测绘通报 ›› 2025, Vol. 0 ›› Issue (6): 43-48.doi: 10.13474/j.cnki.11-2246.2025.0608

• 学术研究 • 上一篇    

融合InSAR与信息量-逻辑回归耦合模型的滑坡动态危险性评价

汪晨煜1, 彭军还1, 薛跃明1,2, 李旭1, 张炎3   

  1. 1. 中国地质大学(北京)土地科学技术学院, 北京 100083;
    2. 中国地质环境监测院, 北京 100081;
    3. 中国地质大学(武汉)信息工程学院, 湖北 武汉 430074
  • 收稿日期:2024-11-28 发布日期:2025-07-04
  • 通讯作者: 彭军还。E-mail:pengjunhuan@163.com
  • 作者简介:汪晨煜(2000—),男,硕士生,主要研究方向为InSAR技术与应用。E-mail:2112220022@email.cugb.edu.cn
  • 基金资助:
    国家自然科学基金(42074004);国家重点研发计划(2023YFC3007205);湖北省自然资源科技项目(ZRZY2024KJ07)

Landslide dynamic hazard assessment based on InSAR and information-logistic regression coupling model

WANG Chenyu1, PENG Junhuan1, XUE Yueming1,2, LI Xu1, ZHANG Yan3   

  1. 1. School of Land Science and Technology, China University of Geosciences (Beijing), Beijing 100083, China;
    2. China Institute of Geo-Environment Monitoring, Beijing 100081, China;
    3. Faculty of Information Enginerring, China University of Geosciences (Wuhan), Wuhan 430074, China
  • Received:2024-11-28 Published:2025-07-04

摘要: 针对当前不同滑坡危险性评价模型的结果存在差异性较大、时效性较差及假阴性误差的问题,本文以云南省祥云县为研究区,首先对信息量(I)、信息量-逻辑回归(I-LR)、长短时记忆神经网络(LSTM)和支持向量机(SVM)4种模型的性能进行对比分析,以确定最适滑坡危险性评价模型;然后采用SBAS-InSAR方法处理2019年8月—2023年7月的升降轨Sentinel-1数据,并解算地表坡向形变速率;最后通过校正矩阵将滑坡危险性分级并与坡向形变速率相结合,生成滑坡动态危险性图。研究结果表明,I-LR模型相比其他3种模型具有更好的预测精度和稳定性。结合I-LR模型滑坡危险性评价分级和InSAR坡向形变速率分级生成的滑坡动态危险性图对不稳定形变区具有更好的识别效果,低危险性占比降低10.35%,较低、中和高危险性的比例分别提高6.30%、2.45%和1.60%,提高了滑坡危险性评价结果的时效性和准确性。

关键词: 滑坡, 危险性制图, 信息量-逻辑回归模型, SBAS-InSAR

Abstract: In view of the issues of significant discrepancies, poor timeliness, and false negative errors in the results of various landslide hazard assessment models.This paper takes Xiangyun county, Yunnan province as the study area.Firstly, compares the performance of four models:information(I), information-logistic regression(I-LR), long short-term memory(LSTM) and support vector machine(SVM), to determine the most suitable model for landslide hazard assessment. Then, the SBAS-InSAR method is used to process the Sentinel-1 data from both ascending and descending orbits between August 2019 and July 2023, and the surface slope deformation rate is calculated. Finally, the landslide hazard classification and slope deformation rate are combined using a correction matrix to generate the dynamic landslide hazard map.The results show that the I-LR model has better prediction accuracy and stability than the other three models.The landslide dynamic hazard map, generated by combining the I-LR model landslide hazard classification and the InSAR slope deformation rate classification, demonstrates better identification of unstable deformation areas. The proportion of low hazard areas is reduced by 10.35%, while the proportions of lower, medium, and high hazard areas increase by 6.30%, 2.45%, and 1.60%, respectively. This enhances both the timeliness and accuracy of landslide hazard assessment results.

Key words: landslide, hazard mapping, informatics-logistic regression model, SBAS-InSAR

中图分类号: