G. R. Rakhshandehroo - Dept. of Civil and Environmental Engineering, Shiraz University, Shiraz, Iran.
H. Damangir

In this study, RBF neural network was utilized to estimate groundwater level in an unsampled piezometric well. Six scenarios were performed using the data collected from piezometric wells distributed around the unsampled well to estimate the missing groundwater level. The results showed that the estimated groundwater level time series closely follows the entire fluctuation trend in water table with a mean absolute error of 0.13 meter. The distance between the sampled zone and unsampled area affected the correlation in the data drastically. Besides the distance, station elevation played an important role in correlating groundwater changes in adjacent stations.

Groundwater level estimation, Empirical models, Artificial neural network, Radial basis function.