Zhao Fancheng, Miao Fasheng, Wu Yiping, Gong Shunqi, Zheng Guyue, Yang Jing, Zhan Weiwei
Faculty of Engineering, China University of Geosciences, Wuhan 430074, China.
State Grid Jingzhou Electric Power Supply Company, Jingzhou 434000, China.
Sci Total Environ. 2024 Nov 1;949:175059. doi: 10.1016/j.scitotenv.2024.175059. Epub 2024 Jul 29.
Landslides pose a noteworthy threat in urban settlements globally, especially in areas experiencing extreme climate and rapid engineering. However, researches focusing on the long-term uninterrupted land use and land cover change (LULCC) impacted on landslide susceptibility mapping (LSM) in rapid urban expansion areas remains limited, let alone different temporal scenarios adjacency thresholds. This work aims to refine the temporal LSM considering spatiotemporal land use and land cover (LULC) and to provide decision makers with governing factors in landslides control during urbanization in mountainous areas. Herein, annual LULC data and landslide inventory spanning from 1992 to 2022 were utilized to map dynamic landslide susceptibility in Wanzhou District of the Three Gorges Reservoir Area, China. Initially, the landslide-related factors were filtered as input features of random forest (RF) model before diagnosis via multicollinearity test and Pearson Correlation Coefficient (PCC). The advanced patch-generating land use simulation (PLUS) model was then invited to fuel temporal susceptibility prediction powered by LULCC projections. Finally, the performance of various scenarios was evaluated using Receiver Characteristic Curve (ROC) curves and Shapley Additive exPlanation (SHAP) technique, with discussions on LULCC temporal adjacency thresholds and mutual feedback mechanism between territorial exploitation and landslide occurrences. The results indicate that the precision of LSM is positively correlated with the time horizon, acted by incorporating the latest LULC and LULCC achieving an area under the curve (AUC) of 0.920. The transition of land from forest to cropland and impervious areas should be avoided to minimize the increase in landslide susceptibility. Moreover, a one-year adjacency threshold of LULCC is recommended for optimal model accuracy in future LSM. This dynamic LSM framework can serve as a reference for decision makers in future landslide susceptibility mitigation and land resources utilization in rapid urban expansion areas worldwide.
滑坡在全球城市住区构成了显著威胁，尤其是在经历极端气候和快速工程建设的地区。然而，针对快速城市扩张地区长期不间断的土地利用和土地覆盖变化（LULCC）对滑坡易发性制图（LSM）影响的研究仍然有限，更不用说不同时间情景的邻接阈值了。这项工作旨在考虑时空土地利用和土地覆盖（LULC）来优化时间LSM，并为山区城市化过程中滑坡控制的决策者提供控制因素。在此，利用1992年至2022年的年度LULC数据和滑坡清单，绘制了中国三峡库区万州区的动态滑坡易发性图。首先，在通过多重共线性检验和皮尔逊相关系数（PCC）进行诊断之前，对与滑坡相关的因素进行筛选，作为随机森林（RF）模型的输入特征。然后引入先进的斑块生成土地利用模拟（PLUS）模型，以推动由LULCC预测驱动的时间易发性预测。最后，使用接收器特征曲线（ROC）和夏普利加法解释（SHAP）技术评估各种情景的性能，并讨论LULCC时间邻接阈值以及土地开发与滑坡发生之间的相互反馈机制。结果表明，LSM的精度与时间跨度呈正相关，通过纳入最新的LULC和LULCC，曲线下面积（AUC）达到0.920。应避免土地从森林转变为农田和不透水区域，以尽量减少滑坡易发性的增加。此外，建议LULCC的一年邻接阈值可实现未来LSM的最佳模型精度。这种动态LSM框架可为全球快速城市扩张地区未来滑坡易发性缓解和土地资源利用的决策者提供参考。
