Title: Spatiotemporal groundwater level forecasting and monitoring using a neural network-based approach in a semi arid zone

Authors: Soumaya Hajji; Wafik Hachicha; Salem Bouri; Hamed Ben Dhia

Addresses: Water-Energy-Environment Laboratory (LR3E), Engineering School of Sfax, University of Sfax, B.P. 1173, Sfax 3038, Tunisia; Higher Institute of Sciences and Techniques of Water of Gabès (ISSTEG), Cite Erryadh 6072 Gabès, Tunisia ' Unit of Mechanic, Modelling and Production (U2MP), Engineering School of Sfax, B.P. 1173, Sfax 3038, Tunisia; Department of Industrial Management, Higher Institute of Industrial Management of Sfax (ISGI), B.P. 954, Sfax 3018, Tunisia ' Water-Energy-Environment Laboratory (LR3E), Engineering School of Sfax, University of Sfax, B.P. 1173, Sfax 3038, Tunisia; Faculty of Sciences of Sfax, BP no. 1171 – 3000 Sfax, Tunisia ' Water-Energy-Environment Laboratory (LR3E), Engineering School of Sfax, University of Sfax, B.P. 1173, Sfax 3038, Tunisia

Abstract: Forecasting the groundwater level (GWL) of an aquifer is one of the most important tasks in hydrogeology. Such forecasts prove vital to develop better exploitation strategies of water resources and to carry up-to-date groundwater monitoring. In this paper, an artificial neural network (ANN)-based approach, which was based on linear regression model, is presented to simulate and predict both temporal and spatial GWL. The novelty of the proposed ANN model lies in the use of a non-linear model toward the input variables and a linear one toward the coefficients. The proposed approach was applied to the Middle Miocene aquifer groundwater (MMAGW) of Hajeb Elyoun Jelma (HJ) multilayered hydro-system, central Tunisia. The investigation was based on a limited database set of 22 wells during the period between 1994 and 2010. The design of the appropriate ANN model over third-order linear regression model and the reverse GWL prediction were assisted by the neuro software. Then, an iterative water exploitation strategy for the period 2011 to 2013, was performed based on this GWL spatiotemporal ANN model and through various three-dimensional piezometric maps performed by kriging method. Promising results were found for forecasting and monitoring MMAGW in HJ basin.

Keywords: groundwater levels; GWL; water management; linear regression modelling; groundwater resources; spatiotemporal; Neuro One software; water exploitation scenarios; kriging method; piezometric maps; semi arid zones; Tunisia; hydrology science; groundwater level forecasting; groundwater level monitoring; artificial neural networks; ANNs; aquifers; hydrogeology.

DOI: 10.1504/IJHST.2012.052366

International Journal of Hydrology Science and Technology, 2012 Vol.2 No.4, pp.342 - 361

Published online: 16 Aug 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article