This work discusses the theory of PM2.5 source apportionment and presents an integrated modeling approach, which combined receptor models based on PM2.5 component measurement data and CTM air quality models using emission inventory. This approach obtains refined source apportionment results through a three-step modeling process. Firstly, this approach takes advantage of the CMB receptor model's ability to resolve the contributing source sectors of the primary PM2.5 components. This approach also used relatively accurate PM2.5 component measurement data, especially for secondary components and dust of PM2.5, which present high uncertainty for the CTM model due to the challenges with model mechanisms, emission inventory, and model resolution. Secondly, this approach utilizes CTM’s source apportionment tools to estimate the contributions of specific emission sources by simulating atmospheric physical and chemical processes and tracking the emission sources groups of specific spatial locations for complete components of PM2.5. Finally, the CMB receptor model source apportionment results for primary PM2.5 components, along with the measured secondary components PM2.5 concentrations, and the source apportionment results including local and long-range contributions estimated by the CTM model are used to obtain the refined integrated source apportionment. The integrated approach developed in this work was applied to conduct source apportionment in the 2+26 Cities Region during the autumn and winter of 2017-2018. This work provided refined source apportionment results for all PM2.5 components, including the local contributions and mutual contributions among different cities and regions, which are crucial for policy and action plans. This work directly supported the PM2.5 pollution control action plans in the region.