1 Possibility for an infrared sounder as IASI to document the HCOOH chemistry in biomass burning plumes

Formic acid (HCOOH) concentrations are often underestimated by models and its chemistry is highly uncertain. HCOOH is, however, among the most abundant atmospheric volatile organic compounds and it is potentially responsible for 15 rain acidity in remote areas. HCOOH data from the Infrared Atmospheric Sounding Interferometer (IASI) are analyzed from 2008 to 2014, to estimate enhancement ratios from biomass burning emissions over seven regions. Fire-affected HCOOH and CO total columns are defined by combining total columns from IASI, geographic location of the fires from MODerate resolution Imaging Spectroradiometer (MODIS) and the surface wind speed field from the European Centre for MediumRange Weather Forecasts (ECMWF). Robust correlations are found between these fire-affected HCOOH and CO total 20 columns over the selected biomass burning regions, allowing the calculation of enhancement ratios equal to 7.30×10 3 ±0.08×10 -3 mol/mol over Amazonia, 11.10×10 -3 ±1.37×10 -3 mol/mol over Australia, 6.80×10 -3 ±0.44×10 -3 mol/mol over India, 5.80×10 -3 ±0.15×10 -3 mol/mol over Southern East Asia, 4.00×10 -3 ±0.19×10 -3 mol/mol over Northern Africa, 5.00×10 3 ±0.13×10 -3 mol/mol over Southern Africa, and 4.40×10 -3 ±0.09×10 -3 mol/mol over Siberia, in a fair agreement with previous studies. The comparison with other studies highlights a possible underestimation by 60% of emission or a secondary 25 production of HCOOH by Siberian forest fires while the studied fire plumes originating from Southern African savanna could suggest a limited secondary production of HCOOH or a limited sink. In comparison with data set characterizing emissions, it is also shown that the selected agricultural burning plumes captured by IASI over India and Southern East Asia correspond to recent plumes where the chemistry or the sink do not occur.


Introduction 30
Formic acid (HCOOH) is one of the most abundant carboxylic acids present in the atmosphere.HCOOH is mainly removed from the troposphere through wet and dry deposition, and to a lesser extent by the OH radical.It is a relatively short-lived species with an average lifetime in the troposphere of 3-4 days (Paulot et al., 2011;Stavrakou et al., 2012).HCOOH contributes for a large fraction to acidity in precipitation in remote areas (e.g.Andreae et al., 1988).
More generally there are still large uncertainties on the sources and sinks of HCOOH, and on the relative contribution of anthropogenic and natural sources, despite the fact that recent progress has been made possible by using the synergy between atmospheric models and satellite data (e.g.Stavrakou et al., 2012;Chaliyakunnel et al., 2016).These uncertainties have an 45 impact on our understanding of the HCOOH tropospheric chemistry, as on the oxidizing power of the atmosphere or the origin of the acid rains.One of the large uncertainties in the HCOOH tropospheric budget seems to be the underestimation of the emissions from the forest fires, as recently suggested by Stavrakou  One way to estimate the atmospheric emissions of pyrogenic species is the use of emission factors.The emission factors are 50 often obtained from ground and airborne measurements or from small fires burned under controlled laboratory conditions.
The emission factors can also be derived from enhancement ratios of the target species relative to a reference species, which is often carbon monoxide (CO) or carbon dioxide (CO 2 ) due to their long lifetime (e.g.Hurst et al., 1994) and are based on the characteristic of the combustible and hence depends on the type of biomass burning.However, the difference between an emission ratio and an enhancement ratio is that emission ratios are calculated from measurements at the time of emission and 55 enhancement ratios are related to the ongoing chemistry.To correctly convert these enhancement ratios to emission ratios, the decay of the chemical species need to be taken into account or assumptions need to be made, suggesting that the enhancement ratios are equivalent to emission ratios, hence measured at the source and not impacted by chemistry.
Compilations of numerous enhancements ratios, emission ratios and emission factors for several trace gases from measurements at various locations in the world are published regularly (e.g.Akagi et al., 2011) in order to facilitate their use 60 in Chemistry Transport Models.
There has been a recent interest in calculating enhancement ratios and emission factors from satellite data (e.g.Rinsland et al., 2007;Coheur et al., 2009;Tereszchuk et al., 2011).Related to the above difficulty of inferring emission factors using the satellite observations comes from the fact that these observations are indeed typically made in the free/upper troposphere and further downwind of the fires.The fact that satellite mainly probe transported plumes where chemistry modifies the original 65 composition explains why the use of enhancement ratio is more relevant than emission ratio.This work aims to provide a list of enhancement ratios relative to CO for HCOOH over several biomass burning regions.For this, we analyzed seven years of IASI measurements, between 2008 and 2014.Section 2 describes the IASI satellite mission and the retrieval characteristics for the CO and the HCOOH total columns.Section 3 presents the fire product used from the MODerate resolution Imaging Spectroradiometer (MODIS) to identify the fire locations.Section 4 details the methodology 80 used to identify of the IASI fire-affected observations.In Section 5 we describe and analyze the enhancements ratios obtained from the IASI measurements and we compare these values to those available in the literature.Finally, the conclusions are presented in Section 6.

