Groundwater pollution source identification using a hybrid Metropolis-Hastings and particle filter algorithm.

Area of Science:

• Environmental Science
• Hydrogeology
• Computational Science

Background:

• Groundwater pollution source identification (GPSI) is crucial for environmental protection.
• Existing Metropolis-Hastings (MH) and particle filter (PF) algorithms have limitations in efficiency and stability for probabilistic parameter estimation.
• High computational costs and uncertainties challenge traditional GPSI methods.

Purpose of the Study:

• To develop an efficient and stable hybrid algorithm for GPSI.
• To reduce the computational burden of numerical simulations in GPSI.
• To improve the accuracy of pollutant release and hydrogeological parameter estimation.

Main Methods:

• Proposed a hybrid Metropolis-Hastings and particle filter (PF-MH) algorithm.
• Integrated a residual neural network (ResNet) surrogate model to replace numerical simulators.
• Validated the method using two hypothetical aquifer cases with varying complexity.

Main Results:

• The ResNet surrogate model demonstrated superior performance over the multilayer perceptron (MLP) model.
• The PF-MH algorithm achieved higher accuracy and stability compared to the standard MH algorithm.
• Reduced average relative errors in pollutant release and hydrogeological parameter estimates were observed.

Conclusions:

• The integration of ResNet surrogate models with the PF-MH algorithm offers an efficient and reliable solution for GPSI.
• This approach effectively addresses the challenges of high computational costs and uncertainties in groundwater remediation.
• The developed method shows significant promise for practical applications in groundwater protection.