Review of “Duff burning from wildfires in a moist region: different impacts on PM2.5 and Ozone”

The authors are addressing a much-needed topic in fire emission calculations and air quality modeling. When ground fuels burn they become a large source of trace gases and aerosols into the atmosphere and there is a great lack of data needed to quantify these emissions. This manuscript highlights the need for more information, and how the lack in the current available information hinders air quality analyses. Below I have specific comments designed to help make this analysis more robust.

Specific Comments 

Fire Events: Four of the largest wildfire events in the southeastern US were selected for air quality modeling with WRF-Chem for two domains. One domain simulates fires in the southern Appalachian mountains and one domain simulates fires in southern Georgia. The fires ranged from 91K acres to > 500K acres. One thing I found lacking was consideration of the Evans Road and Pains Bay wildfires.  They were smaller than the top 4 criteria (41K, and 5K acres respectively), but were significant in terms of the emissions from the burning of organic/ground fuels and subsequent air quality impacts. Rappold et al. 2011 conducted a health impact analysis from the Evans Road wildfire and Tinling et al. 2016 conducted a similar analysis for the Pains Bay wildfire. The first two sentences of the Abstract state that “Wildfires can significantly impact air quality and human health. However, little is known about how duff and peat burning contributes to these impacts.” Given the goals of this paper, I would expect these studies/impacts be part of the introduction and also a consideration in this study.

Duff Flaming Phase Emissions: Why only focus on flaming emissions of duff? Smoldering phase emissions are important in terms of air quality impacts (as noted by Rappold et al. 2011 and Tinling et al. 2016). Smoldering of ground fuels does not always occur on a time scale of months to year, it can occur on the scale of hours/days, and while the plumes do not necessarily loft high, during the day they mix near the surface (where people breath) under the mixing height and can be transported further distances. At night they can transport along terrain features often impacting small towns in rural areas closer to the fires. This becomes an environmental justice issue as well. Limiting the work here to only flaming phase duff emissions unnecessarily limits the utility of this study.

Duff Consumption: Related to this is how much duff consumption actually went into each of the scenarios? I see that 4.6 cm of duff burned in the 2016 Rough Ridge fire which went into the App16 case (fuel loading 3.15 kg/m2). How many centimeters of duff burned in the Oke07, Oke11 and Oke17 cases? And was the same fuel loading from App16 (3.15 kg/m2) assumed for the Oke scenarios? The end of section 2.4.2 discusses how regrowth was handled, but again, what actual data went into the scenarios? I recommend adding the duff depth burned and fuel loading estimates to a table. Further, were all duff estimates assumed to burn in the flaming phase? Or were some estimated to burn in the smoldering phase (and thus eliminated if I am interpreting the discussion regarding the focus on flaming phase emissions)?

Duff Emissions: Section 2.4.2 indicates that “we estimated duff emissions and added them to FINN.” Was a full suite of trace gas and aerosol species added? Or were only PM2. 5, NO and NO2 species added to the simulations? At a minimum I would expect a full suite of species using default above-ground fuel emission factors would be added to represent the duff fuels, and ideally those emission factors be adjusted based on available literature for duff fuels. Recent studies for the SE in George et al. 2016 and Black et al. 2016 may be useful. They both conducted lab experiments based on peat from North Carolina. Many of these trace gases have implications for ozone and secondary aerosol formation.

VOC-limited: Section 3.3. Is the SE (App16 domain) really in a VOC-limited scenario in the winter (Nov)? I recommend showing estimates to support this.

NOX and Ozone Generation: Much of the discussion focuses on how the duff did not add much ozone to the model simulations, which is attributed to the NO/NO2 emission factors being low. There needs to be more discussion about the variability in NO/NO2/NOX emissions from duff. Yokelson et al. 2013 is just a single experiment and Urbanski 2014 applies an uncertainty of 100% to the data. Studies such as Burling et al. 2010, McMeeking et al. 2009, Selimovic et al. 2018 and Clements and McMahon 1980 are all studies that measured emission factors for ground fuels. NO values range from 0.56 to 2 g/kg, and NO2 values range from 0.23 to 2.7 g/kg. These data argue for perhaps using greater EF values for NO and NO2 and also (especially) sensitivity runs that vary the NOX EF’s by more than just 20%. Recommendation: I recommend an additional sensitivity run using 100% per Urbanski 2014 (e.g. 2x duff).