95
The HCOOH and CO columns from IASI are used hereafter to determine the enhancement ratios of HCOOH.CO is chosen as reference due to its longer tropospheric lifetime (several weeks) as compared to HCOOH.In our study we use CO as the reference and not CO 2 since variations in CO 2 concentration are difficult to measure with sufficient accuracy from IASI.

The CO retrieval characteristics
The CO concentrations are retrieved from IASI using the FORLI-CO software (Hurtmans et al., 2012), which uses an 100 optimal estimation method based on Rodgers (2000).The spectral range used for the retrieval is between 2143 and 2181.25 cm -1 .The CO total columns have been validated for different locations and atmospheric conditions (e.g.In order to keep only the most reliable retrievals, the selected data used have a root-mean-square error lower than 2.7×10 -9 W/(cm 2 cm -1 sr) and a bias ranging between -0.15 and 0.25×10 -9 as recommended in Hurtmans et al. (2012).

The HCOOH retrieval characteristics
The retrieval is based on the determination of the brightness temperature difference (ΔT b ) between spectral channels with 110 and without the signature of HCOOH.The reference channels used for the calculation of ΔT b were chosen on both sides of the HCOOH Q-branch (1105 cm -1 ), i.e., at 1103.0 and 1109.0 cm -1 .These ΔT b were converted into total columns of HCOOH using conversion factors compiled in look-up tables.This simple and efficient retrieval method is described in more details in Pommier et al. (2016).
As shown in Pommier et al. (2016), the vertical sensitivity of the IASI HCOOH total column ranges between 1 and 6 km.

115
This study also showed that large HCOOH total columns were detected over biomass burning regions (e.g.Africa, Siberia) even if the largest values were found to be underestimated.This underestimation, which is lower than 35% for the columns below 2.5×10 16 molec/cm 2 (Pommier et al., 2006), will propagate in the enhancement ratios calculated in this work.
On the other hand, a large overestimation of the IASI HCOOH columns was shown in comparison with ground-based FTIR.This overestimation was larger for background columns (expected to reach 80% for a column close to 0.3×10 16 molec/cm 2 ) 120 and it can also impact our enhancement ratios.developed by the University of Maryland (https://earthdata.nasa.gov/data/near-real-time-data/firms/active-fire-data#tabcontent-6)which for each detected fire pixel includes, the geographic location of the fire, the fire radiative power (FRP), the confidence in detection, and the acquisition date and time.The FRP provides a measure of fire intensity that is linked to the 130 fire fuel consumption rate (e.g.Wooster et al., 2005).

MODIS
Only data presenting a high confidence percentage are used, i.e. higher or equal than 80% as recommended in the MODIS user's guide (Giglio, 2013).

