Validation of ACE-FTS N 2 O measurements

The Atmospheric Chemistry Experiment (ACE), also known as SCISAT, was launched on 12 August 2003, carrying two instruments that measure vertical profiles of atmospheric constituents using the solar occultation technique. One of these instruments, the ACE Fourier Transform Spectrometer (ACE-FTS), is measuring volume mixing ratio (VMR) profiles of nitrous oxide (N2O) from the upper troposphere to the lower mesosphere at a vertical resolution of about 3–4 km. In this study, the quality of the ACE-FTS version 2.2 N2O data is assessed through comparisons with Correspondence to: K. Strong (strong@atmosp.physics.utoronto.ca) coincident measurements made by other satellite, balloonborne, aircraft, and ground-based instruments. These consist of vertical profile comparisons with the SMR, MLS, and MIPAS satellite instruments, multiple aircraft flights of ASUR, and single balloon flights of SPIRALE and FIRS-2, and partial column comparisons with a network of groundbased Fourier Transform InfraRed spectrometers (FTIRs). Between 6 and 30 km, the mean absolute differences for the satellite comparisons lie between −42 ppbv and +17 ppbv, with most within ±20 ppbv. This corresponds to relative deviations from the mean that are within ±15%, except for comparisons with MIPAS near 30 km, for which they are as large as 22.5%. Between 18 and 30 km, the mean absolute differences for the satellite comparisons are generally Published by Copernicus Publications on behalf of the European Geosciences Union. 4760 K. Strong et al.: Validation of ACE-FTS N 2O within ±10 ppbv. From 30 to 60 km, the mean absolute differences are within ±4 ppbv, and are mostly between −2 and +1 ppbv. Given the small N 2O VMR in this region, the relative deviations from the mean are therefore large at these altitudes, with most suggesting a negative bias in the ACEFTS data between 30 and 50 km. In the comparisons with the FTIRs, the mean relative differences between the ACEFTS and FTIR partial columns (which cover a mean altitude range of 14 to 27 km) are within ±5.6% for eleven of the twelve contributing stations. This mean relative difference is negative at ten stations, suggesting a small negative bias in the ACE-FTS partial columns over the altitude regions compared. Excellent correlation ( R=0.964) is observed between the ACE-FTS and FTIR partial columns, with a slope of 1.01 and an intercept of −0.20 on the line fitted to the data.


Introduction
Nitrous oxide (N 2 O) is an important atmospheric constituent, as it is the primary source gas for nitrogen oxides in the stratosphere, a useful dynamical tracer, and an efficient greenhouse gas. N 2 O has many surface and near-surface sources, with approximately equal contributions from natural and anthropogenic emissions. Natural sources include biological nitrogen cycling in the oceans and soils and oxidation of NH 3 , while anthropogenic sources include chemical conversion of nitrogen in fertilizers into N 2 O, biomass burning, cattle, and some industrial activities (IPCC, 2007). It is the only long-lived atmospheric tracer of human perturbations of the global nitrogen cycle (Holland et al., 2005). There are large uncertainties in N 2 O source strengths derived from emissions inventories, with estimates of the total source strength varying by ±50% (McLinden et al., 2003, and references therein). Tropospheric N 2 O is transported through the tropical tropopause into the stratosphere, where approximately 90% is destroyed by photolysis at wavelengths from 185 to 220 nm, which creates N 2 and O. The remaining 10% is destroyed by reaction with O( 1 D). The latter has two channels, one of which generates two NO molecules and serves as the source for stratospheric nitrogen oxides, which participate in catalytic destruction of ozone (Bates and Hays, 1967;Crutzen, 1970;McElroy and McConnell, 1971).
While N 2 O is well-mixed in the troposphere, its concentration decays with altitude in the stratosphere due to the reactions noted above. Its photochemical lifetime varies from approximately 100 years at 20 km and below, to 1 year at 33 km and 1 month at 40 km (Brasseur and Solomon, 2005). As these lifetimes are longer than dynamical time scales, the global distribution of N 2 O is primarily governed by the Brewer-Dobson circulation. This makes it a useful tracer in the stratosphere, both as a diagnostic tool in atmospheric models (Mahlman et al., 1986;Holton, 1986;Bregman et al., 2000;Plumb and Ko, 1992;Avallone and Prather, 1997;Sankey and Shepherd, 2003) and for the interpretation of observational data. For example, N 2 O has been used in numerous studies of polar vortex dynamics and chemistry (e.g., Proffitt et al., 1989Proffitt et al., , 1990Proffitt et al., , 1992Müller, 1996;Bremer et al., 2002;Urban et al., 2004), the tropical pipe (e.g., Plumb, 1996;Murphy et al., 1993;Volk et al., 1996;Avallone and Prather, 1996), transport and chemistry in the upper troposphere/lower stratosphere (e.g., Boering et al., 1994;Hegglin et al., 2006), and global transport processes (e.g., Randel et al., 1993Randel et al., , 1994. Radiatively, N 2 O is a long-lived greenhouse gas (Yung et al., 1976;Ramanathan et al., 1985). It has a global warming potential of 289 over 20 years, and a global average radiative forcing due to increases in N 2 O since the pre-industrial era of 0.16±0.02 Wm −2 , making it the fourth most important trace gas contributing to positive forcing (IPCC, 2007). Global surface concentrations of atmospheric N 2 O are currently increasing at about 0.26% per year, and have risen from a preindustrial value of about 270 ppbv to 319 ppbv in 2005, due to an increase of 40-50% in surface emissions over that period due to human activities (Battle et al., 1996;Flückiger et al., 1999;Zander et al., 2005;Hirsch et al., 2006;WMO, 2006;IPCC, 2007, and references therein). There is a hemispheric difference in N 2 O, with about 0.8 ppbv more in the northern hemisphere, which is the source of approximately 60% of the emissions (Brasseur and Solomon, 2005).
Global distributions of N 2 O have been measured from space since 1979, when the Stratospheric and Mesospheric Sounder (SAMS) on Nimbus 7 began operations (Drummond et al., 1980;Jones and Pyle, 1984;Jones et al., 1986). SAMS used an infrared pressure modulator radiometer to measure thermal emission from the limb at 7.8 µm, from which stratospheric N 2 O profiles were retrieved until 1983. This was followed by the Improved Stratospheric and Mesospheric Sounder (ISAMS) and the Cryogenic Limb Array Etalon Spectrometer (CLAES) on the Upper Atmosphere Research Satellite (UARS). ISAMS also used pressure modulator radiometers, operating from 4.6 to 16.3 µm, and provided N 2 O profiles between October 1991 and July 1992 (Taylor et al., 1993;Ruth et al., 1994;Remedios et al., 1996). CLAES also measured N 2 O using thermal limb emission, from 3.5 to 13 µm, between October 1991 and May 1993 . The Atmospheric Trace Molecule Spectroscopy (ATMOS) instrument, flown on four Space Shuttle missions, first on Spacelab-3 in 1985 and subsequently on Atmospheric Laboratory for Applications and Science (ATLAS)-1, -2, and -3 in 1992, 1993, and 1994, made the first infrared solar occultation measurements of N 2 O from space Michelsen et al., 1998;Irion et al., 2002). Also flown on ATLAS-3, in 1994, was the CRyogenic Infrared Spectrometers and Telescopes for the Atmosphere (CRISTA), which used four spectrometers to measure emission in the limb at mid-infrared (4-14 µm) and far-infrared (15-71 µm) wavelengths Riese et al., 1999). The Improved Limb Atmospheric Spectrometer (ILAS) and ILAS-II instruments on the Advanced Earth Observing Satellite (ADEOS) and ADEOS-II, respectively, both measured N 2 O using infrared solar occultation. ILAS made measurements from September 1996 to June 1997 (Kanzawa et al., 2003;Khosrawi et al., 2004), while ILAS-II operated for eight months in 2003 (Ejiri et al., 2006;Khosrawi et al., 2006).
There are currently four satellite instruments in orbit measuring N 2 O. One of these is the Atmospheric Chemistry Experiment Fourier Transform Spectrometer (ACE-FTS) on SCISAT, launched in 2003 . The others are the Sub-Millimetre Radiometer (SMR) on Odin, launched in 2001 (Murtagh et al., 2002;Urban et al., 2005aUrban et al., ,b, 2006, the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) on Envisat, launched in 2002 , and the Microwave Limb Sounder (MLS) on the Aura satellite Lambert et al., 2007), launched in 2004. These are described in more detail below.
