Main Comments:

Improving methods for estimating SBTEX emissions and consequently concentration fields is important for studies on the potential health risks posed by exposure to these pollutants. The HAPI tool developed and released as part of this dataset represents a straightforward and seemingly robust way to make progress on that front by leveraging and combining relevant information currently present in different underlying emission inventories. The other parts of the new dataset, i.e. the new gridded emission files including the imputed SBTEX emissions and the SBTEX concentration fields calculated from these emissions with the CAMx RTRAC tool serve as a nice example of what can be accomplished after applying the HAPI tool. The associated figures and tables documenting features of the gridded datasets for this 5 month case study are well done. That said, given the limited spatial and temporal scope of the gridded emission and concentration fields, the direct use of these aspects of the dataset in future studies may be somewhat limited. Put differently, unless any follow-up studies requiring SBTEX data specifically focus only on this 5-month period in 2012 over this specific area, I expect to see little follow-up use of these portions of the dataset by other groups. The model evaluation R and python scripts used to generate the manuscript figures are nice and easy to follow for anyone familiar with these languages and could easily be adapted to perform similar analyses for other periods or regions.

With a strong background in emissions and air quality modeling over the U.S., the gridded data sets are usable and understandable, but without such a background, I would not consider them to be described and documented in sufficient detail. For example, the zenodo description lists these files as being in netCDF format, but omits the important fact that the spatio-temporal metadata information stored in these files follow I/O API conventions (https://www.cmascenter.org/ioapi/). These conventions, while often used in U.S. emissions and air quality modeling applications, are virtually unknown outside that community, making the interpretation of time and geolocation information of the values stored in these files impossible to many users. At a minimum, the documentation on both zenodo and within the article should make reference to the use of I/O API conventions and link to these conventions, but given the niche nature of I/O API and the learning curve associated with correctly interpreting its geospatial metadata, preferably the time and spatial coordinate and attribute conventions of these files should follow more widely used conventions such as netCDF-CF with explicit latitude, longitude, and time variables included in each file. Likewise, there is no explicit documentation on the meaning of variable names in the emission and concentration files. The “var_desc” attribute could be used to provide such documentation, but it currently just repeats the variable name without providing any further insight into, for example, the differences between “ALD2” and “ALD2_PRIMARY” which are not likely to be known to most readers / users. Documentation should be provided for all variables in all datasets. Because of these shortcomings in metadata and variable documentation, I do not consider the gridded datasets in their current form to be of high enough quality.

I am confused by the description of RTRAC as a “post-process feature” / “post-analysis feature”, and “post-processing step” in the main article and supplement, and its depiction as a separate box in Figure S1. The RTRAC documentation in the CAMx user guide describes it as a built-in probing tool that, if enabled for a given CAMx simulation, is being applied simultaneously with the base model, and not after the CAMx simulation. Moreover, the description in the text and Figure S1 state that RTRAC tracer concentrations are affected by emissions and physical and chemical decay, but do not mention transport (advection and diffusion). Are the RTRAC SBTEX species not transported from their emission sources? Based on the CAMx user guide, I think they are, but based on the RTRAC method description provided in this article, I was left with the impression that they weren’t.

The method description in Section 2.2 of the main article should be improved as it would be critical for any user trying to replicate the CAMx-simulated concentration fields. The details of the air quality model description provided in Section 2 of the supplement should be included in Section 2.2 of the main article (e.g. information on the version of WRF, photolysis rates, and the generation of boundary conditions). There appears to be a contradiction between the main manuscript and the supplement about how WRF fields were prepared for CAMx – the main manuscript states “They were converted to SMOKE- and CAMx-ready gridded hourly meteorology through the Meteorology Chemistry Interface Processor (MCIP)” while the supplement states that “They are converted to the CAMx-ready format using the WRFCAMx version 4.8.1 program developed by the CAMx development team”.

The method description should also define “flexi-nesting”, a term not likely to be familiar to most readers or users of the dataset. Based on my read, both meteorology and emissions were processed for a 12 km grid, and the 4 km “flexi nest” grid was solely defined during the CAMx simulation with CAMx interpolating 12 km inputs to that finer grid during the simulation. Or were there any inputs actually prepared for the 4 km grid? (the distributed emissions dataset is for 12 km). If none of the inputs were prepared for the 4 km grid, what is the rationale for having CAMx interpolate 12 km inputs to that finer grid? I also cannot reconcile the fact that CAMx RTRAC simulations apparently were performed for that 4 km grid (implying that having 4 km concentration fields is desirable) with the fact that for model evaluation, the 4 km grid results are aggregated back up to 12 km by considering all nine grid cells surrounding a monitor (implying that modeled gradients at 4 km are not expected to be realistic).

The structure and language of the article and supplement is generally good, but there are a number of instances with somewhat awkward wording, with a few examples noted in the specific comments below. During revision, the article would benefit from a careful editorial review.

Specific Comments:

Line 49: Here and in all subsequent references to (USEPA, [year]) in the main article and supplement: please double check that all EPA citations listed in the references section actually have a publication year associated with them. At first glance, only the “Guidance on the Use of models and other analyses for demonstrating attainment of air quality goals for ozone, PM2.5 and Regional Haze” [which was actually updated in 2018] and “2014 Fire NEI Workshop Emissions Processing-SmartFire Details” references currently do, and yet there are many “(USEPA, [year])” citations throughout the text that don’t have a clear connection to an item in the reference list.

