基于不同采样策略的武义县滑坡易发性评价对比分析

Comparative analysis of landslide susceptibility evaluation in Wuyi County based on different sampling strategies

摘要

滑坡是一种常见的地质灾害，影响地区经济的发展，危害人民的生命安全。本文以浙江省金华市武义县为例，利用GIS平台提取高程、坡度、坡向等13种环境因子，通过信息量法分析各个环境因子对滑坡灾害的影响，基于信息量值生成滑坡易发性图并由此划分出易发性等级区域，根据不同采集策略从极低易发区、低易发区采集不同比例的非滑坡样本点，形成数据集，训练小龙虾算法(crayfish optimization algorithm，COA)耦合反向传播(back propagation，BP)神经网络模型(COA-BP)，通过准确率、模型内部权重影响分析、滑坡易发性评价精度分析，对不同采样策略进行评价。结果表明，滑坡灾害发生区间与人类活动区间重合，测试集准确率与训练集准确率不存在强关系，不同采样策略以及样本量比例对模型内部权重影响不同，并且权重的正负性也不能定性，1∶4不处理样本的采样策略在该地区综合采样能力最佳，为当地防灾减灾提供一定的参考。

Abstract

Landslides are a common geological hazard that impacts economic development and endangers people’s lives. This paper takes Wuyi County, Jinhua City, Zhejiang Province as an example. Using a GIS platform, 13 environmental factors, including elevation, slope, and aspect, are extracted. The impact of each environmental factor on landslide hazards is analyzed using the information content method. A landslide susceptibility map is generated based on the information content values, and susceptibility levels are delineated. Different proportions of non-landslide sample points are collected from extremely low-susceptibility and low-susceptibility areas according to different collection strategies to form a dataset. A crayfish algorithm coupled with a back propagation (COA-BP) neural network model is trained. Different sampling strategies are evaluated through accuracy, internal model weight influence analysis, and landslide susceptibility assessment accuracy analysis. The results show that the landslide hazard occurrence area overlaps with the human activity area. There is no strong correlation between the accuracy of the test set and the accuracy of the training set. Different sampling strategies and sample size ratios have different impacts on the internal model weights, and the positive or negative nature of the weights cannot be qualitatively interpreted. The 1∶4 sampling strategy with no sample processing has the best overall sampling capability in this area, providing a certain reference for local disaster prevention and mitigation.

关键词

小龙虾算法 / BP神经网络 / 采样策略 / 滑坡易发性评价 / 信息量法

Key words

crayfish algorithm / BP neural network / sampling strategy / evaluation of landslide susceptibility / information value

参考文献

[1] CAO Y F, ZHAO Z F, WEN M C, et al. Identification and susceptibility assessment of landslide disasters in the red bed formation along the Nanjian-Jingdong Expressway[J]. Ecological Indicators, 2025, 170: 113002.

[2] SONG Y, SONG Y X, WANG C N, et al. Landslide susceptibility assessment through multi-model stacking and meta-learning in Poyang County, China[J]. Geomatics, Natural Hazards and Risk, 2024, 15(1): 180-192.

[3] PARK C Y, CHOM G, PARK C, et al. Landslide risk on photovoltaic power stations under climate change[J]. Geomatics, Natural Hazards and Risk, 2024, 15(1): 2286904.

[4] 杨永鹏, 郭雅, 陈浩, 等 . 青藏高原西部滑坡易发性评价对比研究[J]. 自然灾害学报, 2025, 34(1): 95-104.

[5] YANG Yongpeng, GUO Ya, CHEN Hao, et al. Comparative study on evaluation of landslide susceptibility in western Qinghai-Xizang Plateau[J]. Journal of Natural Disasters, 2025, 34(1): 95-104. (in Chinese)

[6] 张云, 资锋, 曹运江, 等 . 基于MaxEnt模型的沅陵县滑坡易发性评价[J]. 自然灾害学报, 2025, 34(1): 105-116.

[7] ZHANG Yun, ZI Feng, CAO Yunjiang, et al. Evaluation of landslide susceptibility in Yuanling County based on MaxEnt model[J]. Journal of Natural Disasters, 2025, 34(1): 105-116. (in Chinese)

[8] 肖波, 赵良军, 郑莉萍, 等 . 融合时序InSAR与多因子耦合模型的长宁县滑坡灾害易发性评价[J]. 自然灾害学报, 2025, 34(1): 117-126.

[9] XIAO Bo, ZHAO Liangjun, ZHENG Liping, et al. Landslide hazard susceptibility assessment in Changning County through fusion of time-series InSAR and multi-factors coupling model[J]. Journal of Natural Disasters, 2025, 34(1): 117-126. (in Chinese)

[10] 崔婷婷, 安雪莲, 孙德亮, 等 . 基于SHAP的可解释机器学习的滑坡易发性评价模型[J]. 成都理工大学学报(自然科学版), 2025, 52(1): 153-172.

[11] CUI Tingting, AN Xuelian, SUN Deliang, et al. Explainable machine learning models for landslide susceptibility mapping based on SHAP[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2025, 52(1): 153-172. (in Chinese)

[12] 郭飞, 吴迪, 葛民荣, 等 . 连续变量因子分级和机器学习模型对滑坡易发性评价精度的影响[J/OL]. 武汉大学学报(信息科学版), 2024: 1-17. (2024-09-20). https://link.cnki.net/doi/10.13203/j.whugis20230413.

[13] GUO Fei, WU Di, GE Minrong, et al. The influence of continuous variable factor classification and machine learning model on the accuracy of landslide susceptibility evaluation[J/OL]. Geomatics and Information Science of Wuhan University, 2024: 1-17. (2024-09-20). https://link.cnki.net/doi/10.13203/j.whugis20230413. (in Chinese)

[14] 许烈, 巨能攀, 邓明东, 等 . 不同评价单元的滑坡易发性评价精度对比分析[J]. 工程地质学报, 2024, 32(5): 1640-1653.

[15] XU Lie, JU Nengpan, DENG Mingdong, et al. Comparative analysis of landslide susceptibility assessment accuracy in different evaluation units[J]. Journal of Engineering Geology, 2024, 32(5): 1640-1653. (in Chinese)

[16] 张军以, 丁悦凯, 孙德亮. 基于不同样本比例与超参数优化的滑坡易发性评价: 以重庆市武隆区为例[J]. 重庆师范大学学报(自然科学版), 2022, 39(5): 47-57.

[17] ZHANG Junyi, DING Yuekai, SUN Deliang. Landslide susceptibility evaluation based on different sample proportion and superparameter optimization: Take Wulong District of Chongqing municipality as an example[J]. Journal of Chongqing Normal University (Natural Science), 2022, 39(5): 47-57. (in Chinese)

[18] 黄发明, 曾诗怡, 姚池, 等 . 滑坡易发性预测建模的不确定性: 不同“非滑坡样本”选择方式的影响[J]. 工程科学与技术, 2024, 56(1): 169-182.

[19] HUANG Faming, ZENG Shiyi, YAO Chi, et al. Uncertainties of landslide susceptibility prediction modeling: Influence of different selection methods of “non-landslide samples”[J]. Advanced Engineering Sciences, 2024, 56(1): 169-182. (in Chinese)

[20] 武雪玲, 杨经宇, 牛瑞卿. 一种结合SMOTE和卷积神经网络的滑坡易发性评价方法[J]. 武汉大学学报(信息科学版), 2020, 45(8): 1223-1232.

[21] WU Xueling, YANG Jingyu, NIU Ruiqing. A landslide susceptibility assessment method using SMOTE and convolutional neural network[J]. Geomatics and Information Science of Wuhan University, 2020, 45(8): 1223-1232. (in Chinese)

[22] 吴宏阳, 周超, 梁鑫, 等 . 基于样本优化策略的滑坡易发性评价[J]. 武汉大学学报(信息科学版), 2024, 49(8): 1492-1502.

[23] WU Hongyang, ZHOU Chao, LIANG Xin, et al. Evaluation of landslide susceptibility based on sample optimization strategy[J]. Geomatics and Information Science of Wuhan University, 2024, 49(8): 1492-1502. (in Chinese)

[24] JIA H M, RAO H H, WEN C S, et al. Crayfish optimization algorithm[J]. Artificial Intelligence Review, 2023, 56(2): 1919-1979.

[25] 薛正海, 冯文凯, 韩靖楠, 等 . 基于信息量及层次分析关联法和BP神经网络耦合的滑坡易发性评价[J]. 桂林理工大学学报, 2022, 42(4): 809-819.

[26] XUE Zhenghai, FENG Wenkai, HAN Jingnan, et al. Landslide susceptibility evaluation based on information value and AHP correlation method coupled with BP neural network model[J]. Journal of Guilin University of Technology, 2022, 42(4): 809-819. (in Chinese)

[27] 翁铭锴, 肖桂荣. 训练样本采样优化与机器学习结合的滑坡易发性评价方法[J]. 地球信息科学学报, 2025, 27(5): 1113-1128.

[28] WENG Mingkai, XIAO Guirong. A landslide susceptibility assessment method integrating training sample optimization and machine learning[J]. Journal of Geo-Information Science, 2025, 27(5): 1113-1128. (in Chinese)

[29] 麻晟, 陈剑, 李强, 等 . 负样本选择和评价模型对滑坡易发性预测不确定性的影响[J]. 现代地质, 2025, 39(5): 1373-1391.

[30] MA Sheng, CHEN Jian, LI Qiang, et al. Effects of non-landslide sample selection methods and evaluation models on the uncertainty of landslide susceptibility prediction[J]. Geoscience, 2025, 39(5): 1373-1391. (in Chinese)

[31] 赵恒梓, 李素敏, 郭庆, 等 . 样本优化的集成学习模型在滑坡易发性评价中的应用[J]. 测绘科学, 2024, 49(12): 132-141.

[32] ZHAO Hengzi, LI Sumin, GUO Qing, et al. Application of ensemble learning model based on sample optimization in landslide susceptibility evaluation[J]. Science of Surveying and Mapping, 2024, 49(12): 132-141. (in Chinese)

基金资助

吉林省科技发展计划项目(YDZJ202501ZYTS563)

吉林省教育厅科学研究项目(JJKH20190874KJ)