.
[3] 王成龙, 宋强, 李文峰, 张仕民. PAM-DETR: 基于改进RT-DETR的医用手套小目标缺陷检测算法[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 79-95.
[4] 王燕, 徐婕, 牛梦圆. 自适应归一化的多尺度水下图像增强网络[J]. 广西师范大学学报（自然科学版）, 2026, 44(3): 89-106.
[5] 田晟, 冯帅涛, 李嘉. 一种基于复合框架的城市道路场景车辆轨迹提取方法[J]. 广西师范大学学报（自然科学版）, 2026, 44(2): 31-51.
[6] 吕辉, 司可. 基于改进RT-DETR的光伏板缺陷检测[J]. 广西师范大学学报（自然科学版）, 2026, 44(2): 52-64.
[7] 王旭阳, 梁宇航. 多尺度非对称注意力遥感去雾Transformer[J]. 广西师范大学学报（自然科学版）, 2026, 44(2): 77-89.
[8] 罗缘, 朱文忠, 王文, 吴宇浩. 基于改进PatchTST的多步水质预测模型[J]. 广西师范大学学报（自然科学版）, 2026, 44(2): 115-131.
[9] 谢翔, 江林峰, 杨凤莲. 基于自适应权重两阶段PINNs方法求解偏微分方程[J]. 广西师范大学学报（自然科学版）, 2026, 44(2): 175-189.
[10] 黄艳国, 肖洁, 吴水清. 基于D2STGNN的双向高效多尺度交通流预测[J]. 广西师范大学学报（自然科学版）, 2026, 44(1): 10-22.
[11] 韩华彬, 高丙朋, 蔡鑫, 孙凯. 基于HO-CNN-BiLSTM-Transformer模型的风机叶片结冰故障诊断[J]. 广西师范大学学报（自然科学版）, 2025, 43(6): 13-28.
[12] 魏梓书, 陈志刚, 王衍学, 哈斯铁尔·马德提汗. 基于SBSI-YOLO11的轻量化轴承外观缺陷检测算法[J]. 广西师范大学学报（自然科学版）, 2025, 43(6): 80-91.
[13] 刘廷汉, 梁艳, 黄鹏升, 闭金杰, 黄守麟, 李廷会. 基于改进YOLOv8s的人脸痤疮小目标检测[J]. 广西师范大学学报（自然科学版）, 2025, 43(5): 114-129.
[14] 易见兵, 张裕贤, 曹锋, 李俊, 彭鑫, 陈鑫. 基于时空注意力的3D人体姿态估计网络设计[J]. 广西师范大学学报（自然科学版）, 2025, 43(5): 130-144.
[15] 黎宗孝, 张健, 罗鑫悦, 赵嶷飞, 卢飞. 基于K-means和Adam-LSTM的机场进场航迹预测研究[J]. 广西师范大学学报（自然科学版）, 2025, 43(4): 15-23.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 广西师范大学学报（自然科学版）编辑部
地址：广西桂林市三里店育才路15号　邮编：541004
电话：0773-5857325　E-mail: gxsdzkb@mailbox.gxnu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发