Global distributions of methanol and formic acid retrieved for the first time from the IASI/MetOp thermal infrared sounder

Methanol (CH3OH) and formic acid (HCOOH) are among the most abundant volatile organic compounds present in the atmosphere. In this work, we derive the global distributions of these two organic species using for the first time the Infrared Atmospheric Sounding Interferometer (IASI) launched onboard the MetOp-A satellite in 2006. This paper describes the method used and provides a first critical analysis of the retrieved products. The retrieval process follows a two-step approach in which global distributions are first obtained on the basis of a simple radiance indexing (transformed into brightness temperatures), and then mapped onto column abundances using suitable conversion factors. For methanol, the factors were calculated using a complete retrieval approach in selected regions. In the case of formic acid, a different approach, which uses a set of forward simulations for representative atmospheres, has been used. In both cases, the main error sources are carefully determined: the average relative error on the column for both species is estimated to be about 50%, increasing to about 100% for the least favorable conditions. The distributions for the year 2009 are discussed in terms of seasonality and source identification. Time series comparing methanol, formic acid and carbon monoxide in different regions are also presented. Correspondence to: A. Razavi (arazavi@ulb.ac.be)


Introduction
Volatile Organic Compounds (VOCs) includes thousands of different carbon-containing gases present in our atmosphere at concentrations ranging from less than a pptv to more than a ppmv (only methane exceeds 1 ppmv and is usually excluded from the VOC definition). Emitted from a large variety of processes (biogenic or anthropogenic) at the Earth's surface, they have an important influence on the atmospheric composition and climate. VOCs are precursors to tropospheric ozone (Houweling et al., 1998), they play an important role on the oxidizing capacity of the troposphere (Atkinson and Arey, 2003;Monks, 2005), they lead to the formation of secondary organic aerosols (Tsigaridis and Kanakidou, 2007;Heald et al., 2008) and they impact on climate change in different indirect ways (Meinshausen et al., 2006;Feingold et al., 2003;Charlson et al., 1987). In order to better understand and quantify their emissions and the role they play in the Earth's system, it is important to assess their atmospheric distribution at the global scale.

Observation of VOCs from space
The first steps in measuring volatile organic compounds by infrared satellite sounders were made during the last years, mostly using high-sensitive limb-viewing instruments. Methanol and formic acid, along with a series of other compounds, have been observed in young or aged biomass burning plumes with the ACE-FTS instrument Dufour et al., 2006;Coheur et al., 2007;Rinsland et al., 2007;Herbin et al., 2009). Large datasets of that sounder, covering several years, have been gathered to provide quasiglobal distributions of these two species and to study their seasonal variability  Published by Copernicus Publications on behalf of the European Geosciences Union.

858
A. Razavi et al.: CH 3 OH and HCOOH observations with IASI 2007;Abad et al., 2009). Similarly the observations of the MIPAS limb emission sounder have enabled mapping upper tropospheric distributions of several organic species such as PAN (Glatthor et al., 2007;, acetone  as well as HCN and C 2 H 6 (Glatthor et al., 2009); very recently global distributions of formic acid have also been gathered and analyzed (Grutter et al., 2010). The possibility of probing these VOCs lower in the atmosphere using nadir infrared sounders was suggested based on local observations, both by TES (Beer et al., 2008) and IASI , in the latter case in large biomass burning plumes.
Other VOC observations from space have been performed using ultraviolet (UV) sounders. The GOME and OMI instruments provide measurements of formaldehyde (Chance et al., 2000;Millet et al., 2008b), providing constraints on the emissions of isoprene (Shim et al., 2005;Palmer et al., 2006) and other non-methane volatile organic compounds (NMVOCs) (Fu et al., 2007;Stavrakou et al., 2009). With the complementary use of SCIAMACHY (De Smedt et al., 2008) and with the GOME follow-on instrument onboard MetOp-A (GOME-2), a long-term dataset (14 years) of formaldehyde observations is now available. In addition, measurements of glyoxal have also been made from the SCIAMACHY instrument (Wittrock et al., 2006;Vrekoussis et al., 2009).
This study provides the first global distributions of methanol and formic acid observed by the IASI (Infrared Atmospheric Sounding Interferometer) instrument , briefly described in Sect. 2.1. This infrared nadir-looking sounder has already demonstrated its potential for the monitoring of different trace gases such as CO Turquety et al., 2009), O 3 , CH 4  and HNO 3 . IASI has also demonstrated its high sensitivity to weak absorbers such as NH 3 , for which global and local distributions have been obtained . The same method as the one used for retrieving global NH 3 , and which relies on a simple difference of brightness temperatures, has been adapted to retrieve methanol and formic acid total columns. The method is described in Sect. 2.2 and a critical analysis of the resulting methanol field is provided in Sect. 3. The latter includes a first interpretation of the global distributions and seasonality of methanol as well as an error analysis. Section 4 provides similar analysis for formic acid. In addition, time series comparing the two species and CO for selected regions are presented in Sect. 5. Prior to this, a review of sources, sinks and previous measurements of methanol and formic acid is provided.

