First characterization and validation of FORLI-HNO 3 vertical profiles retrieved from IASI / Metop

Knowing the spatial and seasonal distributions of nitric acid (HNO3) around the globe is of great interest and allows us to comprehend the processes regulating stratospheric ozone, especially in the polar regions. Due to its unprecedented spatial and temporal sampling, the nadir-viewing Infrared Atmospheric Sounding Interferometer (IASI) is capable of sounding the atmosphere twice a day globally, with good spectral resolution and low noise. With the Fast Optimal Retrievals on Layers for IASI (FORLI) algorithm, we are retrieving, in near real time, columns as well as vertical profiles of several atmospheric species, among which is HNO3. We present in this paper the first characterization of the FORLI-HNO3 profile products, in terms of vertical sensitivity and error budgets. We show that the sensitivity of IASI to HNO3 is highest in the lower stratosphere (10–20 km), where the largest amounts of HNO3 are found, but that the vertical sensitivity of IASI only allows one level of information on the profile (degrees of freedom for signal, DOFS; ∼ 1). The sensitivity near the surface is negligible in most cases, and for this reason, a partial column (5–35 km) is used for the analyses. Both vertical profiles and partial columns are compared to FTIR ground-based measurements from the Network for the Detection of Atmospheric Composition Change (NDACC) to characterize the accuracy and precision of the FORLI-HNO3 product. The profile validation is conducted through the smoothing of the raw FTIR profiles by the IASI averaging kernels and gives good results, with a slight overestimation of IASI measurements in the upper troposphere/lower stratosphere (UTLS) at the six chosen stations (Thule, Kiruna, Jungfraujoch, Izaña, Lauder and Arrival Heights). The validation of the partial columns (5–35 km) is also conclusive with a mean correlation of 0.93 between IASI and the FTIR measurements. An initial survey of the HNO3 spatial and seasonal variabilities obtained from IASI measurements for a 1-year (2011) data set shows that the expected latitudinal gradient of concentrations from low to high latitudes and the large seasonal variability in polar regions (cycle amplitude around 30 % of the seasonal signal, peak to peak) are well represented by IASI data.


20
Nitric acid is the main form of oxidized nitrogen, in both the stratosphere and the troposphere, and constitutes the principal chemical sink/reservoir for NO x (≡NO+NO 2 ) (Austin et al., 1986;Crutzen, 1979;Wespes et al., 2009). Being directly coupled to NO x , HNO 3 impacts the ozone (O 3 ) budgets in the two layers (e.g. Neuman et al., 2001).
In the troposphere, the main sources of NO x are fossil fuel combustion and biomass burning (~70%). Natural sources exist, such as lightning and microbial activity in soils, but their contribution to the total tropospheric NO x is smaller than the 25 anthropogenic one, especially in industrialized areas (Cooper et al., 2014;Kasibhatla et al., 1993;Logan, 1983;Wespes et al., 2007). The distribution of the NO X sources influences directly that of HNO 3 , which in the troposphere has a residence time of a few days to several weeks, depending on the latitudes (Logan et al., 1981;Wespes et al., 2007).
In the stratosphere, the main source of NO x is nitrous oxide (N 2 O) which is emitted at the surface by a variety of sources, including agricultural activities (Chipperfield, 2009;McElroy et al., 1976), and is then transported to the stratosphere, where 30 it photodissociates or reacts with O( 1 D) to form two NO molecules (Fischer et al., 1997;Muller, 2011;Portmann et al., 2012). The formed NO x catalyse stratospheric ozone destruction through several cycles (Mohanakumar, 2008;Solomon, 1999). Apart from being an important reservoir species for NO x , HNO 3 is a key species for the formation of polar stratospheric clouds (PSCs, type I) during the polar winter (Höpfner et al., 2006;Lambert et al., 2012;Tabazadeh et al., 2000). PSCs, which develop at very low temperatures (195 K) and which are composed mainly of HNO 3 , sulphuric acid and 35 ice (Drdla & Müller, 2010;Lowe & MacKenzie, 2008), allow heterogeneous reactions and lead to the activation of chlorinated compounds in the gas phase, which in turn induce the subsequent massive destruction of ozone in the low to middle stratosphere of the polar regions in mid-spring (von Clarmann, 2013;Wegner et al., 2012). The process is amplified by the denitrification accompanying the sedimentation of the HNO 3 -rich particles at the end of winter, which prevents the reformation of chlorine nitrate, one of the stable chlorine reservoirs (Gobbi et al., 1991;Solomon, 1999). In a denitrified 40 stratosphere, most often observable in the Antarctic due to lower temperatures, very low HNO 3 concentrations are observed inside the polar vortex and a high concentration collar remains at the edge of the vortex (Santee et al., 1999;Staehelin et al., 2001;Wespes et al., 2009).
Note that the sedimentation is not the main sink for stratospheric HNO 3 at a broader scale; the principal degradation pathways are oxidation with the hydroxyl radical and photodissociation (Austin et al., 1986).

