Nitrate pollution is one of the major environmental issues in reservoirs. Small and medium-sized reservoirs are the sources of drinking water more sensitive to seasonal variations in diffuse pollution. This study aims to analyze the changes in nitrate sources in a reservoir in a hilly watershed under agricultural cultivation in different periods. The Qiaodian Reservoir basin was selected as the study area. 16 sites were set to collect the water samples in January (freeze-up period), March (ablation period), June (pre-flood period), August (high water period), and November (low water period) in 2023. These samples were examined for water quality indicators, major ion compositions, and nitrogen and oxygen isotopes. Different sources of nitrate contamination in the water and their contributions were identified using hydrochemistry analysis, the nitrogen and oxygen isotope tracer technique, and the Bayesian stable isotope mixing model (MixSIAR). Hydrochemical analysis showed that the hydrochemistry was dominated by the HCO₃·SO₄-Ca type in the study area, where the dissolution of rock weathering seriously controlled the ionic composition. The ionic sources were enhanced by water-rock interaction during the abundant water period. There were relatively consistent trends in the TN and NO₃^--N concentrations, with NO₃^--N emerging as the primary form of dissolved inorganic nitrogen. Hydrometeorological conditions, land use patterns, and anthropogenic activities primarily contributed to the fluctuations in the nitrate concentrations. There was dry land (35.8% of the watershed area) with various crops. Livestock farming existed in the villages, where more animal manure was applied to the farmland. The rest types of land use were forest land and grassland, which accounted for 29.6% and 28.5% of the watershed area, respectively. Temporally, the nitrate concentrations declined in the descending order of freeze-up period (3.83 mg/L), high water period (3.57 mg/L), ablation period (3.51 mg/L), low water period (2.54 mg/L), pre-flood period (1.90mg/L). At the spatial scale, NO₃^--N concentrations were more variable in the upper and middle reaches of the watershed, while the downstream NO₃^--N concentrations were close to those in the reservoir area. The δ¹⁵N-NO₃^- mean values of nitrate were 9.61‰, 9.11‰, 8.1‰, 7.18‰, and 6.04 ‰ in the pre-flood, ablation, low water, freeze-up, and high water periods, respectively. The δ¹⁸O-NO₃^- mean values of nitrate were 9.52‰, 4.25‰, 3.74‰, 3.46‰, and 1.96 ‰ in the pre-flood, low water, high water, freeze-up, and ablation periods, respectively. The range of δ¹⁵N-NO₃^- and δ¹⁸O-NO₃^- values varied outstandingly in the different periods, indicating the multiple sources of nitrate. Various analyses showed that soil nitrogen, manure, and sewage were the pivotal contributors to nitrate concentration shifts within the reservoir basin. MixSIAR model was used to quantitatively assess the contribution rates of different nitrate sources. Nitrate was derived mainly from the soil nitrogen, manure, and sewage during the freeze-up, ablation, and low water periods. The proportions of nitrate sources were more consistent between the freeze-up and the low water period. Specifically, soil nitrogen contributed the highest proportion of nitrate to the watershed, 37% and 36%, respectively. Nitrate depended on the atmospheric deposition during the pre-flood period, accounting for 13%. There was the most severe loss of soil nitrogen during the high water period, when the highest contribution rates of nitrate from soil nitrogen and chemical fertilizer were 41% and 31%, respectively. This finding can provide a scientific basis for preventing and controlling surface pollution in the small and medium-sized reservoir watersheds in hilly agricultural areas.