Groundwater contamination source identification using improved differential evolution Markov chain algorithm

The groundwater contamination source identification (GCSI) can provide important bases for the design of pollution remediation plans. The Bayesian theory is commonly used in the GCSI problem. Usually, we use the Markov chain Monte Carlo (MCMC) method to realize the Bayesian framework. However, due to the ill-posed nature of the GCSI and the system model’s complexity, the conventional MCMC algorithm is time-consuming and has low accuracy. In this study, we proposed an adaptive mutation differential evolution Markov chain (AM-DEMC) algorithm. In this algorithm, the Kent mapping chaotic sequence method, combined with differential evolution (DE) algorithm, was used to generate the initial population. In the iteration process, we introduced a hybrid mutation strategy to generate the candidate vectors. Moreover, we adaptively adjust the essential parameter F of the AM-DEMC algorithm according to the individual fitness value. For further improving the efficiency of solving the GCSI problem, the Kriging method was used to establish a surrogate model to avoid the enormous computational load associated with the numerical simulation model. Finally, a hypothetical groundwater contamination case was given to verify the effectiveness of the AM-DEMC algorithm. The results indicated that the proposed AM-DEMC algorithm successfully identified the contamination sources’ characteristics and simulation model’s parameters. It also exhibited stronger search-ability and higher accuracy than the MCMC and DE-MC algorithms.