General Remarks:

This study develops an optimized emission estimate of crop residue burning in North India in October and November 2022. This study develops this emissions inventory by using a top-down approach, which according to the authors, has not been done for post monsoon crop residue burning in North India. I find this to be an important contribution to the understanding of crop residue burning emissions and the impact of these emissions on PM_2.5 in North India. However, I believe the results could be made more clear and concise. In the present state, the scientific significance of the paper is inhibited by the quality of the discussion and presentation. I recommend major revisions.

Specific Comments:

Introduction

1. I find the introduction to be a bit too brief. There could be more discussion on the estimates of PM from CRB in prior studies and reasons for why a consensus has yet to been reached. Additionally, it would be helpful to summarize the emissions estimates for global biomass burning inventories (eg. GFAS, GFED, QFED, FINN) during the post monsoon period in this region as this would help to provide additional context for the inventories used in this study (both GFAS and the developed top-down CRB inventory).
2. I also think it would be beneficial to further discuss specific advantages of using the top-down approach over the bottom-up approach in terms of estimating biomass burning emissions, and to the extent prior studies allow in the context of CRB or agricultural burning in other parts of the world.
3. I think expanding the final paragraph of the introduction to layout the format of the rest of the paper would improve the ability to follow the results. This could be accomplished through stating what general topics are discussed in each section of the results.

Methods

1. Figure 1 could be improved. The colorbar for the terrain could be the same between the left and right panels. The coloring of the observation sites could be more informative for the interpretation of later figures by using different shapes to indicate CPCB and CUPI-G sites and coloring to indicate Source, Intermediate, and Delhi NCR sites.
2. Aspects of the emission optimization using tagged simulations need to be clarified or explained further
   1. Equation 1 needs to be better explained in terms of what the variables are in this specific study. For example, O and S are observed and simulated data of CO, or PM₅.
   2. The description of the method of optimization to minimize the cost function is too vague.
   3. Is the No_CRB simulation simply a simulation with no GFAS emissions? This needs to be clear.
   4. Was a threshold of PM₅ bias during the period not affected by CRB used to exclude observation sites? If so what was this threshold and how was it selected?
   5. Was the discussion on spatial temporal representativeness of an observation used to exclude sites?

Results

I believe the results could be restructured to better highlight the impact of the optimized emission estimates developed in this study. Also, ensure that simulation names are consistent with Table 1 and throughout the results section.

1. Section 3.1 Time series comparisons: This section would better be entitled “Whole Period GFAS Simulation Evaluation”, or something similar. However, I find this section to be too long. Figures 3 is better discussed in the context of the impact of the optimized section (Figures 7 and 8 show the GFAS simulation). While the point of Figure 5 and 6 to explain reasons for potential biases in the GFAS inventory during the plume events is valuable, it could be an SI figure that is discussed with the results of Figure 9.
2. Table 2: The discussion states that x_m values were obtained for different temporal tags, could these be presented in the SI?
3. I think the discussion in Section 3.2 could be improved by having Figure 7 be altered to have 3 panels: Source Region PM₅, Intermediate Region PM_2.5, Delhi NCR Region PM_2.5. On each panel would be a line for observation, GFAS run, CPCB+CUPI-G, CPCB+CUPI-G+CUPI-G Punjab and CPCB+CUPI-G+CUPI-G P89.
4. Figure 8 could then be changed to show the simulated contribution of different emission sources in each of the four simulations.
5. There needs to be better discussion and quantification of the PM₅ biases. What is meant by the optimized run substantially improved biases discussed in the first paragraph of page 21. Table 3 could be introduced here.
6. Table 3 could be improved by including mean biases of PM₅.
7. The discussion of the contribution of different tagged source regions on page 23 could benefit from more quantitative discussion rather than vague substantially increased or became higher language currently used.
8. Table 4 would benefit from including FINN or GFED data even though it was not used for the optimization.
9. Section 3.4: would it be possible to plot simulated AOD in the first three rows of Figures 10 and 11 instead of surface PM₅?

Technical Corrections:

Abstract

• Line 8: accounting for 80% of the CRB emissions
• Line 9: “OFEv1.0 substantially boosted emissions,” relative to what? GFAS?
• Line 16-17: Last two sentences of the abstract seem unnecessary, could be moved or re-worded.

Introduction

• Page 3, Line 8: Delhi’s air quality worsens during the post monsoon to winter period because of weaker wind speeds than other times of the year (Citation?), increased emissions from space heaters (Guttikunda and Gurjar, 2012; Chowdhury et al., 2017, 2019; Guttikunda et al., 2023), use of fireworks during Diwali festivities (Singh et al., 2019), and crop residue burning (CRB) upwind of Delhi (CITATION).
• Page 3, Line 20: list other uncertainties in source apportionment. Is there a citation to support that the largest uncertainty is in the biomass burning inventories?
• Page 4, Line 2: “which resulted in a 109% increase compared to MODIS” Increase in what? CO emissions? PM₅ emissions?

