مرکز منطقه ای اطلاع رساني علوم و فناوري - Investigation of relationship between groundwater level and meteorological variables with the artificial neural network at unconfined aquifers

Abstract :

Determination of the distribution of water in a hydrological cycle and the ground and surface water potential correctly in terms of sustaining the water sources are very important. In this study, prediction of the relationship between meteorological data in Develi-Yeşilhisar sub-basin and groundwater levels is aimed with the help of artificial intelligence and multiple linear regression techniques. In this study, daily data of a groundwater observation well of Develi Yeşilhisar sub-basin between 2007-2010 and daily data of Develi meteorology station are used. The available data are divided into two subsets to test the models accuracies. The data of 2007 and 2008 are used to train the model, the data of 2009 and 2010 are used for testing. By trying the various combinations of antecedent meteorological data, the groundwater levels of the ensuing month are estimated. Variables affecting is primarily determined with Principal Component Analysis (PCA). PCA results of linear regression obtained after determining the meaningful variables are compared with Multi-Layer Perceptron (MLP) and Radial Basis Neural Network (RBNN). The findings show that the best way which yields the most accurate results among other techniques is Multi-Layer Perceptron (MLP). The findings also suggest that there is a strong relation between groundwater levels in unconfined aquifers and meteorological data.