Formic acid (HCOOH) is one of the most abundant volatile
organic compounds in the atmosphere. It is a major contributor to rain
acidity in remote areas. There are, however, large uncertainties on the
sources and sinks of HCOOH and therefore HCOOH is misrepresented by global
chemistry-transport models. This work presents global distributions from 2008
to 2014 as derived from the measurements of the Infrared Atmospheric Sounding
Interferometer (IASI), based on conversion factors between brightness
temperature differences and representative retrieved total columns over seven
regions: Northern Africa, southern Africa, Amazonia, Atlantic, Australia, Pacific, and
Russia. The dependence of the measured HCOOH signal on the thermal contrast
is taken into account in the conversion method. This conversion presents
errors lower than 20 % for total columns ranging between 0.5 and
Formic acid (HCOOH) is among the most abundant volatile organic compounds (VOCs) present in the atmosphere. Along with acetic acid it is a major contributor to the acidity of precipitation, especially in remote regions (Keene and Galloway, 1988; Andreae et al., 1988). HCOOH has small direct emissions from vegetation (Keene and Galloway, 1984, 1988), ants (Graedel and Eisner, 1988), biomass burning (e.g., Goode et al., 2000), soils (Sanhueza and Andreae, 1991), agriculture (e.g., Ngwabie et al., 2008), and motor vehicles (Kawamura et al., 1985; Grosjean, 1989) but is mainly a secondary product from other organic precursors. The largest global chemical sources of HCOOH include isoprene, monoterpenes, other terminal alkenes (e.g., Neeb et al., 1997; Lee et al., 2006; Paulot et al., 2011), alkynes (Hatakeyama et al., 1986; Bohn et al., 1996), and acetaldehyde (Andrews et al., 2012; Clubb et al., 2012). HCOOH is mainly removed from the troposphere through wet and dry deposition and to a lesser extent through oxidation by the OH radical.
HCOOH is a short-lived species and its lifetime is mainly determined by the
precipitation rate. The lifetime ranges between 2 days during the rainy
season and 6 days in the dry season in the boundary layer (Sanhueza et al.,
1996). The global lifetime in the troposphere is 3–4 days (Paulot et al.,
2011; Stavrakou et al., 2012). Photochemical loss is relatively slow (
Our knowledge about sources and sinks of HCOOH is still incomplete despite the numerous studies that have been published during the last decade. In current emissions inventories, such as the biogenic emission inventory MEGAN-MACC (Sindelarova et al., 2014), the main source regions are located in tropical regions as presented in Fig. 1 for the period between 2008 and 2010. A recent work shows a possible source of HCOOH over the Arctic Ocean (Jones et al., 2014). The study by Stavrakou et al. (2012) highlights a misrepresentation of emissions from tropical and boreal forests in models compared to total columns retrieved from Infrared Atmospheric Sounding Interferometer (IASI) observations by Razavi et al. (2011). Millet et al. (2015), Paulot et al. (2011), and Stavrakou et al. (2012) point to the existence of one or more large missing sources. These studies suggest an important gap in our current understanding of hydrocarbon oxidation and/or the existence of unknown direct fluxes of HCOOH.
MEGAN-MACC HCOOH emissions for the period between 2008 and 2010 on a
0.5
Nadir-looking atmospheric sensors allow us to derive global distributions for trace gases, with a limited vertical sensitivity as compared to airborne or ground-based measurements. Their extended spatial coverage allows the observing of remote regions which are sparsely studied by field campaigns. Only a few satellites provide tropospheric HCOOH observations, such as the nadir-viewing instrument IASI (e.g., Razavi et al., 2011) and the Tropospheric Emission Spectrometer (TES) (e.g., Cady-Pereira et al., 2014). The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) limb instrument provided monthly global distributions of HCOOH around 10 km (Grutter et al., 2010), and the solar-occultation Atmospheric Chemistry Experiment (ACE) provides seasonal global distribution in the upper troposphere (e.g., González Abad et al., 2009).
The data used in this study are provided by IASI. This instrument has two
important advantages: a low radiometric noise and a high spatial coverage.
