基于遥感图像阈值分割的震后次生地质灾害风险源自动判识及致灾过程数值模拟

Earthquake-induced loose soil deposits are the major sources for the secondary geo-disasters. They distribute widely in the earthquake-stricken mountainous regions, and consequently are difficult to detect manually. These loose deposits are prone to induce the secondary geo-disasters, the process of which is also difficult to predict or simulate. As such, this research focuses on the pressing need of hazard mitigation work, and a geo-disaster chain of “earthquake – loose soil – debris flow” is studied. This research is beneficial by proposing an integrated technique, with which the earthquake-induced loose soils can be detected automatically and rapidly, and subsequent debris flows initiated from the loose soils can be predicted. Two issues regarding this topic should be addressed, (i) the embedded algorithm for the interpretation of remote sensing images, which considers the spectrum features of the loose deposits and addresses the problems arisen from the boundary effect in the local thresholding approach, (ii) the numerical algorithm of the developed Cellular Automata (CA) Model, which combines the influence of topography and debris-flow inertial. The highlight of this integrated technique rests on the fact that the detected loose soil deposits from remote sensing interpretation can be automatically imported into the so-called developed CA model, and possible inundated area of the induced debris flow can be further simulated and predicted. The expected technique requires little computational capability and manual manipulation, and the generated results are consistent with the observed features of this kind of geo-hazards. It is believed that the proposed technique would be instrumental for hazard mitigation in the earthquake-stricken region.

震后产生的松散物质是次生地质灾害的主要风险源，其分布广、范围大、人工调查方法难以适用，且易被触发形成泥石流等次生灾害，危害范围难以预测与模拟。为此，本项目着眼于震区地灾应急响应工作需求，围绕“地震—松散物质—泥石流”灾害链，结合遥感、数值模拟方法，旨在提出一套能够自动地、快速地判识震后产生的松散堆积物及其转化为次生泥石流灾害的危险区域的综合技术。为此，本项目主要研究：（1）针对地物光谱特征的遥感自动解译算法，解决震后松散物质判识中局部阈值分割方法的窗口尺寸选择及边界效应问题；（2）考虑泥石流运动惯性与地形因素的综合作用，改进元胞自动机数值算法。其特色在于：基于遥感的地灾风险源解译结果可自动导入改进的元胞自动机模型，进一步模拟、预测松散物质转化为次生泥石流的可能危害范围。整套技术计算负荷小，适用区域大、人为干预少，计算结果符合地质灾害成灾特性，能够满足震后灾区安置规划及灾害应急响应工作需求。

震后次生地质灾害的链式效应危害显著、作用持久，对灾区重建等民生工作意义重大。为此，本项目围绕“地震—松散物质—泥石流”灾害链，结合遥感、数值模拟方法，开展了如下研究工作：.（1）结合遥感影像数据特点及震后产生的松散物质的光谱特征，构建一套基于窗口尺度选择的遥感影像局部阈值分割方法；（2）考虑泥石流运动惯性与地形因素的综合作用，对震后地质灾害风险源转化为泥石流进行数值模拟；（3）补充开展强侵蚀泥石流对沟床松散堆积物的侵蚀机制分析方法。通过上述工作开展，取得如下创新性成果：.1) 提出了基于局部阈值与蒙特卡洛模拟的二值化分割方法，能够有效识别位于山体阴影处的低灰度滑坡。同时将二值化分割结果与地形数据融合，以地形坡度为指标，排除了河流、城镇、道路等高灰度地物干扰，提高了解译精度。通过改进这一遥感影像对比检测技术，能够有效提高巨灾过后遥感影像及航片的解译效率，为巨灾链型地质灾害物源形成及积累圈定合理范围。.2）提出了基于蒙特卡洛方法的改进元胞自动机模型算法，综合考虑泥石流沟道地形因素及流动惯性因素，确定物源滑坡转化为泥石流的可能泛滥区域。该方法通过概率学手段，分析泥石流路径堆积范围的形成概率。因此其计算效率相比以往方法大为提高，适合于区域性的多沟泥石流模拟，可为震后松散物源滑坡转化泥石流危害区域的判别及风险评价提供依据。.3）提出了基于动量守恒原理的泥石流侵蚀速率计算物理模型。同时，结合课题组前期在复杂地形泥石流沟道断面内的流速场及泥深分析方法，将上述泥石流侵蚀动力学计算模型进行集成，提出了基于黎曼积分与侵蚀动力学模型的复杂地形泥石流沟道断面侵蚀过程分析方法，可实现对复杂地形下泥石流侵蚀作用的时程演化开展量化分析。.上述研究成果共形成SCI/EI论文9篇（其中SCI 7篇，1篇已接收待发表，标注该基金6篇），实用新型专利1项，受理发明专利5项。依托项目开展，课题组成员于2018年入选湖南省普通高校青年骨干教师培养对象。