| 许铜建,祁亚科,李正.基于时间惩罚决策树算法的时序 InSAR动态滑坡易发性评估——以河南郑州市为例[J].矿产勘查,2026,2(2):403-416 |
| 基于时间惩罚决策树算法的时序 InSAR动态滑坡易发性评估——以河南郑州市为例 |
| Dynamic landslide susceptibility assessment using time-penalized trees algorithm with time-series InSAR data: A case study in Zhengzhou City, Henan Province |
| 投稿时间：2024-10-13 |
| DOI：10.20008/j.kckc.202602016 |
| 中文关键词: 易发性评价 时间惩罚树 时变协变量 DS-IPTA InSAR |
| 英文关键词: susceptibility evaluation time-penalized trees time-varying covariates DS-IPTA InSAR |
| 基金项目:本文受郑州市自然资源和规划局项目“郑州市多要素城市地质调查(DKSHT-2020-0116)”资助。 |
|
| 摘要点击次数: 193 |
| 全文下载次数: 28 |
| 中文摘要: |
| 滑坡灾害的精准评估对防灾减灾至关重要。针对传统方法处理动态因子时存在的过拟合与泛化能力不足问题，本文提出一种基于时间惩罚决策树算法的时序 InSAR动态滑坡易发性评估方法。通过融合地质、地形、水文等静态因子与 DS-IPTA InSAR技术获取的地表形变动态数据，构建兼顾时空特征的滑坡易发性模型，并引入时间惩罚机制优化决策树分裂准则，以抑制时变协变量引起的模型复杂度过高问题。以郑州市为例的实验表明，该方法在历史灾害数据和实地隐患点验证中均表现出较高的评估精度与泛化能力，其 AUC值显著优于随机森林、支持向量机等传统模型。本研究为动态滑坡风险评估提供了新的技术路径，尤其适用于 InSAR数据支持的复杂地质区域。 |
| 英文摘要: |
| The precise assessment of landslide disasters is crucial for disaster prevention and mitigation. In or-der to address the problems of overfitting and insufficient generalization ability in traditional methods for processingdynamic factors, this study proposes a time-series InSAR dynamic landslide susceptibility assessment methodbased on a time-penalty decision tree algorithm. By integrating static factors such as geology, topography, and hy-drology with dynamic data on surface deformation obtained through DS-IPTA InSAR technology, a landslide suscep-tibility model that takes into account spatial and temporal characteristics is constructed. A time penalty mechanismis introduced to optimize the decision tree splitting criterion to suppress the problem of excessive model complexitycaused by time-varying covariates. Experiments conducted in Zhengzhou City demonstrate that the method exhibitshigh evaluation accuracy and generalization ability in both historical disaster data and field hazard verification, withits AUC value significantly surpassing traditional models such as random forest and support vector machine. Thisstudy provides a new technical path for dynamic landslide risk assessment, especially for complex geological areassupported by InSAR data. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |