This manuscript investigated the sources of black and brown carbon in Northeast China, by integrating observational and simulation results. An agreement between observed and modeled PM_2.5, especially with respect to the source-resolved chemical compositions, is essential to design efficient air pollution control strategies. Such comparisons have rarely been made for Northeast China, which possessed distinct primary sources and meteorological conditions compared to other regions in China. The authors observed efficient SOA formation at low temperatures during winter and abundant open-burning POA in spring, both of which could not be properly reproduced by CMAQ. The results contribute to the understanding of haze pollution in China. I think this manuscript could be considered for publication after addressing my following concerns.

My major concern is that for the comparison of observational and modeling results, an interesting point was missing, i.e., the relationship between ECmod and ECobs during the fire episodes. The authors argued that open burning emissions and thus ECmod were underestimated, whereas ECobs were biased high due to fire-induced BrC. Then it would be expected that ECmod and ECobs should have larger differences for the fire episodes compared to other periods, but the authors did not show the comparison. If this inference did not hold, the major conclusions would be questionable.

Minor comments

1. Lines 16-18. I guess what the authors want to say is “the understanding on the abundances and sources of light-absorbing carbon is still subject to non-negligible uncertainties”. In other words, the “abundances and sources” themselves don’t have uncertainties. This sentence needs to be re-organized.
2. Line 92. It should be “light-absorbing”.
3. Line 206. I assume the authors mean “2020-2021”, not “2021-2022”.
4. Line 223. Provide quantitative result for “close to zero”.
5. Line 287. Why did TC decreased after extraction for the blank filters? Is the difference significant?
6. Line 392. Provide quantitative description for “a considerable number”.
7. Line 467. Clarify whether the mean bias was calculated as model ‒ observation.
8. Figure 9. It would be better to use log scale for the y-axis of the upper panel.
9. Caption of Figure S12. OC* and EC* should be used for the equation of EC-tracer method. In addition, the firework periods should be clearly shown in the figure, i.e., clarify whether all the events without SOC results were associated with fireworks.

The manuscript by Cheng et al. explored the characteristics and sources of carbonaceous aerosols during three successive winters in Harbin. Samples were collected and analyzed for a variety of species, e.g., brown carbon, elemental carbon and levoglucosan. The authors evaluated the loss of EC during methanol extraction of filter samples, a long-lasting debate on the measurement method for BrC mass concentration. This artifact was suggested to be unimportant based on indirect evidences, providing valuable implications for future studies. The authors also explained the EC discrepancies between different analytical methods, and identified the OC/EC ratios (i.e., OC and EC results) that were in reasonable accordance with secondary aerosol formation. The authors then performed source apportionment of BrC and EC using the measurement results, and finally compared the observation-based attributions with those predicted by an air quality model. Overall, the results were properly interpreted and presented. However, as listed below, there are some major concerns on top of writing problems.

Major points

(1) The authors only broadly stated that the high RH conditions in winter should favor heterogeneous formation of secondary aerosols. This statement could be more specific. For example, did the heterogeneous reactions occur in aerosol water like Beijing or on the surface of frozen particles due to the low temperature?

(2) Line 300. Suggesting toning down the statement, as this study did not have robust evidence for heterogeneous chemistry. Please note that the RH-dependent increases of SOR + NOR (Figures 1b and 1c) and SOC/EC (Figures 3 and 4) should only be considered indirect evidences.

(3) Please clarify whether it was acceptable to approximate MSOC as untreated OC. This point is important, given that thermal-optical analysis of the extracted filters could be laborious and time-consuming.

Minor points

(1) Line 18. Suggest clarifying that EC is a measure of black carbon.

(2) Lines 25 and 27. Change SOA to SOC, as no SOA result was presented throughout the manuscript.

(3) Line 77. Suggest providing an example for the “possible artifacts”.

(4) Line 102. Full name should be given for PM_2.5.

(5) Line 142. Use “Results and discussion”.

(6) Line 160. Suggest adding a “the” before “filter”.

(7) Line 183. Suggest changing the first “as” to “since”.

(8) Line 249. Suggest re-writing this sentence as “Importantly, as shown in Figure 2a, ΔATN were negligible……”

(9) Line 281. Suggest adding an “in turn” before “supported”.

(10) Line 325. Change “is” to “was”.

(11) Line 334. Suggest adding a “the” before “winters”.

(12) Lines 394-396. The statements did not hold for all the Harbin samples.

(13) Line 480. Remove the “the” before “seasonal”.

(14) Line 542. Suggested adding a “(i.e., MSOC)” after “BrC mass”.