Meteorology/Transport: Section 3.2 discusses the PM2.5 emissions and transport. I recommend making the discussion more robust by including references to support the statement “Both biases in fire emission calculation and smoke transport simulation should be the contributors.” I recommend Li et al. 2020 and Garcia-Menendez et al. 2013. 

Technical Corrections

Table 1 caption needs to include more information such that the information in the table is understandable independent of the paper. Also, I recommend adding fire size (acres) to the table as well.

References cited here

Robert R. Black, Johanna Aurell, Amara Holder, Ingrid J. George, Brian K. Gullett, Michael D. Hays, Chris D. Geron, Dennis Tabor. Characterization of gas and particle emissions from laboratory burns of peat, Atmospheric Environment, Volume 132, 2016, Pages 49-57, ISSN 1352-2310, https://doi.org/10.1016/j.atmosenv.2016.02.024.

Burling, I. R., Yokelson, R. J., Griffith, D. W. T., Johnson, T. J., Veres, P., Roberts, J. M., Warneke, C., Urbanski, S. P., Reardon, J., Weise, D. R., Hao, W. M., and de Gouw, J. 2010. Laboratory measurements of trace gas emissions from biomass burning of fuel types from the southeastern and southwestern United States. Atmospheric Chemistry and Physics 10: 11115-11130. doi.org/10.5194/acp-10-11115-2010 https://www.fs.fed.us/rm/pubs_other/rmrs_2010_burling_i002.pdf

Clements, H. B. and McMahon, C. K. 1980. Nitrogen oxides from burning forest fuels examined by thermogravimetry and evolved gas analysis. Thermochimica Acta 35: 133-139. doi.org/10.1016/0040-6031(80)87187-5 https://www.srs.fs.usda.gov/pubs/ja/ja_mcmahon015.pdf

Garcia-Menendez, F., Y. T. Hu, and M. T. Odman. 2013. Simulating smoke transport from wildland fires with a regional-scale air quality model: Sensitivity to uncertain wind fields. J. Geophys. Res.-Atmos. 118 (12):6493-6504. doi: 10.1002/jgrd.50524.

Ingrid J. George, Robert R. Black, Chris D. Geron, Johanna Aurell, Michael D. Hays, William T. Preston, Brian K. Gullett. Volatile and semivolatile organic compounds in laboratory peat fire emissions, Atmospheric Environment, Volume 132, 2016, Pages 163-170, ISSN 1352-2310, https://doi.org/10.1016/j.atmosenv.2016.02.025. 

Li, Y., D. Q. Tong, F. Ngan, M. D. Cohen, A. F. Stein, S. Kondragunta, X. Zhang, C. Ichoku, E. J. Hyer, and R. A. Kahn. 2020. Ensemble PM2.5 forecasting during the 2018 camp fire event using the HYSPLIT transport and dispersion model. J. Geophys. Res.-Atmos. 125 (15):e2020JD032768

McMeeking, G. R., Kreidenweis, S. M., Baker, S. Carrico, C. M., Chow, J. C., Collett, J. L., Hao, W. M., Holden, A. S., Kirchstetter, T. W., Malm, W. C., Moosmüller, H., Sullivan, A. P., and Wold, C. E. 2009. Emissions of trace gases and aerosols during the open combustion of biomass in the laboratory. Journal of Geophysical Research 114: D19210. doi:10.1029/2009JD011836 https://www.fs.fed.us/rm/pubs_other/rmrs_2009_mcmeeking_g001.pdf

Rappold, A. G., S. L. Stone, W. E. Cascio, L. M. Neas, V. J. Kilaru, M. S. Carraway, J. J. Szykman, A. Ising, W. E. Cleve, J. T. Meredith, H. Vaughan-Batten, L. Deyneka, and R. B. Devlin. 2011. Peat bog wildfire smoke exposure in rural North Carolina is associated with cardiopulmonary emergency department visits assessed through syndromic surveillance. Environ. Health Perspect. 119 (10):1415-1420. doi: 10.1289/ehp.1003206

