山区降雨型滑坡定量化预警方法研究

Research on quantitative early warning method for rainfall-induced landslide in mountainous areas

摘要

为提高降雨型滑坡预警精度，以陕西省略阳－宁强县为研究区，联合贝叶斯公式和机器学习方法，构建了滑坡灾害定量化预警模型(quantitative landslide early warning model，QLEWM)。首先，采用统一标准识别致灾与非致灾降雨事件，采用信息增益确定最优降雨变量组合和超参数，并联合贝叶斯公式和三阶逻辑回归，拟合得到区域降雨致灾概率模型(时间概率模型)。然后，以高危险性边坡为样本，以9个孕灾因子为特征训练随机森林模型，得到了滑坡空间概率分布。最后，采用贝叶斯公式耦合滑坡时空概率得到QLEWM，对2018-2023年的雨季进行逐日模拟预警，并与传统启发式预警模型(heuristic based landslide early warning model，HLEWM)进行对比。研究结果表明，研究区构建降雨阈值/时间概率模型的最优变量组合为当日－前3日有效降雨量(R₀-AE₃)，最优衰减系数为0.72。R₀-AE₃空间中的时间概率等值线为一系列S形的曲线，可用三阶逻辑回归进行拟合。HLEWM较为离散，对不同降雨事件的致灾区分度较差，使得模型对强降雨事件不敏感，导致漏报现象较为严重，模拟预警结果表明HLEWM的漏报率达到51.1%，预警成功率为47.7%，而时间概率模型能够精确预测降雨的实际致灾概率，使得QLEWM预警性能有显著提升，预警成功率达到75.4%，漏报率仅为22.9%。结果表明，通过优化时间概率模型、提高预警模型的连续性，可以有效提升预警模型的预警成功率，减小漏报率。

Abstract

To improve the accuracy of rainfall-induced landslide early warning, this paper built a quantitative landslide early warning model (QLEWM) by integrating Bayesian methods and machine learning algorithms, with a case application in Lveyang-Ningqiang County, Shaanxi Province. First, disaster-causing and non-disaster-causing rainfall were identified using standardized criteria. Then, the optimal rainfall variable combination and hyperparameters was selected by information gain, and a rainfall-induced landslide probability model (temporal probability model) was developed with Bayesian inference and third-order logistic regression. Subsequently, high-risk slopes were selected as samples, nine geo-environment factors were used to extract features, and a random forest model was trained to generate landslide susceptibility map. Finally, the QLEWM was developed by coupling landslide spatiotemporal probabilities with Bayesian inference. Daily early warning simulations and early warnings were conducted for the rainy seasons from 2018 to 2023, and the results were compared with the traditional heuristic landslide early warning model (HLEWM). The results show that the optimal variable combination for constructing the rainfall threshold/temporal probability models in the study area is the effective rainfall amount of the current day plus the previous three days (R ₀-AE ₃), with an attenuation coefficient of 0.72. The temporal probability counterlines of in R ₀-AE ₃ space were a series of sigmoidal curves, which can be fitted using third-order logistic regression. HLEWM is relatively discrete and has poor distinguishability between different rainfall events, making the model insensitive to heavy rainfall events and resulting in a significant underreporting rate. Simulation results show that HLEWM has an underreporting rate of 51.1% and a warning success rate of 47.7%. In contrast, the temporal probability model can accurately predict the actual disaster probability of rainfall, significantly improving the warning performance of QLEWM, achieving a warning success rate of 75.4% and an underreporting rate of only 22.9%. These results indicate that optimizing the temporal probability model and improving the continuity of the warning model can effectively improve the warning success rate and reduce the underreporting rate.

关键词

降雨型滑坡 / 气象风险预警 / 降雨阈值 / 滑坡易发性

Key words

rainfall-induced landslide / meteorologic risk early warning / rainfall threshold / landslide susceptibility

参考文献

[1] 刘传正, 王建新. 自然灾害的基本型式及防控对策研究[J]. 岩石力学与工程学报, 2023, 42(2): 275-291.

[2] Liu Chuanzheng, Wang Jianxin. Basic patterns of natural disasters and some countermeasures for risk mitigation[J]. Chinese Journal of Rock Mechanics and Engineering, 2023, 42(2): 275-291. (in Chinese)

[3] 强菲, 赵法锁, 段钊. 陕南秦巴山区地质灾害发育及空间分布规律[J]. 灾害学, 2015, 30(2): 193-198.

