Daily groundwater level fluctuation forecasting using soft computing technique

TLDR: The results obtained in this study suggest that GLF monitoring can be conducted by a forecasting model with considering time-lag as inputs, and that the best accuracy was for one-day-ahead prediction.

Abstract: The study presented here deals with forecasting daily groundwater level fluctuation (GLF) for monitoring of GLF pattern. The calculation model is based on the adaptive neuro-fuzzy inference system (ANFIS) and two algorithms of artificial neural networks (ANN) models, namely Levenberg- Marquardt (LM) and radial basis function (RBF). The objective in this study is to predict daily GLF for monitoring purposes. The first step was to investigate the effect of the number time lags as inputs for one- day-ahead prediction using the ANFIS algorithm. It was found that three input nodes containing three time- lag of well studied gave good prediction results. The second experiment was to predict the GLF one to seven steps ahead using the three input nodes. In this experiment, the three soft computing techniques were applied. The results indicate that the performances were decreasing by increasing the time step ahead, and in general there was no significant difference between the three techniques used. The best accuracy was for one-day-ahead prediction. The results obtained in this study suggest that GLF monitoring can be conducted by a forecasting model with considering time-lag as inputs. (Nature and Science. 2007;5(2):1-10).