Selimovic, V., Yokelson, R.J., Warneke, C., Roberts, J.M., de Gouw, J., Reardon, J., and Griffith, D.W.T. 2018. Aerosol optical properties and trace gas emissions by PAX and OP-FTIR for laboratory-simulated western US wildfires during FIREX. Atmospheric Chemistry and Physics 18: 2929–2948. doi.org/10.5194/acp-18-2929-2018 https://www.atmos-chem-phys.net/18/2929/2018/acp-18-2929-2018.pdf

Tinling et al. Environmental Health (2016) 15:12. DOI 10.1186/s12940-016-0093-4

The purpose of this study is to assess the importance of including duff in simulation of wildfire impacts on air quality. The authors conduct WRF-Chem simulations for four large wildfire events in the southeastern US using two fire emission scenarios – duff and no duff, as well as a control simulation with no fire emissions. The main findings of the study are:

1) relative to the no duff fire scenario (surface and understory fuels only), the increase in wildfire emissions through the inclusion of duff burning resulted in large increases in simulated surface PM2.5 concentrations near the fire locations (< 300km) and at remote urbans areas

2) while the no duff fire scenario increased regional O3 levels, the impact of additional emissions from duff burning were negligible for O3

3) relative to the control scenario (no fire) both the no-duff and duff fire simulations generally increased agreement between the WRF-Chem simulated PM2.5 and observations at surface air quality monitoring sites in fire impacted areas

4) relative to the no duff fire scenario, inclusion of duff emissions generally improved agreement of the WRF-Chem PM2.5 and observations at surface air quality monitoring sites in fire impacted areas

The authors conclude that modeling of regional air quality in the southeastern US can be improved by adding duff burning emissions to existing fire emission datasets and that emissions from duff burning have much garter impact on PM2.5 than O3.

 The topic addressed in this study, contributions of duff and peat burning emissions to regional air quality, is certainly an important and of interest to air quality modelers and atmospheric scientists and is relevant to biomass burning in many regions.  However, in the manuscript requires major revisions before it is suitable for publication. The paper is missing key methodological details and a couple important choices in the study design are not well justified. Additionally, the presentation and discussion of results and lacks definition and focus, making it difficult to evaluate the authors’ conclusions and the overall broader relevance of the study.     

Here I provide my most important concerns regarding the paper, followed by less crucial, specific comments.

1. Estimation of duff consumption by flaming combustion

In Zhao et al. (2019), post-fire field measurements at 4 sites (2 pair: 2 burned, 2 unburned) at the location of the ~11,700 ha Rough Ride Fire, indicated duff depth consumption of 4.5 cm and 4.9 cm. Based on undocumented and unreferenced information, “In fact, whereas a duff layer is typically consumed during the smouldering phase of combustion, the monitoring and images taken during the RRF indicated that a large portion of the duff layer burned during the flaming phase of combustion.”, Zhao et al. (2019) assume that nearly all of the duff consumption occurred during flaming combustion in one day. In the current study, the authors use the 4.5 cm duff depth consumed by flaming consumption claimed by Zhao et al. (2019) and apply it to the four fire cases. This choice does not seem justified based on the less than robust information presented in Zhao et al. (2019).   

2. Duff PM2.5 emission factors

The study is simulating the impacts of flaming duff consumption on air quality, but they use PM_2.5 EF factors for smoldering duff (Urbanski 2014; Geron & Hays, 2013). The high PM_2.5 and VOC emission factors for duff burning result in large part because duff burns primarily by smoldering combustion. The authors should have used a reduced PM2.5 EF to represent flaming combustion.

3. Temporal emissions profile

The authors do not describe how the daily fire emissions were converted into hourly emissions for the WRF-Chem simulation. The appear seems to imply they were not:

L408-409: “The daily variations are different between observations and simulations because the observed fire emission dataset was at daily rather than hourly intervals.”

