Development and Application of an Artificial Neural Network Model to Forecast Ground-water Flooding Events

Brent E. Huntsman and Daniel J. Wagel
Terran Corporation
4080 Executive Drive
Beavercreek, Ohio 45430

Abstract

Fluctuations of groundwater levels in glacial-fluvial derived aquifers are highly dependent upon recharge, particularly from river or stream channel leakance and precipitation infiltration. Temporal changes in recharge together with the heterogeneity of the aquifer and existing groundwater use create complex interactions between these variables. To address the non-linear nature of the correlations between parameters that effect groundwater levels, an artificial neural network (ANN) model was developed to simulate and forecast water level changes.

The neural network model built to evaluate this relationship utilized long-term river discharge measurements, groundwater elevations, precipitation and temperature records for a portion of the Great Miami River buried valley aquifer in Dayton, Ohio. All data for the initial model were compiled from published online databases and required parsing or averaging to standardize the measurement periods. Using the hydrologic records for the previous twenty years, groundwater levels were simulated, on average, within a few percent of actual measurement values. Based upon river stage, precipitation, temperature and antecedent groundwater levels, the ANN model forecasts are used to predict ground-water flooding events. Depending upon the projected magnitude of these events, dewatering systems may be activated to prevent or minimize the flooding of subsurface structures in the downtown Dayton area.

Presentation Slides (PDF)

Presented at:
The 5th International Conference on Environmental Informatics - ISEIS 2006
August 1-3, 2006
Bowling Green, Kentucky, U.S.A.

Sponsored by:
International Society for Environmental Information Sciences