Faculty of Land and Resources Engineering, Kunming University of Science and Technology, Kunming 650093, China.
Application Engineering Research Center of Plateau and Mountainous Spatial Information Surveying and Mapping Technology, Yunnan Universities, Kunming 650093, China.
Sensors (Basel). 2022 Jul 26;22(15):5587. doi: 10.3390/s22155587.
Landslide susceptibility maps (LSM) are often used by government departments to carry out land use management and planning, which supports decision makers in urban and infrastructure planning. The accuracy of conventional landslide susceptibility maps is often affected by classification errors. Consequently, they become less reliable, which makes it difficult to meet the needs of decision-makers. Therefore, it is proposed in this paper to reduce classification errors and improve LSM reliability by integrating the Small Baseline Subsets-Interferometric Synthetic Aperture Radar (SBAS-InSAR) technique and LSM. By using the logistic regression model (LR) and the support vector machine model (SVM), experiments were conducted to generate LSM in the Dongchuan district. It was classified into five classes: very high susceptibility, high susceptibility, medium susceptibility, low susceptibility, and very low susceptibility. Then, the surface deformation rate of the Dongchuan area was obtained through the ascending and descending orbit sentinel-1A data from January 2018 to January 2021. To correct the classification errors, the SBAS-InSAR technique was integrated into LSM under the optimal model by constructing the contingency matrix. Finally, the LSMs obtained before and after correction were compared. Moreover, the correction results were validated and analyzed by combining remote sensing images, InSAR deformation results, and field surveys. According to the research results, the susceptibility class of 66,094 classification error cells (59.48 km) was significantly improved in the LSM after the integration of the SBAS-InSAR correction. The enhanced susceptibility classes and the spectral characteristics of remote sensing images are highly consistent with the trends of InSAR cumulative deformation and the results of field investigation. It is suggested that integrating SBAS-InSAR and LSM is effective in correcting classification errors and further improving the reliability of LSM for landslide prediction. The LSM obtained by using this method plays an important role in guiding local government departments on disaster prevention and mitigation, which is conducive to eliminating the risk of landslides.
滑坡易发性图（LSM）常被政府部门用于土地利用管理和规划，为城市和基础设施规划提供决策支持。传统滑坡易发性图的准确性往往受到分类错误的影响，因此其可靠性降低，难以满足决策者的需求。因此，本文提出通过整合小基线集-合成孔径雷达干涉测量（SBAS-InSAR）技术和 LSM，减少分类错误，提高 LSM 可靠性。利用逻辑回归模型（LR）和支持向量机模型（SVM），在东川地区生成 LSM，并将其分为五类：极高易发性、高易发性、中易发性、低易发性和极低易发性。然后，通过 2018 年 1 月至 2021 年 1 月的升轨和降轨 Sentinel-1A 数据获取东川地区的地表变形速率。为了纠正分类错误，通过构建列联表，在最优模型下将 SBAS-InSAR 技术集成到 LSM 中。最后，对比校正前后的 LSM。此外，结合遥感图像、InSAR 变形结果和实地调查，对校正结果进行验证和分析。研究结果表明，在集成 SBAS-InSAR 校正后，LSM 中 66094 个分类错误单元（59.48km）的易发性等级得到了显著改善。增强的易发性等级与遥感图像的光谱特征与 InSAR 累积变形趋势和实地调查结果高度一致。表明整合 SBAS-InSAR 和 LSM 可有效纠正分类错误，进一步提高 LSM 对滑坡预测的可靠性。该方法得到的 LSM 对指导地方政府部门防灾减灾具有重要作用，有利于消除滑坡风险。