The selected areas 135
The determination of the biomass burning regions is based on the MODIS fire product.Figure 1 highlights the main areas that contributed to the biomass burning for the period between 2008 and 2014.Seven regions were selected for this work: Amazonia (AMA, 5-15°S 40-60°W), corresponding mainly to the Brazilian Cerrado, Australia (AUS, 12-15°S 131-135°E), Northern Africa (NAF, 3-10°N 15-30°E), Southern Africa (SAF, 5-10°S 15-30°E), Southern East Asia (SEA, 18-27°N 96-105°E), India (IND, 15-27°N 75-88°E) and Siberia (SIB, 55-65°N 80-120°E).Among these regions, India and Siberia were 140 not the most actives in terms of number of fires but it seemed important to also investigate them.One first reason for this is that Pommier et al. (2016) showed a misrepresentation of the fire emissions of HCOOH over India.Secondly, India also encounters excess of acidity in rainwater, which could be partly attributed to biomass burning (e.g.Bisht et al., 2014).
Concerning Siberia, this region and the surrounding areas experienced intense fires over the past years such as during the summer 2010 (Pommier et al., 2016;and R'honi et al. 2013 for the region close to Moscow).145

The IASI data used
For this work, both the daytime and nighttime IASI data were used.We have verified that using only the daytime retrievals did not change the results.Figure 2 presents the time-series of the monthly mean for the HCOOH and CO total columns over the seven selected regions.The number of fires and their FRP are also indicated.The variation in the total columns of HCOOH and CO matches relatively well with the variation of the number of fires.It is also worth noting that these 150 variations in the total columns do not depend on the intensity of the fires as shown by Fig. 2 and by the scatterplots with the values characterizing each fire as described below (not shown).
The monthly HCOOH and CO total columns are found to be highly correlated over the selected biomass burning regions (correlation from 0.75 to 0.91), except over India (r=0.34) and Siberia (r=0.58).Over both regions, the impact of other sources than biomass burning is thus not negligible.Over India, the CO budget is influenced by long-range transport (e.g.155 Srinivas et al., 2016) and the anthropogenic emissions have also a large impact (e.g.Ohara et al., 2007).This could explain that the variation in CO does not follow perfectly the variation in the number of fires and that the difference between the background level and the CO peaks is less marked than for the HCOOH.Over Siberia, a temporal shift between the highest peaks for CO and for HCOOH is noticed for few years as for 2009, 2010 and 2011.For these years, the variation in CO does not follow the variation in the number of fires.The large selected region over Siberia is known to be also impacted by CO- (e.g.Pochanart et al., 2003).These external influences interfere with the CO plumes originating from forest fires measured over this region.
Despite the overall good match between the number of fires and the variation in HCOOH and CO, we are not certain that the HCOOH and the CO were emitted solely by fires, and the discrimination between a natural and an anthropogenic origin for 165 each compound is challenging.This assessment is particularly obvious for IND and SIB.To isolate the HCOOH and CO signals measured by IASI, potentially emitted by a fire, we decided to only use the data in the vicinity of each MODIS hotspot.To do so, we co-located the IASI data at 50 km around each MODIS pixel and between 0 and 5h for each detected fire, so that each MODIS pixel is associated with a value of HCOOH and CO total columns from IASI.
With these co-location criteria good correlation coefficients, calculated by linear least-square fitting, are found between the 170 HCOOH and the CO total columns as shown in Tab. 1.The lower coefficients, i.e. less than 0.7 are found for India, Australia, Siberia and Northern Africa.It is also important to note that the HCOOH and the CO columns are better correlated for India and Siberia compared to the monthly time-series shown in Fig. 2. The three other regions present a large correlation, around 0.8.The high correlation suggests that IASI sampled the same biomass burning air mass for these compounds.When filtering out the data associated with a large surface wind (higher than 1.44 m/s), new correlations between the 185 HCOOH and the CO total columns from IASI are calculated (Tab.2).This value of 1.44 m/s for the surface wind speed corresponds to the 25 th percentile of the distribution of the three regions characterized by the lowest surface wind speed (Fig. 3).
The correlation coefficients, shown on the scatterplots in Fig. 4 and summarized in Tab. 2, increase for all regions except over NAF, where the coefficient is found to be slightly lower than the previous correlation (Tab.1).The correlation 190 coefficient is significantly improved over IND and SIB (Tab.2).These results confirm a robust correlation between the HCOOH and the CO total columns measured by IASI in the vicinity of each MODIS fire location.