[4] Qiang Fei, Zhao Fasuo, Duan Zhao. Development and spatial distribution of geological disasters in Qinling-Daba Mountains of south Shaanxi[J]. Journal of Catastrophology, 2015, 30(2): 193-198. (in Chinese)

[5] 王雁林. 陕南地区滑坡灾害气象预报预警及其防范对策探析[J]. 地质灾害与环境保护, 2005, 16(4): 345-349.

[6] Wang Yanlin. Study on the early-warning and prediction of land-slide hazards and countermeasures in the southern Shaanxi Province[J]. Journal of Geological Hazards and Environment Preservation, 2005, 16(4): 345-349. (in Chinese)

[7] 张雅斌, 杜继稳, 蔡蕊, 等 . 陕西省精细化地质灾害气象预报预警研究[J]. 灾害学, 2011, 26(3): 28-34.

[8] Zhang Yabin, Du Jiwen, Cai Rui, et al. Meticulous meteorological forecasting and alarming for geological hazard in Shaanxi Province[J]. Journal of Catastrophology, 2011, 26(3): 28-34. (in Chinese)

[9] 赵世发, 王俊, 杜继稳, 等 . 秦巴山区地质灾害成因及预报预警[J]. 气象科技, 2010, 38(2): 263-269.

[10] Zhao Shifa, Wang Jun, Du Jiwen, et al. Causes of geological disasters in Qinling-Bashan area and their forecast and warning[J]. Meteorological Science and Technology, 2010, 38(2): 263-269. (in Chinese)

[11] Guzzetti F, Gariano S L, Peruccacci S, et al. Geographical landslide early warning systems[J]. Earth-Science Reviews, 2020, 200: 102973.

[12] Piciullo L, Calvello M, Cepeda J M. Territorial early warning systems for rainfall-induced landslides[J]. Earth-Science Reviews, 2018, 179: 228-247.

[13] Park J Y, Lee S R, Lee D H, et al. A regional-scale landslide early warning methodology applying statistical and physically based approaches in sequence[J]. Engineering Geology, 2019, 260: 105193.

[14] Xu Qing, Peng Dalei, Zhang Shuai, et al. Successful implementations of a real-time and intelligent early warning system for loess landslides on the Heifangtai terrace, China[J]. Engineering Geology, 2020, 278: 105817.

[15] Caine N. The rainfall intensity: duration control of shallow landslides and debris flows[J]. Geografiska Annaler Series A, Physical Geography, 1980, 62(1/2): 23-27.

[16] Cannon S H, Gartner J E, Wilson R C, et al. Storm rainfall conditions for floods and debris flows from recently burned areas in southwestern Colorado and southern California[J]. Geomorphology, 2008, 96(3/4): 250-269.

[17] 李铁锋, 丛威青. 基于Logistic回归及前期有效雨量的降雨诱发型滑坡预测方法[J]. 中国地质灾害与防治学报, 2006, 17(1): 33-35.

[18] Li Tiefeng, Cong Weiqing. A method for rainfall-induced landslide prediction based on Logistic regression and effective antecedent rainfall[J]. The Chinese Journal of Geological Hazard and Control, 2006, 17(1): 33-35. (in Chinese)

[19] 李长江, 麻土华, 孙乐玲, 等 . 降雨型滑坡预报中计算前期有效降雨量的一种新方法[J]. 山地学报, 2011, 29(1): 81-86.

[20] Li Changjiang, Ma Tuhua, Sun Leling, et al. A new approach to calibrate antecedent precipitation for rainfall-triggering landslide forecast[J]. Mountain Research, 2011, 29(1): 81-86. (in Chinese)

[21] Glade T, Crozier M, Smith P. Applying probability determination to refine landslide-triggering rainfall thresholds using an empirical “antecedent daily rainfall model”[J]. Pure and Applied Geophysics, 2000, 157(6): 1059-1079.

[22] Brunetti M T, Peruccacci S, Rossi M, et al. Rainfall thresholds for the possible occurrence of landslides in Italy[J]. Natural Hazards and Earth System Sciences, 2010, 10(3): 447-458.

[23] Melillo M, Brunetti M T, Peruccacci S, et al. A tool for the automatic calculation of rainfall thresholds for landslide occurrence[J]. Environmental Modelling & Software, 2018, 105: 230-243.

[24] Baum R L, Godt J W. Early warning of rainfall-induced shallow landslides and debris flows in the USA[J]. Landslides, 2010, 7(3): 259-272.

[25] 刘传正, 温铭生, 唐灿. 中国地质灾害气象预警初步研究[J]. 地质通报, 2004, 23(4): 303-309.