Methods

• Page 5, Line 6-Line10: “The CTM part of offline-coupled NHM-Chem can be employed…” This sentence seems to describe possible couplings not used in this study; it could be deleted.
• Page 7, Line 10: “Open biomass burning emissions were assumed to be constant over time.” Does this mean over the entire simulation, or that there was assumed to be no diurnal cycle?
• Page 9, Line 21: To optimize CRB emission fluxes, we used…
• Page 10, Line 10: The list of sensitivity simulations performed herein are summarized in Table 1.

Results

• Page 12, Line 16: Fig. 3 presents PM₅ not emissions
• Figure 3: Legend entry should say GFAS to avoid confusion with other simulations presented in the study, y-axes labels should say what variable is plotted
• Figure 4: y-axes labels should say what variable is plotted not just the units
• Figure 7e: GFSS run should say GFAS run
• Figure 7 caption: (g-l) is not the same as (a-f), say it’s the fractional contribution of different emission sources to PM₅
• Figure 7/8 axes labels: the y-axes labels should be informative as to which variable is plotted not just the units of the variable
• Page 23, Line 6: “even though its magnitude was small,” this sentence could be reworded to be clearer
• Page 26, Line 17: Reword the sentence, maybe “maps” is more informative than “horizontal distribution”

Authors have used chemical transport model and observations from dense network of low cost sensor to derive emission fluxes at resolution of 12 hours over a region known for crop residue burning (CRB) to find their impact on downwind mega city Delhi. The analysis is very important for the policy makers and air pollution researchers who try to see impact of such activity on air quality, health and cost-benefit analysis. Overall paper is well written, I suggest few minor correction which I believe will make manuscript better and are necessary.

• Authors use emissions from inventories to calculate non-CRB contribution. One can see that simulation and observation for non-CRB period compare well from Fig. 3. However, authors have used selected stations to prepare those plots. Authors do mention criteria for selection and list of stations selected vis-a-vis not selected in main text and supplementary but I think more details are necessary to replicate their finding by others as well as to understand subjectivity vis-a-vis objectivity in the criteria for selection of stations. As authors indicate in the absence of these selection, the comparison may not be as good, and in that case CRB fluxes which are interpreted based on difference between observed and simulated concentration will also be different.
• Emission flux estimates in top down approach are highly sensitive to simulation of vertical distribution of species which in turn is highly sensitive to simulation of boundary layer dynamics. Authors should show either using previously published studies for their model or from this study, how good were boundary layer simulation over South Asia and preferably over North India.
• Top down approach for flux estimate is also sensitive error in observation. Why authors have not mentioned errors in observations explicitly, they imply temporal standard deviation of 3 hours exceeds error in observation. This may be true for random error but if the observation had systematic error then it will still affect the calculation of emission flux. Authors should provide information on error in observation and discussion on how they would affect emission flux estimate.
• The discussion regarding emission optimization is not very clear -- at least to me. Based on the discussion that follows in and after section 2.3, x_m simply appears to be ratio of observed concentration to simulated concentration or a multiplier when multiplied to simulated concentration excluding non-CRB concentration, it matches simulated concentration to observed concentration. I might be wrong to interpret xm in this manner but If I am correct about x_m as ratio, then authors should explain
  
  
  • Are optimized emission fluxes are simply scaled up emission fluxes based on that ratio?
  • How would second term in the equation 1 would imply importance of apriori in the cost function?
• Authors write that the denominator in the second term of Eq 1 (u_m, l_m)  were taken as (2.0 and 0.5) and in an multistep process calculations were repeated until x_m values were between 0.009 and 1.001. Later-on in the result section authors show values of xm ranging from 0 to 69 with several of them more than 1. The text is not clear enough to describe how the values of xm get these high values. Authors should clarify xm calculations and how they can have such high value in spite of their set criteria.
• Page 19 lines 5-10: The description of calculation of anthropogenic contribution is not clear.  Why the 20% emission reduction, why not 25% or 40%? Which emissions -- All the emissions or the emission that are newly calculated over and above non emissions? What does multiply by 5 implies? Is the simulated concentrations multiplied by 5? Why would difference between control and reduced/multiplied concentrations would imply anthropogenic contribution?

Minor Issues

• Through out the text and in abstract and in conclusion authors mention their model as NHM-Chem. However, in the methodology section, authors mention that only the Chem part is used from NHM-Chem whereas meteorology is simulated using WRF. Since, the WRF model is also available with its own Chem version known as WRF-Chem. It is imperative that authors should briefly describe why instead of WRF's Chem they chose the different Chem model and Instead of NHM's meteorology, they chose WRF's meteorology. Also they need to come-up with better nomanclature then NHM-Chem since WRF-Chem would be misundestood as default WRF-Chem model and NHM-Chem would be misunderstood as default NHM meteorology.
• Page 3 Line 29: Authors write "polar orbiting satellites travel once during the day ..". Most polar orbiting satellites including Terra and Aqua travel twice over a place in 24 hours, once during daytime and once during nighttime. Satellite use thermal channels to find fire hot-spot which in principle should work better in night. Authors should clarify this point in the text and explain why there are no night time data for the fire hotspots?
•  Page 9 Eq 1: There is an extra plus sign in the equation. Also, the symbol  σ₀ should be σ_n if separate sigma was used for each observational data. Or otherwise mention how the σ₀ is calculated.
• Fig 7e: GFAS is written as GFSS