A comparison of statistical methods for evaluating missing data of monitoring wells in the Kazeroun Plain, Fars Province, Iran

Groundwater is the most abundant freshwater resource and it needs protection and sustainable management. Therefore, knowing the groundwater level is important for many reasons. For instance, groundwater level monitoring provides essential data for decision-makers to adopt policies in order to provide a plan for controlling discharge rate and contaminant sources in aquifers. Many regions have missing data in long-term groundwater-level observations that can have a significant effect on the conclusions that can be drawn from the data. The aim of this study is evaluating missing data in monitoring wells with statistical methods. Statistical analysis was conducted on the monitoring wells data in the Kazeroun Plain, Fars Province, Iran. The results were compared with traditional methods for filling the gaps. For this purpose, groundwater-level in 39 monitoring wells were considered as variables with 6552 individual samples from 1995 to 2009. One well located in the middle of the plain was selected and its 15 years-groundwater table data were recalculated based on traditional (such as mean estimation approach) and modern methods (such as interpolation and multivariate analysis). According to the results, traditional methods showed a high level of inaccuracy and decreased data quality. Conversely, modern methods solved the missing values problem with high possible accuracy, minimal error and acceptable R-square. In particular, multivariate analysis has been carried out using k-means algorithm, principal components analysis and regression analysis. K-mean analysis as a clustering procedure classified data in five clusters and then the regression analysis method was applied on data. Finally, we present a simple formula that shows a numerical relationship between the selected monitoring wells in a cluster.