[26] Liu Chuanzheng, Wen Mingsheng, Tang Can. Meteorological early warning of geo-hazards in China based on raining forecast[J]. Geological Bulletin of China, 2004, 23(4): 303-309. (in Chinese)

[27] 刘艳辉, 刘传正, 温铭生, 等 . 中国地质灾害气象预警模型研究[J]. 工程地质学报, 2015, 23(4): 738-746.

[28] Liu Yanhui, Liu Chuanzheng, Wen Mingsheng, et al. Study of early warning models for regional geo-hazards in China[J]. Journal of Engineering Geology, 2015, 23(4): 738-746. (in Chinese)

[29] 张钰洁, 段平, 王扶卷, 等 . 顾及不同易发区的降雨型滑坡发生时间概率研究[J]. 自然灾害学报, 2025, 34(3): 133-144.

[30] Zhang Yujie, Duan Ping, Wang Fujuan, et al. Study on the induced time probability of rainfall-induced landslides considering different prone regions[J]. Journal of Natural Disasters, 2025, 34(3): 133-144. (in Chinese)

[31] 李环禹, 陈朝晖, 范文亮, 等 . 区域降雨型滑坡风险分析统计模型研究[J]. 自然灾害学报, 2018, 27(4): 103-111.

[32] Li Huanyu, Chen Zhaohui, Fan Wenliang, et al. The statistical risk analysis model of rainfall-induced landslide in large areas[J]. Journal of Natural Disasters, 2018, 27(4): 103-111. (in Chinese)

[33] 夏梦想, 李远耀, 吴吉民, 等 . 基于I-D统计模型的张家界市滑坡灾害降雨预警阀值研究[J]. 自然灾害学报, 2021, 30(4): 203-212.

[34] Xia Mengxiang, Li Yuanyao, Wu Jimin, et al. Research on rainfall early warning threshold of landslide disaster in Zhangjiajie City based on I-D statistical model[J]. Journal of Natural Disasters, 2021, 30(4): 203-212. (in Chinese)

[35] Shu Heping, He Jiale, Zhang Fangu, et al. Construction of landslide warning by combining rainfall threshold and landslide susceptibility in the Gully Region of the Loess Plateau: a case of Lanzhou City, China[J]. Journal of Hydrology, 2024, 645: 132148.

[36] Liu Shuhao, Du Juan, Yin Kunlong, et al. Regional early warning model for rainfall induced landslide based on slope unit in Chongqing, China[J]. Engineering Geology, 2024, 333: 107464.

[37] DB61/T589- 2013 地质灾害预报技术规程[S]. 西安: 陕西省质量技术监督局.

[38] DB61/T589- 2013 Forecasting of geological disasters technical regulations[S]. Xi’an: Shaanxi Provincial Bureau of Quality and Technical Supervision. (in Chinese)

[39] Berti M, Martina M L V, Franceschini S, et al. Probabilistic rainfall thresholds for landslide occurrence using a Bayesian approach[J]. Journal of Geophysical Research: Earth Surface, 2012, 117(F4): 2012JF002367.

[40] Robbins J C. A probabilistic approach for assessing landslide-triggering event rainfall in Papua New Guinea, using TRMM satellite precipitation estimates[J]. Journal of Hydrology, 2016, 541: 296-309.

[41] Jiang Wanyu, Chen Guan, Meng Xingmin, et al. Probabilistic rainfall threshold of landslides in data-scarce mountainous areas: a case study of the Bailong River Basin, China[J]. Catena, 2022, 213: 106190.

[42] Song Yufei, Fan Wen, Yu Ningyu, et al. Rainfall induced shallow landslide temporal probability modelling and early warning research in mountains areas: a case study of Qin-Ba Mountains, western China[J]. Remote Sensing, 2022, 14(23): 5952.

[43] Xiao Te, Zhang Limin. Data-driven landslide forecasting: methods, data completeness, and real-time warning[J]. Engineering Geology, 2023, 317: 107068.

[44] 黄发明, 陈佳武, 范宣梅, 等 . 降雨型滑坡时间概率的逻辑回归拟合及连续概率滑坡危险性建模[J]. 地球科学, 2022, 47(12): 4609-4628.

[45] Huang Faming, Chen Jiawu, Fan Xuanmei, et al. Logistic regression fitting of rainfall-induced landslide occurrence probability and continuous landslide hazard prediction modelling[J]. Earth Science, 2022, 47(12): 4609-4628. (in Chinese)

[46] Xiao Te, Zhang Limin, Cheung R W M, et al. Predicting spatio-temporal man-made slope failures induced by rainfall in Hong Kong using machine learning techniques[J]. Géotechnique, 2023, 73(9): 749-765.