The objective of this study is to assess the quality of the ACE-FTS version 2.2 N 2 O data, prior to its public release, through comparisons with coincident measurements. The paper is organized as follows. In Sect. 2, the ACE mission and the N 2 O retrievals are briefly described. Section 3 outlines the methodology used to compare and present the validation results. In Sect. 4, the results of vertical profile comparisons with the SMR, MLS, and MIPAS satellite instruments are discussed. Section 5 focuses on the results of comparisons with data from the ASUR (Airborne SUbmillimeter wave Radiometer) aircraft flights and from the SPI-RALE (SPectroscopie Infra-Rouge d'Absorption par Lasers Embarqués) and FIRS-2 (Far-InfraRed Spectrometer-2) balloon flights. Partial column comparisons with a network of ground-based Fourier Transform InfraRed spectrometers (FTIRs) are presented in Sect. 6. Finally, the results are summarized and conclusions regarding the quality of the ACE-FTS version 2.2 N 2 O data are given in Sect. 7.

The ACE mission and ACE-FTS N O retrievals
The Atmospheric Chemistry Experiment has been in orbit since its launch on 12 August 2003. ACE is a Canadianled satellite mission, also known as SCISAT, which carries two instruments, the ACE-FTS  and the Measurement of Aerosol Extinction in the Stratosphere and Troposphere Retrieved by Occultation (ACE-MAESTRO) (McElroy et al., 2007). Both instruments record solar occultation spectra, ACE-FTS in the infrared (IR), and MAE-STRO in the ultraviolet-visible-near-IR, from which vertical profiles of atmospheric trace gases, temperature, and aerosol extinction are retrieved. The SCISAT spacecraft is in a circular orbit at 650-km altitude, with a 74 • inclination angle , providing up to 15 sunrise and 15 sunset solar occultations per day. The choice of orbital parame-ters results in coverage of the tropics, mid-latitudes and polar regions with an annually repeating pattern, and a sampling frequency that is greatest over the Arctic and Antarctic. The primary scientific objectives of the ACE mission are: (1) to understand the chemical and dynamical processes that control the distribution of ozone in the stratosphere and upper troposphere, particularly in the Arctic; (2) to explore the relationship between atmospheric chemistry and climate change; (3) to study the effects of biomass burning on the free troposphere; and (4) to measure aerosols and clouds to reduce the uncertainties in their effects on the global energy balance Bernath, 2006, and references therein).
ACE-FTS measures atmospheric spectra between 750 and 4400 cm −1 (2.2-13 µm) at 0.02 cm −1 resolution . Profiles as a function of altitude for pressure, temperature, and over 30 trace gases are retrieved from these spectra. The details of ACE-FTS processing are described in Boone et al. (2005). Briefly, a non-linear least squares global fitting technique is employed to analyze selected microwindows (0.3-30 cm −1 wide portions of the spectrum containing spectral features for the target molecule). Prior to performing volume mixing ratio (VMR) retrievals, pressure and temperature as a function of altitude are determined through the analysis of CO 2 lines in the spectra. Forward model calculations employ the spectroscopic constants and cross-section measurements from the HITRAN 2004 line list (Rothman et al., 2005). First-guess profiles are based on ATMOS measurements, but the retrievals are not sensitive to this a priori information.
The ACE-FTS instrument collects measurements every 2 s, which yields a typical altitude sampling of 3-4 km within an occultation, neglecting the effects of refraction that compress the spacing at low altitudes. The altitude coverage of the measurements extends from the upper troposphere to as high as 150 km, depending on the constituent. Note that the altitude spacing can range from 1.5 to 6 km, depending on the geometry of the satellite's orbit for a given occultation. The actual altitude resolution achievable with the ACE-FTS is limited to about 3-4 km, a consequence of the instrument's field-of-view (1.25-mrad-diameter aperture and 650-km altitude). Atmospheric quantities are retrieved at the measurement heights. For the purpose of generating calculated spectra (i.e., performing forward model calculations), quantities are interpolated from the measurement grid onto a standard 1-km grid using piecewise quadratic interpolation. The comparisons in this work use the ACE-FTS VMR profiles on the 1-km grid.
N 2 O is one of the 14 primary target species for the ACE mission. A total of 69 microwindows are used in the version 2.2 ACE-FTS retrievals for N 2 O. They are in the wavenumber ranges 1120-1280, 1860-1951, 2180-2240, 2440-2470, and 2510-2600 cm −1 . The altitude range for the retrieval extends from 5 to 60 km. The primary interfering species in the microwindow set are CO 2 , O 3 , and CH 4 . These interfering species are retrieved simultaneously with N 2 O. The precision of the ACE-FTS N 2 O VMRs is defined as the 1σ statistical fitting errors from the least-squares process, assuming a normal distribution of random errors . We have examined these fitting errors for the ACE-FTS N 2 O profiles used in the comparisons with MLS (Sect. 4.2), and found that the median value is <3% from 5-45 km, increasing to 17% at 60 km, while the mean value is <4% from 5-35 km, and oscillating above this due to some outliers in the individual percent fitting errors. While these fitting errors provide some useful information regarding the precision of the ACE-FTS measurements, they do not represent complete error estimates. Work is in progress to generate full error budgets for all of the ACE-FTS data products, but that information is not currently available.
To date, ACE-FTS N 2 O profiles have been compared with MLS data Lambert et al., 2007;Toohey and Strong, 2007), and partial columns have been compared with those retrieved using the Portable Atmospheric Research Interferometric Spectrometer for the Infrared (PARIS-IR), a ground-based adaptation of ACE-FTS, during the spring 2004 Canadian Arctic ACE validation campaign (Sung et al., 2007).

Validation approach
The comparisons shown in this work include ACE-FTS data from 21 February 2004 (the start of the ACE Science Operations phase) through to 26 February 2007. The coincidence criteria for each correlative dataset were determined in consultation with the teams involved, striving for consistency insofar as possible. The location for each ACE occultation is defined as the latitude, longitude and time of the 30-km tangent point (calculated geometrically), and it is this value that was used in searching for coincidences. Because N 2 O is a long-lived and well-mixed constituent, it was possible to use relatively relaxed temporal and spatial coincidence criteria, thereby providing good statistics for the comparisons. For the global satellite datasets, available for SMR and MLS, the coincidence criteria were defined as ±12 h, ±1 • latitude and ±8 • longitude, as used by Lambert et al. (2007) in the validation of MLS N 2 O measurements. Correlative data for MIPAS were only available for a two-month period in early 2004 for northern mid-and high latitudes, and for these, slightly tighter criteria were defined. For the ASUR aircraft measurements, obtained during several flights, the coincidence criteria were defined as ±12 h and 1000 km. For the statistical comparisons, unless otherwise noted, multiple counting of profiles was allowed, so that if n validation measurements met the criteria with respect to a single ACE-FTS occultation, then these would be included as n coincidences and the ACE-FTS measurement would be counted n times. Balloonbased single profile measurements by SPIRALE and FIRS-2 obtained within ±26 h and 500 km of ACE occultations were included in the comparisons. Finally, for the ground-based FTIRs, the criteria were set at ±24 h and 1000 km for all but two stations (see Sect. 6), to provide a meaningful number of coincidences. Table 1 summarizes the correlative datasets, comparison periods, temporal and spatial coincidence criteria, and number of coincidences for the statistical and individual profile comparisons. Information about the FTIR stations and comparisons is provided in Tables 2 and 3 in Sect. 6.
The SMR, MLS, MIPAS, and FIRS-2 VMR profiles all have vertical resolutions that are similar to that of ACE-FTS, and so no smoothing was applied to these data. These correlative profiles were linearly interpolated onto the 1-km ACE altitude grid. MLS profiles, reported on pressure levels, were mapped onto the 1-km altitude grid of ACE by interpolating in log pressure each MLS profile onto the retrieved pressure profile of the coincident ACE-FTS observation. The aircraftbased ASUR instrument has lower vertical resolution than ACE-FTS, so the ACE-FTS profiles were convolved with the ASUR averaging kernels. The balloon-borne SPIRALE VMR profile was obtained at significantly higher vertical resolution than ACE-FTS, and so was convolved with triangular functions having full width at the base equal to 3 km and centered at the tangent height of each occultation (see Eq. 1 of Dupuy et al., 2008). This approach simulates the smoothing effect of the ACE-FTS field-of-view, as discussed by Dupuy et al. (2008). The resulting smoothed profiles were then interpolated onto the 1-km grid. Finally, for the comparisons with the ground-based FTIR measurements, which have significantly lower vertical resolution, the ACE-FTS profiles were smoothed by the appropriate FTIR averaging kernels to account for the different vertical sensitivities of the two measurement techniques. The method of Rodgers and Connor (2003) was followed and Eq. (4) from their paper was applied, using the a priori profile and the averaging kernel matrix appropriate for each FTIR. Partial columns over specified altitude ranges were then calculated for both ACE-FTS and the FTIRs, as described in Sect. 6, and these were used in the comparisons.
