Inversion Of Hydrogeological Parameters Based On The Nested Bayesian Method

Groundwater simulation has been focused on the uncertainty analysis of groundwater numerical simulation for a long time, which based on Bayesian theory has become focus of the research. Nested sampling is a Bayesian sampling algorithm that can estimate the Bayesian evidence from posterior distribution. By coupling constrained sampling and M-H algorithm make the likelihood values of the samples evolves to the high likelihood region along with the corresponding sample sets.In this paper, NS algorithm was exploited based on GFModel program. The inversion of hydrogeological parameters was carried out for the ideal model and hydrogeological test field in Sanpu. The main content and conclusions are as follows:(1) An ideal groundwater flow model was adopted. While both the number of iterations of the nested sampling algorithm m and the number of samples in active set N were not same, the posterior distribution of parameters and the water level fitting situation of each observation hole that corresponding to the parameter samples in the active set was analysised. It was shown that when N remained equal while m became bigger, the variance of samples in active set became smaller as well as the uncertainty of hydrogeological parameters. The sampling process was more concentrated and the water level of observation wells fitted well. But it was contrary when m remained equal while N became bigger.(2) The underground water level data of the observation wells was obtained through hydrological condition investigation, data collection, indoor test analysis, pumping test and et al. The numerical model of groundwater flow was established. The hydrogeological parameters were inversed by NS algorithm. The posterior distribution of parameters obtained by NS algorithm was relatively concentrated, and the uncertainty of posterior parameters was small. Finally, the 95% confidence intervals of the parameters were obtained according to the mean and variance of the parameter samples, which was [8.29,8.65] of the permeability coefficient K and [0.042,0.090] of the water exchange coefficient C of GHB.