45
HNO 3 has been measured by a variety of instruments since its first observation from infrared solar absorption spectra in 1968 (Murcray et al., 1968). Ground-based instruments (Fiorucci et al., 2013;Rinsland et al., 1991;Wood et al., 2004), sounding instruments on board balloons or aircrafts (Jucks et al., 1999;Neuman et al., 2001) or embarked on satellites (Austin et al., 1986;Orsolini et al., 2009;Wespes et al., 2007) or aboard the space shuttle (Rinsland et al., 1996) have all contributed to the characterization of the HNO 3 distributions throughout the low atmosphere. One of the most complete data sets has been 50 acquired by the Microwave Limb Sounder (MLS) first on the Upper Atmosphere Research Satellite (UARS) from 1991 to et al. 1999; 2004) but at a coarse horizontal resolution due to the viewing mode. The vertical resolution of MLS ranges between 3 and 5 km, and the instruments probes the entire altitude range from the ground to 90 km. However, the HNO 3 measurements are considered reliable only in a narrow altitude range between 11 and 30 km, where the precision on the 55 retrieved volume mixing ratio is 0.6-0.7 ppbv (Santee et al., 2007). HNO 3 distributions have also been obtained by the MIPAS instrument on ENVISAT, in the range 14-43 km with a sampling of 3-4 km and a reported accuracy of 0.2-0.6 ppbv (Piccolo & Dudhia, 2007;Vigouroux et al., 2007), and by the ACE-FTS on-board SCISAT with even better accuracy (3%) between 10 and 37 km (Wang et al., 2007). Measurements from the ODIN instrument made at high vertical resolution (1.5-2 km) but with a precision of only 1.0 ppbv over the altitude range 18-45 km Wang et al., 2007) have been 60 somewhat less used so far.
The Infrared Atmospheric Sounding Interferometer (IASI) on the MetOp satellite series operates in a different geometry (nadir) than all aforementioned instruments, and allows monitoring HNO 3 using spectral information from its 5 and 2 9 vibrational bands. The IASI measurements have a limited vertical resolution because of the integrated view of the atmospheric column but they are made at exceptional spatial and temporal sampling. Specifically, as detailed in section 2, IASI provides global measurements with a particularly good spatial and temporal sampling of the polar regions at all seasons. Other key features of IASI for HNO 3 are the fact that the instrument provides simultaneous measurements of O 3 and other trace gases, allowing studying the coupled HNO 3 -O 3 cycles, and that it will operate on a long-term (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022), allowing the identification and monitoring of trends. The potential of using the HNO 3 measurements of IASI was first shown by Wespes et al. (2009) using a two-year data set. Although important conclusions and perspectives for the capability of 70 IASI to sound HNO 3 were drawn from this study, it suffers from the fast operational retrieval method restricted to a total column calculation for the purpose of saving time, and from a missing quantitative validation due to the absence of archived data from ground-based FTIR measurements. Moreover, the total column retrieval combined with its maximum sensitivity in the stratosphere was shown to largely mask the potential of IASI to sound HNO 3 in lower layers. After this study, the Fast Optimal Retrieval on Layers for IASI (FORLI) software was adapted to allow retrieval of HNO 3 vertical profiles from IASI 75 . A full 8-year data set of global profiles is now available (2008-2015) but has not yet been used for extensive analysis (a subset was used in Cooper et al. (2014) to constrain lightning NO x emissions in models). The product characterization and validation are also still lacking. This paper uses a full year (2011) of HNO 3 profiles retrieved from IASI on Metop-A to  Fully characterize the HNO 3 retrieved concentrations in terms of vertical sensitivity and errors (section 3) on the 80 profiles and partial columns,  Validate the profiles and columns using correlative data from the NDACC FTIR network, which are also detailed here (section 4 and 5),  Provide an overview of how the product can be used to analyse spatial and temporal variability (section 6).

IASI instrument
The first IASI instrument (IASI-A) was launched in 2006 on the Metop-A platform in a polar orbit Hilton et al., 2012). It is still operating nominally at the time of writing, in parallel with IASI-B that was launched on Metop-B in 2012. IASI is a nadir-viewing infrared Fourier transform spectrometer measuring the radiation emitted by the Earth's surface and the atmosphere in the 645-2760 cm -1 spectral range . The spectral 90 resolution is 0.5 cm -1 after apodization over the entire spectral range (Cayla, 2001;Hilton et al., 2012). The apodized radiometric noise is low, around 0.2 K in the atmospheric window of interest to this work, which includes the 5 + 2 9 band of HNO 3 (860-900 cm -1 ), mostly suitable for the retrievals, as described in Wespes et al. (2007Wespes et al. ( , 2009. IASI collects 120 views every 8s along the 2200 km swath across to the satellite track and provides this way global coverage twice a day , with one overpass in the morning and one in the evening, at 9:30 equator crossing time. The spatial 95 resolution varies from 113 km 2 at nadir to 400 km 2 at the end of the swath. Note that IASI measures in addition to HNO 3 a series of greenhouse (carbon dioxide (CO 2 , Crevoisier et al., 2009), methane (CH 4 , Crevoisier et al., 2013)) and reactive trace gases (carbon monoxide (CO, George et al., 2009), O 3 (Boynard et al., 2016;Wespes et al., 2016), ammonia (NH 3 , Van Damme et al., 2014)), which altogether provide an extensive monitoring of the atmospheric system.

