Performance Comparison of ANN and Geo statistics Methods for Estimating Spatial Distribution of Sodium Adsorption Ratio (SAR) in Groundwater

ABSTRACT:Water is an important compartment of the environment that is particularly easy to compromise, sensitive to short and long-term pollution and directly affects sustainability of ecosystems and human health. The chloride content and the sodium adsorption ratio (SAR) are among the most important water quality parameters, that is, the estimation of these two parameters and, consequently, Na+, Ca2+ and Mg2+ concentration are much more time-consuming and expensive than water salinity measurement. Thus, it seems necessary to find a guideline to estimate SAR and chloride using salinity extent.In the present study were used ANN and Geostatistics Models. Results show that between 34 studied wells in Birjand region during (1989- 2010) 30 of them had high rate of sodium. Cokriging with semivariorum as a spherical model is the best method for estimation spatial distribution of sodium adsorption ratio in groundwater’s in Birjand plain. (MAE= 83/37, RMSE= 105/12 and R2= 0/732). Estimated outputs from Cokriging with the networking on the surface area combined with optimized ANN by Genetic Algorithm. Evaluation criteria of this combination method is (MAE: 63/72, RMSE: 73/45 and R2= 0/896). In general, the two methods have suitable accuracy to estimate SAR and CL based on water salinity parameter.