【作者】 李军霞；

【导师】 王常明；

【作者基本信息】 吉林大学 ， 地质灾害防治工程， 2011， 博士

【摘要】 滑坡作为一种全球性突发地质灾害,影响和危害程度重大,常造成人民生命财产严重损失、严重破坏人类赖以生存的生态地质环境和阻碍区域经济社会的可持续发展,如瓦伊昂滑坡导致数千人死亡和水库失效、新滩滑坡致使长江断流数日、“5·12”汶川地震诱发大量滑坡次生灾害。因此,认清滑坡形成机理和科学有效地评价与预测滑坡灾害具有极大的社会意义、经济价值,是防灾、减灾决策的重要理论基础。论文遵循“地质过程机制分析”和“量化评价”的学术思想,以西藏隆子县列麦乡-加玉乡研究区段内的滑坡灾害为研究对象,以遥感影像和地理信息系统空间分析为信息技术源、以三维电子沙盘可视滑坡场景为辅、野外现场调查与核实为主的综合集成研究途径,对滑坡典型地质特征、发育特征、空间分布特征、形成条件及影响因素进行了相关分析,宏观探讨了滑坡灾害形成的共有规律性和研究区滑坡灾害特有的形成机理,基于模糊物元理论、突变理论分别建立起滑坡危险性评价计算模型,实现了滑坡危险程度的预测和可靠性对比验证,基于经验模型进行滑坡水平最大滑移距离预测,采用BP神经网络模型进行了滑坡特征参数与滑坡水平最大滑移距离的非线性映射与预测,取得以下主要研究结论：(1)借助遥感影像和GIS空间分析,以研究区遥感影像为地表纹理贴面与相应地数字高程模型镶嵌融合,制作出三维电子沙盘纹理地形,形象地展示出滑坡灾害体的空间分布,提高了灾害特征识别的时效性,是行之有效的分析方法。(2)现场调查得知,隆子县列麦乡-加玉乡段内滑坡灾害主要分布在隆子雄曲、觉拉雄曲、伦巴支沟河谷两岸、重要交通干线和人口居住密集区,具有不同程度的潜在危险性,给当地居民生命财产安全带来的威胁不容忽视。(3)研究区滑坡发育特征、空间分布特征表现为：滑坡物质以土质滑坡为主,规模以大型滑坡为主,易滑地层以侏罗系日当组为主,滑坡为老滑坡,运动形式属牵引式,滑坡空间分布主要在3500～4300m高山区,阳坡比阴坡利于滑坡分布,侏罗系日当组、维美组、遮拉组、陆热组为主要分布地层,滑坡分布面积随断层影响距、道路影响距、河流影响距的增加呈减少态势。(4)区域性滑坡灾害形成机理的宏观共有规律性表明：基本因素(地形地貌、地质构造、地层岩性等)是滑坡灾害产生及发展演化的内在因素,诱发因素(降雨强度、河流冲刷、人类工程活动等)是促进滑坡灾害发生的外在原因,滑坡形成是基本因素和诱发因素共同耦合作用的结果,是个复杂的非线性动态系统。(5)研究区滑坡灾害特有的形成机理表现为：区内滑坡为堆积土滑坡,堆积体表层物质经多次改造滑坡堆积而成,灾害形成经历滑坡体堆积阶段→变形扩展阶段→滑动破坏阶段的演化过程,未来变形破坏模式为滑塌-拉裂式和蠕滑-拉裂式,最终导致滑坡灾害形成。(6)以喜马拉雅山地区隆子县列麦乡-加玉乡段内的地质环境条件和滑坡工程地质特征为研究基础,建立起“滑坡形成机理分析→主要评价指标选取→评价指标量化→数学模型构建→模型计算分析评价”的基本评价思路,可为研究区滑坡危险性评价服务。(7)以基本因素、滑坡体特征、诱发因素为出发点,选取高程、坡向、断层影响距、滑体平均坡度、河流影响距、土地利用等13项评价指标,基于偏好系数法组合赋权的模糊物元理论、突变理论分别建立起滑坡危险性评价模型,判定滑坡危险性分为高度危险、中度危险、低度危险,并与现场调查结果对比验证,滑坡潜在危险以中度危险性最为发育,提出采用突变级数判别滑坡危险性的新依据。(8)以典型滑坡滑移距离进行经验模型预测的适用性检验,采用适宜经验模型予以研究区滑坡滑距预测,基于BP神经网络基本理论,以滑体高度、滑坡体积、滑体平均厚度、滑坡平均坡度作为输入量,以滑坡水平最大滑动距离为输出量,建立起4-9-1三层BP人工神经网络模型,实现滑坡特征参数与滑坡水平最大滑移距离的非线性映射与预测,综合确定出研究区滑坡危害范围并进行潜在危害程度评估,滑坡潜在危害程度有特重级、重级、中级、轻级。更多还原

