Integrated surface and ground water modeling utilizing artificial neural networks and three-dimensional finite element strategies

Interactions between surface water and ground water were studied. A neural network system was developed to predict the groundwater elevation at monitoring wells from river discharge data. This neural network required 6 months of field data for training and verification of the system. It successfully predicted the groundwater elevation in wells within a mile of the river. This neural network system reduces groundwater monitoring costs substantially and facilitates the establishment of initial and boundary condition data for numerical solvers of flow in porous media.; A three-dimensional, transient, site-specific, finite-element model was established for the integrated surface-ground water system of the study area. Various methods and algorithms were developed to construct the geometry model and deploy the finite element initial and boundary conditions. The finite element model was calibrated with field data and verified using independent field measurements.; The developed neural network system coupled with the finite element numerical model facilitated the establishment of long-term ground-water prediction strategies. A contaminant fate and transport model used this flow field model to predict contaminant plume distributions. The simulated results compared very favorably with existing field data. This integrated surface and porous flow system provides an efficient and economic strategy for water resource management and water quality control.