HCOOH is a weak absorber, so it is a challenge to retrieve total columns
from the IASI radiance. Global distributions of HCOOH over land were
initially derived using a method based on brightness temperature difference
and using forward simulations (Razavi et al., 2011). More recently, R'Honi
et al. (2013) developed a specific method to study extreme events occurring
during the large fires in Russia during the summer 2010. These studies,
however, highlighted discrepancies between the retrieved distributions and
especially within enriched HCOOH air masses as, for instance, over large
forest fires. Indeed, the total columns from R'Honi et al. (2013) were on
average a factor of 2 lower than in Razavi et al. (2011) (around a factor of
1.5 for columns higher than
Selected regions used for the retrieval. The localization of each area and the number of spectra retrieved during the studied period are provided. The numbers correspond to the total number of successfully retrieved spectra and those given in parentheses to the total number of spectra in each region.
IASI is a nadir-viewing Fourier transform spectrometer instrument. Currently
two instruments are in orbit. The first model was launched on board the
MetOp-A platform in October 2006 providing now more than 8 years of
observations. The second instrument was launched in September 2012. Owing to
its wide swath, each instrument delivers near global coverage twice per day
at around 09:30 and 21:30 local time. IASI measures in the thermal infrared
part of the spectrum, between 645 and 2760 cm
Mean normalized Jacobians of all retrieved spectra (over the seven selected regions) as a function of altitude.
Analysis of the mean of the normalized Jacobians (Fig. 2) over the spectral
range used by IASI for the HCOOH retrievals (1095–1114 cm
Processing 7 years (2008–2014) of IASI data using an iterative method
such as the OEM (Rodgers, 2000) is computationally demanding. Hence we have
chosen a fast approach based on brightness temperature differences (
Retrieval parameters used in this work.
In the current work, the main difference with the previous IASI HCOOH
determination in Razavi et al. (2011) is the use of retrieved total columns
over selected regions to determine conversion factors instead of the use of
forward simulations. The OEM implemented in the line-by-line radiative
transfer model Atmosphit (Coheur et al., 2005) has been used as in Razavi et
al. (2011). In Razavi et al. (2011), the reason invoked for performing only
forward simulations for HCOOH was the unstable character of the retrievals.
With the retrieval settings chosen here, we relied on the retrieved columns
since 82 % of the selected spectra were successfully retrieved (Table 1)
and the mean root mean square (RMS) between the observed and fitted spectra
is about
The conversion factors allowing the calculation of total columns based on
In the OEM-based retrieval, only cloud-free scenes (when the cloud coverage
for the pixel is below 2 %) have been used. The total columns of
interfering species in the studied spectral range, such as ozone, ammonia,
and CFC-12, in addition to the partial columns of water vapor, were
retrieved simultaneously. The details of the retrieval parameters are given
in Table 2. The EUMETSAT L2 operational data were used, and daytime and
nighttime data with a positive thermal contrast (TC) were taken into
account. The TC was defined as the temperature difference between the
surface and the air just above. Negative TC data were excluded as these
were found to deteriorate the correlation
The OEM-based retrieval has been performed over seven geographical regions shown as blue boxes in Fig. 1. These include five source areas (in southern Africa, Northern Africa, Amazonia, Australia, Russia) and two remote areas (over the Atlantic and Pacific oceans). These seven regions are representative of different conditions: emission sources, remote areas, areas influenced by long-range transport, over land, and over sea. We have retrieved the 5 first days of each month in 2009 over the seven regions, allowing the characterization of the seasonal variation. The localization of each area and the number of retrieved spectra are given in Table 1.
From these retrievals, we derived a linear regression between the retrieved
total columns and the
To account for the TC dependence in the
Note that this conversion could lead to negative total columns. If we eliminated all the negative values and kept only all the positive values, we would introduce an artificial bias in the average. For comparisons with zonal or temporal averages, the negative total columns were included in the average. However, when the average was found to be negative, it was filtered out.
This technique has a low computational cost but the drawbacks of the method are the difficulty to characterize the retrieval in terms of vertical sensitivity (averaging kernels not available) and the lack of an error budget.
The total error of the
A Gaussian distributed random noise (with
Histogram of the relative differences between the calculated total
columns (derived from the
Figure 5 shows the histogram of the relative difference (RD) between the
calculated total columns and the true total columns used as input in the
forward simulations. The RD was defined as the difference between the
calculated total columns and the true total columns, divided by the latter.
