wAbstract
@A stepwise logistic regression (LR) model was generated to evaluate
the association between contamination of groundwater by nitrates
with several risk factors such as soil types, farm facilities
and practices, and well characteristics. The odds ratio was calculated
to estimate the degree of impact that the associated variables
had on the risk of contamination in a semiconfined multilayer
aquifer underlying rural areas of Buenos Aires, Argentina. Duplicate
farm groundwater samples (n = 160) were taken and nitrate was
analyzed. Data, involving various farm factors, was gathered via
two questionnaires concerning farm's general and productive aspects,
and well characteristics. Statistical tests were run between nitrates
and each variable present in the survey. A 96.25 of the samples
presented detectable nitrate levels, 40.91 of which had more
than 45 ppm nitrates. The final LR model involved five of the
variables under study: well age, soil permeability, depth of water
table, location, and distance from well to contamination sources.
Cross validation proved to be a good estimator of nitrate water
contamination. Suspicions about how these characteristics influence
groundwater contamination by nitrates were confirmed, and as these
five factors represent a higher risk for this type of aquifer,
their proper management may contribute to a better resource protection.
Keywords: Logistic regression; Nitrates; Groundwater contamination;
Well construction; Argentinax
Introduction
Materials and methods
@Study area
@Sampling and surveys
@Statistical treatment
Results and discussion
Conclusions
Acknowledgments
References