Markov chain Monte Carlo sampling for accurate estimation of spatial distribution of naturally occurring radionuclides radiation (232Th, 40K, and 238U) in Senegal

An inaccurate quantification of the reference level of background radiation introduces an inherent statistical bias. Thus, it is necessary to develop a flexible and easy method for accurate characterization of naturally occurring radionuclides (NOR) reference background level in the mining area. In this paper, we propose the use of Bayesian modeling as an alternative statistics technique to study the spatial distribution of NOR. The Markov chain Monte Carlo (MCMC) approach is used to infer the statistical parameter of naturally occurring gamma-induced radionuclides such as 232 Th, 40 K and 228 Ra. We used a bootstrapping method to obtain an accurate sub-sample and then exclude all potential outliers which are out of the Highest Density Interval (HDI). With the resampled sample, we build a model with a Bayesian statistics method with MCMC to draw an inference of the posterior distribution of the gamma-induced radionuclides.