Validation and data characteristics of methane and nitrous oxide profiles observed by MIPAS and processed with Version 4.61 algorithm

The ENVISAT validation programme for the atmospheric instruments MIPAS, SCIAMACHY and GOMOS is based on a number of balloon-bone, aircraft and ground-based correlative measurements. In particular the activities of validation scientists were coordinated by ESA within the ENVISAT Stratospheric Aircraft and Balloon Campaign of ESABC. As part of a series of similar papers on other species [this issue] and in parallel to the contribution of the individual validation teams, the present paper provides a synthesis of comparisons performed between MIPAS CH 4 and N 2 O profiles produced by the current ESA operational software (Instrument Processing Facility version 4.61 or IPF v4.61) and correlative measurements obtained from balloon and aircraft experiments as well as from satellite sensors or from ground-based instruments. The MIPAS-E CH 4 values show a positive bias in the lower stratosphere of about 10%. In case of N 2 O no systematic deviation with respect to the validation experiments could be identified. The individual used MIPAS data version 4.61 still exhibits some unphysical oscillations in individual CH 4 and N 2 O profiles caused by the processing algorithm (with almost no regularization). Taking these problems into account, the MIPAS CH 4 and N 2 O profiles are behaving as expected from the internal error estimation of IPF v4.61.


Error budget
The MIPAS L2 products contain estimates of random error derived from the propagation of the radiometric noise through the retrieval. The noise itself varies with time, steadily rising between decontamination events, but its contribution to the L2 random 5 error also depends on the atmospheric temperature, which controls the total radiance received. Hence, for all species, the random error varies latitudinally/seasonally with atmospheric temperature, with a superimposed time dependence on decontamination events.
The main source of the random error of the ESA L2 Offline MIPAS profiles is the 10 noise error due to the mapping of the radiometric noise in the retrieved profiles. This predicted random error is proportional to the NESR (Noise Equivalent Spectral Radiance) and inversely proportional to the Planck function (therefore atmospheric temperature), but it does not directly depend on the VMR of the gases. In the ESA retrieval processing, first, temperature and tangent pressure are retrieved 15 simultaneously, then the 6 "key species" (H 2 O, O 3 , NO 3 , CH 4 , N 2 O and NO 2 ) VMR profiles are retrieved individually in sequence. The effects of temperature and pressure errors on the VMR retrievals are taken into account in the predicted random error estimation. The MIPAS noise error is the covariance matrices given in the MIPAS level 2 prod-20 ucts. The systematic errors are described in Dudhia et al. (2002) and can be find in the Oxford web page (www.atm.ox.ac.uk/group/mipas/err) where errors are divided into systematic errors with random variability and in purely systematic errors, with one exception: the altitude shift has been taken as a systematic error with random variability. The total error is the root sum square of systematic error and random error com-25 ponents. The random errors take into account the propagation of instrument noise through the retrieval. The definition of systematic error here includes everything which is not propagation of the random instrument noise through the retrieval. However, to 18048 Introduction EGU use these errors in a statistically correct manner for comparisons with other measurements is not straightforward. Each systematic error has its own length/time scale: on shorter scales it contributes to the bias and on longer scales contributes to the SD of the comparison. Fortunately, two of the larger systematic errors (propagation of error due to pressure and temperature retrieval, and spectroscopic database errors) can be 5 treated properly. The p/T propagation error is uncorrelated between any two MIPAS profiles (since it is just the propagation of the random component of the p/T retrieval error through the VMR retrieval). Spectroscopic database errors are constant but of unknown sign, so will always contribute to the bias of any comparison. Of the other significant errors, the calibration-related errors should, in principle, be uncorrelated be- 10 tween calibration cycles however analysis of the residuals suggests that these errors are almost constant so could be included in the bias. Figure 1 presents for CH 4 and for N 2 O the vertical distribution of random, systematic and total errors for a global composite of the five reference atmospheres, with twice the weight given to results from the polar winter case.

Validation experiments and analysis methods
The correlative measurements for MIPAS CH 4 and N 2 O profiles considered here (see Table 1) have been obtained from a large number of in situ and remote sensing instruments carried out from ground, balloon, aircraft and satellite platforms participating in the ENVISAT Stratospheric Aircraft and Balloon Campaign (ESABC) coordinated by 20 (Wursteisen, 2003). The coincidence criteria recommended for the intercomparison were set to 300 km and 3 h. However, some individual research groups involved in the validation work presented here have used more relaxed criteria whenever justified on the basis of previous experiences. Representation of CH 4 and N 2 O volume mixing ratio (VMR) vertical pro-25 files is preferred versus pressure than altitude. Another requirement to be considered for intercomparison of polar winter measurements has been a recommended maximum 18049 smoothing is applied.
The use of trajectory calculations to increase the number of coincidences (with the same baseline collocation criteria adopted for direct coincidences) has been used.

Comparison with validation balloon campaign data
The balloon experiments for which CH 4 and/or profiles N 2 O (as well as the correspond- 10 ing MIPAS data) were available, include FTIR remote sensing instruments operating in limb thermal emission such as IBEX (Biancini et al., 2003) in the far-infrared and MIPAS-B (Friedl-Vallon et al., 2004) or in solar occultation such as LPMA (Camy-Peyret et al., 1995) as well as in situ samplers such as the Bonbon cryosampler (Engel et al., 1998) and in situ diode laser spectrometers such as SPIRALE (Moreau et al., 2005). 15 They are discussed in sequence, a priority being given to the balloon experiments of the 2002 campaigns for which IPF v4.61 MIPAS CH 4 and N 2 O profiles are available.

IBEX
The IBEX (Infrared Balloon Experiment, Istituto di Fisica Applicata "Nello Carrara", IFAC CNR, Firenze, Italy) (Bianchini, 2003) is a far-infrared Fourier transform spec-20 trometer, which was flown during the first campaign of ESABC from Sicily 38 N,12 E) over the Mediterranean to Spain on 28-29 July 2002. Because there was no coincidence between the period when IBEX was at float and an overpass of EN-VISAT, the data used for comparison was taken from MIPAS-E limb scans performed over the Mediterranean within a ±1 day window covering the IBEX measurements. combined precision (blue line) and combined total (green line) errors. The data plotted in Fig. 2 shows a reasonable agreement in the mid stratosphere with some dispersion of the balloon data. The MIPAS-E values in the very lower stratosphere present a positive bias with respect to IBEX values, a situation which is also seen in other correlative measurements (see below). parameters that MIPAS is covering. Essential for the balloon instrument is the sophisticated line of sight stabilization system, which is based on an inertial navigation system and supplemented with an additional star reference system. Averaging several spectra during one single elevation angle yields to a reduction of the noise equivalent spectral radiance (NESR) and therefore to an improvement of the signal to noise ra-20 tio. The MIPAS-B data processing including instrument characterization is described in Friedl-Vallon et al. (2004) and references therein. Retrieval calculations of atmospheric target parameters were performed with a least squares fitting algorithm (using analytical derivative) of spectra simulated by the Karlsruhe Optimized and Precise Radiative transfer Algorithm (KOPRA; Stiller et al., 2002;Höpfner et al., 2002). A Tikhonov-

25
Phillips regularization approach constrained with respect to the shape of an a priori profile was adapted. The resulting vertical resolution lies typically between 2 and 3 km and is therefore comparable to the vertical resolution of MIPAS-E. Target  sequence (sequence S of flight 11) is given in Fig. 3.
The mean deviations between MIPAS-B and MIPAS for all balloon flights together are shown in Fig. 4. The differences MIPAS-B minus MIPAS-E v4.61 have been compared with the combined (root sum squares) error and demonstrate the impact of the remaining "oscillations": the mixing ratio values of MIPAS-E around 100 and 300 hPa 20 are clearly overestimated and underestimated, respectively for both species.