Methanol
Methanol (CH 3 OH) is the most abundant organic species in the Earth's atmosphere after methane and is also the main non-methane organic volatile compound in the mid to up-per troposphere (Heikes et al., 2002). Because its main removal process is oxidation by OH (Atkinson, 1986), CH 3 OH has a noticeable impact on the oxidizing capacity of the troposphere and on the global budget of tropospheric ozone (Tie et al., 2003). Additional sinks include removal by dry and wet deposition and uptake by the ocean (Jacob et al., 2005;Millet et al., 2008a). Its mean lifetime is evaluated to be about 10 days in the free troposphere. The main emission sources of methanol are biogenic processes, involving plant growth (MacDonald and Fall, 1993;Nemecek-Marshall et al., 1995;Harley et al., 2007;Galbally and Kirstine, 2002) and plant decay to a lesser extent (Warneke et al., 1999). Other sources include biomass burning (Holzinger et al., 2005), oxidation of methane and other VOCs, as well as anthropogenic emissions from vehicles and industrial activities . There still exists large uncertainties in the relative source strengths, the atmospheric distribution and budget of CH 3 OH Tie et al., 2003;Jacob et al., 2005).
During the past few years, coordinated measurement campaigns have been conducted which provide, among other trace species, in situ and aircraft determination of methanol abundances, and from there information about its different emission sources. The diurnal and seasonal cycle of CH 3 OH has been studied from in situ measurement in different types of environments such as forest (Karl et al., 2003(Karl et al., , 2005, rural (Schade and Goldstein, 2006;Brunner et al., 2007;Jordan et al., 2009) or urban areas (Filella andPeņuelas, 2006;Nguyen et al., 2001) with a variety of techniques. The concentrations usually increase during daytime due to the light induced release of methanol by plants. Moreover, CH 3 OH abundances are found to be higher during spring because of high plant growth emissions during that season. Methanol concentrations of about 4 ppbv have been reported in the amazonian region (Karl et al., 2007;Eerdekens et al., 2009). Aircraft measurements were also carried out over oceans (Singh et al., , 2004 where background concentrations of a few hundreds of pptv were found. Airborne observations led also to the study of CH 3 OH in several biomass burning plumes (Yokelson et al., 1999;Fischer et al., 2003;Yokelson et al., 2003;Holzinger et al., 2005) where volume mixing ratios (vmr) of up to several tens of ppbv were found. Observations of CH 3 OH with ground-based infrared spectrometers were also recently conducted in Australia (Paton-Walsh et al., 2008) and in Arizona ). The latter study provides an unprecedented 22 years time series of free tropospheric CH 3 OH; it does not report any significant trend over the years but shows a clear seasonality with a maximum in early July and a minimum during January.
The quasi-global distributions obtained from ACE-FTS  have further revealed that the surface sources of methanol have a significant impact on its upper tropospheric concentrations, which are mostly driven by biogenic and biomass burning emissions in the Northern and Southern Hemisphere, respectively. Beer et al. (2008) has firstly demonstrated the possibility to measure methanol from a nadir infrared sounder and reports lower CH 3 OH concentrations in California than near Beijing, where emissions could be caused by local sources.
The strongest absorption band of methanol in the infrared is the ν 8 CO-stretching mode centered at 1033cm −1 . We have used line parameters of Xu et al. (2004), implemented in the HITRAN database for which a precision of 6% for line intensities is quoted.

