Overall this is a useful addition to the literature on fire emissions of VOCs in the western US. After addressing my comments, it should be published:

Major comments:

• Lines 31-33: Why is the underestimation likely due to DM underestimates and not other causes? 
• 206 - 224: Ideally all inventories will be used with the same vertical injection scheme no matter what the baseline in GEOS-Chem is. Can you clarify that you used the same across all three or four - either by putting all into the boundary layer or using the GFAS or QFED default schemes for all of them? Although the baseline in GEOS-Chem may use different approaches, that’s just because different researchers have used them for different purposes, not because that is a scientifically appropriate approach. Furthermore, these default schemes have changed over time and used to all emit into the boundary layer with the option to turn on the Fisher et al. (2014) scheme for all/any of them. It should not make much of a difference as you show later in the manuscript, but it would be better to ne consistent.
• Also the diurnal representation should really be standardized across the inventories no matter what is in the baseline GEOS-Chem - otherwise that’s another variable that’s not consistent across datasets, but which could be. It would likely make sense to apply the WRAP diurnal cycle to GFED for consistency sake.
• Line 288-289: How does this plot show that regionally averaged ERs make up the majority of disagreement and not DM burned? The back of the envelope calculation that you discuss seems to indicate a large DM underestimate.

Minor comments:

• Lines 57-58 - please add a citation.
• Line 70: the comma before “even though” should be a semicolon
• Line 92: missing a semicolon between “estimates” and “thus”
• Line 99: I’m not sure that I agree with the parenthetical “(more accurate)” here. Perhaps the top-down estimates match observations better in some regions, but it may be “correct” for the wrong reasons, such as how QFED scales up emissions to match AOD, which itself may be biased high for other reasons, such as optical property uncertainties.
• Figure S7: I’m fairly sure that with GFED4s, GEOS-Chem does include EFs for lumped alkanes and also alkenes per the van Der Werf spreadsheet that helps us calculate them. Please confirm.
• Line 324: “being” should be “is”
• Line 363: should be “Campaigns targeting plumes…” And adding “like WE-CAN” would also be helpful context here.
• Line 392: please don’t use “efficient” twice in one sentence.
• Figure 7: I’m fairly sure that GFED has EFs for acetaldehyde and xylenes. Please check and add.

In this manuscript, the authors evaluated model predictions of volatile organic compound emissions from wildfires in the western United States. They performed GEOS-Chem simulations using different inventories and compared predicted VOC concentrations with those measured in 2 recent field campaigns, WE-CAN and FIREX-AQ. They found that the model systemically underpredicted concentrations regardless of the inventory used, and tripling one inventory can reconcile the model-measurement difference. There are more intricate differences when comparing individual VOC species, and the authors examined those in detail as well and proposed potential reasons. The manuscript is well written and easy to follow, and the results are significant and insightful. I recommend the manuscript be published. My comments are mostly suggestions and not required changes for publication.

In Line 146, the authors mentioned that xylenes in PTR could potentially be overestimated due to fragmentation in the mass spectrometer. I think this issue is very likely and should be explored further. The authors indicated that they would explore the measurement issue in Sect 4, but this issue was not specifically looked at in Sect 4. I wonder if the abundance of oxygenated aromatic compounds from BB emissions could factor into the overestimation of xylenes.

In Lines 355-360, the authors investigated why oxygenated compounds like formaldehyde, acetic acid are underestimated, even more so than other VOCs. While I agree that secondary in-plume production is very likely, I wonder if there could be other factors that cannot be ruled out. For example, these are all water soluble compounds. Could the wet deposition be overestimated in the model? Also, could there be production of these secondary species from other non-BB species? As mentioned earlier, this is more a suggestion for a deeper investigation, and I am just curious.

I noticed that there is always a jump in concentrations at two heights (530 hPa and 650 hPa) in the measured data. For some species, the model can reproduce this trend, but not always. Does this have to do with the injection heights? I am wondering if this will help diagnose the underestimation as well.

Line 421-425: I do not fully understand this. How would aromatic hydrocarbon/hydrocarbon relationship help diagnose the OH issue? Is this based on looking at the ratio of two hydrocarbons with different reactivities to look at OH exposure (i.e. hydrocarbon clock)? If not, can the hydrocarbon clock be used to assess OH exposure?

Minor comments:

Line 106: I cannot tell if it should be “though” or “through”. Both could potentially make sense.

Line 318: typo in xylene

Figure 1: it may be useful to indicate the ground site locations in the biomass burning panel too to give an idea of how far these ground sites are from the wildfires

Figure 2: for the species plotted with no bars, they are presumably not considered by each inventory? Might be helpful to indicate with n/a.