The manuscript attempts to explore the influence of adopted emission source profiles in CTMs on the simulated results of PM2.5 components by sensitivity analysis. The extent of the influence for different components were quantitatively analyzed, the impact laws and pathway were identified. The topic is interesting and their findings highlight the importance of effective utilization of emission source profiles in CTMs. Although the description of experiments is complete to allow their reproduction by fellow researchers, some explanations and discussions are not clear. I recommend its publication subject to the following amendments.

Major concerns:

1. What is the design basis for the perturbation of emission source profile in the sensitivity experiments?
2. The discussion of the results should be extended. The authors mentioned that emission source profile adopted in CTMs has a significant impact on the simulation results of PM₅ components, so how to select the appropriate source profiles in the simulation? In the section of conclusion (Line 549-551), the author concluded that “the representativeness and timeliness of the source profile should be considered”. How to understand the “representativeness” and “timeliness” here?

Minor concerns:

1. Line 21 and Line 27, there are two notes for CTM in one paragraph, which appear to be repetitive.
2. Line 57-59, the references are verbose.
3. Line 111-113, It is not clearly explained the role of source profiles in CTMs.
4. Line 257: “The detailed information on” should be “The information of…”
5. Line 259: “Coefficient Divergence (CD)” would be appropriate
6. In the supplementary material, Fig. S1, the author selected code 91041, 900162.5, 91155, 91022 and 91162 as SPECIATE source profiles for simulation. Detailed information of these source profiles need be provided by authors.

The manuscript investigates the sensitivity of simulated PM_2.5 and its components’ concentrations to the uncertainties in the component-specified PM_2.5 source emission inventories using the CMAQ chemical transport model. The relatively-complete chemical components, including Al, Ca, Cl, EC, Fe, K, Mg, Mn, Na, OC Si, NH₄⁺, NO₃^-, SO₄^2-, and others, are taken into account in the emission inventory used. The authors showed that the influence of the relative contributions of different components to the total PM_2.5 emission (denoted as source profile changes in the manuscript) on simulated PM_2.5 concentration was insignificant, but its impact on PM_2.5 components could not be ignored. They also showed that these source profile changes caused the variations in simulated gaseous pollutants’ concentrations. While such kind of model experiment should be a welcome addition to the literature on air quality model simulation, I do have concerns that the data and methodology used in this study would be sensible (or well introduced) and the conclusions applicable to the simulations done by other chemical transport models with different chemical and physical modules. Therefore, I cannot recommend publication the current version of this manuscript in GMD.

The major issues are follows:

1. What is the grid resolution of the MEIC emission inventory that was used for the model simulation in this study? Is the resolution sufficiently fine for the Dom3 (4 km× 4km) simulation? What does the area marked in green in Fig. 1 refer to? No information on the regional distributions of either PM_2.5 emission sources or their simulated concentrations is provided in the manuscript. Are all the 10 monitoring sites located in the cities of Dom3? Is there any site that is located near the desert area? Were the mineral dust emissions taken into account in the simulation?
2. At the beginning of Sect. 2.2 it is stated that in addition to SPA and SPE, the PM₅ emission source profile database from published literature was used. Where and what are the final, merged emission source profiles used in this study? The simulated PM_2.5 and its components’ concentrations using CMAQ_SPA are compared with those using CMAQ_SPE. However, no comparison with observed PM_2.5 components’ concentrations at the monitoring sites has been made to show the advantage of the SPA over the SPE.
3. While the MEIC inventory includes four categories, i.e. power plants (PP), industrial processes (IN), residential emission (RE) and transport sector (TR), the SPA and SPE are shown to have different categories (perhaps more than the MEIC does). How were these chemical PM₅ emission source profiles combined to match the MECI categories? For instance, the residential emission should include not only coal burning but also straw burning, and the latter was seemly not considered in the simulations. Also, the chemical profiles for gasoline and diesel oil in the transport sector might be different.
4. How are the dynamic, microphysical and chemical processes of aerosols treated in the CMAQ model used for this study? Are the size distribution, mixing state, aging and solubility taken into account for different aerosol components? By which molecular form are the chemical components (Al, Ca, Cl, EC, Fe, K, Mg, Mn, Na, OC Si, NH₄⁺, NO₃^-, and SO₄^2-) emitted from the sources? Taking elemental Ca as an example, it should be emitted by CaO, CaCO₃, CaSO₄, or other compound, rather than merely by the cation Ca²⁺. The similar principle applies for anions (NO₃^- and SO₄^2-). The difference in the exiting form of these emitted aerosol components might have large impacts on the thermodynamic equilibrium of ions in liquid aerosols and clouds.
5. In Sect. 1 and Table S1, the deviations of PM₅ components simulated by CMAQ are presented. All these components (NH₄⁺, NO₃^-, SO₄^2-, and part of OC), except for EC and part of OC, are second aerosols, and their loadings in the atmosphere are controlled primarily by the emissions of gaseous precursors, instead of the emission of aerosols. The presentation here and associated arguments seems to be misleading as the effect of uncertainties in the gaseous emissions is not considered in this study.