Co-located pairs of vertical VMR profiles from ACE-FTS and each validation experiment (referred to as VAL in text and figures below) were identified using the appropriate temporal and spatial coincidence criteria. Then the following procedure was applied to the vertical profile measurements used in this assessment, with some modifications for the individual profile comparisons (SPIRALE and FIRS-2) and the FTIR partial column comparisons (see Sect. 5 and Sect. 6 for details).
(a) Calculate the mean profile of the ensemble for ACE-FTS and the mean profile for VAL, along with their standard deviations calculated from the individual profiles for each altitude. These mean profiles are plotted as solid lines, with ±1σ as dashed lines, in panel (a) of the comparison figures discussed below. The standard error on the mean, also known as the uncertainty in the mean, is calculated as σ (z)/ √ N (z), where N (z) is the number of points used to calculate the mean at a particular altitude z, and is included as error bars on the lines in panel (a). Note: in some cases, these error bars, as well as those in panels (b) and (c) (see below) may be small and difficult to distinguish.
(b) Calculate the profile of the mean absolute difference, ACE-FTS − VAL, and the standard deviation of the individual differences of all coincident pairs as a function of altitude. (Note that the term "absolute", as used in this work, refers to differences between the compared values and not to absolute values in the mathematical sense.) To do this, the differences are first calculated for each pair of profiles at each altitude, and then averaged to obtain the mean absolute difference at altitude z: where N (z) is the number of coincidences at z, ACE i (z) is the ACE-FTS VMR at z for the ith coincident pair, and VAL i (z) is the corresponding VMR for the validation instrument. This mean absolute difference is plotted as a solid line in panel (b) of the comparison figures below, with ±1σ as dashed lines. Error bars are also included in these figures. For the statistical comparisons involving multiple coincident pairs (SMR, MLS, MIPAS, ASUR), these error bars again represent the uncertainty in the mean. For individual profile comparisons (SPIRALE, FIRS-2), these error bars represent the combined random error, computed as the root-sumsquare error of the ACE-FTS fitting error and the error provided for VAL. (c) Calculate the profile of the mean relative difference, as a percentage, defined using: where MEAN i (z) is the mean of the two coincident profiles at z for the ith coincident pair. Panel (c) of the comparison figures presents the mean relative difference as a solid cyan line. In addition, the relative deviation from the mean is calculated for the statistical comparisons using: where MEAN(z) is the mean of all pairs of coincident profiles at z, which is equivalent to the mean of the average ACE-FTS VMR at z and average VAL VMR at z. This is plotted as the solid dark blue line in panel (c). The relative standard deviation is calculated as the standard deviation on abs (z) from step (b) divided by MEAN(z), and is plotted as dashed lines (±1σ ), with the corresponding relative standard error on the mean included as error bars. In the discussions of relative comparisons below, it is mean (z) that is primarily used; this reduces the impact of very small denominators and noisy data in Eq. (2), which can make rel (z) very large (von Clarmann, 2006).
(d) For the statistical comparisons, calculate the relative standard deviations on each of the ACE-FTS and VAL mean profiles calculated in step (a). For individual profile comparisons, the relative values of the ACE-FTS fitting error and the error for VAL are determined instead. These results are plotted in panel (d) of the comparison figures, with selected values of the number of coincident pairs given as a function of altitude on the right-hand y-axis for the statistical comparisons. For clarity, numbers are not given for all levels.

SMR
The Sub-Millimetre Radiometer (SMR), launched on Odin in February 2001, has four tunable heterodyne radiometers that are used to detect thermal limb emission from atmospheric molecules between 486 and 581 GHz. Odin is in a sun-synchronous, near-terminator orbit at an altitude of ∼600 km and an inclination of 97.8 • (Murtagh et al., 2002). SMR observes a thermal emission line of N 2 O in the limb at 502.3 GHz, and measurements of near-global fields of N 2 O are performed on a time-sharing basis with other observation modes on roughly one day out of three, based on 14-15 orbits per day and 40-60 limb scans per orbit. Algorithms based on the optimal estimation method (Rodgers, 2000) are used for SMR profile retrievals. The latest level 2 version is Chalmers v2.1. N 2 O profile information is retrieved in the stratosphere between ∼12 and ∼60 km with an altitude resolution of ∼1.5 km (in the lower stratosphere, degrading above) and a corresponding single profile precision smaller than 30 ppbv (10-15% below 30 km) .
The horizontal resolution is of the order of 300 km, determined by the limb path in the tangent layer. The satellite motion leads to an uncertainty of the mean profile position of similar magnitude. The SMR N 2 O data are validated in the range ∼15-50 km. The systematic error is estimated to be better than 12 ppbv at altitudes above ∼20 km and increases up to values of 35 ppbv (∼10-15%) below (Urban et al., 2005b), consistent with results obtained in the validation studies showing, for example, a good overall agreement within 4-7 ppbv with data from MIPAS (European Space Agency (ESA) operational processor version 4.61) (Urban et al., 2005a. For this study, only SMR profiles of good quality (assigned Quality flag=0 or 4) were used. The measurement response, provided in the SMR level 2 files for each retrieval altitude, was required to be larger than 0.9 as recommended by Urban et al. (2005a), in order to exclude altitude ranges where a priori information used by the retrieval algorithm for stabilization contributes significantly to the retrieved mixing ratios. The comparisons used coincidence criteria of ±12 h, ±1 • latitude, and ±8 • longitude, and included data from 21 February 2004 to 30 November 2006. This yields 1099 multiple coincident pairs, allowing investigation of the latitudinal behaviour of ACE-FTS-SMR comparisons. In order to exclude extreme outliers, relative differences over 1000% were not included when deriving the mean of the relative differences. This excluded about 2% of the data from the comparison, removing 984 altitudes between 21 and 59 km, and leaving 45 690 altitudes for which the relative differences were less than 1000%.
The results of the comparison between ACE-FTS and SMR profiles between 83 • S and 83 • N (nominally 90 • S-90 • N) are shown in Fig. 1. Excellent agreement is seen between the mean N 2 O VMR profiles (panel (a)) and in the mean absolute differences (panel (b)) between 15 and 50 km, which is the validated altitude range for SMR N 2 O. From 15-50 km, the mean absolute difference is better than −10 ppbv, and is better than −5 ppbv for all but four levels in this altitude range, with ACE-FTS values generally being the smaller of the two, by −2.4 ppbv on average. Comparisons are also shown outside the validated range for SMR (13-15 km and 50-57 km): between 50 and 57 km, the SMR profiles decrease rapidly, leading to larger differences relative to ACE-FTS, varying from −1.5 ppbv at 50.5 km to +1.0 ppbv at 56.5 km. Figure 1c illustrates the difficulty of obtaining useful information from the mean relative difference defined in Eq. (2) for a species such as N 2 O, whose VMR decreases to very small values (typically a few ppbv in the upper stratosphere), and for which there are some coincident profiles whose values for each instrument are of the same magnitude but opposite sign. In some cases, the denominator in Eq. (2) is zero or close to zero, resulting in very large values. These values strongly affect the mean relative difference, although the number of these cases is relatively small; thus these extremely large values are excluded as stated above. However, the mean relative difference is still affected by the noisiness of SMR data in the upper stratosphere and lower mesosphere, as seen in the relative standard deviation on SMR in Fig. 1d. In the upper stratosphere and lower mesosphere where the ACE-FTS N 2 O VMR is small and the SMR N 2 O VMR is noisy, the denominator in the expression is close to half of the SMR VMR and the numerator is close to the SMR VMR, making the ratio approach 200%. As a consequence, the mean relative difference is not a good indicator of the agreement between ACE-FTS and SMR at higher altitudes, although it is better than −7% between 15 and 30 km. Figure 1c thus also includes the relative deviation from the mean, as defined in Eq.
(3); this shows better agreement between ACE-FTS and SMR, to better than −20%, and to −4% on average, between 15 and 40 km. Between 40 and 50 km, the relative deviation from the mean is as large as −44%, with ACE-FTS consistently smaller. Above 50 km, as the SMR N 2 O VMRs decrease, the negative bias of ACE-FTS decreases, becoming a positive bias at 54 km, with a maximum value of +127% at 56.5 km. These large values of the relative deviation from the mean at high altitudes are due to the noisy data in this region, particularly for SMR, as can be seen in the large relative standard deviations on the mean profiles plotted in Fig. 1d.