Retrieval method and settings
The IASI data are processed every day in near-real time at ULB, by the FORLI algorithm which relies on a fast radiative transfer and on a retrieval methodology based on the Optimal Estimation Method (OEM, Rodgers, 2000) to solve the inverse problem in the retrieval . FORLI provides twice-daily vertical distributions of three species, namely O 3 , CO and HNO 3 . The FORLI methods have already been largely described  so only a brief 105 reminder will be presented here, focusing on the retrieval parameters for HNO 3 .
The forward model can be written in a generic way as where Y is the measurement vector (the IASI calibrated and apodized radiances in our case), x is the retrieved state vector, b includes all parameters influencing the measurement, and η is the measurement noise. F is the forward function, which 110 describes the complete physics of the measurement Rodgers, 2000).
The inverse problem consists in finding a state vector x approximating the true state vector x, in accordance with the measurement Y and with a prior knowledge of the state of the atmosphere, characterized by an a priori profile x and the corresponding variance-covariance matrix S . The solution of the above equation for a linear problem is expressed as where K is the Jacobian of the forward model F, and S is the measurement error covariance. For a non-linear problem as ours, the solution is found iteratively. The optimal estimation provides a very appropriate framework for characterizing the retrieved profiles, in terms of vertical sensitivity (analysed with the averaging kernel functions) and errors. The way those quantities are calculated are described in Hurtmans et al. (2012) and are not repeated here. FORLI-HNO 3 in its latest version (v.20140922) provides profiles on 41 layers (from surface up to 40 km). The retrieval parameters, adapted from Wespes et al. (2009) andHurtmans et al. (2012), are detailed in Table 1. The spectral range for the retrieval of HNO 3 profiles is 860-900 cm -1 (see Fig.2 in Wespes et al. 2009), in which only water vapour significantly interferes. The a priori profile (x ) is defined as the mean of a combination of daily profiles from the LMDz-INCA chemistry-transport model (from the ground up to 15.6 km) and of all profiles obtained from ACE-FTS (from 6 to 60 km). The resulting mean profile is constant over time and does not depend on latitude or longitude, i.e. the same a priori profile is used for all observations around the globe. A variance-covariance matrix from the ensemble of profiles is then calculated (S ) and yields high variability in the boundary layer (170%) and in the UTLS region (80%) and lower variability in the troposphere and the stratosphere (50% and 20%, respectively). The uncorrelated noise varies around the value of 2.10 -8 W/(cm 2 cm -1 sr). The FORLI-HNO 3 retrieval performances in terms of root mean square (RMS) and bias values of the spectral residuals calculated after the retrievals, of retrieval total errors, and of degree of freedom for signal (DOFS calculated as the trace of the averaging kernel matrix -

130
DOFS=trace(A)) are detailed in Hurtmans et al. (2012). Identically to Wespes et al. (2009), posteriori filtering of the data has been performed to remove some strongly biased HNO 3 observations. For instance, only HNO 3 observations with a good spectral fit (RMS of the spectral residual lower than 3.10 -8 W/(cm 2 cm -1 sr)) have been analysed. Additional quality flags rejecting biased or sloped residuals, suspect averaging kernels and maximum number of iterations exceeded are also applied.
Cloud contaminated IASI scenes are also filtered out, based on cloud information from the Eumetcast operational processing 135 (fractional cloud cover below 25%).

