关键词:
滑坡
机器学习
易发性
大别山区
摘要:
滑坡作为一种重大地质灾害,对人类生活和基础设施构成了严重威胁。因此,深入研究滑坡易发性,尤其是利用机器学习模型进行预测,具有重要的学术和实践意义。基于此,本文通过收集相关地质、气象和地形数据,构建滑坡易发性评价指标体系,并利用多种机器学习算法来评价滑坡易发性,通过比较不同算法的结果,探讨各算法的优缺点并绘制滑坡易发性区划图。研究结果表明:(1) 大别山区植被覆盖NDVI在0.7~0.8之间其滑坡点分布最密集,且在林地滑坡点分布最多。(2) 随机森林是滑坡易发性拟合最优的模型。在诸多因子分析中,高程、坡度、剖面曲率和多年平均降水对滑坡易发性有显著影响。大别山区的滑坡易发性高于周边地区,且南坡的滑坡易发性风险最大。Landslides, as a major geological hazard, pose serious threats to human life and infrastructure. Therefore, in-depth research on landslide susceptibility, especially using machine learning models for prediction, holds significant academic and practical value. Based on this, this study collected relevant geological, meteorological, and topographical data to establish a landslide susceptibility evaluation index system. Multiple machine learning algorithms were employed to assess landslide susceptibility, and by comparing the results of different algorithms, the advantages and disadvantages of each were discussed, accompanied by the creation of a landslide susceptibility zoning map. The findings revealed that: (1) Landslide points are most densely distributed in the Dabie Mountains when vegetation cover NDVI ranges from 0.7 to 0.8, with the majority of landslides occurring in forested areas. (2) Random forest was identified as the optimal model for landslide susceptibility fitting. Among various factors analyzed, elevation, slope, profile curvature, and long-term average precipitation had significant impacts on landslide susceptibility. The landslide susceptibility of the Dabie Mountains is higher than that of surrounding areas, with the southern slopes exhibiting the highest landslide susceptibility risk.