Evaluation of Groundwater Quality by the RAGABP Based on AM-MCMC

In water resources evaluation, the groundwater quality evaluation is an important matter, which could provide scientific references to protection and water resources recover and water resources management. So groundwater quality evaluation can also give guidance to sustainable groundwater utilization and to build a water-saving society. Back Propagation Artificial Neural Net based on Real coded Accelerating Genetic Algorithm and Markov Chain Monte Carlo was built and used to evaluate groundwater quality. RAGA was used to optimize topology, initialize weights and bias of BP to accelerate the convergence of BP net; MCMC based on Adaptive Metropolis method was used to produce enough simulated samples for training BP net and to consider fuzziness between adjacent grades of water quality. AM was taken as a new sampling method to improve sampling efficiency of MCMC. Results of study case show RAGABP based on AM-MCMC can improve convergence velocity about a fifth than that of standard BP, and evaluation results of RAGABP are more objective, reasonable than that of other methods listed in text. Furthermore, RAGABP based on AM-MCMC can be used to evaluate surface water quality, flood disaster loss, earthquake disaster loss and so on.