Characterization of FORLI-HNO profiles and columns
Figure 1 displays for illustration typical FORLI-HNO 3 retrieved profiles at tropical (Izaña), mid (Jungfraujoch) and polar (Thule) latitudes in July as well as the a priori profile. It can be seen that the tropospheric concentrations are small and remain close to the a priori for all latitudes. Only Izaña stands out in this respect, with the retrieved concentrations 30% smaller than those of the a priori profile. The stratospheric concentrations are much higher, especially between 15 and 25 km altitude, and the better sensitivity of the instrument at these altitudes (discussed next) allows for significant departure from the a priori profile. For these three example profiles, maximum values range between less than 5 ppbv at Izaña and 9.5 ppbv at Thule. This latitudinal difference is in fact a persistent and well-documented feature (e.g. Santee et al., 2007Santee et al., , 1999Wespes et al., 2007), which is mostly associated with reduced HNO 3 photodissociation in the polar regions, mainly in winter.
The averaging kernels, presented in Figure 2 (top panels) for the same three locations as in Figure 1, allow characterizing the sensitivity of IASI to the HNO 3 profile, within the general FORLI framework. We find that the averaging kernels show similar shapes whatever the latitude and cover the whole range of altitudes from the surface to the upper stratosphere. The 150 limited amount of vertical information is well seen from the overlap between the individual layer kernels. Also the quasi absence of sensitivity in the low troposphere below 5 km is obvious, with absolute values of the averaging kernels close to zero. This confirms the conclusion from Wespes et al. (2009) that the IASI instrument does not carry several pieces of vertical information for HNO 3 . From the averaging kernels, we can also conclude that the maximum sensitivity to HNO 3 is at around 15 km, slightly higher (18 km) at tropical latitudes in comparison to polar latitudes (13 km). This translates to a 155 maximum sensitivity in the upper troposphere-lower stratosphere (UTLS) at equatorial latitudes, and in the low to middle stratosphere at polar latitudes, which corresponds generally to the altitude of highest concentrations.
In order to have a global vision of the results, Figure 3 shows the global distributions of the degrees of freedom for signal (DOFS, top panels) separately for January (left) and July (right) 2011. These represent the number of independent pieces of information in the measurements and give an estimation of the vertical sensitivity of the retrievals. On average, all DOFS 160 values are close to 1, further indicating that only one level of information can be extracted from the IASI data for HNO 3 .
However, there are some latitudinal differences, with the DOFS being generally larger in the intertropical belt, with values around 1.1 or slightly more (e.g. in the deserts during the summer) due to larger temperatures inducing a better signal to noise ratio, in comparison to the mid and polar latitudes, where the DOFS is mostly around 0.9. However, it should be noted that the larger values of DOFS, particularly the ones found in the deserts, might also be, at least partly, attributed to the 165 misrepresentation of the emissivity of these surfaces . Indeed, even though we also find high surface temperatures (see Figure 3, middle panels) in most of the intertropical region of the globe, it does not necessarily induce such a large DOFS value. The lowermost panels in Figure 3 depict the altitude of maximum sensitivity of the IASI in January (left) and July (right). We show that the altitude of maximum sensitivity is invariant at equatorial and tropical latitudes, whereas it varies with the season at mid-and polar latitudes. As expected, the variations of the altitude of maximum values of the averaging kernels, that the total retrieval error in the low troposphere reproduces the a priori covariance, meaning that no information on HNO 3 profile is obtained from the IASI measurement at these altitudes. The total error decreases with increasing altitude, reaching minimum values of about 20% around 20 km, where the sensitivity is highest.
Comparing with the a priori covariance, the precision is in fact increased mainly between 5 and 25 km, and especially around 185 14 km where the reduction of error reaches almost 50%. Above 30 km, the gain of information is again minimal. The measurement error is a minor source of error; it reaches a maximum of 30% in the boundary layer and quickly becomes negligible above 10 km. The error on the profile translates to a total retrieval error in the range of 5 to 50% depending on latitude, with a mean value of 10%. When calculated for various partial columns (see box in Figure 4, left), it appears clearly that the tropospheric column carries the largest error (62%), whereas the error is lower for the total (10%) or the 190 stratospheric (8%) column. Following the analysis of the error profiles in Figure 4 (left), a column ranging from 5 to 35 km can be considered as the one carrying most of the retrieved information; it is characterized by a total retrieval error of 3% on a global average and the DOFS is the same as for the total column (ranging from 0.9 to 1.2).
Spatially, we find that the total error is larger at tropical latitudes, with values around 10-15% ( Figure 4, right), mainly due to the higher concentrations of water vapor, which has absorption lines that interfere under large humidity with the ones of 195 HNO 3 in the spectral region used for the inversion Wespes et al., 2009). The mid and polar latitudes are characterized by much lower total retrieval errors, with maximum values of about 3% for the 5 to 35 km column. This corresponds to a reduction by a factor 30 as compared to the prior uncertainty defined by S (90%). The very large errors found in Antarctica are most probably due to a misrepresentation of the surface emissivity above cold regions and to a very poor sensitivity above such cold regions .

4 Validation methodology
The IASI derived HNO 3 profiles/columns are compared here with reference profiles/columns retrieved from measurements made by ground-based infrared Fourier-transform spectrometers (FTIR) for the sake of validation. Note that the FTIR HNO 3 profiles have been used to validate other satellite datasets before (Vigouroux et al., 2007;Wang et al., 2007;Wolff et al., 2008;Wood et al., 2002).

215
Kiruna and Izaña, SFIT2 (Pougatchev al. 1995;Rinsland et al., 1998) for Thule and Jungfraujoch, and SFIT4 (Pougatchev et al., 1995) for Lauder and Arrival Heights. An updated Lauder and Arrival Heights HNO 3 data set was used in this study. The Lauder and Arrival Heights retrievals implemented SFIT4 closely adhering to the prescribed NDACC IRWG HNO 3 retrieval strategy (see http://www.acom.ucar.edu/irwg/IRWG_Uniform_RP_Summary-3.pdf), with the following particularities: at both sites the spectral resolution is 0.0035 cm -1 and pressure and temperature profiles (ZPT) were obtained from NCEP.