Positive RDs imply that the calculated total column is higher than the true
column. This histogram presents a mean of
Considering the detection threshold defined as 2
Mean HCOOH global distributions (averaged on a 0.5
Yearly global distributions between 2008 and 2014 with the updated data set
are also presented in Fig. 8 (on a 1
These distributions highlight well the recurring source regions detected by
IASI such as equatorial Africa, northern Australia, Amazonia, and India,
as well as the long-range transport such as over the Atlantic Ocean from
Africa. The long-range transport over oceans (Atlantic, Indian, and Pacific)
was not investigated in Razavi et al. (2011). The retrieved columns over the
Atlantic Ocean are consistent with the Fourier transform infrared spectroscopy (FTIR) data from ship cruises reported
in the study of Paulot et al. (2011). They showed a gradient of columns from
the poles to the Equator, with the highest values between 0 and
10
Mean HCOOH global distribution between 2008 and 2014, derived using
the IASI radiance observations on a 0.5
Several hotspots and distributions are detected and are numbered from (1) to (10) in Fig. 8.
A particularly striking feature is the large hotspot over Russia (close to
Moscow) in 2010 as documented by R'Honi et al. (2013) due to intense forest
fires during the summer and also in 2012 over Siberia (see label 1). The
current data set presents a mean total column twice lower (
Annual HCOOH global distribution from 2008 to 2014, derived using
the IASI radiance observations on a 1
The monthly means over the 7 years are also presented with an animation
(Fig. S1) in the Supplement. As already observed over eastern Russia with
label (2), in June 2010 and 2012, there were large concentrations, close to
Khabarovsk Krai, compared to the other years in this region. Whereas intense
fires were detected in June 2012 in this region, this was not the case in
June 2010 (see maps on
Large columns were similarly retrieved over a large region encompassing Laos, Thailand, and Myanmar in April 2010, 2012, 2013, and 2014 (see label 6). It matches well with the locations of fire hotspots detected by MODIS.
Over India, the largest total columns are observed from March to June probably due to biomass burning (see label 7). Indeed, those emissions present a marked seasonal variation with a maximum in March–May according to the GFED3 inventory (van der Werf et al., 2010), with 50, 22, and 11 % of annual emissions occurring in March, April, and May, respectively.
Larger total columns were retrieved in August 2010 along the Euphrates River compared to the other years (see label 8).
The monthly distributions also highlight hotspots over the USA, besides those shown in the annual distributions (see label 9). In summer 2011, large signatures over the USA were not confined to coastal regions; high total columns were also detected in the Midwestern USA such as over Kansas, Mississippi, Missouri, or Oklahoma. These states are flagged as biogenic emission regions of VOCs by Millet et al. (2015), acting as secondary source of HCOOH. In July 2012, the emissions over the USA were mostly confined to the eastern part.
The Asian HCOOH outflow is well captured over the western Pacific (see label 10). The range of values of the IASI total columns, from 2008 to 2014,
broadly agrees with our estimation of total columns using the measurements
from the aircraft PEM-West-B campaign conducted in February–March 1994
(Talbot et al., 1997a, b) over a large region covering the latitudes
0–60
Overall, these monthly means highlight the seasonal variation of the HCOOH distribution around the world. The animation reveals clear variations in the HCOOH distribution due to the seasonality of biomass burning and vegetation growth. This is shown well with the large total columns observed during September and October 2008, 2012, and 2014 in the Southern Hemisphere (over Amazonia, Africa, and Australia). In 2010, the same features were noted except for Australia.
The IASI HCOOH retrieved columns in this work have been compared with
ground-based FTIR measurements. This comparison was done without smoothing
the data since the averaging kernels (AKs) were not provided by our
retrieval method. This comparison is presented at four sites: Jungfraujoch
(46.55
Time series of HCOOH daily means over Jungfraujoch
A complete description of the FTIR instruments and the retrieved HCOOH data
can be found in Zander et al. (2010), Paton-Walsh et al. (2005), and
Vigouroux et al. (2012) for the Jungfraujoch, Wollongong, and Saint-Denis
stations, respectively. For the Jungfraujoch, spectra were typically recorded
at spectral resolutions of 0.004 and 0.006 cm
The current IASI retrieved columns were also compared with a set of columns retrieved by OEM around the sites. For each OEM-based retrieved column, the corresponding column using the conversion factors was calculated, showing that the current data set and the OEM-based retrieval are in agreement (correlation ranging from 0.7 to 0.8, with an underestimation of the columns calculated with the conversion factors between 6 and 15 %) (Fig. S2). It is also worth noting that similar biases were found between the columns retrieved by OEM around the ground-based locations and the FTIR columns as between the columns retrieved in this work and the FTIR ones (Table S1).
