Comparison of Standard Drastic and Nonparametric Models Instance-Based Learning with parameter K (IBK) and the Tree Decision M5 in Determination of Groundwater Pollution Potential (Case study: Kuchesfahan- Astane plain)

Abstract

Background and Aim:Due to the increasing demands of the human population to groundwater, protection and prevention of these water resources from pollution are necessary. The purpose of this study was to evaluate the vulnerability of groundwater aquifer in Kuchesfahan- Astane plain located in Gilan province using DRASTIC method and nonparametric models. Materials and Methods:In this study, seven layers were prepared for parameters in GIS software, and after weighting and combining standard ranks, the groundwater vulnerability maps for the study area were prepared. Nitrate data were used to validate the model in this region. Subsequently, by using the nonparametric models, Instance-Based Learning with parameter K (IBK) and the Tree Decision M5, the amount of nitrate was estimated. Meanwhile, Gamma test was conducted to find the best combination of input parameters. ResultsThe results revealed that the vulnerability of groundwater aquifer in this plain has 4 classes including 18.56 % in low vulnerability, 51.29 % in low to medium vulnerability, 28.46% in medium to high vulnerability, and 1.67% in high vulnerability classes. Also, the results showed that both of the nonparametric models have suitable estimates of the nitrate content, but the M5 decision tree model yielded the best results (R2=0.98). Conclusion:The results showed that nonparametric models are efficient method to estimate the aquifer vulnerability and provide accurate results to estimate the potential of contamination in the study area.This demonstrates the superiority of the M5 model over other aquatic vulnerability assessment methods.

Keywords

• decision tree
• geographic information system
• groundwater
• population