[47] 宋宇飞, 范文, 左琛, 等 . 基于敏感性分析的最优降雨阈值选择[J]. 工程地质学报, 2024, 32(2): 529-544.

[48] Song Yufei, Fan Wen, Zuo Chen, et al. The optimal rainfall thresholds selection based on sensitivity analysis[J]. Journal of Engineering Geology, 2024, 32(2): 529-544. (in Chinese)

[49] 冯杭建, 周爱国, 唐小明, 等 . 浙江省泥石流灾害发育分布规律及区域预报[J]. 地球科学, 2016, 41(12): 2088-2099.

[50] Feng Hangjian, Zhou Aiguo, Tang Xiaoming, et al. Development and distribution characteristics of debris flow in Zhejiang Province and its regional forecast[J]. Earth Science, 2016, 41(12): 2088-2099. (in Chinese)

[51] 王伟, 袁雯宇, 邹丽芳, 等 . 基于滑坡敏感性评价的库区水动力型滑坡区域综合预警研究[J]. 岩石力学与工程学报, 2022, 41(3): 479-491.

[52] Wang Wei, Yuan Wenyu, Zou Lifang, et al. Comprehensive regional-scale early warning of water-induced landslides in reservoir areas based on landslide susceptibility assessment[J]. Chinese Journal of Rock Mechanics and Engineering, 2022, 41(3): 479-491. (in Chinese)

[53] 刘谢攀, 殷坤龙, 肖常贵, 等 . 基于I-D-R阈值模型的滑坡气象预警[J]. 地球科学, 2024, 49(3): 1039-1051.

[54] Liu Xiepan, Yin Kunlong, Xiao Changgui, et al. Meteorological early warning of landslide based on I-D-R threshold model[J]. Earth Science, 2024, 49(3): 1039-1051. (in Chinese)

[55] 陈贺, 付有旺. 贡山县进藏通道群发性滑坡地质特征与降雨阈值[J]. 岩石力学与工程学报, 2024, 43(增刊2): 3861-3871.

[56] Chen He, Fu Youwang. Geological feature and rainfall threshold of cluster landslides along roads toward Xizang in Gongshan County, Yunnan Province, China[J]. Chinese Journal of Rock Mechanics and Engineering, 2024, 43(S2): 3861-3871. (in Chinese)

[57] 周志华. 机器学习[M]. 北京: 清华大学出版社, 2016.

[58] Zhou Zhihua. Machine learning[M]. Beijing: Tsinghua University Press, 2016. (in Chinese)

[59] DZ/T0449- 2023 地质灾害气象风险预警规范[S].

[60] DZ/T0449- 2023 Specification of geological hazard risk early warning based on meteorological factors[S]. (in Chinese)

[61] 黄发明, 曹中山, 姚池, 等 . 基于决策树和有效降雨强度的滑坡危险性预警[J]. 浙江大学学报(工学版), 2021, 55(3): 472-482.

[62] Huang Faming, Cao Zhongshan, Yao Chi, et al. Landslide hazard warning based on decision tree and effective rainfall intensity[J]. Journal of Zhejiang University (Engineering Science), 2021, 55(3): 472-482. (in Chinese)

[63] 吴益平, 张秋霞, 唐辉明, 等 . 基于有效降雨强度的滑坡灾害危险性预警[J]. 地球科学, 2014, 39(7): 889-895.

[64] Wu Yiping, Zhang Qiuxia, Tang Huiming, et al. Landslide hazard warning based on effective rainfall intensity[J]. Earth Science, 2014, 39(7): 889-895. (in Chinese)

[65] 胡涛, 樊鑫, 李荻, 等 . 贵州省思南县降雨型滑坡危险性预警研究[J]. 安全与环境工程, 2020, 27(3): 74-81.

[66] Hu Tao, Fan Xin, Li Di, et al. Rainfall-induced landslide hazard warning study in Sinan County of Guizhou Province[J]. Safety and Environmental Engineering, 2020, 27(3): 74-81. (in Chinese)

[67] Calvello M, Piciullo L. Assessing the performance of regional landslide early warning models: the EDuMaP method[J]. Natural Hazards and Earth System Sciences, 2016, 16(1): 103-122.

[68] Piciullo L, Dahl M P, Devoli G, et al. Adapting the EDuMaP method to test the performance of the Norwegian early warning system for weather-induced landslides[J]. Natural Hazards and Earth System Sciences, 2017, 17(6): 817-831.

基金资助

国家自然科学基金项目(42502291)