Bonbon
The flight of the cryosampler Bonbon (Engel et al., 1998)  these two flights a detailed analysis of the vertical structure of the stratosphere based on the N 2 O and CH 4 measurements obtained has been made by Huret et al. (2006). Figure 6 presents the comparison of SPIRALE and MIPAS-E profiles, for CH 4 and N 2 O respectively, measured on 21 January 2003. In order to take into account the large difference between SPIRALE and MIPAS-E vertical resolution (150 m and 3 km 15 respectively), the CH 4 and N 2 O SPIRALE profiles have been smoothed using MIPAS-E averaging kernels. A good agreement is obtained from 12 km up to 24 km. Above 24 km for CH 4 the absolute difference between the two set of data is increasing. It can be noticed that the SPIRALE instrument has intercepted a thin PV filament at 28 km, in this layer the volume mixing ratios of each species is enhanced (Huret et al., 2006).

20
This thin layer is not observed by MIPAS because of its coarser vertical resolution.
Since MIPAS was not operating on 2 October 2002 when SPIRALE was launched for its second flight, the comparison is only possible with backward trajectories starting from MIPAS measurements on 26, 27 and 28 September and ending at the SPIRALE location on 2 October. The SPIRALE flight took place in pre-vortex formation condi-25 tions when air mass exchanges between tropics region and polar region occur. The abundance of long lived species is largely modified by these exchanges leading in particular to non monotonic profiles. Air mass origin discussed using N 2 O-CH 4 correlation 18053

EGU
in Huret et al. (2006) are very different as a function of altitude. Then before comparing the MIPAS data to SPIRALE measurements we must check for the consistency of dynamical conditions using a potential vorticity analysis. This is performed with the MIMOSA PV contour advection model (Hauchecorne et al., 2002 et al., 1995). Its high spectral resolution and sensitivity allow the retrieval of vertical profiles of trace species having stratospheric mixing ratios as small as 0.1 ppbv. The measurements of three flights have been used for the validation of MIPAS CH 4 and N 2 O vertical profiles. As an example of LPMA measurements, during the flight performed on 24 March 2004, the Sun was acquired above a rather elevated cloud deck 5 at about 10 km. The first complete interferograms (after proper setting of the gains of the preamps for each channel) have been obtained just above 10 km. From that point on, the primary pointing system, the heliostat, the interferometer and all the ancillary equipments performed nominally during ascent, float and occultation up to loss of sun, again behind the high cloud cover (∼10 km). The 180 recorded spectra show sufficient 10 absorption by CH 4 and N 2 O for precise retrieval in the appropriate micro-windows. The LPMA flight observations started at 14:31 UT (the balloon was at an altitude of 10 km during its ascent), the float was reached at 16:03 UT and occultation measurements (conventionally distinguished from ascent measurements as pertaining to negative solar elevation angles) have been recorded until loss of Sun at 17:29 UT.

15
The slant column density (SCD) retrieval of N 2 O, CH 4 , O 3 , NO 2 , NO, HNO 3 , H 2 O, HCl, CO 2 and ClONO 2 is performed simultaneously using a multi fit of 11 micro-windows. The target micro-window for N 2 O and CH 4 are around 1240.38 to 1243.65 cm −1 . In addition CH 4 appears as an interfering absorber in the O 3 , NO 2 , HCl and HNO 3 target windows whereas N 2 O contribute in the HNO 3 target window. These 20 contributions need to be included for a reliable SCD retrieval. Based on absorption line parameters from HITRAN 2004 (Rothman et al., 2005) and a reasonable a priori guess for the trace gas profiles, a forward model calculates synthetic spectra which are fitted to the measured ones by a non-linear Levenberg-Marquardt algorithm. The calculation of the synthetic spectra relies on atmospheric parameters taken from nearby 25 radiosonde launches and climatological and meteorological model data. Fitting parameters include a polynomial of up to third order, a small additive wavenumber shift and several parameters to adjust the instrumental line shape (ILS). All auxiliary ILS parameters are determined separately in various test runs and finally set to a fixed value for 18055 EGU all spectra during a balloon flight.
The error bars comprise the statistical error of the fitting routine (1σ), the uncertainty in determining the instrumental line shape, the error coming from the ambient atmospheric parameters and their impact on the spectroscopic parameters and the stated error bars of the latter (in total 10% systematic contribution for both gases). Each spec-5 trum yields an N 2 O and CH 4 SCD according to the specifications described above. Vertical trace gas profiles are then inferred during balloon ascent and solar occultation. For more details on LPMA retrieval and data analysis see Payan et al. (1998Payan et al. ( , 1999 and Dufour et al. (2005).
The vertical mixing ratio profiles of CH 4 and N 2 O and the corresponding errors have 10 been plotted as a function of pressure for the MIPAS IPF v4.61 together with the balloon profile. An example is given in Fig. 8

