Evaluation of Groundwater Vulnerability Using Data Mining Technique in Hashtgerd Plain

Due to simple operation and no needs for expensive infrastructure construction to convey water from a source to farm lands, groundwater becomes the most important sources of agricultural water supply in Iran.
However, the contamination of aquifers is a major concern in many countries, specifically in areas without effective groundwater protection and management. Therefore, groundwater vulnerability assessment would be one of the effective informative methods to provide a basis for determining source of pollution. Assessment of groundwater vulnerability is often done by intrinsic vulnerability, which considers hydro-geological conditions. The concept of vulnerability of aquifers was introduced for first time by Marget in 1986. The first definition of vulnerability was proposed by Marget and it means the degree of groundwater contamination by pollution reaching an groundwater system. Overlay and index method could be mentioned as existing method to assess intrinsic vulnerability of groundwater. Moreover the vulnerability index is relatively, dimensionless and immeasurable and depends to hydrogeology and geology of aquifer characteristics. Since then, many researchers applied many methods and techniques to provide a standard way for evaluation of vulnerability. As it mentioned before, it should be noted that all of methods is relatively and dimensionless, using various data depended to sort of aquifer.
Compared to other models, DRASTIC model, which is an overlay and index method, is the most popular index that used in many researches. The DRASTIC model is easy to implement and provides a good basis for assessment of groundwater vulnerability in facing contamination. Also it needs a relatively small amount of data that is often available in many aquifers.
Vulnerability maps applied as one of the effective management ways to qualitative management. Several models like DRASTIC were applied to this end and many researchers have tried to introduce approaches to provide more realistic constant ranks and weights using in the models. Meanwhile application of DRASTIC model is highly influenced by assigning the weights and ranks. Therefore, it is necessary to use a model depended on variable weights and ranks according to the aquifer features.
In this study, clustering technique is employed in regionalization of groundwater flow zone so as to vulnerability assessment of a groundwater case study. To this end, K-Means clustering as unsupervised pattern recognition technique is applied. Thanks to intelligent algorithm of the clustering in finding similarities out of the dataset, the proposed method of this research is capable to be used in each aquifer without considering calibration. The method is employed in a large-scale aquifer in center of Iran and finding results are compared with vulnerability maps of the regions created by DRASTIC approach.
In this paper clustering algorithm as one of the applicable data mining methods is used taking the advantage of be independent from constant ranks and weights. In another word, vulnerability map of each region is provided by clustering according to the specific features of each region. Optimum number of clusters is obtained 4 numbers by applying cluster validity index. Cluster map is then created based on spatial location of the created clusters representing the vulnerability map of the region. The map shows that upstream parts of the field are in high risk of groundwater vulnerability. Here, just four influential parameters are used by the clustering method in providing maps. Other parameters could be considered by clustering based on specific characteristics of each aquifer.