Welcome to NOAH
Groundwater Level Prediction
Tampa Bay, Florida

Problem Definition

High groundwater extractions in Tampa Bay, Florida have produced severe aquifer overdraft, resulting in adverse environmental impacts, including wetlands dewatering, streamflow depletion, and land subsidence (see Water Follies, 2002, Island Press). An artificial neural network (ANN) model was developed to accurately predict highly dynamic head values (i.e. groundwater levels) at 12 monitoring wells located in a complex multi-layered groundwater system under variable pumping and weather conditions. Of the twelve monitoring wells, located in varying proximity to a large wellfield consisting of seven high-capacity production wells, four monitor the unconfined surficial sediment aquifer with the remaining eight monitoring the underlying hydraulically connected limestone aquifer. Figure 1 depicts the hydrogeology of the study area.



The trained ANN was validated with ten sequential 7-day prediction periods, and the results were compared against both measured head and model simulated head generated by an extensively calibrated numerical flow model developed by the Water Company. The absolute mean error between the ANN predicted and the measured head is 0.54 feet, compared to the 2.80 feet absolute mean error achieved with the numerical model over a 71-day validation period. Unlike the numerical model, the ANN accurately mimicked the dynamic head responses to pumping and recharge in the complicated multi-layered groundwater system. In addition, when tested, the ANN's were able to accurately predict head changes over the 10 consecutive stress periods without using measured water levels for ANN re-initialization. That is, measured head values were used only for initialization of the ANN at the start of the 71-day validation period. For the remaining nine-stress period predictions, the head values predicted by the ANN for the previous stress period were used to re-initialize the network.

Figures 2 and 3 compare measured head data (Measured) for two representative wells against the numerical model predictions (Numerical) and the ANN predictions, where the ANN is reinitialized with measured data (ANN-Rei) for each stress period prediction, versus self-initialization of the ANN model (i.e. extended continuous simulation) with its previous predicted value (ANN-Con). As shown, there is minimal degradation for the extended continuous simulation; the mean absolute error for all wells increased from 0.54 to 0.59 feet, demonstrating that the ANN can be used to accurately forecast highly dynamic water level changes over an extended simulation horizon consisting of multiple stress periods.

As depicted by Figures 4 and 5, a sensitivity analysis conducted where select input variables were excluded as predictors also demonstrated that head in the shallow surficial aquifer is most affected by precipitation (P) while head in the deeper limestone aquifer is most affected by pumping (Q) of the municipal supply wells.

The ANN could be used to help optimize pumping of the wellfield. Using the ANN derived state-transition equations as part of the constraint set, formal optimization methodology can be used to maximize total pumping without violating imposed water level constraints at the observation wells. The higher predictive accuracy achieved with the ANN that with the traditional physical-based model would result in identification of superior management solutions that maximize water supply to the extent possible while protecting sensitive environmental resources.

A more detailed overview of this work can be found in the following two journal articles:

Coppola, E., F. Szidarovszky, M. Poulton, and E. Charles. (2003). Artificial Neural Network Approach for Predicting Transient Water Levels in a Multilayered Groundwater System Under Variable State, Pumping, and Climate Conditions.
Journal of Hydrologic Engineering, 8, no. 6: 348-359.

Coppola, E., M. Poulton, E. Charles, J. Dustman, and F. Szidarovszky. (2003). Application of Artificial Neural Networks to Complex Groundwater Management Problems. Journal of Natural Resources Research, 12, no. 4: 303-320.