滑坡易发性评价中的空间多尺度问题研究

Landslide susceptibility evaluation refers to the forecast to the spatial distribution and occurrence probability of the landslide, and the result provides important reference for the risk management of the landslide disaster, formulation of monitoring plans and planning layout of towns. However, as the existing landslide susceptibility evaluation methods ignored the influence of different spatial scales to the selection of the landslide evaluation factors and evaluation model parameters, there is big difference in the precision of the evaluation results obtained under different spatial scales, which further leads to the limitation in the accuracy and reliability of the evaluation results. This project quantitative expression the spatial scale concepts, and scientifically introduces it into the study of the landslide susceptibility evaluation by studying the selection of landslide evaluation factors and evaluation model parameters under different spatial scales via the method of dividing the research area. Furthermore, this project builds a landslide susceptibility evaluation coupling model based on the geographical weighted regression, the particle swarm optimization and the support vector machine, so as to analyze the root reasons for the influence of spatial scales to the landslide susceptibility evaluation results, and further to deeply reveal the relationship between the spatial scales, the evaluation results and the evaluation model parameters. The research result can provide scientific theoretical explanation for the important problem of the unbalance in the precision of the existing landslide susceptibility evaluation results and build a spatial multi-scale landslide susceptibility evaluation models, thereby improving the scientificity, accuracy and reliability of landslide susceptibility evaluation.

滑坡易发性评价是对滑坡空间分布和发生概率进行预测，其预测结果是滑坡灾害风险管理、监测规划制定和有关城镇规划布局的重要参考依据。现有的滑坡易发性评价方法，由于没有顾及不同的空间尺度对选择滑坡评价因子和评价模型参数的影响，在不同空间尺度下得到的评价结果的精度之间存在较大差异，导致评价结果的精确性和可靠性受到局限。本项目将空间尺度概念进行定量化表达，并科学地引入滑坡易发性评价研究中，采用对研究区进行分割的方法来研究不同空间尺度下滑坡评价因子和评价模型参数的选择问题，建立基于地理加权回归、粒子群优化和支持向量机的滑坡易发性评价耦合模型，分析空间尺度对评价结果造成影响的根本原因，深刻揭示空间尺度、评价结果和评价模型参数之间的关系。研究成果能够从理论上为目前不同空间尺度下得到的评价结果的精度之间存在不平衡性这一重要问题提供科学解释，并建立空间多尺度评价模型，提高滑坡易发性评价的科学性、精确性和可靠性。

本项目以三峡库区秭归到巴东段为研究区，针对滑坡易发性评价结果的空间不平衡性问题，提出了滑坡易发性评价具有空间尺度特征，并且在研究滑坡易发性评价的空间尺度问题时，将空间尺度概念进行定量化表达，计算对滑坡发生起到重要作用相关因子的地理加权回归系数，根据总结出的三条规则对研究区进行空间尺度分割，并进一步建立了地理加权回归-粒子群优化-支持向量机滑坡易发性评价耦合模型，通过对比不同空间尺度下滑坡易发性评价模型和结果的差异，揭示了空间尺度对于滑坡易发性评价结果造成影响的根本原因，解释了滑坡易发性评价结果的空间不平衡性问题，即滑坡易发性评价因子本身具有空间不平衡性，再以此为新的研究内容，深入分析了岩土特性因子对于滑坡易发性评价结果的影响。.本项目的顺利实施，提高滑坡易发性评价的科学性、精确性和可靠性，对滑坡易发性评价研究提供了新的思路。此外，将空间尺度概念应用到高分辨率遥感影像大范围主要地物信息提取中，取得了良好的效果。