Determination of the enhancement ratios
Based on scatterplots in Fig. 4, an enhancement ratio can be calculated for each region.These enhancement ratios defined as A good agreement is however globally found with previous studies even if it is important to keep in mind that an underestimation of our ER (HCOOH/CO) is possible due to the underestimation in the highest values of HCOOH as over the forest fires (see Section 2.3).On the other hand, the overestimation of the background column can also impact the 205 calculation of our ER (HCOOH/CO) .Both biases is however limited since most of HCOOH total columns used in our analysis over the selected regions are higher than 0.3×10 16 molec/cm 2 and lower than 2.5×10 16 molec/cm 2 as explained in Section 2.3.
Nevertheless, in order to investigate the possible impact of the overestimation in the lower columns on the calculated ratio, a test was performed, by using only HCOOH columns with a thermal contrast larger than 10K.Indeed, the increase in the thermal contrast (i.e. the temperature difference between the surface and the first layer in the retrieved profile) leads to 210 improve the detection limit and thus to minimize the bias with the lowest columns, as shown in Pommier al. (2016).Similar results were generally found in terms of slope and correlation coefficient, suggesting a negligible effect of the low column biases.The only exception being an increase in ER (HCOOH/CO) over Siberia (6.5×10 -3 ± 0.19×10 -3 mol/mol against 4.4×10 -3 mol/mol ± 0.09×10 -3 in tab.3).This gives us confidence that the results presented in Tab. 3 are reliable.
When compared with other studies, the best agreement for the values presented in Tab. 3 is found over Southern Africa ± 7.6×10 -3 mol/mol) since our ratio is 55% lower.One possible explanation is the multi-origin of the plumes studied by 230 Rinsland et al. (2006), since, based on their backward trajectory, their plumes could be influenced by biomass burning originating from Southern Africa and/or from Southern America.It is worth noting the ACE-FTS instrument used in their study, works in a limb solar occultation mode.This means the atmospheric density sampled by the instrument is larger than the one measured by the nadir geometry as with IASI.However, the difference in geometry is not a reliable assumption in this case, since it is does not explain the reasons of the agreement with the ACE-FTS measurements used by Coheur et al.

235
(2007) and the disagreement with those from Rinsland et al. (2006).The impact of the difference in the geometry of sampling in the HCOOH retrieval needs however to be estimated in a proper comparison between both instruments.Such comparison could also be extended with the difference in the assumptions used in the retrieval (e.g. the a priori).The ER (HCOOH/CO) from our work is also 15% lower than the EmR (HCOOH/CO) in Yokelson et al. (2003) (5.9×10 -3 ± 2.2×10 -3 mol/mol).With this difference we can also suggest the presence of a sink of HCOOH within the plumes detected by IASI or 240 this slight underestimation is simply related to the faster decay of HCOOH than the one of CO.In opposite, the ER (HCOOH/CO) retrieved from IASI is also twice higher than the ER (HCOOH/CO) in Chaliyakunnel et al. (2016) (2.6×10 -3 ± 0.3×10 -3 mol/mol).
(2016), who report however a larger bias over Amazonia.Over this region, our ER (HCOOH/CO) is higher than the one obtained by González Abad et al. ( 2009) with ACE-FTS in the upper troposphere (5.1×10 -3 ± 1.5×10 -3 mol/mol).This difference with the study done by González Abad et al. ( 2009) may be explained by the difference in the altitude of the detection of the forest fire plume between IASI (mid-troposphere) and ACE-FTS (upper-troposphere) and thus by a difference in the ongoing 260 chemistry within their respective sampled plumes.The geometry of the sampling (nadir vs limb) or the difference in the retrieval may also have an impact in the retrieved HCOOH.
The Siberian ER (HCOOH/CO) (4.4×10 -3 mol/mol ± 0.09×10 -3 ) is found to be in good agreement with the wide range of values obtained by Tereszchuk et al. (2013) and Viatte et al. (2015).This ER (HCOOH/CO) is however lower than the ratios calculated by R'Honi et al. ( 2013) who focused on the extreme fire event that occurred in 2010.

265
For India and Southern East Asia, the comparison is not possible since no previous studies were reported.The comparison is performed next, based on the emission factors.