220
Using LINEFIT (Hase et al., 1999) to diagnose the instrument lineshape (ILS), an ideal ILS is assumed at Lauder whilst a parametrised linear ILS from 1.0 at ZPD to 0.95 (max OPD) is used in Arrival Heights retrievals. A single beam channelling fit is also implemented in the Arrival Heights retrievals. The signal-to-noise ratio (SNR) is calculated per spectrum and yields typical values of 180 and 195 for Lauder and Arrival Heights, respectively.
The three algorithms (PROFFIT, SFIT2 and SFIT4) use the optimal estimation method (OEM) developed by Rodgers (2000) 225 which facilitates the comparison with IASI. All stations use microwindows in the region 866-875.2 cm -1 for the retrieval of HNO 3 profiles and the DOFS range between 1.9±0.5 (Jungfraujoch) and 3.1±0.4 (Thule). The a priori profiles are described individually at each station, regardless of seasonality. It should be noted that, due to the lower degrees of freedom in the Lauder and Arrival Heights retrievals, there were numerical artefacts in the smoothing operation between the IASI and the FTIR ground-based measurements (see section 4.2). To reduce these artefacts, a smoothed IASI a priori profile was used as 230 the a priori in the ground-based retrievals in this study. The associated ground-based retrieval a priori covariance matrices (S a ) were also constructed from a smoothed IASI covariance data set. Spectroscopic line parameters were all taken from HITRAN 2008 database (Rothman et al., 2009). Note that IASI retrievals use spectroscopic line parameters taken from HITRAN 2004 database (Rothman et al., 2005). The update consists mainly in an improvement of the line positions and intensities (Flaud et al., 2006;Gomez et al., 2009;Rothman et al., 2009) and the differences it might induce should be kept 235 in mind when analysing the comparison between the two instruments. Additional differences between the two instruments might also come from the uncertainty of the FTIR measurements themselves, which is dominated by the temperature and the spectroscopy information in the retrieval.

Co-location criteria and comparison method
The validation is performed for the year 2011 of IASI data. For the spatial co-location, the line of sight of the FTIR location of the maximum sensitivity along the line of sight that determines the reference for the co-location with IASI measurements. The co-location criteria chosen is such that the IASI measurements should be within 0.5° in latitude and 1° in 245 longitude from the FTIR reference point.
The co-location in time has been chosen after several tests (not shown) as ≤ 12h. If more than one IASI measurement was satisfying these criteria, then all IASI retrieved profiles inside that spatial and temporal window were averaged. Each pair (FTIR, co-located IASI) undergoes the validation steps described below and in Rodgers & Connor (2003).
The raw profiles retrieved from IASI and FTIR cannot be compared adequately because of the difference in vertical 250 sensitivity (Rodgers & Connor, 2003). To illustrate this, Figure 2 depicts both IASI and FTIR typical averaging kernels (top and bottom panels, respectively) as well as the IASI and FTIR so-called "sensitivities" (red curves). The sensitivity at altitude is calculated as the sum of the elements of the corresponding averaging kernel, ∑ (with the averaging kernels matrix) and represents the fraction of the retrieval that comes from the measurement rather than from the a priori profile (Vigouroux et al., 2007). As opposed to this sensitivity, the total averaging kernel (dashed black lines) is calculated as 255 ∑ and represents the contribution of each level to the sensitivity at a given altitude .
We find that the FTIR instrument provides better vertical resolution, with each averaging kernel maximum close to its corresponding layer, and a total DOFS within the known values (see Table 2 over-sensitive to HNO 3 concentrations in that region of the atmosphere (Vigouroux et al., 2007), and that it might overcompensate the lack of sensitivity in other regions of the atmosphere, yielding large HNO 3 concentrations.

265
In order to compare an FTIR-IASI profiles pair and to reduce the smoothing uncertainties on the comparison, we use Rodgers and Connor (2003). First we insert the IASI a priori in the FTIR profile using where x is the FTIR profile aligned with the IASI a priori x , x is the raw FTIR profile, x the a priori FTIR profile, and A is the FTIR averaging kernel matrix.

270
Next, this aligned FTIR profile is interpolated to the IASI altitude grid and the standard smoothing equation is applied, using the IASI averaging kernel and a priori: The comparison is carried out next for the profiles (section 5.1) or partial columns (5-35 km) (section 5.2). For the discussion, we rely on the relative difference between the quantities retrieved from IASI and the corresponding ones from the smoothed FTIR data following: with being the relative difference. Standard deviations are also calculated as: with the number of observations, the difference between IASI and FTIR values for the th observation and ̅ the mean of the differences for all measurements, and allow the characterization of the variability of the data set.