Formic acid
Formic acid (HCOOH) is one of the most abundant organic acid and has a strong influence on pH-dependent chemical reactions in clouds (Keene and Galloway, 1988). It has also been identified as a major sink for OH reactions in cloud water (Jacob, 1986). Although its sources and sinks are still poorly quantified, HCOOH is known to be emitted by various processes: biomass burning (Goode et al., 2000;Worden et al., 1997;Yokelson et al., 1997), biogenic emissions from vegetation Galloway, 1984, 1988), emissions from soil , from ants (Graedel and Eisner, 1988), as a secondary product from organic precursors (Rasmussen and Khalil, 1988;Arlander et al., 1990) and also from motor vehicles (Kawamura et al., 1985;Grosjean, 1989). Formic acid can be also produced from the aqueous oxidation of formaldehyde in cloud and rain water (Chameides and Davis, 1983) as well as from the oxidation of formaldehyde by HO 2 radicals in the cold tropopause region (Hermans et al., 2005). It is mainly removed from the troposphere through wet and dry deposition but also through oxidation by the OH radical to a lesser extent. The resulting lifetime of HCOOH, estimated to be a few days in the boundary layer, increases in the free troposphere because of the scarcity of precipitation (Sanhueza et al., 1996). HCOOH has been found to be a product of the isoprene oxidation by ozone (Jacob and Wofsy, 1988;Martin et al., 1991) and by OH radicals, in particular through the OH-oxidation of glycolaldehyde (Butkovskaya et al., 2006a) and hydroxyacetone (Butkovskaya et al., 2006b), which are two important isoprene oxidation products, but also through the OH-oxidation of isoprene nitrates (Paulot et al., 2009).
In situ measurements of formic acid in the boundary layer have been carried out using different techniques in various parts of the world, from rural sites (Talbot et al., 1988;Puxbaum et al., 1988;Talbot et al., 1990;Hartmann et al., 1991;Helas et al., 1992) to urban areas (Grosjean, 1989;Khwaja, 1995;Granby et al., 1997;Souza et al., 1999). The observed HCOOH volume mixing ratio in the boundary layer ranges from 0.01 to 10 ppbv. Its diurnal cycle shows larger concentrations in the mid-to late afternoon (Martin et al., 1991;Hartmann et al., 1991) indicating larger sources during daytime (biogenic emission, photochemical reactions) and dry deposition at nighttime. HCOOH measurements in the upper troposphere, performed during several aircraft campaigns (Reiner et al., 1999;Jaeglé et al., 2000;Singh et al., 2000), have reported mixing ratios from about 30 to 215 pptv. HCOOH has also been probed in different biomass burning plumes (Yokelson et al., 1999;Goode et al., 2000;Herndon et al., 2007) with concentrations of around ten ppbv; Worden et al. (1997) reported HCOOH total columns of 8.6 and 11.2 × 10 16 molec cm −2 above two fire events in the USA using optical measurements in the infrared. Similar spectral measurements, which used the HCOOH ν 6 absorption band in the IR, were performed from a balloon (Goldman et al., 1984;Remedios et al., 2007), with mixing ratios up to 600 pptv at about 7 km height, or from the ground . The latter study gives insight on the seasonal variation of HCOOH, with a mean mixing ratio in the free troposphere from about 300 pptv in October-December to about 800 pptv in July-September, likely resulting from higher biogenic emissions during the growing season. Other groundbased measurements above Jungfraujoch were very recently analyzed over a 22 years period and show a similar seasonal cycle with a maximum occurring during summer as well as significant diurnal and day-to-day variability (Zander et al., 2010).
The first satellite observations in the upper troposphere were reported from the ACE-FTS instrument Coheur et al., 2007) and were correlated to biomass burning events. Recent work performed with ACE-FTS on quasi-global observations of HCOOH reported an average mixing ratio of about 0.3 ppbv in the free troposphere with hot spots of up to 0.59 ppbv in tropical regions (Abad et al., 2009). Very recent global distributions were also assessed by the MIPAS sounder (Grutter et al., 2010) over a 6 years period. They report seasonal variations that may be associated to biogenic emissions with higher mixing ratios at 8 km during summer (about 100 pptv) than in winter (about 45 pptv). High concentrations were also observed in biomass burning plumes in the Southern Hemisphere. The nadir-viewing IASI sounder has also recently demonstrated the possibility to observe the HCOOH spectral signature in fire plumes . In this study, we provide the first distributions of formic acid total columns above land, differentiated into four seasons for the year 2009. The retrieval method is slightly different from the one used for methanol because the conversion factor between brightness temperature differences and total columns is derived from a set of forward simulation. Details and error estimation are provided in Sect. 4.1.
We have used a new set of HCOOH spectroscopic line parameters of Vander Auwera et al. (2007), implemented in the latest versions of HITRAN and GEISA databases. The accuracy on the absolute line intensities is evaluated to be about 7%. The updated data set reports HCOOH line intensities larger by about a factor 2 compared to previous studies . This implies that concentrations obtained from infrared retrievals before 2007 were likely too high by the same factor.

Description of IASI
The IASI instrument, launched onboard the MetOp-A platform in October 2006 in a polar sun-synchronous orbit, is a nadir-looking Fourier transform spectrometer that records the Earth's outgoing radiation from 645 to 2760 cm −1 without gaps at an apodized resolution of 0.5 cm −1 . Its field of view is a 2×2 matrix of circular pixels which have a 12 km footprint diameter at nadir. IASI provides two global Earth coverages per day (about 1 280 000 spectra) thanks to the wide scans across its track (2200 km swaths). It offers also a very good signal-to-noise ratio, with a Noise Equivalent Delta Temperature (NEDT) at 280 K of about 0.2 K in the spectral region of interest (which extends from 975 to 1150 cm −1 ). Moreover, with the successive launch of two other identical instruments, the IASI mission will provide consistent measurements over a 15 years period. A technical description of IASI and examples of applications for chemistry can be found in the review of Clerbaux et al. (2009). The IASI calibrated radiance spectra are disseminated in nearreal time by the EumetCast system along with temperature, humidity profiles and cloud information (coverage, temperature and altitude). Only cloud free observations (when the cloud coverage for the pixel is below 2%) are taken into account for this study.
The measurement of methanol and formic acid concentrations from IASI is quite challenging due to their weak absorption and the interference of other molecules in the same spectral range (see Fig. 1). CH 3 OH is observable using its C−O stretching absorption band located around 1033 cm −1 . The spectral range also covers the ν 6 absorption band of HCOOH located near 1105 cm −1 and the ν 3 band of trans−HCOOH at 1777 cm −1 (Perrin et al., 2009), which cannot be detected in the IASI spectra because of strong water vapor interferences.

Retrieval approach
In order to take advantage of the very good spatial coverage of IASI we have chosen a simple, fast and robust approach based on brightness temperature differences ( T b ), similar to the method already used for the retrieval of sulfur dioxide (Clarisse et al., 2008) and ammonia . It consists in two steps: (i) the determination of T b globally and (ii) the conversion of T b to total column amounts using one (global) or two (continental and oceanic) conversion factors.
The T b corresponds to the difference between the brightness temperatures measured in one or several channels selected in the absorption signature of the target molecule (T b,target ) and the brightness temperature of nearby channels selected in a spectral region where the least absorptions are found (T b,0 ). This is expressed in the following equation as This quantity gives a handy estimate for the strength of absorption and hence of the concentration. However, the physics of the radiative transfer is not accounted for and the conversion of T b to total columns requires a full radiative transfer treatment. For this purpose, we have retrieved CH 3 OH total columns at various locations of the world with an inversion model based on the Optimal Estimation Method (OEM) (Rodgers, 2000) that includes a line-by-line radiative transfer model. This is implemented in the Atmosphit software developed at the Université Libre de Bruxelles (for more information see Clarisse et al., 2008 andCoheur et al., 2005). The method is only applied in selected regions due to computational limitations. The conversion factor derived by matching the retrieved columns on the corresponding T b is applied globally to derive the total column distributions. For HCOOH, we used an alternative conversion method only based on forward simulations (see Sect. 4.1).
In the next two sections we describe the different retrieval parameters (the chosen channels, spectral intervals and a priori information) and present the resulting global distributions for CH 3 OH and HCOOH, respectively.

Retrieval settings
Methanol a priori profiles and covariance matrices were derived from distributions calculated by the IMAGESv2 global chemistry-transport model (Stavrakou et al., 2009). Monthly averaged model profiles over the whole year 2007 were used globally, on a 4 • × 5 • latitude-longitude grid to account for the seasonal and spatial variability of the model. Two different vertical profiles were selected: an average continental and an average oceanic profile. This choice is justified by the existence of strong emission over continents resulting in enhanced concentrations in the boundary layer. These two profiles are illustrated on Fig. 2 together with their associated covariance matrix (where the diagonal elements represent the model variability for the profile, and the off-diagonal elements represent the correlation between values at different altitudes). The continental a priori profile has a surface mixing ratio of about 2.5 ppbv, about four times larger than in the oceanic profile. Above 4 km, the two profiles are similar with only slightly higher concentrations for the continental profile. The covariance matrices both show higher variabilities in the lower (0 to 2 km) and upper (14 to 18 km) troposphere. Over continental surfaces, the variabilities are larger by about 25%. In both cases, the correlation length is large.
As can be seen from Fig. 1, the methanol spectral signature is fully overlapped by the much stronger ozone band at 10 µm. A large spectral range is therefore needed in order to properly account for this strong interference in the retrieval process. The spectral range extending from 981.25 to 1038 cm −1 was selected after several tests. CH 3 OH partial columns are retrieved in 4 km thick layers from the ground to 16 km. Partial columns of O 3 in 6 layers up to 42 km and the total columns of H 2 O and NH 3 are simultaneously adjusted. The regions selected for the retrievals are described in Table 1. These regions/periods were selected because of their high T b values and no significant seasonal differences have been noticed. An example of inversion is presented on Fig. 3 (top panel) for a case where the CH 3 OH signature is unambiguous. The RMS of the residue when CH 3 OH is not included in the retrieval is significantly higher (3.1 × 10 −6 W/(m 2 sr m −1 )) than when it is taken into account (2.5×10 −6 W/(m 2 sr m −1 )), this value being very close to the theoretical noise. This specific retrieval leads to a methanol total column of 5.57×10 16 molec cm −2 . Retrieved methanol mixing ratios in the lowermost atmospheric layer (0 to 4 km) range from about 0.1 to 7.0 ppb for the different selected regions. Comparisons between model data optimized with the IASI methanol product and previous CH 3 OH measurements (in situ and aircraft) were carried out and are detailed in the latest work by Stavrakou et al. (submitted to ACP). On the bottom panel of Fig. 3, representative total column averaging kernels differentiated for continental and oceanic retrievals are illustrated. They correspond to  the mean averaging kernels for all retrievals performed in the selected regions. In both cases, the sensitivity is maximum in the mid to upper troposphere from about 5 to 11 km, but retrievals over land show much higher sensitivity near the ground, largely due to the higher thermal contrast . The resulting DOFS (Degrees Of Freedom for Signal, given by the trace of the averaging kernel matrix) for these retrievals are given in Table 1. DOFS values range from 0.29 to 1.06 over land and are found lower above the ocean (between 0.12 and 0.87) where the dependence on the a priori is therefore larger. The observed (blue curve) spectrum is shown together with the fit residue (in cyan) and the dashed horizontal lines delimit this residue by its RMS value. The dark green curve is the fit residue when CH 3 OH is not taken into account in the retrieval, the red curve represents the calculated CH 3 OH contribution to the spectrum and the dashed vertical line indicates the detectable CH 3 OH absorption band. Bottom panel: Mean total column averaging kernels presented for retrievals performed over land (green curve) and over ocean (blue curve).
For the global T b calculation, the three target channels, all chosen in the Q-branch of CH 3 OH (at 1033.25, 1033.5 and 1033.75 cm −1 , see Fig. 1), are also contaminated by O 3 . Therefore, the baseline channels were also chosen inside the O 3 absorption band (at 1019, 1019.5, 1036.25, 1038, 1047 and 1048.5 cm −1 , see Fig. 1) in a way that maximizes the sensitivity of T b to the CH 3 OH amount. To deal with the remaining contribution of O 3 , the relationship between T b and O 3 concentrations has been derived using a set of forward simulations. For different columns of O 3 (ranging from 185 to 407 DU), the spectrum and T b were calculated with a fixed amount of CH 3 OH (4 × 10 16 molec cm −2 ). The results, presented on Fig. 4 assuming a typical midlatitude summer atmosphere (1976 US standard model), show an approximately linear correlation between the total column of ozone and the calculated T b . Similar results were obtained for other typical atmospheres and an average linear dependence was computed (with a slope of 9.02 × 10 −4 K DU −1 ).

Fig. 4.
Illustration of the influence of water vapor and ozone concentrations on the methanol T b . Simulations were performed for the midlatitude summer model with varying concentrations of H 2 O (black squares) and O 3 (red squares) while the CH 3 OH amount was fixed. In both cases, a linear dependence is found.
The O 3 correction is assumed here to be altitude independent although our analysis indicates a slightly different behavior for ozone variations in the first kilometers near the surface. For very low O 3 concentrations, it is possible that the linear assumption introduces errors in the retrieved methanol columns. However, as concentrations below 200 DU for ozone only occur during the antarctic ozone hole period, this will not affect the distribution discussed here.
The same type of simulations were performed to derive the influence of water vapor on the CH 3 OH T b (gray curve in Fig. 4). From these relationships a corrected T b for methanol which minimizes the dependence on O 3 and H 2 O is calculated as follows where C O 3 and C H 2 O are the total columns of O 3 in Dobson unit and of H 2 O expressed in molec cm −2 , respectively. This correction is applied to all observations using the total columns of O 3 and H 2 O retrieved from IASI with a near-real time algorithm based on the OEM. The next step of the method, i.e. the determination of a suitable conversion factor between T b and CH 3 OH total columns, is performed based on the retrievals in the selected regions and dates described in Table 1. Only retrievals results with a DOFS higher than 0.75 and a RMS of the residual lower than 4 × 10 −6 W/(m 2 sr m −1 ) are taken into account. This translates to a total of 5147 and 2849 observations above land and oceans, respectively. The corresponding scatter plots are shown in Fig. 5. In both cases, a linear fit shows good correlation coefficient (about 0.75). The slope for the retrievals over land is found to be much larger than More details about the selected regions for the retrievals can be found in Table 1. over oceans. Two different conversion factors are therefore used and applied to the T b calculated globally: where the CH 3 OH total columns (C CH 3 OH ) are expressed in 10 16 molec cm −2 . Due to its noise, the minimum methanol total column which can be detected by IASI has been evaluated from simulations using the 1976 US standard atmosphere to be about 1.60 × 10 16 molec cm −2 . It is important to note that for the derivation of the global distributions, only cloud free measurements recorded during daytime were taken into account. This is justified because daytime measurements are generally characterized by a positive thermal contrast and are therefore more sensitive to the lower troposphere.

