A combined Time Series – Wavelet Model For Prediction of Ground Water Level (Case Study: Firuzabad Plain)

Recently, the phenomenon of climate change, drought, the exploitation of the ground water has caused a sharp decline in groundwater levels, which has also led to groundwater related subsidence and desertification. Therefore, reliable prediction of groundwater level for managing these resources is of prime importance. Nowadays wavelet transforms through signal decomposition to time and frequency has created an exceptional method for signal processing, with the help of the transformation of the wavelet that has the ability to split the time series into a multi-dimensional substrate with different scales.
One of the useful characteristics of wavelet transforms is the filtering algorithm, which divids the data into two groups of approximation and details by passing them through the wavelet filter. In this study, monthly groundwater surface, rainfall, and temperature data were used. Using the program code written in MATLAB software, it is necessary to analyze the time series data of all three parameters of temperature, precipitation, and water level. The values ​​of the parameters were selected as inputs and placed in the wavelet function. For analyzing all the parameters, according to various mother-wavelet experiments, and considering the above mentioned point, five mother-wavelets (Haar, Coif, Symlet, Db, Db4) were selected. For this purpose, the program was first implemented for each of the following wavelets with 4 different decomposition levels. After several software runs under specific conditions and scenarios and then comparing them with each other, results were obtained.