Groundwater plays a pivotal role in mediating nitrogen transfer to aquatic ecosystems, particularly in arid regions. Water scarcity, coupled with intensive agricultural activities, has placed the groundwater systems under significant pressure from non-point source pollution, underscoring the need for targeted investigation. Focusing on the Chinese Loess Plateau (CLP), we combined dual-isotope analysis (delta 15N-NO3 -, delta 18O-NO3 -) with water isotopes (delta D-H2O, delta 18O-H2O) and implemented a dual-framework approach to investigate nitrate dynamics. Specifically, we applied the MixSIAR model to quantify nitrate source contributions and employed the Geographical Detector model to identify spatial and seasonal drivers. The results showed that local piston-flow recharge predominates beneath the thick vadose zone. Nitrate concentrations decreased with increasing well depth (0.04 mg/L/m), accompanied by a convergence of nitrate isotopic signatures toward soil organic nitrogen (SN). Nitrate was derived primarily from SN (43 %) and ammonium NH4+ fertilizer (NHF) (34 %), underscoring the dominance of agriculture-related sources. Seasonal patterns revealed minor denitrification during the wet season. Spatial analysis identified land use, precipitation, and the normalized difference vegetation index (NDVI) as key factors controlling nitrate variability. Notably, nitrate leaching was strongly driven by precipitation in regions with sparse vegetation cover. These findings demonstrate that, although nitrate transport and transformation in the CLP are governed by its uniquely deep vadose zone and arid hydrogeological conditions, the integrated isotope and spatial framework developed here provides a transferable approach for investigating nitrate dynamics in other vulnerable deep groundwater systems worldwide.