基于多参数随机场与数据同化的降雨型滑坡动态概率预测

Rainfall-induced landslides occurs frequently in the southwestern mountainous regions, southeastern coastal areas and northwestern Loess areas of China, causing severe damage to the nearby communities and infrastructures. Therefore, it is very important to accurately estimate and predict the transient probability of landslides triggered by rainfall. However, if the prediction model itself is not realistic, the probability estimations of landslides would not be accurate, thereby leading to ineffective measures for landslide mitigation. It is proposed here to include soil spatial variability in slope stability prediction models, to investigate the influence of spatial variability of multiple soil parameters on the advancing mechanisms of the wetting front and its effect on landslide probability estimate under rainfall conditions, using the random finite element method. In the meanwhile, for some critical projects and areas, there are often various on-site monitoring programs such as pore water pressure measurements which is particularly relevant for rainfall-induced slope instability due to the suction reduction mechanism in unsaturated soils. It is therefore proposed to assimilate these monitoring data to update the uncertain random fields and thereby reducing the uncertainty of the parameters and improving the prediction accuracy of landslide probability, by using inverse modelling techniques. Based on a more realistic forward prediction model and more accurate model parameter estimates by inverse analysis, the transient probability of rainfall-induced landslides will be investigated for a real engineering slope by high performance implementation based on cloud or grid computing.

降雨诱发的滑坡，在我国西南山区、东南沿海地区和西北黄土地区等频繁发生，给这些地区的人们及基础设施造成严重危害。因此，准确的降雨型滑坡动态概率预测尤为重要。然而，如果滑坡灾害预测模型本身不够准确的话，则无法进行滑坡概率的准确预测，从而无法采取有效的备灾行动。本项目将采用随机场有限元模型分析降雨入渗条件下多个参数空间变异性对湿润锋推进的影响规律及其对滑坡概率估计的影响，为降雨型滑坡稳定性分析提供更加准确的正分析模型。同时，对一些关键工程和地区，由于滑坡后果严重，往往采取多种监测手段，而对于降雨型滑坡，由于其非饱和基质吸力减小机制，动态孔压监测尤为重要，本项目将利用数据同化技术与反分析方法，融合监测数据，对参数随机场进行更新，从而减小模型参数的不确定性，提高预测精度。基于以上正反两方面核心内容的研究，做边坡工程实例分析，通过云计算或网格计算技术，高效地为降雨型滑坡的动态概率提供准确的预测。

降雨诱发的滑坡，在我国西南山区、东南沿海地区和西北黄土地区等频繁发生，给这些地区的人们及基础设施造成严重危害。单体降雨型滑坡的主要分析方法长期以来以确定性方法为主，采用单一的均匀土体参数，未能全面考虑降雨过程中水力、力学参数固有的空间变异性，因而基于此类方法的滑坡预警预测不准。为采取有效的备灾行动，考虑降雨过程中相关水力力学参数空间变异性的降雨型滑坡动态概率预测尤为重要。本项目采用随机场有限元模型分析了降雨入渗条件下多个参数空间变异性对湿润锋推进的影响规律及其对滑坡概率估计的影响，为降雨型滑坡稳定性分析提供了更加准确的正分析模型。同时，利用集合卡尔曼滤波数据同化技术与地质统计克里金方法，融合现场测试和岩土结构响应监测数据，对水力力学参数随机场进行更新，从而减小了模型参数的不确定性，提高了滑坡概率预测精度。基于以上两方面核心内容的研究，结合边坡工程实例分析，通过云计算或网格计算技术，为高效的降雨型滑坡的动态概率预测提供了正反两个方面的方法支撑。