Identification of illegal groundwater pumping in semi-confined aquifers

Abstract

Illegal pumping from underground reservoirs is approached as an inverse problem. The procedure seeks among a set of suspected potential areas the ones from which water is effectively pumped out (or in) and at what rate does this take place? Influence coefficient and finite element are combined in order to simulate the aquifer response to hypothetical pumping scenarios. Then the optimal scenario which makes of simulated head distribution the best fit to the observed headset is retained as solution to the foregoing question. This optimal scenario is derived by minimizing the classical least squares error using the Levenberg-Marquardt Algorithm (LMA). Under the conditions of the treated aquifer example it is observed that LMA is non-convergent unless a regularizing measure is taken during the calculation of the correction applied to the parameter-vector between two successive iterations. The procedure is tested with uniform and non-uniform pumping. In both cases, it yields good identified pumping rates within the observation period, but its performance deteriorates markedly for the rates corresponding to the days when no monitoring well is yet active.

Citation Saffi, M. & Cheddadi, A. (2010 Saffi, M. 2008. Contribution of Influence Coefficients in Solving Groundwater Problems, Rabat, Morocco: Mémoire d'Habilitation Universitaire, Ecole Mohammadia d'Ingénieurs.  Google Scholar]) Identification of illegal groundwater pumping in semi-confined aquifers. Hydrol. Sci. J. 55(8), 1348–1356.