Forecasting Groundwater Depth Using Time series Spectral Analysis

Nowadays, the maximum operation of groundwater resources has been achieved in Iran. Also, the majority of extractable water resources are utilized and the managing of water resources in the future is depended on more extracting of water resources. For better basin management, forecasting the groundwater depth fluctuations in particular in arid areas is more necessary. In this study, time series spectral analysis is used to forecast the groundwater depth fluctuations of Chamchamal plain. In this regard, the monthly groundwater depth time series during 1995 to 2009 years are used for calibration periods and the periodogram diagrams are depicted. Data periodicity is analyzed by using Fourier spectral analysis and the deterministic term of data periodicity is eliminated. In the next step, stationary and normality in the data are considered. After that, the different time series models are fitted for the prepared data and accuracy of them were assessed by Akaike (AIC) criterion. The results show that ARMA (2, 1), ARMA (1, 1), ARMA (1, 1) models are the best fitted models for the measured data in Bazanabad, Gheshlaghabad and Gavkol piezometers, respectively. Finally, the residuals stationarity assumption test is used to check for the correct diagnosis of the fitted pattern. In this study, the results represent the high performance and accuracy of the applied new approach to the time series spectral analysis for forecasting groundwater depth by application of the regression coefficient amount of 0.78 and SI- Index of 4% to 14% of piezometer's data. Using spectral analysis, as has been provided in this study, is very useful for forecasting groundwater depth.