【Abstract】 The influence and damage of landslide taken as a global and sudden geological disaster is significant, which often causes the great loss of life and property, seriously destroys the geological ecology environment of human livelihood and hinders the sustainable development of regional economies society. Taking an example, Vajont landslide led to thousands of persons’s death and reservoir expired, the new beach landslide raised the Yangtze River to be blocked and unflowed by several days, and 5.12-Wenchuan earthquake induced massive landslides which is the secondary disasters of the earthquake. Therefore, recognizing formation mechanism clearly and appraising of landslide disaster sciencely and effectively have enormous social significance, economic value, which is also an important theoretical basis for the decision-making of prevention and mitigation of disaster.Following academic thought of "mechanism analysis of the geological process" and "the quantitative appraisal", the landslides within the zone of Liemai-Jiayu town, Longzi county in Tibet is taken as the research object, remote sensing and geographic information system spatial analysis can be taken as information technology sources supplemented by three-dimensional electronic sandbox of landslide visualization scenarios, spot investigation and checking primarily as metasynthesis research way. Typical geological feature, growth characteristics, spatial distribution characteristics, formation conditions and influencing factors are analyzed correspondingly. The common law and specific formation mechanism of landslide in the study area is macroscopically discussed. Based on fuzzy matter-element theory and catastrophe theory, the appraisal computation models of landslide risk are suggested separately, and the reliability of landslide hazard prediction are contrastly checked. The maximum sliding distance of the landslide is predicted based on empirical model. The nonlinear mapping and prediction of landslide parameters and maximum sliding distance is completed by BP neural network. The main research conclusions are as follows:(1) The terrain texture of remote sensing image and the digital elevation model are correspondingly mosaic fusion by GIS spatial analysis. The three-dimensional electronic sandbox of terrain texture is made, which demonstrate the landslide disaster body’s spatial distribution vividly and improve the timeliness of disaster feature’s recognition. It is an effective method of analysis.(2) Spot investigation shows that landslide disasters within the zone of Liemai-Jiayu town, Longzi county in Tibet are mainly distributed at both banks of Longzi river, Juela river and Lunba tributary valley, important transportation route areas and densely populated areas. Landslides in study area have different degrees of potential danger, which bring the threat to the local resident personal safety and property which can not be ignored.(3) Landslide development, spatial distribution, characteristics of the formation conditions and factors:The developmental scale of the landslides is large and the soil landslides in study area are formed by general materials. Generally speaking, the Jurassic layer is the slippery layer. The landslides here are mainly old and traction slope. They are mainly developed in the elevation range of 3500～4000m. The sunny slopes are more fitable to the distribution of landslides. The main layer is Ridang group of Jurassic layer, Weimei group, Zhela group, Lure group. The number of landslides decreases as the influence distances of faults, rivers, roads increase.(4) The Regional regularity of Landslide formation mechanism indicates that based factors (topography, geological structure, lithology, etc.) are intrinsic factors of landslide generation, development and evolution. Induced factors (rainfall intensity, river erosion, human engineering activities, etc.) are external factors of landslide promotion and occurrence. The formation mechanism of landslide is the based and induced factors coupled together, which forms a complex non-linear dynamic system.(5) The formation mechanism of landslides in study area:the landslides in study area are formed by accumulating soil that is made up by many times landslide accumulations. The formation can be concluded as stage of landslide accumulation→stage of deformation developing→stage of sliding. The future deformation styles are mainly slump-crack-type and creep-crack-type which finally lead to the formation of landslide.(6) Based on unique geological environment condition and landslide geological characteristic of Liemai-Jiayu town, Longzi county in Himalayan area, landslide risk assessment system of "analysis of landslide formation mechanism→selection of main evaluating indexes→quantification of evaluating indexes-establishment of mathematical model→calculation and evaluation of model" are established. This system can be used for the analysis of landslide risk in study area.(7) Starting from geological factors, landslide characteristics, environmental factors, altitude, aspect, fault affecting distance, landslide average slope, river affecting distance, land utilization and so on 13 indexes are selected. Based on fuzzy element theory and catastrophe theory, the landslide risk models are separately suggested and the results are verified and compared with field survey results. Landslide risks are divided into high risk, moderate risk, low risk, and the potential risk level of landslides is mainly middle. A new judgment of landslide risk is proposed by catastrophe progression method.(8) The applicability of empirical model was verrified by classical sliding distance of landslide which can choose a proper empirical model to predict the sliding distance in the study area. Based on theory of BP neural network, taking landslide height, landslide volume, landslide average thickness, landslide average slope as the input and landslide maximum sliding distance as output, the author establishs a three-layer BP artificial neural network of 4-9-1 type which builds a relationship between landslide parameters and maximum sliding distance very well. The range of landslide hazards in study area was confirmed and the landslide potential hazard was estimated. The potential hazard of landslides is mainly classified as level of very heavy, heavy, media and light in this study.更多还原