EGU
HITRAN spectral line data for the radiative transfer calculation in the forward model, and this error is estimated to be below 10%. Effects such as non-LTE, uncertainties in the pointing of the instrument, horizontal atmospheric inhomogeneity along the line of sight, or the error of the used correlation can cause further errors, which were considered of minor importance. As the dominating error sources are independent, they sum 5 up to below 11%. The N 2 O profiles of MIPAS-STR are plotted in Fig. 9 as a function of tangent pressure, together with the coinciding profiles of MIPAS-E. The vertical mixing ratio profiles of CH 4 are plotted in Fig This kind of problems can not be explained only by the omitted regularization.

SAFIRE-A
SAFIRE-A (Spectroscopy of the Atmosphere by using Far-Infrared Emission -Air-20 borne, IFAC-CNR, Firenze, Italy) is also a limb viewing FT spectrometer, but measures the far infrared (10-250 cm −1 ) atmospheric emission in narrow bands (1-2 cm −1 ). Its characteristics and performance are described by Bianchini et al. (2004).
The geolocation of the SAFIRE-A limb scans and of the corresponding MIPAS-E tangent points is presented in Fig. 11  EGU a positive bias with respect to correlative measurements as already noticed for other comparisons in the UT/LS.

ASUR aboard the German Falcon
ASUR is a passive heterodyne receiver operating in the frequency range of 604.3 to 662.3 GHz (Mees et al., 1995;von Koenig et al., 2000). It is equipped with two 5 spectrometers, an Acousto Optical Spectrometer (AOS) and a Chirp Transform Spectrometer (CTS). Stratospheric measurements performed with the AOS are used in this comparison study. The total bandwidth of the AOS is 1.5 GHz and its resolution is 1.27 MHz. In order to avoid absorption by tropospheric water vapor, observations are carried out aboard a research airplane. The instrument looks upward at a stabilized 10 constant zenith angle of 78 • . ASUR measures thermal emission around rotational lines of the target molecule. The shape of the pressure broadened line is related to the vertical distribution of the trace gas. Measured spectra are integrated up to 150 s, which leads to a horizontal resolution of about 30 km along the flight path. Vertical profiles of the molecule are retrieved in an equidistant altitude grid of 2 km spacing using the 15 Optimal Estimation Method (Rodgers, 1976). Vertical resolution of the N 2 O measurements is about 8-16 km and vertical range is from 16 to 45 km. The precision of a single measurement is 10 ppb and the accuracy is 15% or 30 ppb, whichever is higher, including systematic uncertainties. Details about the measurement technique and retrieval theory can be found in Bremer et al. (2002) andin Kuttippurath (2005). 20 The ASUR N 2 O measurements performed during the SCIAVALUE (Sciamachy Validation and Utilization Experiment) campaign (Fix et al., 2005) are used here. Data from 14 selected ASUR measurement flights during the campaign are analyzed. Details about the flights are given in Table 2. The MIPAS off line IPF v4.61 are considered. A criterion that the ASUR measurements are within +/-1000 km and in +/-12 h around 25 the satellite observations is chosen for the comparison between datasets. This criterion resulted in 323 coincident measurements (from 14 flights) with the IPF data. The MIPAS volume mixing ratios are convolved with the ASUR N 2 O averaging kernels to 18059 Figure 13 shows the results from the comparison between ASUR and IPF v4.61 profiles. There are 101 coincident measurements in the tropics, 38 in mid-latitudes and 184 in high-latitudes. The differences range from -18 to 48 ppb in the tropics, 2 to 31 ppb in the mid-latitudes and -10 to 13 ppb in the high latitudes. The deviation is largest at 24-28 km altitude for all latitude bands, in which the tropical profile shows the 10 highest deviation of about 48 ppb. It is found that the MIPAS profiles underestimate the ASUR VMRs in the altitude range 25-30 km and overestimate the ASUR values above 34 km. However, agreement between the profiles appears to be very good at mid and high latitudes above 30 km altitude.
In comparison with the MIPAS datasets in the tropics and mid-latitudes, there seems 15 to be a systematic difference. Temporary atmospheric variations and the reduced altitude resolution of ASUR can hardly explain these systematic deviations. We note that the N 2 O values in the tropical lower stratosphere retrieved from ASUR measurements seem relatively high. Comparisons with Odin/SMR have also shown this particular feature of ASUR N 2 O retrievals (Urban et al., 2005). However, for mid and high latitudes 20 and for the lower values of N 2 O, agreement between ASUR and MIPAS profiles is very good. This was also true for comparison between ASUR and SMR profiles (Urban et al., 2005). The differences in these latitude and altitude regions are well within the ASUR error bars. ). In addition to total columns, low vertical resolution profiles are retrieved from the spectra by using the Optimal Estimation Method of Rodgers (2000) in the inversion programs. For the Kiruna data, the inversion code used is PROFFIT (PROFile FIT) (Hase, 2000(Hase, , 2004, based on the forward model KO-PRA (Karlsruhe Optimized Precise Radiative transfer Algorithm) (Höpfner et al., 1998).

