A Multiple Regression Analysis for Predicting Salinity in Shallow Groundwater

Abstract

Groundwater salinity is a severe problem particularly to agricultural lands. Measuring the water quality index at some particular locations might not be easy. The objective of this research is hence to predict the groundwater salinity, in terms of electrical conductivity (EC) in the shallow groundwater in the Northeast of Thailand. Groundwater EC was measured for the period of over 2 years at 14 different locations at different time intervals. The data was interpolated and analyzed for basic statistical properties including autocorrelation and stationarity. Linear regression model and transformed linear regression models were developed. The two models produced high adjusted R2 about 0.8 during the calibration step. However, the transformed model provided a better accuracy during the verification step. This variation can be attributed to unaccounted factors, collinearity, and stationarity. The model can be applied to predict the groundwater salinity using the groundwater quality measured at some surrounding region.