The paper’s focus is on the validation of the Polair3D chemistry transport model over the geographical domain of Quebec, Canada at different spatial and temporal resolutions. The authors highlight that the model was used for the first time over the Quebec domain and for the first time with a temporal period long enough to evaluate its performance at the seasonal level.

The authors present a statistical analysis of the model performance in representing a range of primary and secondary air pollutants comparing model results with ground observation sites. Finally, they conduct a sensitivity test on the primary emissions impact on final concentrations switching off the industrial emissions and evaluating the change in the levels of air pollutants.

The manuscript is, on the one hand, highly focused on the evaluation of the model performance in representing a range of pollutants on a large temporal scale. This represents a challenge for any CTM because of the impact that meteorology and chemical mechanisms can have at seasonal levels and at different resolutions. On the other hand, there is a low focus on the expendability of the model (E.g., scenario analysis). The manuscript would benefit from a clearer statement of the use the authors want to do of the model. They mention at the beginning the impact that industrial air pollution has on human health, and they create a scenario to test the model's performance. Nevertheless, there is no health impact quantification analysis of model outputs or scenarios.

A more careful choice of the use of the model would give higher focus to the validation and the choice and description of proposed scenarios. If the choice of the model is motivated by “scenarios-impact” analysis then the validation should focus on high-resolution simulations and on a temporal scale that would allow to evaluate the model against national or international threshold limits at a daily/hourly level. Contrarywise, if the intention is to use the model for monthly regional simulations then the validation could be limited to a 3x3km resolution and evaluate the model representation of seasonal average values analysing the impact of scenarios at the annual/seasonal level.

Major Comments:

43 – 46: The authors apply the Polair3D model over the domain of Canada highlighting that this particular model has seen only a little use over North America and Canada. It should be made clearer why the use of this particular model should represent a step ahead in air pollution research. Several types of CTMs serve for different purposes from different points of view. These can be related - for example - to the representation (or absence) of particular chemical mechanisms to describe the chemical life cycle of some pollutants, or to minor computational costs that make the simulations quicker or smaller in terms of storage space. The authors should make clearer and stronger the motivations that led them to 1) choose the Polair3D model and 2) which use it besides the pure capability of representing concentrations of air pollutants (e.g., scenarios, forecasts, mitigation policies testing, transport/trajectories analysis).

61 – 64: The performance of a CTM is highly influenced by the levels of primary emissions and by the representation of the regional meteorology. The authors mention that the meteorology used to drive the model in representing the air pollution was taken by WRF, but they don’t mention a validation of this meteorology. Analysing the performance of the model at the seasonal level and focusing on primary and secondary pollutants it would be good to understand the levels of reliability of parameters such as surface temperature, solar radiation, wind speed and temperature, and relative humidity.  The evaluation of these parameters should be also analysed in the same temporal dimension (e.g., seasonal, annual) of the air pollution concentrations.

92 – 93: Anthropogenic emissions come from NEI 2014 and EPA 2017 inventories. The authors mention using the SMOKE pre-processor and a combination of the “SMOKE-ready” format of these inventories. Are these the most up-to-date inventories to represent the emissions in Quebec?  The authors should mention how these two inventories have been processed and speciated before being merged. Is it assumed that – individually – the two inventories have been processed in SMOKE and then merged? If yes, in which way and according to which criteria? Besides the representation of the PM, how the VOCs are represented/speciated in the model?

The choice of anthropogenic emissions can be critical in the representation of the final concentrations, and it would be good to have a quantitative analysis (even only in the supplementary material) that shows the annual totals. This is also in light of the scenarios with reduced emissions from Industrial sources. It’s good to have maps showing their positions but it would be good to know which percentage of reduction this sector, once deleted, gives on the totals.

113 – 115: The authors mention the use of the NAPS observation sites for the evaluation of the model performance. It would be good to know where these observation points are (maybe in Figure 1) and most of all which type of observation sites these are. Are urban backgrounds and or rural sites? Any of these are road traffic sites? The model performance could sensibly change if different sites are computed together or by type. The suggestion here is to divide the sites by type and perform the statistical analysis again. The evaluation of urban background sites for NO, NO2 and O3 could reveal information about how the model represents titration processes in urban environments while in rural areas it could be analysed the impact that biogenic emissions of VOCs have on O3.

149 – 160: The evaluation of the performance in representing NO, NO2 and O3 would benefit from more information about VOCs, and by how the original NOX emissions are partitioned in NO and NO2. For what concern is the performance of PM2.5 Is there any transport pattern that could influence the seasonal variability in the model performance? The authors mention that the model performance in summer is lower for all pollutants except for SO2. Is there any reason related to meteorology or seasonal emissions of SO2 that could give this?

