Using PCA-APCS-MLR model and SIAR model combined with multiple isotopes to quantify the nitrate sources in groundwater of Zhuji, East China

Abstract

Rapid urbanization intensifies nitrate (NO3−) pollution in groundwater. Based on hydrochemical analysis and stable isotopes (δD-H2O, δ18O–H2O, δ15N–NO3- and δ18O–NO3-, a principal component analysis-absolute principal component score-multiple linear regression (PCA-APCS-MLR) model and a Bayesian isotopic mixing model (SIAR model) were used to quantify groundwater pollution sources, especially NO3− sources, in Zhuji, East China. The results showed that the main chemical types in Zhuji were HCO3–Ca in shallow groundwater, HCO3–Ca and HCO3–Ca·Na in deep groundwater. The δD-H2O and δ18O–H2O values in groundwater ranged from −43.4‰ to −32.1% and from −7.0‰ to −4.8%, respectively, and atmospheric precipitation was the main recharge source of groundwater. The calculation of the PAC-APCS-MLR model showed that the potential pollution sources in shallow and deep groundwater were hydrogeological conditions, agricultural activities and domestic sewage/manure, with the total contributions of 77.62% and 71.90%, respectively. The contributions from agricultural activities and sewage/manure to TN, NO3− and NH4+ in shallow groundwater (>82.00%) were higher than those in deep groundwater (>74.00%). The δ15N–NO3-(6.0‰–33.0‰) and δ18O–NO3-(7‰–29.9‰) values of groundwater indicated the existence of clear denitrification in groundwater in Zhuji. The results highlighted that an accurate isotopic fractionation enrichment factor could improve NO3− source apportionment by the SIAR model. When the denitrification isotopic fractionation enrichment factors were εN = −5.3‰ to −8.7‰ and εO = −8.0‰ in the SIAR model, the NO3− contributions from sewage/manure accounted for 51.7%–40.9% in shallow groundwater and 49.3%–40.3% in deep groundwater. The NO3− contributions from soil nitrogen and chemical fertilizer were relatively low, those in shallow groundwater were 25.6%–28.4% and 22.7%–30.7%, respectively; and those in deep groundwater were 27.2%–29.8% and 23.5%–29.9%, respectively. It was demonstrated that PAC-APCS-MLR model coupled with SIAR model was an effective method for source identification and apportionment in groundwater.