Quantification of nitrate sources in groundwater using hydrochemical and dual isotopic data combined with a Bayesian mixing model

To determine the contributions of various sources to nitrate loads in groundwater (n=103) from three farmed rural sites in South Korea, we assessed the dual isotopic composition of nitrate, coupled with hydrochemical parameters. The measured isotopic values of nitrate varied widely, with 10th and 90th percentiles of 2.8‰ and 14.2‰ for δ¹⁵N, and 0.7‰ and 8.9‰ for δ¹⁸O, respectively and suggested that nitrification of anthropogenic N compounds is a predominant process of nitrate pollution. To overcome the known uncertainties of quantitative source apportionment for groundwater nitrate, arising from isotope fractionation during N transformations and the significant overlap in isotopic values of different sources, the combined hydrochemical and isotopic datasets were interpreted with orthogonal regression of a principal component analysis (PCA) and a Bayesian mixing model. The PCA projected the observed δ15NNO3 values onto a mixing subspace in the multivariate variability of the dataset, and the regression fits of the sample data were presumed to be conservative mixtures. This procedure also allowed for an assessment of the sample uncertainty, as influenced by natural nitrate contributions and denitrification. The Bayesian mixing model was used to estimate the probability distributions of the proportional contributions of three anthropogenic sources: chemical fertilizers, composted manure, and manure slurries. Nitrate is largely derived from chemical fertilizers with fractional contributions of 0.35-0.71, and organic fertilizers including composted manure with mixing fractions of 0.39-0.49. The relative contribution of nitrate from composted manure compared to chemical fertilizers increased with increasing nitrate concentrations, suggesting that composted manure significantly increases nitrate pollution and therefore its use should be carefully controlled to manage rural groundwater quality. This study also suggests that PCA and Bayesian isotope mixing models are effective for quantitative assessment of the sources of pollutants, such as nitrate.