178 – 182: The authors mention that the model performance does not increase with the model resolution for CO and NO2. This can be seen from the parson coefficient by going to inspect the MB in Table 1 this is always lower in the 1x1km domain. What is not mentioned in the text is that the evaluation of the “3km to 1km” shows higher MB than the “3km”. This could suggest that the agreement in the model performance decreases in urban background sites (that are supposed to be denser in the 1km domain area) and is higher in rural areas. An analysis of the model performance by site type could explain this better.  

Figure 6: Why the analysis of Ozone is shown only for January when its photochemical activity is lower? Why don’t show the same figure for summer and winter?

Test Scenario: This part of the manuscript would benefit from a deeper understating of the impact of the reduction in industrial emissions on final concentrations and the benefit of the health impact. It’s good to know where the emission points are and where the average difference in concentrations is, but it would be also good to know how much change in terms of concentrations at the receptor’s sites (observation points). This would require an analysis of the meteorology to understand which sites are downwind and would benefit more from the industrial emission reduction. Additionally, the authors mention the health impact that industrial air pollution has but they don’t quantify the impact that the suggested scenario could have on final concentration and health impact.

The focus of this paper is on evaluating the Polair3D CTM for a 3km horizontal grid resolution domain centered over Montreal, and two 1km resolution domains over Montreal and Quebec.  Surface concentrations of CO, PM2.5, NO2, NO, SO2, and O3 from the model are shown and evaluated using ground-based observations for each season, where the model was run for four weeks per season (January, April, July, and October).  A test scenario focusing on the industrial emissions is also shown.

Major Comments:

In general, the purpose of the paper is not very clear, the authors do not say why it is important that the Polair3D CTM be evaluated for Canada or why they selected that model in particular.  How is applying this CTM over Quebec novel, as stated on Line 237?  Has the model been used for other urban cases and at a 1km resolution, if so, what is different about Quebec (emissions, terrain, chemistry)? 

Sections 2.4 and Section 3.4: This appears to be a brute force, source-oriented source apportionment approach focusing on industrial emissions.  However, it is not clear in these sections why this is being done, how does this “enrich the model validation findings”?  For example, Lines 229-230 say that the model captures the spatial variability of the pollutants emitted from the industrial sector, but this was not evaluated directly.

The meteorological modeling will impact the CTM results because that is driving the transport processes.  The WRF configuration and evaluation information needs to be provided.  Also, was two-way nesting done on the WRF simulations?  If so, this means that the results from the 1km domains impact the results from the 3km domain and therefore comparing the model evaluation statistics for these two domains (i.e., Table 1) should be done with caution.

More discussion on the uncertainties related to emissions, meteorology, or chemistry would strengthen the paper.  The authors provide a comparison of their evaluation results to three other studies in Section 3.3, but details about the potential sources of uncertainty in the model results are not provided, including whether or not those are similar to these other studies.  Some examples include:  Are there seasonal biases associated with chemical mechanisms or meteorology?  Are the emissions inventories known to be biased high or low for a specific pollutant from a specific sector?  The CTM results are the final results of an entire modeling framework, where uncertainties can be introduced at each stage. 

Minor Comments:

Line 7: add “horizontal” before “resolutions” here to indicate that these are the horizontal grid resolutions of the model.

Line 8:  The word “accurate” here implies that you are able to know to some level of certainty that the model is correct, does the data and evaluation you have available to you give you this level of certainty?

Line 37-40: It is not clear in here why Polair3D was selected, is there a specific motivation for using another CTM?

Line 70: What is the vertical resolution or number of vertical grids in the model?  Vertical resolution has a large impact on the simulation results.

Lines 75-77:  Which chemical mechanism was used in SMOKE?  Later, on line 103, the aerosol model is mentioned (AE6) but it would also be good to specify which chemical mechanism was used here.  Did it require species mapping to run the SMOKE outputs with the Polair3D CTM?

Line 78: U.S. EPA National Emissions Inventory (NEI) is the name of the inventory and, therefore, the N, E, and I should be capitalized (this occurs in other places in the paper as well).

Line 89-93:  Was the U.S. EPA NEI platform used (e.g., spatial surrogates, species profiles, SMOKE scripts) or only the emissions input files?  Also, specify the platform version number in addition to the year. The Technical Support Document (TSD) for the NEI could also be referenced here for how the different emissions sectors are modeled.

Lines 94-102:  There is a lot of detail in this section about the point sources but what about the mobile emissions, were those modeled using MOVES?

Line 122:  Were only the metrics for evaluation used from Emery et al. (2017), or were the benchmark criteria for these metrics also used?  If not, how are you defining success for the model?

Line 128:  The sentence ending with “kept” feels like it is missing something after that word.

Line 134:  It is interesting to block out the lowest and highest concentrations in the spatial maps (white) because it seems like those concentrations might be the ones of interest, especially the high concentrations with respect to human health. Maybe plotting them on a log scale would help retain all of the pollutant concentrations but allow for visualization of the spatial gradients?