Comparison with emission ratios calculated from the emission factors found in literature
We have complemented our comparison of the enhancement ratios by calculating emissions ratios using emission factors found in literature.The main argument to perform such comparison is the lack of measurements of enhancement ratios over 270 IND and SEA.
Even if our methodology tries to characterize the HCOOH emitted by biomass burning close to the source, our columns are probably not representative of the emission at the origin of the fire.The altitude of the sampling (mid-troposphere) and the age of the plumes (at least a few hours) have a large impact in our enhancement ratios.
To perform a proper comparison with emission ratios, our enhancement ratios should be converted to emission ratios.To do 275 so, it would be essential to take into account the decay of the compounds during the transport of the plume.However, due to the methodology used, i.e. averaging the data collected during a few hours (between 0 and 5h from the time registered by MODIS for each detected fire), the calculation of the decay of each compound is not possible.The comparison presented hereafter is therefore mostly illustrative.
For both, IND and SEA regions, the emission ratios have been calculated from the emission factors provided in Akagi et al.Thus, based on equation (1), EmR (HCOOH/CO) were calculated and compared with our ER (HCOOH/CO) .In this calculation, the vegetation type characterizing each region is important.Some regions are composed by a mix of vegetation types.This is the case for example for AMA and SAF (e.g.White, 1981)  Over Northern Africa, our ER (HCOOH/CO) is twice larger than the value provided by Akagi et al. (2011), probably due to the lower correlation found in our scatterplot.This is highly probable that our presumed fire-affected IASI columns are indeed impacted by other air masses.
For Amazonia, the calculated ER (HCOOH/CO) (7.3×10 -3 ± 0.08×10 -3 mol/mol) is close to the value given in Akagi et al. (2011) 310 for the tropical forest (5.2×10 -3 mol/mol), but it is three times higher than the values derived from Yokelson et al. (2007;2008) for the same vegetation type.For the latter, it is worth noting that their factors have been corrected a posteriori (scaled down by a factor of 2.1), as described in their comment following the paper done by R' Finally, in this comparison, the studied plumes over India and Southern East Asia are certainly related to agricultural fires.This is strongly possible as agricultural residue burning is prevalent in these regions (e.g.Kaskaoutis et al., 2014;Vadrevu et al., 2015).Over India and over Southern East Asia, our ER (HCOOH/CO) (6.8×10 -3 ± 0.44×10 -3 mol/mol for India and 5.8×10

335
The difficulties in performing such comparison is associated with the difference in locations, the difference in the altitude of the sampling and the difference in age of each fire plume studied in previous publications.A fair agreement was however found for the enhancement ratios calculated in this work, in comparison with other studies, using satellite or airborne or FTIR measurements.
In agreement with previous studies, the plumes from Southern African savanna fires may reflect a limited secondary  Finally, the values found over Northern Africa were the more difficult to interpret as this region is characterized by the poorer correlation between our fire-affected HCOOH and CO total columns.
With these findings and by updating the enhancement ratios, an interesting modeling work could be performed to estimate a 355 new tropospheric budget for HCOOH.This IASI data set may also be used in the future to study a single plume at different time.This would be useful for the characterization of the chemistry ongoing in a fire plume outflow.
An inter-comparison with other space-borne instruments as TES and ACE-FTS will be helpful to interpret the difference and the biases between the retrieved HCOOH columns in thus between their respective ER (HCOOH/CO).

Data availability 360
The IASI FORLI CO products are publicly available via the Aeris data infrastructure (http://www.aeris-data.fr/)while the IASI HCOOH columns will be made available soon through the same platform.
Only a few papers have reported on the use of satellite retrievals to study tropospheric HCOOH as with the nadir-viewing Infrared Atmospheric Sounding Interferometer (IASI) (e.g.Stavrakou et al., 2012; R'Honi et al., 2013; Pommier et al., 2016), and with the Tropospheric Emission Spectrometer (TES) (e.g.Cady-Pereira et al., 2014; Chaliyakunnel et al., 2016).Other studies have used the solar occultation Atmospheric Chemistry Experiment (ACE-FTS) which measures the 70 atmospheric composition in the upper troposphere (e.g.Rinsland et al., 2006; Tereszchuk et al., 2011; 2013) and the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) limb instrument which is sensitive to around 10 km (Grutter et al., 2010).These Infrared (IR) sounders have limited vertical sensitivity as compared to the ground-based or airborne measurements but their spatial coverage represents a major advantage, which allows observing remote regions which are sparsely studied by 75 field measurements, like the biomass burning regions.