The data shown in Fig. 1 have been subdivided into five latitude bands in Fig. 2: 60-90 • N, 30-60 • N, 30 • S-30 • N, 30-60 • S, and 60-90 • S. The latitudinal gradients in N 2 O are small at the lower and higher altitudes, as can be seen when comparing the mean profiles for each zonal band. However, a clear latitudinal gradient can be seen in the midstratosphere; for example, at 30 km, the mean ACE-FTS VMR is 155 ppbv for 30 • S-30 • N, dropping in the midlatitudes to 82 (63) ppbv for 30-60 • S (N), and down to 35 ppbv in the polar regions of both hemispheres. Very similar behaviour is seen in the SMR mean profiles. The mean absolute differences are similar in the five bands, with ACE-FTS being consistently slightly smaller than SMR between 15 and 50 km, with the exception of a few levels in each case. These differences are again typically about −2 ppbv, with maximum values of −7 ppbv from 60-90 • N, −18 ppbv from 30-60 • N, −27 ppbv from 30 • S-30 • N (at 15.5 km with only 22 coincident pairs), −11 ppbv from 30-60 • S, and −10 ppbv from 60-90 • S. The mean relative differences remain less than 8% between 15 and 30 km, for all but two levels (−13% at 28.5 km for 60-90 • S and −16% at 29.5 km for 60-90 • N). There is more variability between the latitude bands in the relative deviations from the mean; these are typically better than 5% between 12 and 40 km, with maxima of −24% from 60-90 • N, −29% from 30-60 • N, −8% from 30 • S-30 • N, −40% from 30-60 • S (at 12.5 km with only 25 coincident pairs (not labelled)), and −17% from 60-90 • S. The relative deviations from the mean increase above 40 km, where the relative standard deviations for the individual mean profiles are also seen to reach values of 100% and larger.

MLS
The Microwave Limb Sounder (MLS) was launched on the Aura satellite in July 2004. It is in a sun-synchronous orbit at an altitude of 705 km and an inclination of 98 • , with the ascending node crossing the equator at 13:45 local time . Global measurements are obtained daily from 82 • S to 82 • N, with 240 scans per orbit. Like SMR, MLS measures atmospheric thermal emission from the limb, using seven radiometers to provide coverage of five spectral regions between 118 GHz and 2.5 THz. Volume mixing ratio profiles of N 2 O are retrieved from the thermal emission line at 652.83 GHz using the optimal estimation approach described by . The retrieval is performed on a pressure grid with six levels per decade for pressures greater than 0.1 hPa and three levels per decade for pressures less than 0.1 hPa. The vertical resolution for N 2 O VMR profiles is 4-6 km, the along-track horizontal resolution is 300-600 km, and the recommended pressure range for the use of individual profiles is 100-1 hPa .
For the comparisons in this work, MLS version 2.2 is used. Validation of the v2.2 N 2 O data product is described by Lambert et al. (2007), while Froidevaux et al. (2006) discuss initial validation of MLS v1.5 data products, including N 2 O. The precision of individual v2.2 N 2 O profiles is estimated to be ∼13-25 ppbv (7-38%) for pressures between 100 and 4.6 hPa, while the accuracy is 3-70 ppbv (9-25%) over the same pressure range . Initial comparisons between MLS v2.2 and ACE-FTS v2.2 N 2 O indicated agreement in the mean percentage difference profiles to better than ±5% over 100-1 hPa, with MLS showing a low bias (within −5%) for pressures >32 hPa and a high bias (within +5%) for lower pressures. Analysis of the latitudinal behaviour of the mean absolute difference showed that MLS is consistently smaller than ACE-FTS at most latitudes for pressures between 100 and 32 hPa. Differences were somewhat smaller for ACE-FTS sunrise occultations than for sunset at 46-10 hPa. Lambert et al. (2007) used an initial subset of the MLS v2.2 reprocessed data, which provided 1026 coincidences for the comparisons with ACE-FTS N 2 O. These were obtained over 121 days between September 2004 and October 2006. The present study extends the analyses of Lambert et al. (2007), using data from 16 September 2004 through 26 February 2007, which includes 6876 pairs using coincidence criteria of ±12 h, ±1 • latitude, ±8 • longitude, and mul-tiple counting. The MLS data used in this work are screened based on the recommended parameters: even values of the Status field, Quality values greater than 0.5, Convergence values less than 1.55, positive precision, and pressure levels between 100 and 1 hPa . ACE-FTS data were filtered by removing profiles flagged as Do Not Use (DNU) (see https://databace. uwaterloo.ca/validation/data issues.php). For the period of MLS coincidences, this removed only one DNU occultation. Figure 3 shows the results of the comparison between ACE-FTS and MLS profiles from 82 • S and 82 • N. Excellent agreement is seen over all altitudes, with the mean absolute difference between −3 and +10 ppbv from 15 to 50 km, with differences of 1 ppbv on average over this altitude range, and better than 4 ppbv above 20 km. The mean relative difference exhibits large oscillations, which result from some coincident profiles whose values for each instrument are of the same magnitude but opposite sign, leading to extremely small (or infinitesimal) values of the calculated mean. Dividing by infinitesimal values in (Eq. 2) leads to very large outlying values, as has been confirmed by an examination of all the individual profiles of ACE i (z)−VAL i (z) and of , the latter including some significant outliers. Histograms of the ACE-FTS and MLS N 2 O VMRs, their differences, and their relative means were constructed at particular altitudes, and also confirmed this behaviour. In contrast, the relative deviation from the mean (solid blue line in Fig. 3c) is well behaved, with ACE-FTS agreeing to within ±7% from 15-50 km. Below 24 km, ACE-FTS has a high bias of +5 ppbv on average (10 ppbv maximum), with the relative deviation from the mean better than +3% on average (+5% maximum). Above 24 km, ACE-FTS has a low bias of −1 ppbv on average (−3 ppbv maximum), with the relative deviation from the mean better than −4% on average (−7% maximum). These results are consistent with Lambert et al. (2007), with the exception of the slightly larger relative deviation from the mean between 40 and 50 km. The latitudinal dependence of the ACE-FTS-MLS differences is seen in Fig. 4. In general the results are similar for the five bands, with mean absolute differences better than 10 ppbv between 15 and 50 km, and better than 5 ppbv above 20 km, with the exception of a few of the lowest altitudes seen in the tropics (30 • S-30 • N) and mid-latitudes (30 • -60 • ). The ACE-FTS high bias (better than +10% relative deviation from the mean, except for the lowermost altitudes in the tropics) and low bias (except for the uppermost altitudes in the 30 • -60 • N and 60 • -90 • N) persist below and above 24 km, respectively.

MIPAS
The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) is an infrared limb-sounding Fourier transform interferometer on board Envisat, launched in March 2002 . It acquires spectra over the range 685-2410 cm −1 (14.5-4.1 µm), which includes the vibration-rotation bands of many molecules of interest. It is capable of measuring continuously around an orbit in both day and night, and complete pole-to-pole coverage is obtained in 24 h. From 6 July 2002 until 26 March 2004, MI-PAS was operated at full spectral resolution (0.025 cm −1 ) with a nominal limb-scanning sequence of 17 steps from 68-6 km with 3 km tangent height spacing in the troposphere and stratosphere, generating complete profiles spaced approximately every 500 km along the orbit. However, in March 2004 operations were suspended following problems with the interferometer slide mechanism. Operations were resumed in January 2005 with a 35% duty cycle and reduced spectral resolution (0.0625 cm −1 ). In this section, we describe comparisons between ACE-FTS and MIPAS N 2 O products from the full-resolution mission generated by the ESA operational processor version 4.62 (hereafter referred to as MIPAS ESA) and by the Institut für Meteorologie und Klimaforschung (IMK)/Instituto de Astrofísica de Andalucía (IAA) scientific processor version 9 (hereafter referred to as MIPAS IMK-IAA). Negative values in the ESA data product are set to zero; at altitudes above ∼40 km, where the N 2 O VMR is very small, this can result in a high bias of ESA N 2 O relative to IMK-IAA N 2 O.

MIPAS ESA N 2 O
For the high-resolution mission, ESA has processed pressure, temperature, and six species (H 2 O, O 3 , HNO 3 , CH 4 , N 2 O and NO 2 ). The algorithm used for the level 2 analysis is based on the Optimised Retrieval Model (ORM) (Raspollini et al., 2006;Ridolfi et al., 2000) and uses microwindows at 1233.275-1236.275 cm −1 and 1272.05-1275.05 cm −1 for the N 2 O retrievals. MIPAS-ESA retrievals use the MIPAS dedicated spectroscopic database (see Raspollini et al. (2006) and references therein). Here, MIPAS v4.62 N 2 O data are compared with ACE-FTS version 2.2 data from 21 Febru-ary 2004 to 26 March 2004, when the MIPAS full-resolution mode data ended. The vertical resolution of the MIPAS VMR profiles is 3-4 km and the horizontal resolution is 300-500 km along-track . During the first five months of ACE science operations, only sunsets were measured because of problems with spacecraft pointing at sunrise. Therefore the latitude coverage for this comparison is limited to 20 • N-85 • N for the selected coincidence criteria of ±6 h and 300 km. The intercomparison has been done including all the matching pairs of measurements available in the test period, which yields 141 coincidences (with single counting of profiles). For both ACE-FTS and MIPAS ESA, only profiles associated with successful pressure, temperature and target species retrievals have been considered.