Vertical profiles
The comparison of the HNO 3 mean vertical profiles retrieved from IASI and the FTIRs for each station is presented in Figure   5, with the six panels on the left showing the comparison between IASI profile (red) and the raw (black) and smoothed (green) FTIR profiles, as well as IASI and FTIR a priori profiles (dashed red and dashed black lines, respectively). The right panel shows the mean relative differences (%) between IASI and the smoothed FTIR profile for each station. For all stations, 290 the raw FTIR profile is largely different from the IASI profile, and this difference remains after regridding on the IASI retrieved levels (not shown). The smoothing of the FTIR profile with IASI averaging kernels brings the two profiles much closer together, with differences generally below 50% at all altitudes and maximum differences in the upper tropospherelower stratosphere between 300 and 50 hPa. At the higher latitudes, the differences in the profiles are always below 20%. The statistics of the comparison between profiles is summarized in Table 3. Overall, the troposphere and higher stratosphere 295 record lower values of relative difference between IASI and the smoothed FTIR profile. However, this is in most part related to the low vertical sensitivity in these regions of the atmosphere, forcing the retrieval to rely mostly on the a priori information, i.e. the same information after the smoothing of the FTIR profiles with IASI a priori information. As indicated in the Table 3 and as can be seen in Figure 5 (right), at all stations, the maximum values for the relative differences with the smoothed profile are located in the low stratosphere (around 13 km altitude), where the sensitivity of IASI to the 300 measurement is the largest (see Fig.2, top panels). These differences are always positive, suggesting an overestimation of the IASI concentrations in that region of the atmosphere, compared with the FTIR data, and they vary with the latitude (from 4.7% at Arrival Heights to 47.2% at Lauder). Important differences between IASI and the FTIR are also found in the boundary layer, especially in Kiruna (-11.8%), Jungfraujoch (19.7%) and Arrival Heights (43.9%) and probably do not reveal a real difference, but rather an artefact due to the regridding with potential differences in the altitude taken as ground level in 305 both instruments. It should be noted that an overestimation of IASI measurements compared with ground-based measurements in the upper troposphere-lower stratosphere has also been found for O 3 vertical profiles (Antón et al., 2011;Dufour et al., 2012;Gazeaux et al., 2013). While some hypotheses have been brought forward by Dufour et al. (2012), the exact reason for that particular feature of FORLI for both HNO 3 and O 3 retrievals is not clear.

310
HNO 3 partial columns from 5 to 35 km have then been compared in a similar way as the profiles. We recall here that the is displayed in grey. It shows the exceptional temporal sampling and especially the usefulness of IASI measurements in the winter months at high latitude. The IASI a priori is also represented (grey dashed line) for each station. Note that the slight temporal variability of the IASI a priori is due to its representation in column units, making it dependent on the air column at 320 the time of measurement. The possibility to investigate chemical and physical processes from these time series is briefly explored in section 6. The bottom panels in Figure 6 show the relative differences between the different data sets. The statistics of the comparison are provided in Table 4. For all stations and all measurements, we find that the FTIR column values are within the IASI total retrieval error range (see red error bars), with mean values of HNO 3 partial columns very similar between the two instruments (see Table 4, columns 5 and 6). Considering the smoothed columns, IASI is always 325 positively biased, with bias values between 4.0% in Thule and 21.7% for Lauder, and with an overall bias (all stations together) of 11.5% (Table 4). The smoothing of the FTIR data is particularly efficient for the comparison with IASI observations at Izaña, where the bias decreases from 21.3% to 9.2% from the unsmoothed to the smoothed FTIR. Being the station with the largest difference between the FTIR and the IASI a priori profiles (see Figure 5), Izaña is a good example of the influence of the a priori profiles in the retrievals. Indeed, with initially very different a priori profiles and limited vertical 330 sensitivities, yielding large relative differences between the two data sets (see Figure 6, black dots), the smoothing of the FTIR data set largely decreases the mean bias (green dots).
Looking at all stations, the standard deviation of the differences is larger than the bias at Thule, Kiruna, Izaña and Arrival Heights, suggesting that the bias is non-significant compared to the variability (Kerzenmacher et al., 2012). The reason of the different behaviour for the Jungfraujoch and Lauder mid-latitude stations is unclear at this point.  Table 4, Atmos. Meas. Tech. Discuss., doi:10.5194/amt-2016-207, 2016 Manuscript under review for journal Atmos. Meas. Tech. Published: 6 July 2016 c Author(s) 2016. CC-BY 3.0 License.
last column) showing that IASI captures very well the dynamic range of variability for the column, which varies between 0.9.10 16 and 4.5.10 16 molec.cm -2 . This high correlation coefficient must however be considered cautiously, because of the 340 influence of the remaining a priori information, albeit moderately since the least sensitive regions have been removed by considering the 5-35 km partial column. The relative differences clearly illustrate the positive bias of IASI described above, but also that the bias does not show any particular temporal trend.
6 Spatial and temporal variability
The amplitude of the seasonal cycle thus amounts to 3.0 10 16 molec.cm -2 . The annual cycle is consistent with the known variability of stratospheric polar latitudes (Wespes et al., 2009). The comparison of the seasonality with the FTIR is not possible at Thule since the FTIR instrument, which operates with solar light, does not provide data before March and after October.

