Mountainous urban rivers are subject to prominent water pollution problems due to the combined effects of complex topography and high-intensity human activities. Characterizing water quality spatiotemporal features and identifying pollution sources are therefore essential for refined watershed management. This study focuses on the Yanjin Stream, a first-order tributary of the Chishui River. We characterize its water quality spatiotemporal features, quantitatively evaluate pollutant overloading through environmental capacity calculations, and apply the Positive Matrix Factorization (PMF) model to identify major pollution sources and their contribution variations under different hydrological conditions, thereby elucidating pollution pattern formation mechanisms driven by the joint effects of hydrological processes and land-use structure.The results indicate that: (1) Except for the TN concentration, which generally shows an upward trend with increasing flow, the changes of other water quality indicators in each flow interval are not significant; pollutant fluxes increase with rising flow and exhibit differentiated reduction requirements across different flow intervals. DOC requires reduction throughout all flow intervals, with reduction ratios of 23.76%~30.93% (25.00~427.93 kg/d). CODMn requires reduction under low-to-medium flow conditions, with a reduction amount of 40.93 kg/d (11.93%). No reduction requirements are identified for NH4+?N or TP. The reduction demand for TN is relatively lower compared with that of regional background rivers, ranging from 80.50 to 416.14 kg/d (14.32%~98.21%) in different flow intervals and from 142.56 to 686.25 kg/d (29.77%~41.38%) in different river sections. (2) Pronounced spatial differentiation of water quality is observed, with overall deterioration from upstream to mid-downstream reaches. The middle reaches exhibit the highest pollutant overloading and reduction demand. DOC and CODMn require substantial reductions of 241.67 kg/d and 92.81 kg/d, corresponding to reduction ratios of 92.80% and 84.79%, respectively. NH4+?N and TP require reductions of 51.30 kg/d (86.16%) and 10.27 kg/d (55.58%). The proportion of built-up land significantly enhances TN, NH4+?N, and TP levels, whereas land-use effects on DOC and CODMn are relatively limited at the macro scale. (3) PMF source apportionment shows that during dry season, domestic sewage and industrial wastewater together contribute more than 60% of total pollution. Domestic sewage predominantly contributes NH4+?N (68.97%) and TP (91.66%), while industrial wastewater is mainly characterized by CODMn (60.11%) and electrical conductivity (62.80%). During wet season, point-source contributions weaken while non-point source contributions intensify, resulting in a more complex pollution source structure. Surface runoff contributes 15.37%, primarily characterized by DOC (43.44%) and TN (28.06%), whereas combined industrial and aquaculture wastewater contributes 26.84%, mainly characterized by TP (32.44%) and CODMn (63.34%). Overall, this study demonstrates that water pollution patterns in the Yanjin Stream are jointly regulated by hydrological variability and land-use structure, providing a scientific basis for water environment regulation and management in mountainous urban rivers.