Averaging kernels were also unavailable for the FTIR measurements performed at Jungfraujoch (extended from Zander et al., 2010) and at Wollongong (Paton-Walsh et al., 2005). The measurements at Saint-Denis and Maido reached a maximum sensitivity between about 3 and 12 km as described in Vigouroux et al. (2012) and shown in Fig. 11. The AKs indicate the vertical sensitivity of the retrieval. The Jacobians express the sensitivity of the radiative transfer model and the IASI instrument (through its instrumental function) to the variation of HCOOH in the atmosphere. Both functions then give a good indication of the vertical sensitivity for each data set. These AKs and the Jacobians show that FTIR and IASI were both mostly sensitive to the free troposphere but that the FTIR measurements presented a broader vertical sensitivity, reaching higher altitudes than IASI.
Box and whisker plots showing mean (red central circle), median (red
central line), and 25th and 75th percentile (blue box edges) of the relative
difference between the HCOOH derived using the IASI radiance observations
from this work or using the conversion from Razavi et al. (2011) and the FTIR
measurements for each station: Jungfraujoch, Wollongong, Saint-Denis, and
Maido. The whiskers encompass values from
Mean total column AK for the FTIR ground-based measurements over Maido (black) and Saint-Denis (red) at La Réunion. Both stations are shown by green stars in Fig. 1. Both FTIR stations have a degree of freedom for signal close to 1. As reminder, the mean normalized Jacobian from Fig. 2 is plotted in blue.
A difficulty in comparisons of satellite columns with ground-based
measurements over mountain sites like Jungfraujoch over the Swiss Alps (3.6 km altitude) or Maido at La Réunion (2.2 km altitude) is the difference
of altitude between the FTIR sites and the co-located IASI ground pixel
height. To account for the altitude dependence, both the IASI and the FTIR
total columns were normalized to the sea level altitude using
The time series of the IASI and FTIR columns over the selected sites are
shown in Fig. 9. The comparison used IASI data collocated within
0.5
The comparison between the ground-based measurements and the total column derived from the IASI spectra may be affected by sampling differences associated, e.g., with cloudiness. IASI may be able to measure through clear skies in the vicinity of the station when the FTIR data are not available due to localized cloud. Moreover, despite the use of strict co-location criteria (spatial and temporal), most mismatches in peak values could be a result of mismatches in the spatial and temporal scales of the measurements being compared.
The correlation coefficients and the biases between FTIR and IASI are also
provided for each year in Table 3. Over Jungfraujoch (Fig. 9), large total
columns were measured by the ground-based instrument during the spring and
the summer for each year. These large values were not captured by IASI data,
causing a large bias of
Correlation coefficient (in italics), mean bias in
10
The seasonal cycle obtained from IASI agrees well with FTIR data over both
sites at La Réunion (correlation coefficient
At Wollongong, IASI and ground-based FTIR background levels are in broad
agreement. The correlation is highest in 2008. The peaks in the HCOOH
columns in October 2012, 2013 and in November 2014 observed by both instruments
are also seen in the distributions in Fig. S1. As in the case of La
Réunion data, a larger positive bias is found when the FTIR total
columns are low (<
The FTIR measurements were also used to evaluate the current HCOOH columns
with those using the conversion from Razavi et al. (2011) (Fig. 10). The
colocation criteria have been enlarged to
Overall, the current data set presents higher correlation and lower bias than the columns from Razavi et al. (2011).
Seasonal HCOOH global distribution from 2008 to 2014 derived using
the IASI radiance observations, gridded to IMAGESv2 horizontal resolution
(2
The IMAGESv2 global CTM runs at 2
For the sake of comparison, the IASI HCOOH total columns have been averaged to the horizontal model grid resolution. The IASI and the model total columns have also been averaged by season, defined as December–January–February (DJF), March–April–May (MAM), June–July–August (JJA), and September–October–November (SON) over the 7 years, between 2008 and 2014. Figure 12 presents these global distributions for IASI and the model. For each season, we find that the IASI total columns are higher than those from IMAGESv2 simulations. This highlights the difficulty to predict the measured concentrations by models as found in previous modeling studies such as Stavrakou et al. (2012) and Paulot et al. (2011).