15
For all other stations, the retrievals have been performed using the SFIT2 algorithm (Pougatchev et al., 1995a, b;Rinsland et al., 1998). The PROFFIT and SFIT2 codes have been cross-validated successfully by Hase et al., 2004. In all cases, the synthetic spectra were calculated using daily pressure and temperature data of the National Centers for Environmental Prediction (NCEP). All retrieval parameters (spectral 20 microwindows, spectroscopic parameters, instrumental line shape, a priori information, and model parameters) have been optimized independently for each station. For the N 2 O retrievals, all stations used the spectroscopic line parameters from the HITRAN 2000 database including official updates through 2001 (Rothman et al., 2003). For the CH 4 retrievals, the northern hemisphere stations used the HITRAN 2000 database, 25 while the southern hemisphere stations used the HITRAN 2004 database (Rothman et al., 2005).
The FTIR products are low vertical resolution profiles: the degrees of freedom for 18061 The upper altitude limit for the comparisons is chosen taking into account the groundbased FTIR sensitivity which is reasonable up to around 30 km for both molecules at all stations. spatial distances of, respectively, ± 3 h and ± 300 km maximum at the MIPAS nominal tangent height of 21 km. For Wollongong, the number of coincidences found using these criteria is very small, so we decided to include the results of comparisons using relaxed coincidence criteria of ± 4 h and ± 400 km distance. When the spatial variability of the target gas is high, such as in winter-spring at 20 high latitude stations, the standard deviations of the comparisons would become large and would not represent the agreement between both measurements. This is due to 1) the collocation error of the air masses, and 2) the horizontal smoothing error which corresponds to the gradient of the target gas within the instruments' line of sight (Cortesi et al., 2006, Sect. 4;von Clarmann et al., 2006

