基于窗口粒子滤波算法的土壤水分同化及滑坡灾害预警
Assimilation of soil moisture for landslide disaster warning based on particle batch smoother
林雨珊 1邵伟 1杨宗佶 2董建志 3倪钧钧 4林齐根5
作者信息
- 1. 南京信息工程大学水文与水资源工程学院,江苏南京 210044;水利部水文气象灾害机理与预警重点实验室,江苏南京 210044
- 2. 中国科学院、水利部成都山地灾害与环境研究所,四川成都 610041
- 3. 天津大学地球系统科学学院,天津 300072
- 4. 东南大学交通学院,江苏南京 211189
- 5. 南京信息工程大学地理科学学院,江苏南京 210044
- 折叠
摘要
[目的]在水-力耦合计算中,土壤水力参数通过量化土壤含水量与孔隙水压力的转换关系,决定有效应力及边坡稳定性的计算结果.研究稳健、可靠的数据同化方法,降低土壤水力参数的不确定性,提升土壤水动力模拟的准确性,对降雨型滑坡灾害预警具有重要意义.[方法]通过虚拟算例和实例应用,提出将窗口粒子滤波数据同化方法(简称PBS算法)与渗流-边坡稳定分析模型结合,通过同化土壤含水量数据,达到反演土壤水力参数、模拟土壤孔隙水压力以及预测边坡稳定性的目标.通过虚拟算例,证实了当PBS算法设定大于2d的时间窗口,以及大于80个的粒子(参数样本)时,能够获得较为准确的模拟结果.实例应用选取四川省都江堰市银洞子沟滑坡堆积体,将PBS算法同化三个位置的土壤含水量的野外监测数据,以4d为窗口,更新100个粒子样本的土壤水力参数.[结果]结果表明,土壤含水量的模拟值与实测值基本吻合,且模拟的孔隙水压力及边坡稳定系数能对降雨做出清晰、有效的响应.在经过2～3个窗口更新后,三个探头孔隙水压力模拟值不确定区间大小均小于0.11 m,边坡稳定系数的不确定区间大小分别为0.03、0.01和0.11.针对2017年8月28日的极端降雨诱发的滑坡灾害事件的预警,经PBS算法同化后的土壤含水量、孔隙水压力以及边坡稳定系数都收敛到较窄的集合区间,且当日低于1.0的边坡稳定系数,可警示滑坡风险.[结论]通过虚拟算例及实际应用,证实了 PBS算法可支持稳健、可靠的土壤水力参数估计及渗流过程模拟,在边坡稳定分析及降雨型滑坡灾害预警领域具有广阔的应用价值.
Abstract
[Objective]In coupled seepage-stability analysis,soil hydraulic parameters govern the relation of conversion from soil moisture content to pore water pressure,and therefore dictate the calculation of effective stress and slope stability.The develop-ment of robust and reliable data assimilation method may reduce the uncertainty of soil hydraulic parameters and improve the accuracy of soil hydrological simulation and early-warning of rainfall-triggered landslide disasters.[Methods]In this paper,the particle batch smoother(PBS)is integrated with a seepage-stability model,and the soil moisture content data is assimilated to achieve the objective of inverse estimation of soil hydraulic parameters,simulation of soil pore water pressure,and prediction of slope stability.Both synthetic and real-case numerical experiments were used for validating the purposed method.Through the synthetic numerical experiment,it has been confirmed that the PBS algorithm can achieve more accurate simulation of unsaturated soil hydrology when specifying the time window longer than 2 days and particles numbers(parameter samples)larger than 80.The numerical simulations for selected real-case landslide disaster occurred in Yindongzi Gully,Dujiangyan,Sichuan Province also adopts the PBS algorithm to assimilate in-situ measured soil moisture content at three locations.The posterior soil hydraulic parameters are statistical result of 100 particle samples after 3 times resampling,each with a time window length of 4 days.[Re-sults]Result indicated well agreement between the simulated and measured soil moisture content,and the simulated pore water pres-sure and factor of safety also provide a clear and effective response to rainfall.After 2 to 3 windows of resampling,the uncertainty bands of the simulated pore water pressure of the three locations are all less than 0.11 m,and the uncertainty bands of the factor of safety are 0.03,0.01 and 0.11,respectively.As for the landslide disaster induced by extreme rainfall on August 28,2017,after assimilation by PBS algorithm,the uncertainty bands of simulations of soil water content,pore water pressure and factor of safety are extremely narrow.The PBS algorithm can provide a robust estimation of slope instability(Fs＜1.0)contributing an effective ear-ly-warning of disaster.[Conclusion]In both synthetic and real-case numerical experiment,the PBS algorithm can robustly and reli-ably support estimation of soil hydraulic parameters and soil hydrological processes with sufficient accuracy,and it has great poten-tial and practical value in the field of coupled seepage-stability analysis and early-warning of rainfall-induced landslides.
关键词
渗流-边坡稳定分析/土壤水分数据同化/土壤水动力模拟/窗口粒子滤波/滑坡灾害预警/降雨/滑坡/渗透系数Key words
seepage-stability analysis/soil moisture data assimilation/soil hydrological modeling/particle batch smoother/early-warning of rainfall-triggered landslide disaster/rainfall/landslides/permeability coefficient引用本文复制引用
基金项目
中国科学院成都山地所自主部署创新团队项目(IMHE-CXTD-01)
国家自然科学基金青年科学基金项目(41807286)
国家自然科学基金项目(41877158)
青海省科技厅基础研究项目(2023-ZJ-705)
出版年
2024
国家科技期刊平台
NSTL
万方数据