Spatiotemporal Groundwater Level Forecasting in Davarzan Plain

In this article, a hybrid, artificial neural network-geostatistics (Kriging) methodology is utilized to predict the spatiotemporal groundwater level in Davarzan plain in Khorasan Razavi province in Iran. The data for the study were the groundwater levels of 5 piezometers from September 2003 to April 2012 which were recorded on monthly basis. Neural network was used for predict the groundwater level of the successive months and geostatistic were used to estimate the groundwater level at any desired point in the plain. To determine the accuracy and efficiency of model, the method was tested on a new piezometer (Bagherabad) at the first stage. The results were compared with the actual value. And the results (E=0.812) show the efficiency of model. Then, based on appropriate achieved results, the groundwater level was predicted in the month ahead. The results show that neural network with average coefficient of determination (E=0.688) and Gaussian variogram with (R2=0.657) had high efficiency for predicting the groundwater level in this plain.
[1]- Assistant Prof., Dept., of Civil Eng.; University of Qom; Iran (Corresponding author), Email:taher_rajaee@yahoo.com.
[2]- M.A Student; Hydr
Spatiotemporal Groundwater Level Forecasting in Davarzan Plain
Article 1, Volume 1, Issue 4, Winter 2017, Page 1-19 XML PDF (854 K)
Abstract
Taher Rajayee[1]*
Fatemeh Pouraslan[2]
Abstract
In this article, a hybrid, artificial neural network-geostatistics (Kriging) methodology is utilized to predict the spatiotemporal groundwater level in Davarzan plain in Khorasan Razavi province in Iran. The data for the study were the groundwater levels of 5 piezometers from September 2003 to April 2012 which were recorded on monthly basis. Neural network was used for predict the groundwater level of the successive months and geostatistic were used to estimate the groundwater level at any desired point in the plain. To determine the accuracy and efficiency of model, the method was tested on a new piezometer (Bagherabad) at the first stage. The results were compared with the actual value. And the results (E=0.812) show the efficiency of model. Then, based on appropriate achieved results, the groundwater level was predicted in the month ahead. The results show that neural network with average coefficient of determination (E=0.688) and Gaussian variogram with (R2=0.657) had high efficiency for predicting the groundwater level in this plain.