350
In Kiruna, IASI partial columns range between around 1.5 10 16 molec.cm -2 and 3.5 10 16 molec.cm -2 . The winter months (February to April, approximately) are characterized by higher columns whereas the summer columns are lower (July and August), giving a seasonal amplitude of 2 10 16 molec.cm -2 . Due to the relatively high latitude of the station, January and December data are also missing in the FTIR record. With the data available, we find good agreement on the seasonality between IASI and the FTIR, however with a time-dependent bias and especially a significant positive bias of IASI (~30%)

355
for several days around mid-February and a low bias around -20% throughout March.
At Jungfraujoch and Izaña, IASI partial columns are lower than at higher latitudes, ranging between around 1.3 10 16 and 3.0 10 16 molec.cm -2 for Jungfraujoch, and 0.9 10 16 and 1.7 10 16 molec.cm -2 for Izaña. The amplitude of the seasonal cycle is thus much weaker than at higher latitudes (1.7 10 16 molec.cm -2 at Jungfraujoch and 0.9 10 16 molec.cm -2 at Izaña), with only slightly higher concentrations observed in January and February at Jungfraujoch.

360
In the Southern hemisphere, at Lauder, IASI partial columns range between 1.0 10 16 and 2.6 10 16 molec.cm -2 with slightly higher values recorded during the local winter and spring (August-September, mainly). The annual cycle amplitude reaches 1.5 10 16 molec.cm -2 and reflects well the small annual cycle usually recorded for southern mid-latitudes.
As for Arrival Heights, columns range between 0.9 10 16 and 3.1 10 16 molec.cm -2 , and the day-to-day variability caused by the variability of the vortex itself is well seen in the data (see for example the three data points in April). Due to its very during winter is important. Hence, we find lower concentrations than what could be expected at Thule from January to April (consistent with the few FTIR data available in March) and a strong decrease in HNO 3 concentrations in June at Arrival Heights, which are discussed below. Another important feature highlighted by the complete IASI data set is the large spatial variability recorded at high latitudes. Indeed, with each grey data point representing a daily mean of all IASI observations in a quite large box around the station, and the red dots being the closest IASI observation to the FTIR reference point for the comparison (see section 4.2. for details), the difference between the two shows the spatial variability that can exist within a 375 defined region. Such a feature can be observed at Arrival Heights, Thule and Kiruna, whereas other stations located at more mid-latitudes (Lauder, Izaña and Jungfraujoch) show very little to no spatial variability. The daily variability is also highlighted in Figure 6, with the grey shaded areas representing the standard deviation (3 ) of the daily mean IASI observations. This variability includes the spatial variation of HNO 3 inside the selected boxes around the FTIR stations and the daily variation captured by the IASI daytime and nighttime measurements. It is particularly interesting in the mid-and 380 high latitude stations, especially in winter (Januay-April for Thule, July-September for Arrival Heights), where it is much larger than the error values (~1.7.10 16 molecules.cm -2 for the daily variations vs ~0.5.10 16 molecules.cm -2 for the total errorred vertical error bars) and thus proves that the IASI instrument captures a real daily variability of HNO 3 partial columns at these latitudes. This is mainly linked to the rapid zonal transport of air masses induced by the development of the polar vortex in these regions (Wespes et al., 2009). Regarding the tropical latitudes, the daily variability should be considered 385 more cautiously, since it is of the same magnitude as the retrieval error. Though not in the scope of the present study, the question of day and night variability needs to be further investigated.

Global variability
Beyond assessing the validity of HNO 3 locally, IASI offers the potential of global analysis thanks to its sampling. Monthly With these fluctuations, the polar regions record not only the highest columns but also the largest cycle amplitudes, which are around 1.0 10 16 and 1.4 10 16 molec.cm -2 for North and South poles, respectively, with low columns in the summer and high columns in the winter.

410
In addition to the seasonal cycles, significant daily variability represented by the colored shaded areas in Figure 9 (calculated as 1 sigma of the daily IASI measurements) can be revealed thanks to the IASI sampling, especially at high latitudes during the denitrification periods. This daily variability has already been reported by Wespes et al. (2009). The main reason behind the day-to-day variability at high latitude is the variability of the vortex itself.