Lines 108 – 110: see main comments above, is the reactive tracer function really a “post-process”?

Lines 136 – 140: The first sentence says emissions “from certain facilities” are collected by EIS and then used to develop the NEI while the last sentence then says the NEI has “emissions … from all types of emission sources”. Please clarify

Line 139: CO is misspelled

Line 145 – 146: Does this distinction indeed apply to the NEI, or more broadly to an emissions modeling platform that includes the NEI but also tools like SMOKE and speciation databases? Specifically, does the NEI actually include “model surrogate” species, or are they only computed from NEI information (VOC and/or HAP explicit) during emissions processing for a specific chemical mechanism in a specific modeling platform?

Line 193: should “Evoc” be “Evoc,s,f”?

Lines 198 – 199: unclear writing. Perhaps “Only the emission sources for which the sum of all HAPs is zero (sum Eisf =  0) are considered as the "without HAPs" group” instead?

Line 218: “The emissions sectors modulated by meteorology” instead of “meteorology-induced”?

Line 221: “used” rather than “imported”?

Lines 224 – 230: See main comment above regarding clarification of RTRAC.

Line 245: please define “long-term” and justify why the focus was on long-term evaluation.

Lines 252 – 254: this suggests that only high outliers (2*IQR above Q3) were removed while no screening was performed for low outliers. If so, please state explicitly and justify.

Line 265: I believe this should be “vessel”, not “vehicle”

Line 290: change “some rural area” to “some rural areas”

Line 301: suggest “reduce” instead of “mitigate”

Lines 318 – 324: There are some awkward sentences in this paragraph, please review and revise.

Lines 342 – 345: The first line states that it’s mostly meteorology, but then the third line says both meteorology and more active daytime chemistry and the associated chemical loss of these tracers act to decrease daytime tracer concentrations. Review and revise.

Line 346: “sensitivity of the concentration to emissions” instead?

Line 347: Are nighttime exposures when people are mostly at home a concern, and should lowering concentrations during that time period be a priority?

Line 356: Suggest not starting this sentence with “in contrast” because it actually says that the level of peak SBTEX is the same for both locations. Instead, start the next sentence “In contrast to Channelview, the peak concentration at Bayland Park occurs ….”

Line 361 – 362: awkward wording of the first part of the sentence, review and revise.

Line 371 – 372: change “composition of” to “contribution from”?

Line 420: “applied” rather than “implemented” since no modifications to off-the-shelf CAMx code were made?

Line 429: “underestimated” instead of “missed”?

Line 493 – 495: awkward wording, review and revise.

Line 504: Ramboll, not Ramball.

This study uses observations and modeling to create an emission map of five hazardous air pollutants (HAPs) for May-Sep 2012 in the U.S. Gulf region. This map includes emissions that are missing in the US EPA emission inventory. High emissions were reported in the literature for these five HAPs in the Gulf region and they pose health risks, therefore an accurate emission inventory is beneficial. The analysis methodology is sound. However, I have some concerns listed below.

Major comments:

The authors should put more effort into justifying the use of various datasets. The current time coverage of the emission map generated is only five months in 2012. Why do authors not use more up-to-date datasets to generate an emission map for longer and more recent time periods? The AMTIC ambient measurements are available until 2023, and US EPA NEI has the most recent version for SBTEX emissions in 2023. So I’m confused why the authors used 2012 ambient measurements and 2011 NEI. If authors find 2012 emissions very interesting, I encourage them to extend the method to the most recent (at least until 2022) to have a 10-year monthly emission map for the Gulf region. This will make the study more interesting to the community.

Specific comments:

Line 96-97: why dispersion models cannot support regional scale application? Also add references.

Line 129: provide reference for NEI 2011.

Line 130-133: explain the difference between NEI vs. TRI vs. EIS.

Line 140: lowercase for “Styrene”

Line 197-198: this means that their emissions are assumed to be correct? Having no missing emission doesn’t mean it is correct. Uncertainty and evaluation against observations should also be conducted for them.

Section 2.2: what is the spatial and temporal resolution of CAMx RTRAC?

Line 218-220: should include more details about meteorology-induced emissions sectors, biogenic and wildfire emissions in the main manuscript, I see they are now in SI.

Figure S1: I recommend including it in the main manuscript.

Section 3: should provide more specifics on the confidence intervals, and uncertainty range of the emission rates or concentration reported.

Figure 2, 3: should include time periods in the figure legend.

Section 3.3: as model validation should come as section 3.1.

Table 2: (1) should provide such information for each site, each month and individual pollutant to show how they differ. (2) it seems that including missing emissions (“Adj”) does not improve R values, any reason for that?

Figure S8: there is a > 2-factor difference between observation and model for BTES from 10 pm to 8 am, this should be discussed.

Figure 8: color of diamonds are observational data? It is hard to compare observation with model results in this figure.

Section 4: should compare the new emission map to NEI and discuss how US EPA can improve their estimates, potentially also provide a correction factor for NEI to match the new emission map.

References: some references missed the publication year or the year of last access date.

In Supplementary material: line 67: hydroperoxyl radical is HO2, OH is hydroxyl radical