Premium
Prediction of Groundwater Contamination with Multivariate Regression and Probabilistic Capture Zones
Author(s) -
Lim JeongWon,
Bae GwangOk,
Kaown Dugin,
Lee KangKun
Publication year - 2010
Publication title -
journal of environmental quality
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.888
H-Index - 171
eISSN - 1537-2537
pISSN - 0047-2425
DOI - 10.2134/jeq2009.0336
Subject(s) - regression analysis , groundwater , environmental science , probabilistic logic , tobit model , hydrology (agriculture) , statistical model , soil science , linear regression , regression , multivariate statistics , statistics , mathematics , geology , geotechnical engineering
Probabilistic capture zones are combined with a regression model and used as buffer zones around wells for Tobit regression analysis to predict contaminant concentration of groundwater in an agricultural region. A backward transport equation, which is a mathematical model based on the physical processes of solute transport, is used to delineate probabilistic capture zones. The probabilistic capture zone defines the area where contaminant discharge can have a direct influence, with pertinent probability, on the quality of groundwater pumped from a well. Tobit regression analysis is used to find the relationship between independent regression variables and a dependent variable, which is contaminant concentration in this study. The capture zone and the regression are combined into a model, and its applicability for prediction of nitrate concentration is tested in a small agricultural basin in Chuncheon, Korea, which is occupied mainly by vegetation fields, orchards, and small barns. Three cases of Model 1, Model 2, and Model 3 are compared in which buffer zones are circles, capture zones with probability over 0.1, and capture zones divided into sections with different probabilities, respectively. The resulting regression model describes nitrate concentration in terms of selected independent variables. When the concentrations are calculated with the model, the best fit with the observed concentrations was in Model 3. This result supports the applicability of the method proposed in this study to prediction of contaminant concentration of groundwater.