Improving groundwater level forecasting with a feedforward neural network and linearly regressed projected precipitation Ioannis K. Tsanis, Paulin Coulibaly and Ioannis N. Daliakopoulos ABSTRACT Ioannis K. Tsanis (corresponding author) Paulin Coulibaly Department of Civil Engineering, McMaster University, Hamilton, Ontario, Canada E-mail: tsanis@mcmaster.ca Ioannis N. Daliakopoulos Department of Environmental Engineering, Technical University of Crete, Chania, Greece A module that uses neural networks was developed for forecasting the groundwater changes in an aquifer. A modified standard Feedforward Neural Network (FNN), trained with the Levenberg – Marquardt (LM) algorithm with five input variables (precipitation, temperature, runoff, groundwater level and specific yield) with a deterministic component, is used. The deterministic component links precipitation with the seasonal recharge of the aquifer and projects the seasonal average precipitations. A new algorithm is applied to forecast the groundwater level changes in Messara Valley, Crete, Greece, where groundwater level has been steadily decreasing due to overexploitation during the last 20 years. Results from the new algorithm show that the introduction of specific yield improved the groundwater level forecasting marginally but the linearly projected precipitation component drastically increased the window of forecasting up to 30 months, equivalent to five biannual time-steps. Key words | aquifer overexploitation, artificial neural networks, forecasting, groundwater, Messara Valley, specific yield NOMENCLATURE AI Artificial Intelligence BPTT Backpropagation Through Time FNN Feedforward Neural Network IDNN Input Delay Neural Network LM Levenberg – Marquardt (algorithm) MLP MultiLayer Perceptron RMSE Root Mean Square Error RNN Recurrent Neural Networks A evaluation criterion A forecast forecast criterion A reference reference criterion A perfect perfect fit of criterion E i residual error in results e i percentage error H groundwater level N number of observations R annual aquifer recharge S y specific yield skill skill score t time-lag t m monthly time-step t b biannual time-step W annual aquifer withdrawal y i observed variable ^ y i calculated variable INTRODUCTION Groundwater is an inherent part of the hydrological cycle. While precipitation and surface water bodies recharge the underground water bodies, groundwater steadily flows towards a discharge point or is stored in underground geological formations. Provided that the groundwater is primarily influenced by hydro-meteorological processes, the water table fluctuates periodically. Under natural conditions, the aquifer fluctuates around a multi-annual average and there is a balance between annual recharge and withdrawal. doi: 10.2166/hydro.2008.006 317 Q IWA Publishing 2008 Journal of Hydroinformatics | 10.4 | 2008 Downloaded from https://iwaponline.com/jh/article-pdf/10/4/317/386301/317.pdf by guest on 07 June 2020