This paper provides parameters to derive HDD-based daily temporal factors to distribute PM and NO_x emissions from residential heating for EU-27 countries. Various experiments with different configurations of HDD-based temporal factors were carried out to assess the sensitivity of simulated PM_2.5, PM₁₀ and NO₂ surface concentration to the different experiments and to identify the best parameterization compared to in situ observations on the one hand and to use HDDs as a method to model national emission totals from GNFR_C and to compare them with persistence and uncertainty on the other hand. The paper is well structured, employs appropriate methodology, and presents reasonable results. It aligns with the scope of Atmospheric Chemistry and Physics. Considering the questions, comments and suggestions described below, I recommend the publication of the paper.

Specific questions and comments:

1. [Section 1] You describe here that the residential sector is responsible for 22.7% and 10.3% on average of the mortality in European cities attributed to PM_2.5 and NO₂ pollution and that both the urban centres and, due to air advection, the suburban areas are affected. In this context, it would be interesting to know how long the pollutions remain in the atmosphere at a high concentration that is dangerous for people.
2. [Section 3.1] Table 2 provides a very good overview for understanding the experiments carried out.
3. [Section 4.1.1] In Table 3 the validation scores for the CHIMERE reference simulation calculated from AQ-eReporting observations for PM_2.5, PM₁₀, NO₂ and O₃ species are shown. It is not entirely clear to me why the bias (model-observation) for annual average is mostly negative. This means that the model always underestimates it. What could be the reason for this?
4. [Section 4.1.2] It is not clear to me where the values for the thresholds (25 µg/m³ for PM_2.5, 50 µg/m³ for PM₁₀ and 40 µg/m³ for NO₂) come from (line 394). Maybe I missed it, but I can't understand it at this point in the text.

Technical suggestions:

• line 41: … need to be distributed …
• line 67: This parameter should mainly depend on …
• line 78: … are based on …
• line 86: delete the space before the colon
• line 87: “by comparing” or “in comparison”
• line 104: … assuming that there are no …
• line 129: … and their proportions …
• line 130: … come from very different energy types …
• line 154: …, transmission systems, …
• line 183: … ambient temperature.
• line 239: … PM are split into …
• line 256: … calculates …
• line 260: … from the operational …
• line 287: … projections using the HDD method are …
• line 326: Using an HDD …
• line 341: … the HDD parameterization does not appear …
• line 415: … no reason that …
• line 430: As expected a higher deviation is obtained …
• line 435: delete “can”
• Figure 8, S4, S5, S6, S7 and S8: the unit on the y-axes in the figures is μg m^-2, but μg m^-3 in the text

The manuscript by Guion et al. presents a method to temporally redistribute historical estimates of emissions of air pollutants from the residential sector tied to space heating based on the day-to-day variability in the weather. The contribution of biomass burning is a particularly important source of particulate matter (PM) from residential heating and is well known to significantly contribute to PM in certain urban areas during cold weather, so it could be expected that accounting for the shorter-term variability in weather would improve air quality simulations.

The authors derive a set of country-specific factors for the EU-27 group of countries to account for variations across the region necessary to split apart the general emissions category that includes space heating available for the EU-27 and assign temporal variability. With these factors, the authors proceed to temporally re-allocate annual emissions from space heating to account for the observed variability in weather. The authors find some improvement in the model simulation of PM2.5 in the winter months for 2018, particularly during several cold spells with below average temperatures across large parts of the EU. Effects of the parameterization on NO2 and PM10 were generally more modest, but did show some improvements in certain countries. The authors also investigate the use of their parameterization to improve hindcasts of air quality while using emission inventories from two or three years before the simulation year. The idea here is to investigate whether the method could be used to improve real-time air quality forecasts, which use the most recent available inventory that is generally two or three years old.

The article is well written and clearly presents the effects of the day-specific emissions allocation as compared to the monthly-average allocation. Given that models always include a variety of errors that vary spatially and temporally, I am not surprised that the sensible allocation of day-specific emissions improves the simulation in certain regions while producing little improvement in other areas. Figure 6 nicely shows this, comparing the model simulated PM2.5 against observations and finding that the parameterization produces fairly widespread improvements in the correlation and root mean square error but not for all stations.

The only significant comment I have on the manuscript is on the presentation of the multi-year emissions projections in Section 4.2. I understand that the motivation for this work is to explore whether the day-specific allocation could improve real-time air quality forecasts, but because the parameterization requires the meteorology for the full year to be known it is not clear how the parameterization in its current form could be used for real-time air quality forecasting. The analysis of the effects of the day-specific emission allocation is also confounded with the effects of revisions of the emission datasets from one year to the next, that includes a representation of the real change in emissions with time as well as revisions in the methodology to estimate emissions. The authors do demonstrate some improvement in the simulation applying the weather parameterization to previous year emissions, but it is difficult to clearly see how the projected emissions estimates are necessarily better than persistence and how any improvement is related to the use of the day-specific emission allocation. I am not suggesting the authors remove this part, because I am sure the application is of interest for air quality forecasting, but I would suggest the authors consider clearer ways to demonstrate the day-specific emissions allocations have improved the emissions estimates and to demonstrate how.

Figure 5 presents the difference in Spearman correlation coefficient and RMSE for the different experiments broken down for four different countries. Do the authors have any explanation for why the full parameterization (DayTF_Tbfit_fspec, shown in red) seems not to perform noticeably better, and in some countries worse, than the partial parameterization DayTF_Tbfit?  And while there are differences in results between the parameterizations, the differences are much smaller than the difference between countries. Italy has a much lower correlation coefficient than the other countries and France has a RMSE at least one half the size of the RMSE for the other three countries. For the RMSE in particular, do the authors have any explanation for the country-to-country differences? From Figure 8 it does not appear that PM2.5 concentrations are much lower in France than the rest of Europe, which may be expected to result in a smaller absolute RMSE. Does the small RMSE for France suggest the emissions inventory for France is better than for other countries?

Minor comments:

Line 78: The phrase ‘regional air quality forecasting’ seems to be missing the word ‘system’ 

Lines 102 – 103: What measure of the outdoor temperature is used in the calculation of HDD? Is it the HDD calculated from hourly T2m or calculated from daily average temperature? If it is the daily average, how is the daily average calculated?

Lines 171 – 172: Here it is stated that the surface temperature used in the calculation of the HDD threshold is taken from ERA5. Is this the daily average temperature or some other quantity?

Lines 323 – 325:  The authors state that the comparison of the different experiments against observations will be made ‘for each country for which the Tb parameter has been calculated based on the national gas data.’ But Figure 5 shows a comparison for only four countries, while in Section 2.2.2 it is shown that it was possible to calculate country-specific Tb_fit for eight countries. Why the difference?

Line 435: Typo in ‘it can can provide’