EGU
(summer-autumn for high latitude stations).
To avoid a possible geometric altitude error in the MIPAS data, the comparisons between MIPAS and FTIR measurements are made on a pressure grid. The MIPAS profiles are degraded to the lower vertical resolution of the ground-based FTIR measurements by smoothing the MIPAS profiles with the averaging kernels of the ground-5 based FTIR data. Thus, smoothed MIPAS profiles have been used in the comparisons of profiles and partial columns.
The statistics of the profiles and partial columns comparisons are given (in percentages) in the tables and figures of the next sections. The relative differences between MIPAS and FTIR products are calculated by taking the mean absolute difference be-10 tween MIPAS and FTIR data (MIPAS-FTIR), divided by the mean FTIR value. The means (M) of the statistical comparisons (i.e., the biases) will be compared to the 3σ standard errors on the means (SEM) to discuss their statistical significance. The SEMs are calculated as 3×STD/ √ N, N being the number of coincidences, and STD the standard deviation of the differences. The precision of the instruments will also 15 be discussed by comparing the standard deviations (STD) of the differences with the random error on the difference MIPAS-FTIR.
The random error covariance matrix of the difference MIPAS -FTIR has been evaluated, using the work of Rodgers and Connor (2003) for the comparison of remote sounding instruments and of Calisesi et al. (2005) for the re-gridding between the MI-20 PAS and the FTIR data (see Vigouroux et al., 2006 for more details). The FTIR random error budget has been estimated for a typical measurement at Kiruna (F. Hase, IMK, private communication). There are different contributions to the MIPAS random error covariance matrix. The noise error contribution is the covariance matrix given in the MIPAS level 2 products. The mean of these covariance matrices for the coincident 25 MIPAS profiles has been used as the noise error contribution to the MIPAS random error matrix. Following the approach adopted for MIPAS comparison with other satellite measurements, the systematic errors with random variability have been added to the MIPAS random error budget (see Sect. 2). Table 3 give for every station, the height region of the partial columns (in pressure units), the mean (M) and the standard deviation (STD) of the partial column relative differences, along with the number N of coincident pairs, the estimated random error on the partial column differences and the 3σ standard error on the mean (SEM).

5
From Table 3, we see that there is a statistically significant positive bias in the relative differences of partial columns for all the stations except Ny-Alesund and Arrival Heights. Due to the high standard deviation at Arrival Heights during the whole period of comparison, the bias is not significant. If we limit the comparisons to the summerautumn period, the bias at Arrival Heights appears to be also significant. Table 3 also shows that the statistical standard deviation (i.e. the dispersion) is usu-15 ally slightly larger than the estimated random error which is probably due to collocation and horizontal smoothing errors. We see clearly from Fig. 14 that the standard deviations are higher during winter-spring periods for the high latitude stations, which is confirmed by the statistics in Table 3 for reduced time periods.
In the profile comparison plot (Fig. 15), the means and the standard deviations of 20 the relative differences between the ground-based FTIR and the MIPAS CH 4 profiles at each station in percentage versus pressure are given. The combined random error associated with the mean difference is represented by the shaded grey area. The 3σ standard error on the mean is also reported to facilitate the discussion of the statistical significance of the observed bias. The black solid lines in each plot mark the pressure 25 levels adopted as the lower and upper limits for the calculations of partial columns. The CH 4 difference profiles confirm what has been seen for the partial columns comparisons: a significant positive bias is observed at Jungfraujoch, Kiruna, Lauder

EGU
and Arrival Heights in the lower stratosphere. At Wollongong, the bias is maximum in the middle stratosphere. At Ny-Alesund, no bias was seen in the partial columns.
We can see, however, in Fig. 15 that a positive bias exists in the lower stratosphere but is compensated by a negative bias in the middle and upper stratosphere. These oscillations in the difference of profiles are due to the FTIR products at Ny-Alesund.

5
The constraints on the a priori information (Rodgers, 2000) are probably too small, leading to oscillations in the profiles. This would also explain the larger (and probably non realistic) degrees of freedom for signal at Ny-Alesund, given in Sect. 6.1.

