Groundwater in agriculture-dominated regions of California has historically experienced nitrate pollution due to the application of excess nitrogen fertilizers. This study examines the nitrate pollution vulnerability of groundwater in sedimentary aquifers of California's South Coast Range using stepwise logistic regression (LR) modeling. Our results indicate an overall excellent model fit, but an acceptable statistical significance, according to a Wald statistic (p-Wald) cutoff of 0.1, for only two explanatory variables: (1) the dissolved oxygen (DO) concentration, and (2) the modern (i.e., less than ~60year old) water proportion (MWP) in the groundwater mixture. The latter parameter was estimated via Lumped Parameter Modeling (LPM) of groundwater tritium, helium and radiocarbon data that have been corrected for isotopic dilution and exchange using a modified Fontes and Garnier (F&G) approach. The observation that other explanatory variables on land cover (i.e., percentage of agricultural land use, abundance of septic tanks and leaking underground fuel tanks, etc.) were statistically insignificant points out the limitations of low-resolution land cover data in groundwater vulnerability assessments. Our results highlight the utility of quantitative groundwater age and mixing data to evaluate pollution probability in the saturated zone. The herein presented approach can thus provide valuable results in comparable settings where the availability of fertilizer application, crop nitrogen uptake, and soil texture data is limited.
Keywords: Groundwater vulnerability; Logistic regression (LR); Lumped parameter modeling (LPM); Nitrate contamination; Radiocarbon; Tritium-helium.
Copyright © 2017 Elsevier B.V. All rights reserved.