415
In this paper we have characterized, validated and analysed the first results of the FORLI-HNO 3 vertical profiles data set retrieved from IASI/MetOp. The profiles are retrieved on a 41 km altitude grid twice a day, globally. A 8-year record is now available on request and an implementation in the EUMETSAT IASI-Level2 Product Processing Facility  is foreseen. One year (2011) of data has been investigated here. We have shown that IASI has a maximum sensitivity to the HNO 3 profile in the stratosphere, around 10-20 km altitude and that the vertical sensitivity of the instrument typically 420 allows retrieving a single piece of information on the profile (DOFS varying from 0.9 to 1.2). The altitude of maximum sensitivity corresponds to the region of the atmosphere with the highest concentrations. The averaging kernels and error profiles showed that most of the available information in the IASI measurements originates from the altitude range between 5 to 35 km altitude. In terms of the corresponding partial column (5-35 km), the total retrieval error was calculated to be around 3% at high latitudes, where water vapour does not interfere with HNO 3 absorption lines, increasing to 10-15% at 425 equatorial latitudes.
The validation was conducted by comparing the IASI retrieved HNO 3 profiles or partial columns to those retrieved from ground-based FTIR measurements made at six different stations spread around the globe at representative latitudes, namely Thule, Kiruna, Jungfraujoch, Izaña, Lauder and Arrival Heights. We found good general agreement between IASI and the smoothed FTIR profiles, with, however, an overestimation of IASI data compared with FTIR measurements (11.5% positive 430 bias). In most cases, the differences were not found significant compared to the variability. The correlation between the two data sets is high for all stations (0.93 for all stations together), demonstrating the capability of IASI to capture the spatial and temporal patterns of the HNO 3 variability. However, as could be highlighted by the case of Izaña, the influence of the a Atmos. Meas. Tech. Discuss., doi:10.5194/amt-2016-207, 2016 Manuscript under review for journal Atmos. Meas. Tech. Published: 6 July 2016 c Author(s) 2016. CC-BY 3.0 License.
priori profile to the validation process is quite large, since the application of a common a priori profile to both measurements largely improves the comparison. It should also be noted that the differences observed between the two data sets can also be 435 attributed, at least partly, to the difference in the spectral region used by each instrument, the different line parameters (HITRAN 2004 for IASI andHITRAN 2008 for FTIR) and codes used for the retrievals and the errors of each instrument.
IASI data allows remarkable monitoring of year-round HNO 3 concentrations. The global distribution acquired with IASI showed for instance a clear latitudinal gradient, with low and relatively constant concentrations at tropical latitudes, and much higher and very variable concentrations at mid and polar latitudes. The daily variability also highlighted by the IASI 440 data set can be further investigated thanks to the capability of the IASI sounder to monitor the atmosphere at both day and night time.
The polar processes, including the strong denitrification in Antarctic, are well monitored with IASI, both spatially and temporally. Overall, the results presented here are extremely encouraging with regard to the use of the HNO 3 dataset from IASI to investigate stratospheric processes on local to global scales, with particular interest for the polar regions. The long 445 time series that are available from IASI, which will span more than 15 years and which will be extended with the IASI-NG instrument on EPS-SG (Clerbaux & Crevoisier, 2013;Crevoisier et al., 2014), will be important to monitor longer-term changes in stratospheric composition and its link to climate.       Atmos. Meas. Tech. Discuss., doi:10.5194/amt-2016-207, 2016 Manuscript under review for journal Atmos. Meas. Tech. Published: 6 July 2016 c Author(s) 2016. CC-BY 3.0 License.

Figure 5. (left) IASI (red), raw FTIR (black) and regridded and smoothed FTIR mean profiles (green). Also shown are the IASI and FTIR a priori profiles (red and black dashed lines, respectively). (right) Relative differences between the IASI and the FTIR smoothed profiles for each station.
Atmos. Meas. Tech. Discuss., doi:10.5194/amt-2016-207, 2016 Manuscript under review for journal Atmos. Meas. Tech. Published: 6 July 2016 c Author(s) 2016. CC-BY 3.0 License. Figure 6. (top panels) Comparison of the IASI and the corresponding FTIR partial (5-35 km) columns: the red dots are IASI derived partial columns, the black dots refer to the raw FTIR data and the green dots to the smoothed FTIR. The vertical error bars represent the retrieval total error for each instrument. Also represented (in light grey) is the IASI data set averaged over a small region around the FTIR location (see text for details) and the standard deviation for each data point (grey shaded areas). The IASI a priori partial column is represented by the dashed grey line. (bottom panels) Relative differences (dots), bias (dashed line) and standard deviation (dotted lines) between the two data sets for each station: in black, the difference between IASI and the raw FTIR data, in green, the difference between IASI and the smoothed FTIR data. Atmos. Meas. Tech. Discuss., doi:10.5194/amt-2016-207, 2016 Manuscript under review for journal Atmos. Meas. Tech. Published: 6 July 2016 c Author(s) 2016. CC-BY 3.0 License. Atmos. Meas. Tech. Discuss., doi:10.5194/amt-2016-207, 2016 Manuscript under review for journal Atmos. Meas. Tech. Published: 6 July 2016 c Author(s) 2016. CC-BY 3.0 License.