4. Assessment of smoke impacts 

It is unclear how the authors define air quality (AQ) observation sites as influenced or not influenced by smoke. Is smoke “influenced” defined from the perspective of the model e.g., air quality monitoring sites that were impacted by a conserved smoke in the WRF-Chem simulation or PM2.5 or CO levels greater than non-fire simulation?  Or is smoke influenced defined by AQ observation e.g., PM2.5 > some threshold. The criteria for smoke influenced needs to be clearly defined. And the rational for the criteria explained.

There are too many figures and the accompanying discussions are difficult to follow. I feel the study would be better served had the authors focused on a handful of days using air quality sites that were smoke impacted, from the simulations’ perspective using a clear, well defined definition of smoke impacted (e.g., WRF-Chem conserved smoke tracer, CO levels, etc.)

Specific Comments

5. L9: “The emissions of duff burning were estimated based on a field measurement”

6. L24-26: “Fires contribute 26.9% of total volatile organic compounds (VOC) emissions and 27.5% of PM emissions in the U.S. according to the 2014 US Environmental Protection Agency (EPA) National Emissions Inventory (NEI) (USEPA, 2017).”

The VOC contribution seem high. Please double check.

7. L33-34: “Wildfires produce about 3.5% of global tropospheric ozone production, though ozone production rates of individual fires vary with fuel type, combustion efficiency, etc. (Alvarado et al., 2010; Jaffe and Wigder, 2012).”

Location, time of year, meteorology, and pre-existing atmospheric composition are likely important factors as well.

8. L44-45: “In many regions around the world, including the U.S., wildfires have an increasing trend during recent decades…”

Elaborate on what kind of increasing trend? Frequency of large fires, fire severity, burned area?  

9. L49-55: Paragraph needs rewriting. Introductory sentence of the paragraph is about human health impacts of smoke, but two of three following sentences discuss radiative impacts of smoke aerosol. I suggest dropping the radiative impact sentences and added more information on health impacts.

10. L71-72: “Duff typically represents the detritus or dead plant organic materials fallen at the top layer of soil.”

This is a good location to define the terms “duff”, “peat”, and “organic soil”. They are often used interchangeably when discussing global wildland fire. For example, a couple sentences down the authors use “organic soil”.

11. L82-83: “Besides duff, peat is another burnable organic soil that typically represents the fermentation below the duff layer (Frandsen, 1987).”

See previous comment.

12. L105: Define “prescribed fire”

13. L106: Yokelson et al. (2013) is not a good reference for this statement, more appropriate reference(s) needed.

14. L115-116: Change to: “…estimated to account for approximately 60% of total PM2.5 emitted from the fire.”

15. L131: “…temperate forest duff emission factor of nitrogen oxides (NOx) is 0.67 g/kg…” Citation needed.

16. L137-142: This last paragraph of the Introduction needs to provide a better, but still brief, overview of the study (similar to the abstract).

17. L144: 2.1 Study Region

The authors should provide a map of the study region with polygons of fire perimeters and markers for urban areas of interest in the air quality simulations. Include scale bar.  This is necessary for the reader and will allow the authors to streamline this section which reads very rough.

18. L165: 2.2 Fire cases

Maps of each fire subregion region with fire boundary polygons should be provided. Perhaps a three panel – entire study region (see comment above), southern Appalachian region, and the Okefenokee swamp region.

19. L173-185: It would be interesting if the authors could provide a couple sentences on the fire history of the Okefenokee swamp region. Three large fires in a short period, how does this compare with fire history at the swamp?

20. L209-210: Refer reader to figure for domain. Also, I suggest swapping order of Fig S1 and S2. Present model domains first, then time series of OC emissions related to fire activity.

21. L240-241: “The fire emissions from FINNv1.5 were implemented into WRF-Chem by Pfister et al. (2011a), which contains the daily burned area and emissions of an amount of gas and aerosol species with a spatial resolution of 1 km (Wiedinmyer et al., 2011).”

More explanation is needed here. Did the WRF-Chem installation that you used include FiNNv1.5 fire emissions on an hourly time step that could be included in simulations? I didn’t think that was an option. Or did you download FiNNv1.5 fire emissions with MOZART speciation and ingest these emissions into WRF-Chem? If the latter, how did you convert daily emissions of FiNNv1.5 (https://www.acom.ucar.edu/Data/fire/data/README_FINNv1.5_08112014.pdf ) into hourly input for WRF-Chem?

