基于统计模型与现场试验的白龙江中游滑坡敏感性分析研究

发布时间：2019-06-07 08:47

【摘要】：甘肃南部白龙江流域是我国滑坡泥石流灾害四大高发区之一,滑坡泥石流灾害严重地制约了社会经济发展。然而,白龙江流域基于比较和组合不同统计模型的滑坡敏感性评价研究存在很大不足,特别是缺少利用降雨条件下滑坡观测数据验证敏感性评价的结果,以及缺乏利用现场试验数据弥补敏感性评价对滑坡物理力学过程解释能力不足的研究。为了给白龙江流域土地利用规划和灾害风险管理提供更为可靠的定量化依据,推动滑坡敏感性评价方法学研究的发展,并进一步为建设经济适用和有效可靠的区域滑坡监测预警体系提供理论依据和实际指导,本论文开展了对白龙江中游(舟曲-武都段)"5.12"地震前后滑坡空间分布与地形、地质、水文气象和人类活动等影响要素间关系定量分析研究,进行了基于比较和组合多种不同类型统计模型的区域滑坡敏感性优化评价研究,以及开展长历时、高频率人工降雨事件和多种监测手段下的典型滑坡现场试验的敏感性分析研究。取得了以下三个主要研究成果:(1)软弱地层岩性和发育断裂构造是白龙江中游滑坡最为重要的孕育要素,降雨和地震是滑坡最为重要的诱发因子。"5.12"地震前后滑坡在不同影响因子级别间的分布数量和面积发育率的分布规律基本类似。受地震波在山区斜坡放大效应的影响,地震诱发滑坡在海拔较高和坡度较陡斜坡相对分布较多。(2)为进一步定量表达滑坡与影响要素间关系,深化理解滑坡机理,开展了不同类型统计模型的滑坡敏感性评价的比较研究。结果表明半定量AHP(层次分析法)和多元LR(逻辑回归)具有较好解释能力,而人工智能ANN(人工神经网络)和SVM(支持向量机)具有较好的空间分区和预测能力。为有效结合不同统计模型的优点,提出了在比较不同模型的基础上,将专家知识和统计方法相结合建立组合模型的新策略,进一步考虑敏感性评价中错误分类造成的经济成本损失,提出了将成本曲线和组合模型相结合建立滑坡敏感性优化评价模型的方法。研究结果表明,组合模型可极大程度提高模型的分区能力和评价精度以及降低不确定性,结合成本曲线建立的敏感性优化评价可更好的服务于土地利用规划。(3)为弥补基于统计模型滑坡敏感性评价对降雨条件下的滑坡物理力学过程表达的不足,并填补现场观测数据对敏感性评价结果验证研究的空白,选择典型堆积层滑坡开展现场人工降雨模拟试验研究。结果表明,在降雨条件下,堆积层滑坡内部孔隙水压力、含水量、土压力、深部位移等指标对降雨有较为迅速的响应,这种快速响应主要是受裂缝、裂隙和大孔隙等优势入渗通道控制的,并且水分空间迁移过程可以通过高密度电法手段进行有效监测;堆积层斜坡的变形破坏机制可概化为:降雨入渗诱发变形-加速变形诱发剪胀强化-重新固结使孔隙水压力恢复和强度降低;土压力是对浅层堆积层滑坡变形最为敏感的指示因子,可作为白龙江流域广泛分布的浅层堆积层滑坡的关键监测预警指标;将土压力和位移加速度相结合,建立了堆积层滑坡启动的阈值指标体系,可为该区堆积层滑坡预警提供科学指导。
[Abstract]:The BaiLongjiang river basin in the southern part of Gansu is one of the four major high-risk areas of the landslide and debris flow disaster in China, and the landslide and debris flow disasters severely restrict the social and economic development. However, the study of landslide sensitivity evaluation based on the comparison and combination of different statistical models in the Bailong River basin is not enough, especially the results of the sensitivity evaluation of the landslide observation data under the condition of rainfall. And the lack of the on-site test data to make up for the sensitivity evaluation to the lack of the interpretation ability of the physical and mechanical process of the landslide. In order to provide a more reliable and quantitative basis for the land use planning and disaster risk management of the Bailong River basin, the development of the methodology for evaluating the sensitivity of the landslide is promoted. In order to provide the theoretical basis and practical guidance for the construction of the economic application and the effective and reliable regional landslide monitoring and early warning system, the paper has carried out the spatial distribution of the landslide before and after the "5.12" of the middle reaches of the Bailong River (the Zhouqu-Wudu section) and the terrain and geology. Based on the quantitative analysis of the relationship between the factors such as hydrometeorological and human activities, the sensitivity optimization and evaluation of regional landslide based on the comparison and combination of different types of statistical models are carried out, and the long duration is carried out. Sensitivity analysis of typical landslide field tests under high frequency artificial rainfall events and various monitoring methods. The following three main research results have been obtained: (1) The structure of the weak formation lithology and the development fracture is the most important element of the landslide in the middle reaches of the Bailong River, and the rainfall and the earthquake are the most important inducing factors of the landslide. 5.12 The distribution of the number and area development rate of the landslide before and after the earthquake is similar to that of the area development rate. The earthquake-induced landslide is more and more distributed at higher elevation and steep slope due to the effect of the seismic wave on the slope amplification of the mountain area. (2) To further quantitatively express the relationship between the landslide and the influence factors, deepen the understanding of the landslide mechanism, and carry out the comparative study of the landslide sensitivity evaluation of different types of statistical models. The results show that the semi-quantitative AHP (analytic hierarchy process) and the multiple LR (logistic regression) have better interpretation capability, and the artificial intelligence (artificial neural network) and the SVM (support vector machine) have better spatial partitioning and prediction capability. in order to effectively combine the advantages of different statistical models, a new strategy of combining the expert knowledge and the statistical method on the basis of comparing different models is put forward, and the economic cost loss caused by the error classification in the sensitivity evaluation is further taken into account, The method of combining cost curve and combined model to establish the evaluation model of landslide sensitivity is put forward. The results show that the combination model can greatly improve the partition ability and the evaluation precision of the model and reduce the uncertainty, and the sensitivity optimization evaluation based on the cost curve can better serve the land-use planning. (3) In order to make up for the deficiency of the statistical model landslide sensitivity evaluation on the physical and mechanical process of the landslide under the rainfall condition, and to fill the blank of the field observation data on the verification of the sensitivity evaluation result, a typical pile-up landslide is selected to carry out the on-site artificial rainfall simulation test study. The results show that, under the condition of rainfall, the internal pore water pressure, water content, soil pressure and deep displacement of the accumulative landslide have a relatively rapid response to the rainfall, which is mainly controlled by the dominant infiltration channels such as cracks, cracks and large pores. and the deformation and failure mechanism of the build-up slope can be generalized as follows: rainfall infiltration-induced deformation-accelerated deformation-induced shear-expansion strengthening-reconsolidation, and the pore water pressure recovery and the strength are reduced; The soil pressure is the most sensitive indicator for the landslide deformation of the shallow-layer stack, which can be used as the key monitoring and early warning index of the shallow-layer stack landslide which is widely distributed in the Bailong River basin, and the threshold index system for the initiation of the build-up landslide is established by the combination of the earth pressure and the displacement acceleration. It can provide scientific guidance for the early warning of the accumulation of the deposit in this area.
【学位授予单位】：兰州大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：P642.22

【参考文献】