Nitrogen (N) deposition affects forest biogeochemical cycles worldwide, often contributing to N saturation. Using long-term (>30-year) records of stream nitrate (NO 3 -) concentrations at Fernow Experimental Forest (West Virginia, USA), we classified four watersheds into N saturation stages ranging from Stage 0 (N-limited) to Stage 3 (Nsaturated). We quantified NO 3 -contributions from atmospheric and microbial sources using d 15 N, d 18 O, and D 17 O of NO 3 -and characterized the concentrations and isotopes of NO 3 -in precipitation. Despite receiving identical atmospheric inputs, the proportions of atmospheric NO 3 -in streams averaged from 7 to 10% in the hardwood watersheds (stages 1, 2, and 3) and 54% in the conifer watershed (Stage 0). This suggests that the hardwood watersheds may be less responsive to future reductions in N deposition than the conifer watershed, at least in the short term. As shown in other studies, atmospheric NO 3 -proportions were higher during stormflow. Despite large proportions of atmospheric NO 3 -in the Stage 0 stream, total atmospheric NO 3 --N flux from this watershed (2.9 g ha -1 ) was lower than fluxes in the other watersheds (range = 117.8-338.5 g ha -1 ). Seasonal patterns of d 15 N-NO 3 -in the hardwood watersheds suggest enrichment of the soil NO 3 -pool during the growing season due to plant uptake. In all watersheds, d 18 O-based mixing models over-estimated atmospheric NO 3 -contributions to streams by up to 12% compared to D 17 O-based estimates. Our results highlight the importance of atmospheric deposition as a NO 3 -source in low-concentration streams and demonstrate the advantage of using D 17 O-NO 3 -over d 18 O-NO 3 -for NO 3 -source apportionment.