Global distributions
The method proposed above allows to derive the first global distributions of methanol from IASI. In order to shed light onto seasonal variations, the four seasons have been differentiated as follows: DJF (December 2008, January 2009 and February 2009), MAM (March, April and May), JJA (June, July and August) and SON (September, October and November). The resulting distributions averaged on a 0.5 • × 0.5 • grid are shown in Fig. 6. Note that measurements above sand surfaces, which causes erroneous high CH 3 OH concentrations due to spectrally resolved surface emissivity (Wilber et al., 1999) have also been discarded. Figure 6 shows large seasonal variations in the methanol columns. Higher concentrations are found in the northern hemisphere during spring and summer (up to 4 × 10 16 molec cm −2 in Central and Northern Asia) when vegetation is growing. In the Southern Hemisphere, the highest concentrations are found during the dry season (SON) and may be related to biomass burning. CH 3 OH is also observed over oceans (with values around 2×10 16 molec cm −2 ) mostly between Africa and South America but also in the whole Northern hemisphere in spring and summertime. The presence of CH 3 OH in remote oceanic regions is probably largely due to transport of continental emissions although oceanic emissions (Millet et al., 2008a) and the ubiquitous methane oxidation might also contribute. Methanol total columns range from about 0.01×10 16 molec cm −2 above sea surfaces to 5.40 × 10 16 molec cm −2 over large emission regions.
During the northern hemispheric winter (DJF), methanol hot spots of low intensity are found in the Southern Hemisphere (South America, South Africa and Western Australia) above vegetated areas where CH 3 OH emissions may be related to plant growth. When comparing the distribution with AATSR (Arino et al., 2005) fire count maps, we find a high degree of coincidence in Africa between 5 and 15 • N, suggesting a possible biomass burning contribution. During springtime (MAM), this specific region is subject to enhanced CH 3 OH columns which is possibly due to an increase in the fire numbers and intensities. Strong enhancements are also observed over India and to a lesser extent over Burma, Manchuria and Mexico, which can be at least partly related to biomass burning. In contrast, biomass burning is unlikely to be a dominant source in the Northern Hemisphere except in some regions (Kazakhstan, East Russia, Alaska) during JJA. It can be seen that CH 3 OH concentrations are progressively increasing from winter (DJF) to summertime (JJA). This can be explained by the seasonal cycle of the vegetation source which presents a maximum in late spring (Schade and Goldstein, 2006;Rinsland et al., 2009). During fall (SON), CH 3 OH concentrations decrease again in the Northern Hemisphere and increase in the Southern Hemisphere. Again, biomass burning is likely responsible for the strong enhancements in South America, Congo and Northern Australia.
Although biomass burning is assumed to be a weak emission source of methanol (accounting for less than 5% of total emissions according to current inventories Jacob et al., 2005), the main hot spots in the global distributions could possibly be related to fires. This could be caused by the fact that methanol emitted by fire events is usually transported higher in the troposphere where the IASI sensitivity is larger or by the fact that a better sensitivity near the surface is induced because of higher surface temperatures for burning areas. This is consistent with the fact that biomass burning has been found to be a significant source of methanol in the upper troposphere . The assimilation of IASI data into models should help determining the respective contributions of biogenic and fire emissions to the global methanol budget.

Error assessment
One of the disadvantages of the method described above is that it does not provide an estimate of the errors associated with the retrieved total columns. Because we are deriving the CH 3 OH columns from a weak signal, the associated error is expected to be quite large. An estimate of the error on the CH 3 OH columns derived from IASI can be based on forward simulations. For this purpose, we used a large set of different atmospheres compiled in an ECMWF database (Chevallier, 2001). Different input profiles are used for CH 3 OH, corresponding to the different continental profiles taken from the IMAGESv2 model. The total columns used as input for the simulations are then compared to those retrieved from the simulated spectra using the same method as described above. The difference between the two columns gives a fair estimate of the absolute column errors. We do not expect these errors to be systematic on a global yearly scale but they cannot be excluded in particular circumstances such as in the presence of extreme values of ozone or water vapor.
It turns out that this absolute difference increases when the amount of CH 3 OH in the boundary layer (between the surface and 3 km height) increases. It follows that the high concentrations of methanol in the boundary layer (close to the emission sources) will not be well reproduced by the T b method. This is consistent with the averaging kernel shape and the well known limited sensitivity of the infrared measurements toward low altitudes. Figure 7 shows the histogram of the relative difference between the total columns used as input in the forward simulations and the calculated ones. Negative differences imply that the retrieved total column from T b is lower than the a priori value. Moreover, differences lower than −100% are found but correspond only to methanol total columns which are below the detection limit of IASI (i.e. 1.60 × 10 16 molec cm −2 ). The distribution is similar to a normal distribution with a mean very close to zero (0.6%). The standard deviation (48.9%) provides our best estimate of the relative error on the retrieved methanol total column. This value is, however, clearly a lower bound for regions located close to emission sources and with low thermal contrast. In these regions, based on our analysis, errors as high as 100% are likely.