As far as MIPAS ESA errors are concerned, we refer, in general, to the ESA level 2 products for the random error due to propagation of the instrument noise through the retrieval (see Piccolo and Dudhia (2007)), and to results of the analysis carried out at University of Oxford (see data available at http://www-atm.physics.ox.ac.uk/group/mipas/err) for the systematic error. Some of the components, listed in the Oxford University data set as systematic error on the individual profiles, show a random variability over the longer time-scales involved when averaging different MIPAS scans and/or orbits and tend to contribute to the standard deviation of the mean difference rather than to the bias. Taking this into account, for this intercomparison with ACE-FTS, we have considered the error contribution due to propagation of pressure and temperature random covariance into the retrieval of key species VMR (taken from the Oxford University data set) as a randomly variable component and combined it with the measurement noise -using the root-sum-square method -to obtain MIPAS ESA random error. For the MIPAS ESA profiles used in this work, the random error is less than 20% between 15 and 28 km, increasing to 75% at 6 km and to more than 80% above 36 km. Figure 5 shows the results of the comparison. The mean absolute difference is as large as −38 ppbv at 6.5 km, within ±17 ppbv from 8-60 km, within ±10 ppbv above 15 km, with typical values of ±2 ppbv, particularly above 20 km. ACE-FTS has a low bias relative to MIPAS ESA between 6-10 km, 15-20 km, and 32-60 km. For this comparison, the mean relative difference and the relative deviation from the mean are similar and within ±10% (±4 typical) from 8 to 26 km, then increasing steadily to values greater than −20% in the relative deviation from the mean above 35 km, where the standard deviations on the mean ACE-FTS and MIPAS profiles are also large. The pronounced low bias of ACE-FTS compared to MIPAS ESA at higher altitudes is probably due to the negative values in the ESA data product being set to zero. al. (2005). N 2 O is retrieved jointly with CH 4 from its infrared emission lines in the spectral range from 1230 to 1305 cm −1 . Spectroscopic data are taken from the HITRAN 2004 database (Rothman et al., 2005). The vertical resolution in the case of mid-latitude profiles is about 3-4 km up to altitudes around 40 km, and increases to 6 km at an altitude of 50 km. The noise error is less than or equal to 5% up to 50 km. The systematic errors are within 12% up to 30 km and increase up to 33% above 30 km (Glatthor et al., 2005). The latter include the spectroscopic contribution, because although both MIPAS IMK-IAA and ACE-FTS retrieve N 2 O using spectral microwindows near 1200 cm −1 , ACE-FTS also uses a series of microwindows between 1860 and 2600 cm −1 . It is possible that those bands will have different spectroscopic errors, which will not cancel totally even when using the same version of the HITRAN database.
Here we compare N 2 O profiles from ACE-FTS sunset observations with MIPAS IMK-IAA measurements from 21 February 2004 until 25 March 2004.
For these comparisons, we used as coincidence criteria a maximum time difference of 9 h, a maximum tangent point difference of 800 km, and a maximum potential vorticity (PV) difference of 3×10 −6 km 2 kg −1 s −1 on the 475 K potential temperature level.
Over all matches, this resulted in a mean distance of 296 km (±154 km), a mean PV difference of −0.007 ×10 −6 km 2 kg −1 s −1 (±1.49×10 −6 km 2 kg −1 s −1 ) and a mean time difference of −0.2 h. The distribution of the time differences is bi-modal since MIPAS measurements are either at around late morning or early night, while the ACE-FTS observations used here are made during sunset. Thus, for nighttime MIPAS observations, the time difference (MIPAS-ACE) is 4-5 h, while in the case of MIPAS daytime measurements it is about −6 to −8 h. Since N 2 O shows no diurnal cycle in the sounded altitude range and since there is no significant difference between the daytime and nighttime comparisons, in the follow-ing we show the mean differences for day-and night-time matches together, as was done for the comparisons with the MIPAS ESA product.
Nevertheless, stratospheric N 2 O profiles are affected by the subsidence inside the Arctic polar vortex. Thus, in Fig. 6 we show separately the results of the comparisons outside (372 coincidences with single counting of profiles) and inside (114 coincidences) the polar vortex. We determined the matches outside (inside) the vortex by values of PV of <30×10 −6 (>35×10 −6 ) km 2 kg −1 s −1 on the 475 K potential temperature level. Both instruments nicely detect the typical subsidence of inner vortex N 2 O profiles compared to extra-vortex measurements. In general, the differences between MIPAS and ACE-FTS are similar irrespective of their position relative to the vortex. Over the entire 11-60 km altitude range of the comparison, the mean absolute differences are typically −3 ppbv (maximum difference −30 ppbv) inside the vortex and −5 ppbv (maximum difference −42 ppbv) outside. The corresponding relative deviations from the mean are typically −6% (maximum −43%) inside the vortex, and +3% (maximum +48%) outside, with oscillations about 0 as seen in Fig. 6c. Below about 26 km, ACE-FTS is smaller than MIPAS both outside and inside the vortex. The absolute differences are largest below about 18 km, which can be attributed to a high bias in the MI-PAS data that has also been observed in other comparisons. However, the reason for the bump (+20%) at 30 km in the extra-vortex observations is an open issue. The relative deviations from the mean are largest at the highest altitudes, as expected given the very small N 2 O VMRs in that region. The best agreement between ACE-FTS and MIPAS, taking into account both the mean absolute differences and the relative deviations from the mean, is seen between 18 and 35 km. In this region, on average, the mean absolute differences are −1 ppbv (−6 ppbv maximum) and −3 ppbv (−14 ppbv maximum) inside and outside the vortex, respectively, while the  corresponding relative deviations from the mean are −5% (−13% maximum) and −1% (+22% maximum) inside and outside, respectively. It is also interesting to note the very similar variability observed by ACE-FTS and MIPAS, as seen in the standard deviations in Fig. 6d.

ASUR
The Airborne Submillimeter wave Radiometer from the University of Bremen is a passive heterodyne radiometer operating in the frequency range from 604.3 to 662.3 GHz (von Koenig et al., 2000), which measures a number of species, including N 2 O, O 3 , HNO 3 , and CO. Stratospheric N 2 O measurements obtained with the Acousto Optical Spectrometer are used in this study. The total bandwidth of the spectrometer is 1.5 GHz and its resolution is 1.27 MHz; N 2 O is retrieved using the 652.833 GHz line. This receiver is designed to carry out measurements from a high-altitude research aircraft in order to avoid signal absorption by tropospheric water vapor during the observations. ASUR is an upward-looking instrument at a stabilized constant zenith angle of 78 • . The receiver measures thermal emissions from the rotational lines of the target molecule. The shape of the pressure-broadened lines is related to the vertical distribution of the trace gas. The measured spectra are integrated up to 150 s, which leads to a horizontal resolution of about 30 km along the flight path. The vertical profiles of the molecule are retrieved on an equidistant altitude grid of 2-km spacing using the optimal estimation method (Rodgers, 2000). The vertical resolution of the N 2 O measurements is 8-16 km and the vertical range is 18 to 46 km. The precision of a typical single measurement is 10 ppbv and the accuracy is 15% or 30 ppbv, whichever is larger, including systematic uncertainties. Details about the measurement technique and retrieval theory can be found in Kuttippurath (2005). ACE-FTS profiles, 15 ASUR profiles, and 17 co-located observation pairs. Because the vertical resolution of the ASUR profiles is lower than that of the satellite profiles, the ACE-FTS N 2 O vertical profiles were convolved with the ASUR N 2 O averaging kernels, and compared on the 2-km ASUR altitude grid. Figure 7 shows the results from the comparison. The best agreement between the ASUR and ACE-FTS mean absolute difference profiles is between 30 and 46 km, where they agree to within −4.5 ppbv and on average, to within −3 ppbv. Between 18 and 30 km, the maximum difference is +33 ppbv and typical differences are within ±10 ppbv. The ACE-FTS profiles are consistently smaller than ASUR above 22 km, and larger for the comparisons at 18 and 20 km. The relative deviations from the mean are large, reaching a maximum of +82% at 28 km. In general, the ACE-FTS profiles are in reasonable agreement with the ASUR profiles, as the differences are well within the estimated accuracy of ASUR N 2 O, i.e., 30 ppbv.

