A comprehensive year-round field campaign was conducted in the coastal city of Ningbo, China to examine the driving factors of oxidative potential (OP) of PM2.5 and hence its potential health impacts. Using dithiothreitol (DTT) cell-free assay, OP of water-soluble (OPws) and methanol-soluble components (OPmeth) of PM2.5-based samples were measured. Volume-normalised OP (OPvws and OPvmeth) were higher in heating seasons, and mass-normalised OP (OPmws and OPmmeth) peaked in summer. Five sources were identified to be DTT active in OPvws, with road dust (RD) being dominant, while four sources contributed to OPvmeth, with industrial emission (IE) as the largest contributor. Water extracted more toxic components from RD while toxic components in IE and vehicle emission (VE) were more soluble in methanol. We further compared OP values between PM2.5 of similar concentrations and observed that OPvws prevailed in PM2.5 toxicity during clean days (PM2.5 concentration < 38.5 µg m−3), while OPvmeth became dominant when PM2.5 concentration was larger than 69.1 µg m−3, implying the need to extract whole PM2.5 using both solvents to develop the mitigation strategy for reducing the overall PM2.5 toxicity. Efforts should be devoted to controlling the emissions from dominant sources of both OP indicators in specific episode. © 2022 Elsevier B.V.