During winter (DJF), the model shows large total columns over equatorial Africa and Asia while IASI only detects large values over Africa.
In spring (MAM), the CTM largely underestimates the distributions over Africa compared to IASI.
In summer (JJA), the emissions from fires seem to be the primary cause for the strong HCOOH enhancements in the CTM (Africa, Amazonia, boreal regions of Canada and eastern Siberia) although biogenic emissions (southeastern USA) and anthropogenic activities (eastern China) also produce visible enhancements. In the IASI distributions, enhancements associated with fires are less prominent over North and South America. IASI reveals larger total columns over Africa in JJA than over South America while the CTM shows roughly the same values over both regions. Compared with the CTM, IASI also shows larger HCOOH columns over the Midwestern USA, India, and semiarid regions in southwestern Russia and Kazakhstan.
During fall (SON), the columns from IASI over the southern hemispheric biomass burning emission regions (South America, Africa, and Australia) are larger than over Asia (India and China) and over Indonesia, while those simulated by the model are quite similar for these regions.
The global and seasonal distributions from IASI suggest an underestimation of the modeled columns and a misrepresentation of some emission sources. Figure 13 presents the time series over the areas identified by black boxes in Fig. 1. These regions are AMER (North America), AMAZ (Amazonia), AFRI (Africa), SIBE (Central Siberia), INDI (India), ASIA (Asia), and AUST (Australia).
Over all regions, these time series confirm a large bias of IASI in
comparison to the simulations. The seasonal variation is, however, well
represented, except over India, where the seasonal cycle is out of phase.
This phase difference is coherent with the cycle shown on the global
distribution (Fig. 12). The model underestimation over India suggests that
either the emission factors of HCOOH (or its precursors) are underestimated
in the model or the biomass burnt estimated by GFEDv3 is too low for this
region. The underestimation for forest fire emissions is a severe issue as
shown by Chaliyakunnel et al. (2016) over the tropical forests. The
calculated linear trend is also provided, based on the annual IASI mean
(blue dots in Fig. 13). Over three regions (North America, Africa, and Asia),
the trend is negligible (0.2–0.4 % year
Time series of the monthly HCOOH column means for IASI (blue) and IMAGESv2 (red) over different regions highlighted by black boxes in Fig. 1, between 2008 and 2014. The coordinates of each box (latitude and longitude) are written on the top of each plot. The red and blue shaded areas correspond to the monthly standard deviation. The correlation coefficient, the mean bias, and the normalized mean bias (in parentheses) for the full period are given on each plot. The blue dots correspond to the annual IASI mean. The linear regression on the annual IASI mean and the calculated linear trend are also provided. On the bottom panel, the mean altitude of the maximum in the HCOOH vertical distribution from IMAGESv2 is plotted in black with the corresponding standard deviation represented by the black shade areas.
Over Amazonia, the large peak in 2010 due to large forest fire emissions
(e.g., Hooghiemstra et al., 2012) is well represented in the model and
captured by IASI; however, the peak in the IASI data occurs 1 month later
than in the CTM. A similar shift was already observed between the CO and
NH
The double peaks (
Global distributions of HCOOH were derived from IASI radiance spectra, using
conversion factors between representative retrieved total columns and
selected radiance channels (
IASI provides global distributions of HCOOH, highlighting the long-range transport of tropospheric HCOOH over the Atlantic Ocean and the detection of source regions, e.g., biomass burning areas over Amazonia, Africa, Australia, and Siberia. Other source regions are detected such as the mid-eastern USA in 2011 or over India.
The comparison with an atmospheric model and, to a lesser extent, with
ground-based FTIR observations remains challenging. Despite large biases in
many cases, we show that the interannual and the seasonal variations are
well captured by IASI when compared with ground-based FTIR measurements and
the IMAGESv2 CTM. The best overall correlation with the FTIR is obtained at
Saint-Denis over La Réunion (
The HCOOH columns from IASI will require further evaluation and probably
improvements to narrow down the biases but the data set available now spans
7 years and it will likely contribute to a better understanding of the
tropospheric HCOOH budget. The data set will be made available through the
Ether database (
The IASI data are not available yet but as described in the conclusion, the
data set will be made available through the Ether database
(
IASI was developed and built under the
responsibility of Centre National des Etudes Spatiales (CNES) and flies
onboard the MetOp satellite as part of the EUMETSAT Polar system. We thank
the Aeris data infrastructure (