Estimation of the Groundwater Level by using Combined Geostatistics with Artificial Neural Networks (Case Study: Shahrekord Plain)

Zamaniahmadmahmoodi, Reza; Akhondali, Alimohamad; Samadiborojeni, Hossein; Zareei, Heydar

Estimation of the Groundwater Level by using Combined Geostatistics with Artificial Neural Networks (Case Study: Shahrekord Plain)

Document Type : Research Paper

Authors

Reza Zamaniahmadmahmoodi ¹

Alimohamad Akhondali ²

Hossein Samadiborojeni ³

Heydar Zareei ⁴

¹ M.Sc Student, Dep. of Hydrology and Water Resources, Faculty of Water Science, Engineering Shahid Chamran University of Ahvaz, Iran.

² Professor, Dep. of Hydrology and Water Resources, Faculty of Water Science, Engineering Shahid Chamran University of Ahvaz, Iran.

³ Asistant Professor, Dep. of Water Engineering, Faculty of Agriculture, Shahrekord University, Iran.

⁴ Asistant Professor, Dep. of Hydrology and Water Resources, Faculty of Water Science, Engineering Shahid Chamran University of Ahvaz, Iran.

Abstract

One of the most basical issues in groundwater resources management is the estimation of water table from observation well network data. The purpose of this study is estimate the groundwater level using the combination algorithm of the geostatistics and Artifical Neural Networks method. Shahrekord plain was selected as a case study of this work. After selected February 2007 and September 2009 as the months with the maximum and minimum groundwater level (during the studied period of 2003 to 2009), using Co Kriging, Kriging and Inverse Distance Weights, groundwater level has been estimated. The results showed that Co Kriging with semi-variogram Gousian model had the best statistical validation (R²=0.816 and MAE=16.54 for February 2007, and R²=0.854 and MAE=11.87 for September 2009). So this approach with combination Artificial Neural Networks give the best results. Also two types of Neural Networks layers, Multi Layer Perceptron (MLP) and generalized feed forward (GFF) were used with combination Kriging method. The results showed that Multi Layer Perceptron network was effective to estimate groundwater levels with statistical indicators of (R²=0.906 and MAE=12.73 for February 2007, and R²=0.924 and MAE=8.75 for September 2009).

Keywords

Groundwater

Geostatistics

Artifical Neural Networks

Shahrekord Plain