22. L265: “The duff burning contributed 60% of the total PM2.5 emission”

Should read: “The duff burning was estimated to have contributed 60% of the total PM2.5 emission”

23. L291-295: The authors should note that the Geron & Hays (2013) was a field study that made in-situ measurements of EFPM2.5 from three different peat fires in coastal North Carolina.  Black et al. is a laboratory study that measured EF from peat core samples from two locations in North Carolina. BTW – Black et al. (2016) is missing from bibliography.  

24. L303-305: Please provide reference for NOx EF or refer reader to Table 2. The authors do not mention Table 2 anywhere in the text. Table 2 should be referenced when discussing EF.

25. L306-314: This is a reasonable approach at the Okefenokee fire sites, to estimate duff reduction from previous burns (2007 to 2011, 2007/2011 to 2017).

26. L322-325: “We did not use the commonly used approach to scale up the FINN emissions because we wanted to understand if the missing duff burning contributed to the underestimate FINN emissions to a certain extent. This FINN emission underestimate would lead to uncertainty in quantitatively estimating the contribution relative to the above-ground fuel consumption.”

This is unclear and must be rewritten.

Section 3.1

27. It is unclear how the authors define air quality (AQ) observation sites as influenced or not influenced by smoke. Is smoke “influenced” defined from the perspective of the model e.g., air quality monitoring sites that were impacted by a conserved smoke in the WRF-Chem simulation or PM2.5 or CO levels greater than non-fire simulation?  Or is smoke influenced defined by AQ observation e.g., PM2.5 > some threshold. The criteria for smoke influenced needs to be clearly defined. And the rational for the criteria explained.

28.Figure S3 is not that helpful in discerning agreement between base simulation (No fire) and AQ observations. A time series like Figure 3 would be far more useful (with smoke influence clearly defined, previous comment).

29. State how many air quality sites were used in each domain.

3.2 The PM2.5 emission and transport from duff burning

30. L381-382: “Thus, implementing duff burning doubles the PM2.5 concentrations from App16”

Is this statement based on specific AQ site(s)? Please clarify.

31. L384-385: “The total burned area of Oke07 was 5 times more than that of App16. The emissions were larger from Oke07 and correspondingly the simulated PM2.5 concentrations are greater.”

Since Oke07 lasted > 2 months, it would be more useful to compare area burned and emissions for the periods of the simulations, perhaps as daily average.

32. L386-388: “In the sim_FINN+duff runs, the simulated fire plume effectively approaches the underestimated regions, but the enhancement is still not enough over some regions.”

By approaches, I assume the authors mean the simulated surface PM2.5 in the plume approaches the concentration of the AQ observations showing greatest impact? Please clarify.

33. L408-409: “The daily variations are different between observations and simulations because the observed fire emission dataset was at daily rather than hourly intervals.”

I commented on this topic earlier. How did the authors temporally distribute (daily to hourly) the fire emissions for WRF-Chem input? The temporal distribution of fire emissions is critical to getting realistic simulations. 

Technical

It would make for a more pleasant read if for URLs the authors provided citations in the text and the links in the bibliography, e.g.:

“(http://www.gatrees.net/forest-management/forest-health/alertsand-updates/Wildfire%20Damage%20Assessment%20for%20the%20West%20Mims%20Fire.pdf, last access: December 3,185 2020).”

L33: “Wildfires produce about…” to “Wildfires account for…”

L49: “…when fire plumes are transported…” to “…when smoke plumes are transported…”

L75-76: “…may provoke each other…” Rephrase, be more specific about the physical processes to which you are referring.

L83: change “swamp” to “wetland”, the latter encompasses swamp, bog, march, etc.

L86: “…are also evaluated…” to “have also been evaluated…”

L88: “However, the air quality impacts of emissions from duff fires are very limited…” I don’t think this is what the authors intend to state.

L109: Change “springs” to “spring”

L216: Should refer to Fig S2.

Additional English usage / technical corrections needed at: 154, 175-177, and other places