Quantifying uncertainty associated with non-point source (NPS) pollution loads and identifying its key sources are crucial for enhancing the accuracy of NPS load simulation and prediction, understanding NPS variability under climate change and extreme weather events, and enabling effective watershed pollution control. The objectives of this study are to quantify the uncertainty associated with NPS loads and to determine the impacts of meteorological conditions and underlying surface characteristics on NPS load uncertainty across 74 catchments located in the Mintuo River and Jialing River basins, which are tributaries of the upper Yangtze River. Using eight years of daily monitoring data on flow, water quality, and pollution sources, the relationship between NPS loads and streamflow was established. The non-parametric Alpha Shapes algorithm was employed to characterize flow-dependent variations in NPS loads. The variation range of NPS loads was quantified using three indicators: the upper boundary envelope angle, the lower boundary envelope angle, and the envelope angle. The NPS load of all pollutants—namely ammonia nitrogen (NH3-N), permanganate index (CODMn), total nitrogen (TN) and total phosphorus (TP)—followed consistent right-skewed log-normal distributions. Statistically significant positive correlations were observed between streamflow and both the magnitude and variability of NPS pollutant loads of NH3-N, CODMn, TP, and TN. The variation in maximum pollutant loads was substantially greater than the variation in average loads. Although envelope angles tended to decrease under high-rainfall conditions, both the upper and lower boundary angles generally increased with rainfall overall. The variability in upper boundary envelope angles for NH3-N, CODMn, and TP loads exceeded that of the lower boundary envelope angles. NH3-N and TP loads exhibited greater sensitivity to runoff changes compared to other pollutants. The uncertainty in CODMn and TN loads was predominantly influenced by underlying surface conditions, whereas the uncertainty of TP exhibited dual dependence on both hydrology processes and surface conditions. NH3-N exhibits no significant correlation with underlying surface conditions.