SPIRALE (Spectroscopie Infra-Rouge d'Absorption par
Lasers Embarqués) is a balloon-borne tunable diode laser absorption spectrometer operated by LPCE (Laboratoire de Physique et Chimie de L'Environment, CNRS-Université d'Orléans) (Moreau et al., 2005), which has participated in several European satellite validation campaigns for Odin and Envisat. It can perform simultaneous in situ measurements of about ten chemical species from about 10 to 35 km height, with a high-frequency sampling (∼1 Hz), thus enabling a vertical resolution of a few meters depending on the ascent rate of the balloon. It has six tunable diode lasers that emit in the mid-infrared from 3 to 8 µm, with beams injected into a multi-pass Heriott cell located under the gondola and largely exposed to ambient air. The 3.5-m-long cell is deployed during the ascent when the pressure is less than 300 hPa, and provides a total optical path between the two cell mirrors of 430.78 m. N 2 O concentrations are retrieved from direct infrared absorption of the ro-vibrational line at 1275.49 cm −1 , by fitting experimental spectra with spectra calculated using the HITRAN 2004 database (Rothman et al., 2005). Measurements of pressure (by two calibrated and temperatureregulated capacitance manometers) and temperature (by two probes made of resistive platinum wire) aboard the gondola allow conversion of the measured number densities into VMRs. Uncertainties on these parameters and on the spectroscopic data (essentially molecular line strength and pressure broadening coefficients) are negligible relative to the other sources of error. The uncertainties in the VMRs have been assessed by taking into account random and systematic errors, and combining them as the square root of their quadratic sum. The random errors (fluctuations of the laser background emission signal and signal-to-noise ratio) and the systematic errors (laser line width and non-linearity of the detector) are very low, resulting in an estimated total uncertainty of 3% for N 2 O volume mixing ratios above 3 ppbv (i.e., at altitudes <26 km) and 6% for mixing ratios below 3 ppbv (>26 km).
The SPIRALE balloon flight occurred on 20 January 2006 between 17:46 UT and 19:47 UT, with a vertical profile obtained during ascent between 13.2 and 27.2 km. The measurement position remained rather constant, with the balloon mean location at 67.6±0.2 • N and 21.55±0.20 • E. The comparison is made with ACE-FTS sunrise occultation sr13151, which occurred 13 hours later (on 21 January 2006 at 08:00 UT) and located at 64.28 • N and 21.56 • E, i.e., 413 km away from SPIRALE. Using the MI-MOSA (Modélisation Isentrope du transport Méso-échelle de l'Ozone Stratosphérique par Advection) contour advection model (Hauchecorne et al., 2002), PV maps in the region of both measurements have been calculated each hour between 17:00 UT on 20 January and 08:00 UT on 21 January on isentropic surfaces, every 50 K from 350 K to 800 K  (corresponding to 12.8-30 km height). These PV fields indicated that SPIRALE and ACE-FTS sampled similar air masses within the polar vortex, with PV agreement better than 10%.
Given the very high vertical resolution (on the order of meters) of the SPIRALE N 2 O profile, it was smoothed by a set of triangular weighting functions of 3 km at the base and interpolated onto the ACE-FTS 1-km grid as discussed in Sect. 3. This smoothing truncated the bottom and the top of the SPIRALE profile by 1.5 km. Figure 8 shows that the ACE-FTS and SPIRALE N 2 O profiles agree to within 17 ppbv (and are typically within ±6 ppbv) in the 15 to 26 km altitude range, with relative differences between −15% and +19% (and ±5% on average) except at the highest altitude, where the difference increases to +49%. ACE-FTS is consistently smaller than SPIRALE between 17 and 24 km.

FIRS-2
FIRS-2 (Far-InfraRed Spectrometer-2) is a balloon-borne Fourier transform infrared spectrometer designed and built at the Smithsonian Astrophysical Observatory. It has contributed to numerous previous satellite validation efforts (e.g., Roche et al., 1996;Jucks et al., 2002;Nakajima et al., 2002;Canty et al., 2006). FIRS-2 detects atmospheric thermal emission in limb-viewing mode from approximately 7 to 120 µm at a spectral resolution of 0.004 cm −1 (Johnson et al., 1995). Vertical profiles of about 30 trace gases are retrieved from the float altitude (typically 38 km) down to the tropopause using a nonlinear Levenberg-Marquardt leastsquares algorithm, with pressure and temperature profiles derived from the 15 µm band of CO 2 . Uncertainty estimates for FIRS-2 contain random retrieval error from spectral noise Table 2. The ground-based FTIR stations contributing N 2 O partial columns for comparisons with ACE-FTS. The location (latitude, longitude, and altitude in m above sea level, a.s.l.) of each station is listed, along with the instrument manufacturer and model, the nominal spectral resolution for the measurements used in this study, the retrieval code and microwindows (MW) used to derive N 2 O partial columns, and references that provide additional details regarding the stations and their measurements. Where multiple MWs are listed, these were fitted simultaneously to retrieve N 2 O. and systematic components from errors in atmospheric temperature and pointing angle (Johnson et al., 1995;Jucks et al., 2002). N 2 O profiles are retrieved using the ν 2 band between 550 and 600 cm −1 .
ACE-FTS is compared with the N 2 O profile obtained during a FIRS-2 balloon flight from Esrange, Sweden on 24 January 2007. The average location of the flight was 67.27 • N, 27.29 • E, with some smearing of the longitude footprint as FIRS-2 was observing to the east. The data were recorded before local solar noon, at 10:11 UT, with a solar zenith angle of 86.6 • . The float altitude was just under 28 km, limiting the maximum measurement altitude to 31 km. The 1σ error on the measured N 2 O VMR varied from 5-14% between 13 and 23 km, and increased steadily above 23 km to 117% at 31 km. The closest ACE-FTS occultation was sr18561, obtained on 23 January 2007, at 08:25 UT, 64.70 • N, 15.02 • E, placing it 481 km away from the location of the balloon flight, and almost 26 h earlier. The FIRS-2 footprint was inside the vortex, Table 3. Summary of the results of the N 2 O partial column comparisons between ACE-FTS and the ground-based FTIR stations. N is the number of coincidences, d and t are the mean distance and time, respectively, between ACE-FTS occultations and the FTIR stations, along with the corresponding standard deviations, z is the partial column altitude range, and DOFS is the degrees of freedom for signal for the FTIR partial columns over the given altitude range of the comparison. The mean relative difference is calculated as 100% × the mean of the N differences (ACE-FTS-FTIR)/FTIR, and is given along with the standard error on the mean, and the standard deviation on the ensemble. For Harestua, Kiruna, and Izaña, the mean relative differences given in brackets are corrected for the 1.3% bias due to the use of HITRAN 2000. while the ACE-FTS occultation was nearer the vortex edge. The FIRS-2 N 2 O profile, reported on a 1-km grid, was interpolated onto the ACE-FTS 1-km grid. Figure 9 shows the results of the comparison. The absolute differences vary from −12 to +30 ppbv over the full altitude range of 13-31 km, with typical values of of +8 and +5 ppbv below and above 20 km, respectively. The largest absolute differences are below 15 km, where FIRS-2 reported low values of N 2 O, although the relative differences have maxima at 25 and 28 km (>±100%). It is possible that FIRS-2 is seeing subsidence within the vortex. Below 20 km, the relative differences are between −6% and +17%, but they increase significantly above 20 km. ACE-FTS has a low bias relative to FIRS-2 between 11 and 13 km, and between 27 and 30 km.

Comparisons with ground-based FTIR measurements
In addition to the vertical profile comparisons described above, ACE-FTS N 2 O measurements have been compared with partial columns retrieved from solar absorption spectra recorded by ground-based Fourier transform infrared spectrometers. Twelve such instruments participated in this study; all are at stations of the Network for the Detection of Atmospheric Composition Change (NDACC) (Kurylo and Zander, 2000) and make regular measurements of a suite of tropospheric and stratospheric species. Many have previously provided data for validation of N 2 O measurements by satellite instruments, such as ILAS (Wood et al., 2002), ILAS-II (Griesfeller et al., 2006), SCIAMACHY (Dils et al., 2006), and MIPAS (Vigouroux et al., 2007). Table 2 lists the stations involved, including their location, the instrument type and spectral resolution, and the retrieval code and microwindows used to retrieve N 2 O. More information about the instruments, the retrieval methodologies, and the measurements made at each of these sites can be found in the references provided in Table 2. The participating stations cover latitudes from 77.8 • S to 78.9 • N, and provide measurements from the subtropics to the polar regions in both hemispheres.
