Scholarlink

<p>Landslide run-out evaluation is a necessary prerequisite for risk assessments and disaster prevention. Due to a lack of direct measurement of rheological parameters and various uncertainties (e.g., parameter uncertainty and model uncertainty) associated with dynamic numerical models, it still remains a difficult task to reasonably characterize rheological parameters and accurately evaluate landslide run-out behavior. This paper develops Bayesian approaches to characterize rheological parameters and model uncertainty of numerical simulation for probabilistic landslide run-out prediction using prior knowledge and multiple measurements of the deposition morphology, including the lateral accumulation width and the longitudinal run-out distance. The updated posterior distributions of rheological parameters and model uncertainty are equivalently represented using random samples generated by Markov Chain Monte Carlo Simulation (MCMCS) in the Bayesian framework. Then, MCMCS samples of rheological parameters are used to perform probabilistic run-out evaluation. To remove the computational burden in Bayesian analysis and probabilistic prediction of run-out hazard, Random Forest (RF) algorithm is used to establish surrogate models of numerical simulation for evaluating run-out characteristics (i.e., the lateral accumulation width and the longitudinal run-out distance) at different locations on the depositional area. Finally, the proposed approaches are illustrated and validated using two Zhongzhai landslides occurred on August 19, 2020, in Gansu Province. Results show that the RF-based surrogate models can accurately estimate the run-out characteristics at different locations on the depositional area with high computational efficiency. The proposed approaches not only can properly identify the rheological parameters, but also reasonably characterize the predictive uncertainty. In addition, the probabilistic prediction of landslide run-out characteristics can be improved by using more informative prior knowledge and more measurement data of the final deposition morphology of landslide.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Probabilistic evaluation of landslide rheological parameters and run-out behavior based on lateral and longitudinal morphological characteristics of the depositional area

  • 期刊论文

  • 发表日期：

    2025年10月23日

查看全文
  • Mi Tian,
  • Jiaheng Xu,
  • Chao Ma,
  • Xiaotao Sheng

摘要

Landslide run-out evaluation is a necessary prerequisite for risk assessments and disaster prevention. Due to a lack of direct measurement of rheological parameters and various uncertainties (e.g., parameter uncertainty and model uncertainty) associated with dynamic numerical models, it still remains a difficult task to reasonably characterize rheological parameters and accurately evaluate landslide run-out behavior. This paper develops Bayesian approaches to characterize rheological parameters and model uncertainty of numerical simulation for probabilistic landslide run-out prediction using prior knowledge and multiple measurements of the deposition morphology, including the lateral accumulation width and the longitudinal run-out distance. The updated posterior distributions of rheological parameters and model uncertainty are equivalently represented using random samples generated by Markov Chain Monte Carlo Simulation (MCMCS) in the Bayesian framework. Then, MCMCS samples of rheological parameters are used to perform probabilistic run-out evaluation. To remove the computational burden in Bayesian analysis and probabilistic prediction of run-out hazard, Random Forest (RF) algorithm is used to establish surrogate models of numerical simulation for evaluating run-out characteristics (i.e., the lateral accumulation width and the longitudinal run-out distance) at different locations on the depositional area. Finally, the proposed approaches are illustrated and validated using two Zhongzhai landslides occurred on August 19, 2020, in Gansu Province. Results show that the RF-based surrogate models can accurately estimate the run-out characteristics at different locations on the depositional area with high computational efficiency. The proposed approaches not only can properly identify the rheological parameters, but also reasonably characterize the predictive uncertainty. In addition, the probabilistic prediction of landslide run-out characteristics can be improved by using more informative prior knowledge and more measurement data of the final deposition morphology of landslide.