Prediction of groundwater quality in Khanmirza plain using decision tree method

Groundwater is an important resource for agricultural, industrial and drinking uses in arid and semi-arid regions. Therefore, study and management of these valuable resources is necessary. Since the monitoring and investigation of groundwater quality is a time consuming and costly process, so finding a method which enables us to forecast groundwater quality with the least number of parameters lead to save money and time. The aim of this study is prediction of groundwater quality class, based on the USSL diagram in the Khanmirza plain by decision tree method. The chemical parameters and monthly and cumulative precipitation were used as model inputs. For this purpose, the water quality data of 19 wells which located in the Khanmirza plain, were used for the period 1991-2011. The results showed that decision tree method is able to classify water quality with high accuracy, based on only 4 chemical parameters (i.e. EC, Na, SAR and Cation). In the other words, in order to classify the groundwater quality of Khanmirza plain in the future, measuring these four parameters is enough, which significantly reduces the cost and time of analysis.