2 . 1
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.2. HCOOH and CO columns from IASI 85 The IASI mission IASI is a nadir-viewing Fourier Transform Spectrometer.Two models are currently in orbit.The first model (IASI-A), was launched onboard the METOP-A platform in October 2006.The second instrument was launched in September 2012 onboard METOP-B.Owing to its wide swath, IASI delivers near global coverage twice per day with observation at around 09:30 and 21:30 local time.Each atmospheric view is composed of 2×2 circular pixels with a 12 km footprint diameter, 90 spaced out by 50 km at nadir.IASI measures in the thermal infrared part of the spectrum, between 645 and 2760 cm −1 .It records radiance from the Earth's surface and the atmosphere with an apodized spectral resolution of 0.5 cm −1 , spectrally sampled at 0.25 cm −1 .IASI has a wavenumber-dependent radiometric noise ranging from 0.1 to 0.4K for a reference blackbody at 280K (Clerbaux et al., 2009), and more specifically around 0.15K for HCOOH and 0.20K for the CO the spectral ranges (~1105 cm -1 and ~2150 cm -1 , respectively).
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.To identify the fire locations (hotspot), the fire product from the MODerate resolution Imaging Spectroradiometer (MODIS) on board the polar orbiting sun-synchronous NASA Terra and Aqua satellites (Justice et al., 2002; Giglio et al., 2006) are used.The Terra and Aqua satellites equatorial overpass times are ~10:30 and ~13:30 local time, respectively.Fire pixels are 125 1 km×1 km in size at nadir.For this work, we use more specifically, the Global Monthly Fire Location Product (MCD14ML, Level 2, Collection 5) 160 enriched plumes transported from other regions as by polluted air masses from China (e.g.Paris et al., 2008) or from Europe Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.