N 2 O comparisons
The FTIR datasets used here are the same ones as used already by Vigouroux et 10 al. (2007) for the validation of MIPAS N 2 O v4.61 products, for all the stations except Ny-Alesund. But the coincidence criteria were less strict, which was compensated by the use of the data assimilation system BASCOE. Here we show results obtained with the same criteria as adopted elsewhere in the present paper (±3 h; ±300 km).
Considering the means and their statistical 3σ standard errors (SEM) given in Ta-15 ble 4, there is no statistically significant bias in the relative differences of partial columns for the Kiruna, Jungfraujoch, Wollongong, and Lauder stations. A statistically significant negative bias is seen for the highest latitude stations: Ny-Alesund (-10.1%) and Arrival Heights (-8.5%). For Arrival Heights, we can see in Fig. 16 and Table 4 that the bias is more pronounced during the local spring period, and that it is no longer 20 significant when the comparisons are limited to summer-autumn. For Ny-Alesund, the number of coincidences in the limited time period (3) is too small to draw any significant conclusions. From Table 4, it can be seen that the statistical standard deviations are within the estimated random error for Ny-Alesund, Jungfraujoch,  longong, we see in Fig. 16 that the larger standard deviation for the statistics (with coincidence criteria of ±3 h; ±300 km) is due to one single coincidence only, on the 1 March 2003. Thus, results are better for the relaxed criteria. For Arrival Heights, 18065 EGU considering the whole period, the statistical standard deviation is also larger than the estimated random error, but this is no longer the case in the reduced time period. Indeed, we see in Fig. 16 that the dispersion is larger during local spring for the three highest latitude stations. Figure 17 confirms that, except at Ny-Alesund and to a smaller extent at Lauder, 5 there is no statistically significant bias in N 2 O comparisons in the lower stratosphere where the N 2 O concentration is the highest. At higher altitude, a high positive bias is seen at Wollongong, and a small negative one at Kiruna.

