崩岗是中国南方红壤丘陵区最为严重的侵蚀形式之一,基于GIS开展江西省崩岗发育风险评估,对该地区崩岗侵蚀防治具有重要的科学意义。以江西省为研究区,应用灰色关联度分析法和皮尔逊相关性分析法,计算10个影响因子的重要性和相关性矩阵进行影响因子筛选,利用信息量、加权信息量和频率比3种模型开展江西省崩岗发育风险评估,采用自然断点分级法划分风险等级。结果表明:1)气温、海拔和土地利用类型是江西省崩岗发育的重要影响因子。2)信息量模型、加权信息量模型和频率比模型的受试者工作特征曲线下方面积(Area under curve,AUC值)分别为0.860、0.852和0.833,3种模型评估精度均良好,都适用于江西省崩岗发育风险评估。3)江西省崩岗发育风险等级分布具有显著的空间异质性,呈现由北往南逐渐递增的趋势,极高风险区主要集中分布在江西省南部,高风险区主要分布在江西省南部和中部,这与江西省崩岗实际分布情况基本一致。研究结果可为江西省崩岗侵蚀防治提供参考。
为进一步推进塔吉克斯坦地区的地质灾害评估工作,本文采用信息量模型和随机森林模型对海东市的滑坡灾害进行易发性评价,并通过ROC曲线对这两种模型的性能进行了评估。结果表明,信息量模型与随机森林模型均表现出良好的分类能力,然而随机森林模型的表现更优于信息量模型。信息量模型在对相关因子的分级分析中表现卓越,能够有效识别某因子分级内对滑坡影响最深的区间。而随机森林模型则在衡量因子的相对重要性方面具有优势。两种模型各具特色,结合使用能够更全面地掌握滑坡灾害的成因与易发性,为后续的地质灾害防治措施提供科学依据。To further advance the assessment of geological disasters in the Tajikistan region, this paper employs the information value model and the random forest model to evaluate the susceptibility to landslides in Haidong City, and assesses the performance of these two models using ROC curves. The results indicate that both the information value model and the random forest model demonstrate good classification capabilities;however, the random forest model outperforms the information value model. The information value model excels in the graded analysis of relevant factors, effectively identifying the intervals within a factor’s grading that have the most significant impact on landslides. In contrast, the random forest model has an advantage in measuring the relative importance of factors. Each model has its unique characteristics, and their combined use can provide a more comprehensive understanding of the causes and susceptibility of landslide disasters, offering a scientific basis for subsequent geological disaster prevention and control measures.
