Where there is smoke there is mercury: Assessing boreal forest fire
mercury emissions using aircraft and highlighting uncertainties
associated with upscaling emissions estimates

Abstract. Mercury (Hg) emitted from biomass burning is an important source of the contaminant to the atmosphere and an integral component of the global Hg biogeochemical cycle. In 2018, measurements of gaseous elemental Hg (GEM) were taken on-board a research aircraft along with a series of co-emitted contaminants in the emissions plume of an 88 km2 boreal forest wildfire on the Garson Lake Plain (GLP) in NW Saskatchewan, Canada. A series of four flight tracks were made perpendicular to the emissions plume at increasing distances from the fire each with 3–5 passes at different altitudes at each downwind location. The maximum GEM concentration measured on the flight was 2.88 ng m−3, which represents a ≈2.4x increase in concentration above background. GEM concentrations were significantly correlated with the co-emitted carbon species (CO, CO2, and CH4). Emissions ratios (ERs) were calculated from measured GEM and carbon co-contaminants data. Using the least uncertain of these ratios (GEM : CO), GEM concentrations were estimated at the higher 0.5 Hz time resolution of the CO measurements resulting in maximum GEM concentrations and enhancements of 6.75 ng m−3 and ≈5.6x, respectively. Extrapolating the estimated maximum 0.5 Hz GEM concentration data from each downwind location back to source, 1 km and 1 m (from fire) concentrations were predicted to be 12.9 and 29.9 ng m−3, or enhancements of ≈11x and ≈25x, respectively. ERs and emissions factors (EFs) derived from the measured data and literature values were also used to calculate Hg emissions estimates on three spatial scales: (i) the GLP fires themselves, (ii) all boreal forest biomass burning, and (iii) global biomass burning. The most robust estimate was of the GLP fires (21 ± 10 kg of Hg) using calculated EFs that used minimal literature derived data. Using a Top-down Emission Rate Retrieval Algorithm (TERRA) we were able to determine a similar emission estimate of 22 ± 7 kg of Hg. The elevated uncertainties of the other estimates and high variability between the different methods used in the calculations highlight concerns with some of the assumptions that have been used in calculating Hg biomass burning in the literature. Among these problematic assumptions are variable ERs of contaminants based on vegetation type and fire intensity, differing atmospheric lifetimes of emitted contaminants, the use of only one co-contaminant in emissions estimate calculations, and the paucity of atmospheric Hg species concentration measurements in biomass burning plumes.


The purpose of the extensive discussion of the GEM enhancements is not clear. In addition, it will strongly depend on the meteorology which is omitted from the discussion. Without consideration of the windspeed (dilution) and the distance to the fire, the comparison with measurements published by other authors does not make much sense.
Section 2.1: Measurement of wind speed and direction onboard aircraft is not easy. The reader would like to know how these parameters were measured and with which uncertainties. This information is needed to assess the uncertainty of fluxes calculated by TERRA.
Sections 2.2 and 2.3: What are the estimated uncertainties of the individual GEM, CO, CO2, and CH4 measurements. These uncertainties are needed for assessment of the quality of ERs: e.g. the poorer quality of GEM/CH4 ratio could be caused by higher uncertainty of CH4 measurements? They are also needed for the orthogonal correlations (Cantrell, ACP, 8, 5477-5487, 2008).
Line 233: I presume the GEM background is given as an average of 2 min measurements. What was the number of the GEM measurements used in this average? Lines 237-241: The consideration of only GEM enhancements >125% is probably not justified for several reasons: a) It is arbitrary -why not 115%? b) The selection of >125% GEM data may show only a part of the plume which may not be representative of the whole plume. c) With increasing distance to the fire, the section of plume with >125% would decrease relatively to the whole plume which again poses the question of representativity. d) The authors state that the data below 125% the enhancement are "too variable and too uncertain" to be considered. The concern about the uncertainty should not be the problem if the authors used orthogonal regression with uncertainties of both GEM and X. The variability should also be no problem: the more points the smaller R is significant. Reference to Yokelson et al. (2013) is not quite appropriate for the situation here, i.e. with measurements up to 100 km distance from the fire with nearly constant background mixing ratios.
In addition, the limitation to GEM enhancements >125% even seems to be unnecessary: the ERs in figure S4.1 calculated from all data and in Table 1 calculated with only GEM enhancements >125% are probably the same in statistical terms, i.e. cannot be distinguished taking into account the ERs uncertainties and the number of measurements. An additional table of ERs from Table 1 and ERs from figure S4.1 could be used to illustrate the necessity of the >125% threshold or its absence.
Paragraph lines 245-248: What type of regression was used: LSQF or orthogonal one? The usual LSQF should not be used to calculate ERs!!! The last sentence is difficult to understand because any type of regression automatically adjusts for the backgrounds -it is just a shift in the coordinate system. Probably CO, CO2, CH4, and NMHC measurements were converted to 2 min averages synchronized with 2 min GEM for correlations? If so, it should be mentioned.
Paragraph lines 249-252: How was the integration made? Only for the enhancements >125%?
Paragraph lines 305-327: More information is needed in the description of TERRA calculation? How were GEM measurements (2 min) interpolated? The treatment of the layer above the highest transection and the inversion layer has to be mentioned too. Some of this information is provided in Section 3.4 but the reader would expect it here.
Lines 348 and 349: How was the correlation made: orthogonal? 2min data? What are the regression lines?
Paragraph lines 459-464: The conversion of 2 min GEM measurements into 0.5 Hz data using the GEM/CO ratio was probably made mainly for the TERRA calculations. If so, it should be mentioned.
Paragraph 465-480: The text here is highly speculative because the estimations are C3

Interactive comment
Printer-friendly version Discussion paper strongly dependent on the meteorology which is not mentioned. In addition, it does not contribute much to the purpose of the paper. Table 1: The comparison of maximal measured GEM enhancements is strongly dependent on their temporal resolution, as shown in this work, and on the meteorological parameters (especially wind speed in combination with the distance to the fire, i.e. dilution). Without taking all these factors into account, the comparison does not make much sense and as such should be deleted from the table, and also from the text. Table 2: "Uncertainty" should be used instead of "error" here and throughout the text. The calculation of uncertainties and the terms used in the equation 7 should be given either in the manuscript or in the supporting information. Fig. 2b: I wonder about 2s GEM data derived from 0.5 Hz CO data using the GEM/CO ratio. What is this conversion good for? I find its presentation misleading because it pretends much higher density of GEM data than available. The 0.5 Hz CO, CO2, and CH4 measurements should be converted to the 2 min GEM time stamp, at least for the regressions.  Interactive comment on Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2020-1119, 2020.