The FTIR measurements require clear-sky conditions, but are made year-round, thus providing good temporal coverage for comparisons with ACE-FTS. The data used here were analyzed using either the SFIT2 retrieval code Rinsland et al., 1998) or PROFFIT92 (Hase, 2000). Hase et al. (2004) found that N 2 O VMR profiles retrieved using these two codes showed very good agreement, with total columns agreeing to within 1%. Both algorithms employ the optimal estimation method (Rodgers, 2000) to retrieve vertical profiles from a statistical weighting between a priori information and the high-resolution spectral measurements. Barbe and Marche (1985) and Sussmann and Schäfer (1997) also showed how information on the vertical distribution of N 2 O could be derived from ground-based infrared spectra. Averaging kernels calculated as part of the optimal estimation analysis quantify the information content of the retrievals, and can be convolved with the ACE-FTS profiles, which have higher vertical resolution. For N 2 O, there are typically 3-4 Degrees Of Freedom for Signal (DOFS, equal to the trace of the averaging kernel matrix) in the total column, and 1-2 in the altitude range coincident with ACE-FTS measurements. Given this coarse vertical resolution, we compare partial columns rather than profiles. All participating sites used microwindows in the 2480-2485 cm −1 region for the N 2 O retrievals, with several sites also including microwindows between 2526 and 2541 cm −1 ; these are listed in Table 2. All sites used spectroscopic data from HITRAN 2004, with the exception of Harestua, which used HITRAN 2000, and Kiruna and Izaña, which used HITRAN 2000+ official updates (equivalent to HITRAN 2000 for N 2 O). Recent analysis using Kiruna data coincident with ACE-FTS has shown that N 2 O columns retrieved with HITRAN 2000 are 1.3% larger than those retrieved with HITRAN 2004. This means that the Harestua, Kiruna, and Izaña retrieved N 2 O columns will be biased positive, so that the ACE-FTS -FTIR differences will be biased slightly negative. Other information required for the retrievals, such as a priori profiles and covariances, treatment of instrument lineshape, and atmospheric temperature and pressure are optimized for each site as appropriate for the local conditions.
The coincidence criteria applied for the FTIR comparisons were ±24 h and 1000 km, with two exceptions. For Kiruna, tighter criteria of ±12 h and 500 km were used in order to minimize the potential influence of the polar vortex, and for Reunion Island, the criteria were ±24 h, ±15 • longitude, and ±10 • latitude, resulting in a maximum distance of 1211 km. These relatively relaxed criteria were necessary to obtain a reasonable number of ACE overpasses for each station (between 5 and 39), and are generally used in FTIR-satellite comparisons, particularly for a well-mixed, long-lived species such as N 2 O. In cases where several ACE occultations met the coincidence criterion for one FTIR measurement at a site, only the occultation that was closest (optimized for the combination of temporal and spatial coincidence) was used. For each FTIR station, the mean (and standard deviation) distance and time differences for these closest ACE occultations are listed in Table 3. Overall, considering all coincident pairs of ACE-FTS and FTIR measurements, the mean distance is 637±253 km and the mean time difference (calculating all time differences as positive) is 9.9±6.2 h. Coincidences from February 2004 through December 2006 were included in the comparisons.
For each station, the ACE-FTS profiles were interpolated onto the FTIR retrieval grid and extended below the lowest retrieved altitude using the FTIR a priori VMR values. This combined profile was smoothed using the FTIR averaging kernels and a priori profile, as described in Sect. 3, to minimize the smoothing error (Rodgers and Connor, 2003). For the calculation of partial columns, atmospheric densities were needed; the density derived from the pressure and temperature profiles used in the FTIR retrievals was used for both the ACE-FTS and ground-based measurements. The lower limit of the altitude range of the partial columns at each station was determined by the ACE-FTS altitudes and the upper limit was determined by the sensitivity of the FTIR mea-surements, defined for a given altitude as the sum of the elements of the corresponding averaging kernel (Vigouroux et al., 2007). This sensitivity was required to be 0.5 or greater, indicating that the measurement contributes at least 50% to the retrieved profile, with the remainder coming from the a priori information (Vigouroux et al., 2007). As seen in Table 3, the lower limits ranged between 8.6 and 17.0 km, while the upper limits ranged between 22.0 and 29.2 km, and the corresponding DOFS for the partial columns varied from 0.7 to 1.8. Over these altitude ranges, the uncertainty on the N 2 O partial columns is on the order of 3% (calculated for the Toronto FTIR data), taking into account the measurement noise error (based on the root-mean-square of the noise in the spectrum and assumed to be uncorrelated so as to eliminate numerical ill-conditioning), the state space interference error (due to unphysical correlations between different parameters in the state space), and the temperature error (Rodgers and Connor, 2003).
The time series of the partial column comparisons are shown for all stations in Fig. 10, along with the relative differences as a percentage of the FTIR partial columns. Agreement is good: within ±20% for all but a few cases, and generally better than this. Table 3 summarizes these results, listing the mean relative differences (calculated as 100% × the mean of the N differences (ACE-FTS-FTIR)/FTIR), the standard errors on the mean, and the standard deviations. For Harestua, Kiruna, and Izaña, the values given in brackets are corrected for the 1.3% bias due to the use of HITRAN 2000, thus providing a consistent set of comparisons based on HITRAN 2004 spectroscopic data. The mean differences lie between (−18.6±6.0)% with σ = 29.6% for Harestua (−17.3% if adjusted to HITRAN 2004) and (+3.8±0.6)% with σ = 2.4% for Ny-Ålesund. There are some problems with oscillations in the N 2 O profiles retrieved at Harestua, which may explain the difference in partial columns; this is currently under investigation. Examination of PV maps from the European Centre for Medium-range Weather Forecasts (not shown) suggests that the large relative differences seen in the Kiruna data between 25 January and 7 February 2005 are probably related to the polar vortex. With the exception of Harestua, all the mean relative differences are within ±5.6% (on the basis of HITRAN 2004 spectroscopic data). Given that the mean statistical fitting error for ACE-FTS N 2 O is <4% between 5 and 35 km, and that the uncertainty on the FTIR N 2 O partial columns is on the order of 3%, the mean relative differences are better than, or comparable to, the combined random error estimate of approximately 5%. These results are also consistent with those of Sung et al. (2007), who found that nine ACE-FTS N 2 O partial columns (from 6 to 100 km) measured within 200 km of the PARIS-IR instrument at Eureka, Canada (80.05 • N, 86.42 • W) in spring 2004 agreed to within 5%. At ten of the twelve FTIR stations, the mean relative difference is negative, suggesting that the ACE-FTS partial columns have a small negative bias. , and the right-hand panel shows the relative differences as a percentage of the FTIR partial columns. Note that the y-axis scales on the left-hand panels differ for some stations. Excellent correlation between the ACE-FTS and FTIR partial columns is seen in the scatter plot of the complete data set. Figure 11 shows a tight correlation, with a correlation coefficient (R) of 0.964. The line fitted to the data has slope 1.01, indicating excellent agreement, and intercept −0.20, indicating a small systematic offset between the two datasets consistent with the small negative bias in ACE-FTS noted above. As might be expected, the largest standard deviations in Table 3 and the largest scatter in Fig. 11 are found for the high-latitude stations which may be viewing different airmasses from ACE and seeing (or not) subsidence within the polar vortex.

Conclusions
In this study, we have undertaken an assessment of the quality of the ACE-FTS version 2.2 N 2 O data prior to its public release. N 2 O is one of the 14 baseline species for the ACE mission, and version 2.2 VMR profiles are retrieved from solar occultation measurements in 69 microwindows between 1120 and 2600 cm −1 . The N 2 O retrievals extend from 5 to 60 km at a vertical resolution of about 3-4 km, and have a precision (1σ statistical fitting error) that is typically less than 5%. ACE-FTS N 2 O profiles from the first three years of the mission have been compared with coincident measurements made by the SMR, MLS, and MIPAS satellite instru-ments, multiple aircraft flights of ASUR, and individual balloon flights of SPIRALE and FIRS-2. ACE-FTS N 2 O partial columns have been compared with measurements by twelve globally distributed ground-based FTIRs. In Fig. 12, the mean absolute differences and the relative deviations from the mean for all of the statistical and individual vertical profile comparisons are shown together, while Table 4 provides a summary of the results of these comparisons.
Excellent agreement between ACE-FTS and SMR (Chalmers v2.1) is seen in the mean absolute differences, which are typically −2 ppbv in the global comparison for 15-50 km, with a maximum difference of −10 ppbv at 15 km. ACE-FTS has a low bias relative to SMR at nearly all altitudes up to 54 km, and a small high bias above this. The relative deviations from the mean are typically −4% between 15 and 40 km, with a maximum difference of −20% at 40 km, increasing to more than 100% above 40 km due to the small values of the mean N 2 O VMR. Little latitudinal dependence is seen in the comparisons.
The comparison with MLS (version 2.2) provided the largest number of coincident profiles and also resulted in very good agreement between the datasets. A high bias in ACE-FTS is seen between 15 and 24 km, typically +5 ppbv, and varying between +2 and +10 ppbv. Over this altitude range, the relative deviation from the mean varies from +2% to +5%. From 24-50 km, ACE-FTS has a low bias of −1 ppbv on average, with the relative deviation from the mean between −7% and 0. There is also little latitudinal dependence in the ACE-FTS-MLS differences, with the mean absolute differences consistently better than 10 ppbv between 15 and 50 km, and better than 5 ppbv above 20 km, with the exception of a few points. The ACE-FTS high and low bias persists below and above 24 km, respectively, for all five latitude bands examined.