175 4 . 3
Importance of the meteorological conditionsAs shown by earlier studies, the wind speed can have a large influence on the detection of tropospheric plumes of trace gases from space (e.g NO 2 : Beirle et al., 2011; CO: Pommier et al., 2013; SO 2 : Fioletov et al., 2015).We have chosen to attribute a surface wind speed value for each MODIS hotspot.These meteorological fields were taken from the ECMWF (European Centre for Medium-Range Weather Forecasts) reanalysis data (http://data-portal.ecmwf.int/data/d/interim_full_daily)(Dee et 180 al., 2011).The horizontal resolution of these fields is 0.125° on longitude and latitude with a 6h time step.As shown in Fig 3, the three regions where the HCOOH:CO correlations are found to be high (close to 0.8), correspond to the regions where the surface wind speed was lower, i.e. for AMA, SEA and SAF.IND has also a low mean and median values but the distribution of the surface wind speed is more spread out than for these three regions.
195 ER (HCOOH/CO) , correspond to the value of the slope ∂[HCOOH]/∂[CO] found in Fig 4.This technique to determine the ER (HCOOH/CO) is more reliable than to only use the columns themselves, i.e. by estimating an ER (HCOOH/CO) for each measurements pairs (HCOOH, CO).Indeed, to perform scatterplots helps to identify a common origin for HCOOH and CO.The values of the ER (HCOOH/CO) over each region are summarized in Tab. 3. Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.It is known that trace gas concentrations within smoke plumes can vary rapidly with time and are very sensitive to chemistry, 200 so a comparison with previous works is always challenging, especially if these studies were performed in another altitude range, at a different location or period of the year.
215where the ER (HCOOH/CO) (5×10 -3 ± 0.13×10 -3 mol/mol) is similar to the value calculated byVigouroux et al. (2012) andCoheur et al. (2007).It also agrees with the broad range of values of emission ratios (EmR (HCOOH/CO) ) referenced bySinha et al. (2003).This result reinforces the relevance of the methodology used in this work over this region for the identification of the fire-affected IASI columns close to the source.Vigouroux et al. (2012) sampled biomass burning outflow of Southern Africa, Coheur et al. (2007) calculated their ER (HCOOH/CO) in plumes observed over Tanzania in the upper troposphere while 220 Sinha et al. (2003) did it within plumes over Zambia at the origin of the fire.A few assumptions can be given but the analysis given hereafter is only indicative since these previous studies did not measure the same plume than those presented in this work and our ER (HCOOH/CO) is calculated without making any distinction on the seasonal variation or on the type of biomass burning plumes sampled (e.g.emitted by a savanna fire or by a forest fire).Since these ER (HCOOH/CO) from previous studies and the EmR (HCOOH/CO) from Sinha et al. (2003) agree with our ER (HCOOH/CO) , and since HCOOH has a short lifetime, 225 this may suggest that the selected plumes measured by IASI from 2008 to 2014 and those sampled by Vigouroux et al. (2012) and Coheur et al. (2007), encountered a limited secondary production or a low sink as deposition or reaction with OH in the troposphere during their transport.Moreover, the decay of HCOOH is faster than for CO, suggesting that these plumes are rapidly advected in the troposphere.Our ER (HCOOH/CO) differs however from the value in Rinsland et al. (2006) (11.3×10 -3 Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.Chaliyakunnel et al. (2016) developed an approach allowing the determination of pyrogenic ER (HCOOH/CO) by reducing the impact of the mix with the ambient air, and they concluded that their most reliable value on the amount of HCOOH produced from fire emissions was obtained for African fires.245 Over Northern Africa, the calculated ER (HCOOH/CO) (4×10 -3 ± 0.19×10 -3 mol/mol) is 42% higher than the ER (HCOOH/CO) calculated in Chaliyakunnel et al. (2016) (2.8×10 -3 ± 0.4×10 -3 mol/mol).It is worth reminding that NAF is the region characterized by a scatterplot with the lowest correlation coefficient (Fig. 4).A larger difference is found over Australia where the ER (HCOOH/CO) is 11.1×10 -3 ± 1.37×10 -3 mol/mol.This ER (HCOOH/CO) is roughly the mean of both values reported by Paton-Walsh et al. (2005) and Chaliyakunnel et al. (2016).The difference 250 between our work and the one done by Paton-Walsh et al. (2005) could maybe be explained by the different origin of the probed plume.In our case, the studied area corresponds to the Northern part of the Northern Territory with savanna-type vegetation while Paton-Walsh et al. (2005) sampled bush fire plumes coming from the Eastern Coast of Australia, representative of Australian temperate forest.In the work done by Chaliyakunnel et al. (2016), a quite uncertain value is reported, with an error larger than their ER (HCOOH/CO) .
280 (2011).For the other regions, in addition to the values from Akagi et al. (2011), emission ratios were similarly calculated from emission factors given in other studies.Based on the emission ratios, the emission factors are usually derived by this following equation: EF HCOOH = EF CO × MW HCOOH /MW CO × EmR (HCOOH/CO) (1) Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.EF HCOOH is the emission factor for HCOOH; EmR (HCOOH/CO) is the molar emission ratio of HCOOH with respect to CO; MW HCOOH is the molecular weight of HCOOH; MW CO is the molecular weight of CO and EF CO is the emission factor for CO 285 for dry matter, set to the value taken from Akagi et al. (2011).
studies sampled plumes emitted by savanna fires.Yokelson et al. (2003) and Sinha et al. (2004) used also both the same 300 sampling strategy.They sampled fire plumes by penetrating several minutes old plumes at relatively low altitude (up to 1.3 km for Sinha et al. (2004) and just above the flame front for Yokelson et al. (2003)).This agreement shows, as already described in the previous section, our ER (HCOOH/CO) over Southern Africa is similar to their EmR (HCOOH/CO) .On other hand, it is worth noting that our ER (HCOOH/CO) is also similar to the EmR (HCOOH/CO) from Akagi et al. (2011) but for the tropical forest.A large underestimation compared to Rinsland et al. (2006) is found.This underestimation confirms the disagreement with 305 this study already shown in Tab. 3.