Retrieval settings and errors
In the case of formic acid, the target channel for brightness temperature has been chosen in the Q-branch at 1105.0 cm −1 which is the strongest absorption feature detectable in the IASI spectrum. The reference channels were chosen on both  (Kleb et al., 2011) between 0-5 km, (ii) the arithmetic mean between INTEX-B data and ACE-FTS data (Abad et al., 2009) measured from March to May for altitudes between 6 and 8 km, (iii) and an averaged ACE-FTS profile between 9 and 11 km. The resulting profile, illustrated in Fig. 8, has a surface mixing ratio of about 750 pptv, which smoothly decreases as the altitude increases, down to less than 100 pptv above 7 km. Concentration above 11 km have been linearly extrapolated up to 20 km.
As in the case of methanol, conversion factors between the T b and the HCOOH total columns were tentatively derived based on OEM retrievals in selected regions around the world. Because of the weak signal and the presence of water vapor interferences, the retrievals are unstable and lead to large errors. Forward simulations using the ECMWF database and varying the HCOOH concentrations (by scaling the profile, resulting in a variability of about 350%) were conducted to compute the relative difference between the true columns (input of the forward model) and those retrieved from the T b . Figure 9 illustrates in gray the histogram of these relative differences. The mode of the differences is located at about 60.0% suggesting a strong bias between the calculated and simulated total columns. Therefore, an alternative method has been used to derive the conversion factor between the T b and the total columns. Relying only on the Fig. 8. HCOOH a priori profile derived from the combined aircraft and ACE measurements. See text for details. Fig. 9. Histograms of the relative differences between the simulated total columns of HCOOH and the HCOOH total columns obtained from brightness temperature differences. The gray histogram is derived from the conversion which uses OEM retrievals and the black one corresponds to the conversion obtained with simulated data. The latter, which is much less biased is used to obtain the global distributions (see Sect. 4.2). forward calculations, it minimizes also the dependence upon the water vapor content (C H 2 O expressed in molec cm −2 ) and accounts for the varying sensitivity of the measurement to the local thermal contrast (τ ). The following equation is used with the parameters: and where HCOOH total columns (C HCOOH ) are expressed in 10 16 molec cm −2 . The thermal contrast τ corresponds to the difference between the surface temperature and the air temperature at the first retrieved altitude level, located at about 100 m (both included in the IASI level 2 data). This mapping of T b onto total columns of HCOOH provides much better results. However a significant dependency on the thermal contrast remains, with lower errors found for higher thermal contrast values. We have chosen to consider only the cases for which the thermal contrast is higher than 5 K. This conservative criterion excludes all retrievals above oceans. The resulting histogram of the relative differences between the simulated and calculated HCOOH total columns is shown in black on Fig. 9. It is similar to a normal distribution, with its mean being equal to −0.8%. Our estimation of the error on the formic acid total column is given by the standard deviation of the differences, which is about 60%. This error is only slightly higher than the error found for the CH 3 OH columns, but only applies here to favorable situations with large thermal contrast. Moreover, in the same way as methanol, the negative differences lower than −100% correspond only to HCOOH total columns which are below the detection limit of IASI (and has been evaluated to be about 0.60 × 10 16 molec cm −2 ). This retrieval approach does not provide information about the vertical sensitivity of the formic acid total column. However, the limited set of full profile retrievals performed give a maximum sensitivity between 4 and 14 km.

Global distributions
The global distributions of formic acid for the four seasons are illustrated in Fig. 10, on a 0.5 • ×0.5 • averaged grid. Only cloud free observations recorded during daytime and with a thermal contrast higher or equal to 5 K were taken into account. The latter constraint removes unfortunately many observations at high latitudes (no observations above about 45 • N in winter and about 65 • N during summer). Grid points which include less than ten HCOOH measurements were also filtered out. As for CH 3 OH, clear seasonal variations are observed. The retrieved HCOOH total columns range from background values of less than 0.5 × 10 16 molec cm −2 above Europe and North America to 5×10 16 molec cm −2 above fire events, mainly during summer 2009 in Africa. These high formic acid total columns in fire plumes are in good agreement with the values reported by Coheur et al. (2009) and Worden et al. (1997) if we account for the factor 2 resulting from the use of the improved line parameters (Vander . The 2009 northern hemispheric winter season (DJF) shows the lowest number of observations. Comparing with AATSR fire counts during that period, we found that enhancements of HCOOH in the western-central region of Africa and to a lesser extent in South America might be partly due to biomass burning. During spring (MAM), we observe a decrease in the HCOOH total columns above Africa and South America together with a fire count decrease.
High concentrations are also found in Asia (India, Burma and Manchuria) which are well correlated with CH 3 OH hot spots. The peak of the biomass burning season in South America and Southern Africa happens usually around August-September whereas in Australia most of the burning occurs around October-November (Gloudemans et al., 2006). Together with the minimal washout of HCOOH occurring during the dry season, it probably largely explains the high HCOOH total columns observed during JJA above Congo and Brazil as well as in Northern Australia during the SON period. The overall increase of formic acid in the Northern Hemisphere during JJA is likely caused by the seasonality of its biogenic emissions. However, according to the AATSR fire count distributions, some regions with large HCOOH columns may be associated with boreal fires such as in Eastern Russia and in Alaska. Finally, the very widespread hot spots found during SON above Amazonia and Central Africa do not seem to be only related to fires and very likely points also to biogenic emissions of either HCOOH or HCOOH precursors. Throughout the year (MAM through SON, no available data for DJF), large HCOOH columns are observed in Eastern China which may partly be due to anthropogenic activities. Several common patterns are also found in the distributions of HCOOH and CH 3 OH columns, probably due to their common emission sources (such as biomass burning and plant growth). Also note that the seasonality observed here is in good agreement with that reported from ACE-FTS (Abad et al., 2009) and from MIPAS (Grutter et al., 2010).