ACE-FTS was compared with MIPAS N 2 O profiles generated by the ESA operational processor (version 4.62) and by the IMK-IAA scientific processor (version 9). Comparisons with the ESA product provide validation results to the lowest altitude in this study, extending down to 6 km. At the lowest altitudes, the mean absolute difference is as large as −38 ppbv at 6.5 km, changing to +17 ppbv at 12.5 km and decreasing above. Below 32 km, the mean absolute difference oscillates about 0, with a mean value of −1 ppbv and a standard deviation of 12 ppbv. The corresponding relative deviation from the mean is +3±8%, with values ranging from −12% to +22%. Above 32 km, there is a persistent low bias in the ACE-FTS data, with mean absolute differences of −2 to −1 ppbv, but relative deviations from the mean increasing to −105%. The MIPAS IMK-IAA N 2 O profiles extend from 11-60 km, over which range the mean absolute differences are typically −3 ppbv (maximum −30 ppbv) inside the vortex and −5 ppbv (maximum −42 ppbv) outside. The corresponding relative deviations from the mean are typically −6% (maximum −43%) inside the vortex, and +3% (maximum +48%) outside. The larger negative differences seen below 18 km can be attributed to the known high bias in MI-PAS at those altitudes. The best agreement between ACE-FTS and MIPAS IMK-IAA N 2 O is between 18 and 35 km, where the mean absolute differences are typically −1 ppbv (−6 to +3 ppbv) and −3 ppbv (−14 to +9 ppbv) inside and outside the vortex, respectively, while the corresponding relative deviations from the mean are −5% and −1%. The differences between the comparisons of ACE-FTS with the two MIPAS data products (ESA and IMK-IAA) can be partly explained by the suppression of negative values in the ESA N 2 O VMRs, which leads, at altitudes above ∼40 km, to a high bias of ESA N 2 O relative to IMK-IAA N 2 O, and hence a more negative bias in the ACE-FTS-MIPAS ESA differences.
The fourth set of statistical comparisons involved 17 coincidences between ACE-FTS and ASUR aircraft observations. Between 18 and 30 km, the maximum mean absolute difference is +33 ppbv and typical differences are +2±17 ppbv, while from 30-46 km, the mean differences are better than −4 ppbv and on average, −3 ppbv. Although the relative deviations from the mean are large, the agreement between the ACE-FTS and ASUR profiles is generally well within the 30 ppbv accuracy of ASUR N 2 O measurements.
Comparisons were also made with individual profiles obtained from two balloon flights. The ACE-FTS and SPI-RALE profiles agree to within -17 ppbv (and within −5±7 on average) from 15 to 26 km, with relative differences within 19% except at the highest altitude (+49%). ACE-FTS has a low bias relative to SPIRALE between 17 and 24 km. Larger differences are observed in the comparison with FIRS-2, varying from −12 to +31 ppbv for 13-31 km, with typical values of +8 ppbv below 20 km and +5 ppbv above. Below 20 km, the relative differences are between −6% and +17%, but they increase significantly above 20 km. ACE-FTS is smaller than FIRS-2 between 11 and 13 km, and between 27 and 30 km. The last set of comparisons is with N 2 O partial columns measured by the ground-based FTIRs. Agreement is very good: the mean relative differences are within ±5.6% for eleven of the twelve stations. This mean relative difference is negative at ten stations, suggesting a small negative bias in the ACE-FTS partial columns over the altitude regions compared. Excellent correlation (R=0.964) is observed between the ACE-FTS and FTIR partial columns, with a slope of 1.01 and an intercept −0.20 on the line fitted to the data.
To assess the altitude range over which the ACE-FTS data quality is sufficient to detect the natural variability of N 2 O, we can compare the ACE-FTS VMR statistical fitting errors (i.e., random errors) to the natural variability represented by the relative standard deviations on the mean profiles plotted in panel (d) of Figs. 1 to 7. As noted in Sect. 2, the fitting errors have a median value of <3% from 5-45 km, increasing to 17% at 60 km, with a mean of <4% from 5-35 km, oscillating above this due to some outliers in the individual per-cent fitting errors. Examining Figs. 1d and 3d, as these show results for the largest data sets (1099 and 6876 coincidences respectively), the ACE-FTS relative standard deviations are greater than the median fitting errors at all altitudes. This indicates that the data quality should be sufficient to detect the natural variability of N 2 O over the entire altitude range examined, 5-60 km.
To assess the altitude range over which the ACE-FTS data quality is sufficient to measure the absolute VMR, we must rely on the differences relative to the other instruments, as there is no systematic error budget for the v2.2 data product. However, work is underway to produce an error budget for the next version to be released (v3.0). The mean absolute and relative differences given in Table 4 and Fig. 12, are therefore our best estimates of the absolute data quality.
Overall, the quality of the ACE-FTS version 2.2 N 2 O VMR profiles is good over the entire altitude range of 5 to 60 km, although it is difficult to give an accurate assessment of the data quality at the lowest altitudes due to the small number of comparison datasets available. Between 6 and 30 km, the mean absolute differences for the satellite comparisons lie between −42 ppbv and +17 ppbv, with most within ±20 ppbv. This corresponds to relative deviations from the mean that are within ±15%, except for comparisons with MIPAS near 30 km, for which they are as large as 22.5%.
Between 18 and 30 km, the mean absolute differences are generally within ±10 ppbv, again excluding the aircraft and balloon comparisons. From 30 to 60 km, the mean absolute differences are within ±4 ppbv, and are mostly between −2 and +1 ppbv. Given the small N 2 O VMR in this region, the relative deviations from the mean are therefore large at these altitudes, with most suggesting a negative bias in the ACE-FTS data between 30 and 50 km.
Acknowledgements. Funding for the ACE mission was provided primarily by the Canadian Space Agency (CSA) and the Natural Sciences and Engineering Research Council (NSERC) of Canada. This work was also supported by a grant from the CSA.
Odin is a Swedish-led satellite project funded jointly by the Swedish National Space Board (SNSB), the CSA, the Centre National d'Études Spatiales (CNES) in France and the National Technology Agency of Finland (Tekes).
Thanks to B. Bojkov of the Aura Validation Data Center (AVDC) and the Aura-MLS Data Distribution Team for access to the Aura-MLS dataset (see http://avdc.gsfc.nasa.gov). Work at the Jet Propulsion Laboratory, California Institute of Technology, is carried out under a contract with the National Aeronautics and Space Administration (NASA).
The IMK-IAA team thanks the European Space Agency (ESA) for providing MIPAS level 1 and 2 datasets. We would also like to acknowledge the following people for their work on the MIPAS data relevant to this validation study: T. von Clarmann, U. Grabowski, S. Kellmann, M. Kiefer, A. Linden, M. Milz, and T. Steck from IMK, and B. Funke and M. López-Puertas from IAA.
The ASUR group would like to acknowledge help and support from H. Bremer, A. Kleinböhl and G. Naeveke.
The SPIRALE balloon measurements could only be performed thanks to the technical team (L. Pomathiod, B. Gaubicher, G. Jannet); the flight was funded by ESA and the French space agency CNES for the Envisat validation project; the CNES balloon launching team is greatly acknowledged for successful operations. A. Hauchecorne is acknowledged for making available the use of MIMOSA advection model and F. Coquelet for useful help in the PV calculations and ACE data formatting.
The FIRS-2 work was funded by the NASA Upper Atmosphere Program, and the launch was supported both by the NASA Columbia Scientific Balloon Facility and the Swedish Space Corporation.
All of the ground-based FTIR stations operate within the framework of the Network for the Detection of Atmospheric Composition Change (NDACC, see http://www.ndacc.org), and are nationally funded and supported. The European ground-based FTIR stations have been supported partly by the EU projects UFTIR (http://www.nilu.no/uftir), GEOMON, and HYMN. National support by the Helmholtz Association within the project PEP is acknowledged for Ny-Ålesund and Bremen. The University of Bremen also acknowledges financial support from the ESA project TASTE. The Harestua measurements were also supported by the Swedish Environmental Agency. The support by the local IRF Kiruna staff is highly appreciated. The Belgian contributions to the present effort were partly supported by the ProDEx projects ACE, CINAMON and Envisat Database. Thanks are extended to the "International Foundation High Altitude Research Stations Jungfraujoch and Gornergrat" (HFSJG, Bern, Switzerland) for hosting the Liège FTIR laboratory and for providing accommodation for the observers at the Jungfraujoch site. The National Center for Atmospheric Research is supported by the National Science Foundation. The NCAR FTIR observation program at Thule, Greenland is supported under contract by NASA. Work at the Toronto Atmospheric Observatory was supported by NSERC, CSA, CFCAS, ABB Bomem, the Ontario Research and Development Challenge Fund, the Premier's Excellence Research Award and the University of Toronto.
Edited by: A. Richter