- 3 ±
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.0.15×10 -3 mol/mol for Southern East Asia) are close to the value referenced by Akagi et al. (2011) for cropland fires (6×10 -3 mol/mol).Since our ER (HCOOH/CO) are close to the EmR (HCOOH/CO) derived from the EF HCOOH in Akagi et al. (2011), this may suggests the studied plumes gathered along the 7-yr period correspond to fresh plumes where the chemistry or the sink do not occur.6. Conclusions 330 Seven years of HCOOH data measured by IASI over seven different fire regions around the world were analyzed (AMA=Amazonia, AUS=Australia, IND = India, SEA = Southern East Asia, NAF= Northern Africa, SAF= Southern Africa, SIB= Siberia).By taking into account the surface wind speed and by characterizing each MODIS fire hotspot with a value of HCOOH and CO total columns, this work established enhancement ratios for the seven biomass burning areas and compared them to reported values found in literature.
340production or a limited sink occurring in the upper layers of the troposphere during their transport.Such assumptions remain however difficult to make by comparing individual plumes (from previous studies) with plumes gathering during a 7-yr period (from IASI).Plumes from agricultural fires over India and Southern East Asia probably correspond to fresh plumes as our ER (HCOOH/CO) based on the 7-yr IASI measurements are similar to the EmR (HCOOH/CO) calculated from emission factors provided byAkagi et al. (2011).

345A
very good agreement was found over Amazonia, especially with the work done by Chaliyakunnel et al. (2016) who determinated pyrogenic ER (HCOOH/CO).Fires over Australia and over Siberia are probably underestimated in terms of direct emission or secondary production of HCOOH.The analysis over Australia is however delicate as our ER (HCOOH/CO) approximately corresponds to the mean of the values reported in Paton-Walsh et al. (2005) and in Chaliyakunnel et al. (2016); and it is also 450% higher than the 350 EmR (HCOOH/CO) derived from Akagi et al. (2011).The underestimation by 60% over Siberia is consistent with conclusions given in R'Honi et al. (2013).

Tab. 3
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2017-126,2017 Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 28 February 2017 c Author(s) 2017.CC-BY 3.0 License.Enhancement ratio to CO (mol/mol) for HCOOH with its standard deviation compared to enhancement ratios and emissions ratios reported in the literature for the seven studied regions.

Figure 1 :
Figure 1: Top: Number of MODIS fire hotspots with a confidence percentage higher or equal to 80%, averaged on a 0.5°×0.5°grid, for the period between 2008 and 2014.The blue boxes are the regions studied in this work.Bottom: The IASI CO total columns distribution (left) and the IASI HCOOH total columns distribution (right), for the same period and averaged on the same 745

Figure 3 :Figure 4 :
Figure 3: Box and whisker plots showing mean (red central cross), median (red central line), and 25 th and 75 th percentile (blue box 755 Yokelson et al. (2003) forest and savanna, characterized by an EF CO of 93±27 g/kg and 63±17 g/kg respectively(Akagi et al., 2011).AUS and NAF correspond to a savanna fuel type.SIB is a 290 boreal forest area with an EF CO of 127±45 g/kg.Based also on the maps shown by Fig.9inSchreier et al. (2014)and Fig.13in van der Werf, et al. (2010), the soil for IND is supposed to be mainly composed by cropland (agriculture), which is associated to an EF CO of 102±33 g/kg, and probably also by extratropical forest which is characterized with an EF CO equal to 122±44 g/kg.The fuel type for SEA is supposed to be a mix of extratropical forest and savanna, with EF CO of 122±44 g/kg, and 63±17 g/kg, respectively.Cropland fuel type was also used since large agricultural biomass burning is occurring in this 295 region (e.g.Duc et al., 2016).Despite the assumptions made, a fair agreement is found over Southern Africa.Our ER (HCOOH/CO) (5×10 -3 ± 0.13×10 -3 mol/mol) is indeed similar to the EmR (HCOOH/CO) calculated from Sinha et al. (2004) by using savanna fuel type and the ER (HCOOH/CO) is between both values calculated fromYokelson et al. (2003).This agreement is consistent since both previous