Seasonal variations in relation to biomass burning
In this section we compare the 2009 time series of methanol, formic acid and carbon monoxide for three selected regions subject to biomass burning. The vertical sensitivity profiles of IASI for these three species are all maximum in the free troposphere, i.e. between 4 to 14 km, 6 to 10 km and between 3 to 12 km for HCOOH, CH 3 OH and CO total columns, respectively. The comparisons are therefore most likely to reflect similarities/differences in the free tropospheric columns but fine structures in the respective profiles could be missed. Monthly mean total columns of methanol and formic acid are shown in Fig. 11 together with the total columns of carbon monoxide and the AATSR fire counts for three 10 • ×10 • regions located in Brazil (15 − 5 • S, 60 − 50 • W), Congo (15−5 • S, 20−30 • E) and South-East Asia (20−30 • N, 95− 105 • E). For each regions, the time series of CO, CH 3 OH and HCOOH are similar. An increase in the total columns is observed for the three species just after the month with the maximum fire counts. This delay may be induced by the fact that the probed air masses are located in the free troposphere and are therefore subject to some transport which could result in the spreading of the species or the spatial displacement of the maxima. The highest number of fires (ex- Fig. 11. Time series of methanol (green curve), formic acid (blue curve) and carbon monoxide (black curve) monthly mean total columns for three different areas associated with biomass burning (Brazil, Congo and SE Asia). Carbon monoxide total columns where divided by 100 for comparison. The left scale and gray curves correspond to the sum of the AATSR fire counts for each months.
ceeding 700) is found above Congo where methanol, formic acid and carbon monoxide reach high values, with increases of about 1.6 × 10 16 2.5 × 10 16 and 1.4 × 10 18 molec cm −2 in comparison with their mean total column between January and June, respectively. In each cases, the CH 3 OH maximum lasts longer than for CO or HCOOH. This cannot be explained by its lifetime which is similar to formic acid in the free troposphere (about one week) but suggests an additional source or transport to this region. Overall higher concentrations of CH 3 OH found above Brazil and Congo may also be due to the larger biogenic source in these regions.
In addition to looking into correlations regionally, preliminary global analyses were carried out. Linear correlations (R 2 = 0.7) between CH 3 OH and HCOOH were found during the DJF and SON periods highlighting specific emissions or fate of these two species.

Conclusions and perspectives
In this study, we have retrieved methanol and formic acid from the data provided by the IASI sounder. Using a radiance indexing method and the calculation of brightness temperature differences, first global distributions are computed. They are provided as total columns, after careful conversion of the brightness temperature differences. The conversion has been achieved based on a set of retrievals in selected regions for methanol and on forward simulations for formic acid. Taking full advantage of the IASI spatial and temporal coverage, unprecedented information on the location and origin of sources have been acquired.
The simple retrieval method used here does not allow the determination of individual error for each observations. We have estimated, using a large set of representative forward calculation, global errors of about typically 50% and 60% for the methanol and formic acid total columns, respectively. Although these errors are significant, this robust method takes advantage of the very large number of IASI measurements at low computational cost. In this way, we provide global observations for these two volatile organic compounds.
The global distributions shown in this study highlight strong seasonal variations for methanol and formic acid with maximum total column values evaluated at 5.4 and 5.0 × 10 16 molec cm −2 , respectively. The strong enhancements seen in the global maps might be largely attributed to biomass burning (mostly in tropical regions). The main seasonal patterns, especially at mid-latitudes where we find higher columns in spring and summer, might be explained by variations in biogenic emissions and increased plant growth. Anthropogenic emissions could not been clearly identified in these distributions.
Time series of methanol, formic acid, carbon monoxide and AATSR fire counts were also compared and found to be fairly well correlated for three different regions (Congo, Brazil and South-East Asia) where biomass burning is their likely common source.
It is anticipated that the assimilation of these data in a global chemistry model will help to improve the determination of the emission fluxes for these two species. Mean averaging kernels for methanol (differentiated for land and ocean) are provided in order to account for the vertical sensitivity of the measurements.