1.  This study developed a new method to determine the portion of primary and secondary PM2.5 using some basic measurements and inventory. They evaluated this new approach through the comparison with lots of observations in China and US. In addition, they analyzed the temporal and spatial variation as well as correlation between O3 and PM2.5 using the results from their new method. Although their evaluation looks very well, I think their results were not enough convincing because of unclear statement of their method and defect of this method. I would suggest major revision before reconsideration. My detail comments are following.
2. Eq(1) and Eq(2): These equations are the core of their method. They regarded CO as one tracer to represent the combustion process and assumed the combustion emission sources are same for CO, OC and EC. This assumption is mostly correct, but the emission factor/emission ratio of CO, OC and EC from different combustion sources are different. I think it is unconvincing to use one single coefficient without the influence of diversity of sources to standard for all conditions. I may misunderstand something, please discuss this uncertainty or make this clear.
3.  Eq(2): why did you name b as emission of fine dust? To my knowledge, MEIC does not include the emission of dust even urban dust.
4. I did not understand how you did the sensitivity experiment to examine the uncertainty in the inventories. Page 16, you said you changed the emission coefficient with 10%. If so, how can you keep a+b=100%? According to my understanding on this new method, the results should have large dependence on the inventory of PM2.5, OC, EC even the factor you used to decide OA, SO4 and NO3. I would strongly suggest setting up more comprehensive and scientific sensitivity experiments to discuss the dependence on the inventory.
5. Figure 4, as I saw, the largest concentration is < 60 µg/m3. Why not short the range of axis to spread those dots?
6. P8L7: Why did you remove the heavy pollution cases here as well as in Section 4? As you stated at P10L25, you would like to avoid the influence of extreme high primary emission cases. However, mostly heavy pollution cases are caused by unfavored meteorological condition but not caused by sudden high primary emission (except the biomass burning cases). I would be curious that how your method applied to analyze the heavy pollution cases. In general, it is more important to understand the contribution of secondary particles to heavy pollution cases than the general conditions.
7. P10L30: Could you explain what is regional background cities you defined here? Usually, cities are not background.
8. Section 4.2.1: I think the seasonal variation of PPM and SPM is largely depend on the seasonal variation of emissions you applied.
9. Section 4.2.2: Did you use the emission inventory for specific year here? China conducted a large reduction on PM2.5 emission since 2014. If you did not use the specific inventory, the estimated trend of PPM and SPM would not make sense, even though they agreed with observations. In addition, could you show the correlation coefficient between the observation and estimation here?
10. Section 4.3: The same issue as above. Did you update the inventory to the lockdown condition? If yes, please state the inventory you used here and the decrease in the emission of PM2.5, CO, OC, EC.
11. Section 4.4: How did you decide the diurnal variation of emission? Was your result sensitive to the diurnal pattern? Because the diurnal pattern of O3 concentration is almost constant.
12. Section 4.4: Why did you exclude the wet deposition case here but include in other sections? I would suggest adding the application condition for your method somewhere.
13. The general method to calculate the portion of secondary PM2.5 is chemical transport model using bottom-up inventory. It’s better to examine the difference in the result between your method and CTM with same inventory.

Attached is the PDF version of the comments. 

The manuscript presents a method for estimating the relative contributions of primary and secondary PM by proxy. The observed input parameters are CO, PM10 and PM2.5, however, the metohod also relies on estimated emissions of OA, EC, OC, fine dust, PM2.5, sulfate and nitrate, from emission inventories. The authors develop a proxy for secondary particulate matter on the basis of the observed parameters and estimated emissions. 

The motivation is presented as the need for a low cost, operational method for monitoring the contributions of secondary aerosols to the total PM2.5 levels.

The method appears to have some use for informing operational air quality management or for informing policy, but the scientific value of the method is not convincingly presented. It relies on assumptions and inventories that are not universal, and the manuscript does not present a convicning argument for its use, other than that it is cheaper than source apportionment methods based on chemical speciation. But it does not present comparative estimates of primary-secondary contributions with those methods.

It is questionable if this method has any value. It requires a big body of inputs, as other chemical transport models, but also relies heavily on assumptions and coefficients that are externally adjusted, even tuned to fit the model.

The manuscript describes comparisons between estimated and observed primary particulate matter. Categorisation of measured historical data into secondary and primary aerosols for comparison with the MTEA seems ot be based on chemical compositions, but this process is not clearly described and the criteria are vague.

There has been no attempt to verify the MTEA estimates for ppm by comparing with published estimates based on receptoro modelling, CTMs or AMS studies. There are many studies in the literature that have produced estimates that can be easily compared with the outcomes of the MTEA approach, but that has not been done.

It is true as the authors state that those other methods are labour-intensive and expensive, but they are also scientifically tried and tested and therefore more convincing, so it would make sense to develop the perfomrance of the MTEA against such methods more than has been done in this manuscript.

Also, the manuscript states that the numerical calculations were done on a supercomuting system. It can be argued that if the approach requires a supercomuting facility then it is no less costly or inaccessible than the existing source apportionment methods, but the cost has been shifted from scientific equipment to IT services.

The manuscript does touch on a discussion that has scientific interest, and that is contained in the sections 4.1 and 4.2 on spatial and temporal variation. The discussion on spatial variation has some merit. There is potentially a better motivation for developing the MTEA approach in order to inform a discussionon the spatial and temporal variation where only proxy parameters are available, by leveraging national monitoring networks to learn more about geographical distribution of secpondary aerosols and feed into a discussion on variations in atmospheric processes.