Conclusions
For CH 4 comparisons, we obtain a statistically significant positive bias of 5 to 11% 10 between MIPAS and FTIR lower-middle stratosphere partial columns, and a standard deviation of 4 to 7.5%, when the high variability period (winter-spring) for high latitude stations is not taken into account. For N 2 O comparisons, no statistically significant bias is seen between MIPAS and FTIR lower-middle stratosphere partial columns, and the standard deviation is between 15 2.5 and 6.8%, when the high variability period (winter-spring) for high latitude stations is not taken into account.
When the winter-spring period is included in the comparisons for the high latitude stations, we can reach standard deviations of 9 and 15%, for N 2 O and CH 4 respectively, probably due to collocation and horizontal smoothing errors.Several papers submitted   Satellite-satellite intercomparisons are another method to assess the quality of a new space instrument, once another one, considered to be already validated by independent measurements, is stable and is producing reliable profiles. This is the case for 5 the Halogen occultation Experiment (HALOE on board UARS) providing since 1991 vertical mixing ratio profiles of CH 4 (Park et al., 1996) (and several other species) in the full stratospheric range using solar absorption gas correlation radiometry. The Institute of Environmental Physics (IUP) of University of Bremen has been using HALOE version v19 data for comparison with coincident MIPAS-E measurements. 10 No averaging kernels have been applied because of similar vertical resolution between the two satellite instruments (3 km for MIPAS, 2-3 km for HALOE) The following accuracy/precision are given by Park et al. (1996) : (a) at 0.3 and 50 hPa acccuracy between 6 and 15%, precision between 0 and 14%, (b) at 0.1 and 100 hPa acccuracy between 6 and 27%, precision between 0 and 27%. The validation study per-15 formed by Park et al. (1996) shows an agreement within 10 to 15% of HALOE profiles with balloon-borne (FTS, cryosampler), rocket (cryogenic whole air sampler) and satellite/shuttle (ATLAS1+ATLAS2/ATMOS) measurements from 0.3 to 100 hPa. Figure 18 displays comparisons for a high latitude profile and a tropical profile in good coincidence (distance between HALOE and MIPAS tangent point less than 300 km, EGU systematic errors. The combined random error is given by the root sum square of the averaged random error profiles of the two instruments. The global average of the percentage difference between MIPAS and ODIN-SMR N 2 O values, calculated over the full set of collocated measurements is presented in Fig. 21 (absolute difference) and in Fig. 22 (scaled difference), where the mean profile 5 of the relative difference between MIPAS and ODIN-SMR with respect to the latter is plotted along with error bars representing the standard error on the mean (1σ).
The MRD values are within ±10% from approximately 100 to 10 hPa, with MIPAS mostly underestimating the N 2 O content; the resulting bias is anyhow constantly lower than the combined systematic error in this pressure range. Outside this interval, both in the upper stratospheric layers and in the UTLS, the average N 2 O VMR values retrieved by ODINSMR become increasingly higher than those measured by MIPAS.
This discrepancy could be due to a lack of statistics, not so many points as it can be seen from the standard deviation at altitudes below 60 hPa. We can notice that in the retrieval process, altitudes below 60 hPa might include mainly the a priori information.  Fig. 23), mid-and high latitudes (28- 69 • N, Fig. 24) and arctic latitudes (69-90 • N, Fig. 25). Further the data shown are restricted in time to March/April 2003and November 2002 in samples consisting of 3332 to 13829 values, respectively. These restrictions have been applied to obtain the best possible temporal and latitudinal agreement with the ATMOS data used to derive previous regression curves (Michelsen et al., 1998a,b).
The ATMOS data were obtained on three Spacelab-missions: 25 March to 2 April 1992 (ATMOS-1), 8-16 April 1993 (ATMOS-2) and 4-12 November 1994 (ATMOS-3). Polynomial fits were performed for data from the northern hemispheric tropics, midand high latitudes and from the Arctic vortex. The tropical polynomial was fitted to data 15 obtained on ATMOS-1 and ATMOS-3 between 0 and 10 • N, the mid-and high latitude polynomial to data from AT-3 from 28 to 69 • N and the Arctic vortex polynomial to data obtained on ATMOS-2.
Generally, the MIPAS N 2 O and CH 4 values extend up to about 0.4 and 2.5 ppmv, respectively, which exceed the tropospheric climatological values of 0. 32 and 1.8 ppmv. 20 The mid-latitude and arctic correlations are reasonably compact, whereas the tropical correlations exhibit a somewhat larger scatter. The black curves are 5th order polynomials fitted to the ESA data, and the red curves are third order polynomials fitted piecewise to the ATMOS data (Michelsen et al., 1998a, b). To take into account the difference of about 10 years between ATMOS and MIPAS measurements, the ATMOS-25 polynomials have been trend-corrected by addition of 2.3% (N 2 O) and 3.2% (CH 4 (15-20 N; 40-45 N et 75-80 N) to generate CH 4 /N 2 O correlation plots and to perform comparison with SPIRALE, and reference regression curve from Michelsen et al. (1998a and b). An example is 10 given in Fig. 26. In addition, MIPAS data have been averaged to generate a zonal mean with error bars taking into account the accuracy + SDV/ √ n (where SDV is the standard deviation and where n is the number of measurements in a given latitude bands). These zonal means have then been compared to reference curves. A large dispersion of individual MIPAS data is observed but it is significantly decreased when 15 zonal means are used. The agreement with Michelsen curves is good for N 2 O VMR lower than 200 ppbv. For N 2 O values higher than 330 ppbv, and CH 4 values higher than 2 ppmv, Michelsen curves are outside error bars associated to zonal means. Figure 27 shows N 2 O-CH 4 relationships as measured by MIPAS-E and the balloonborne MIPAS-B instrument. For comparison, trend-corrected correlations observed by 20 ATMOS (Michelsen et al., 1998) and in situ balloon measurements (Engel et al., 1996) are also shown. A polynomial fit has been applied to MIPAS-E and MIPAS-B. The fitted MIPAS-B correlation is very close to the in situ balloon reference. A small bias towards the MIPAS-B data is visible in the fitted MIPAS-E correlation giving a hint that MIPAS-E CH 4 is slightly overestimated and/or N 2 O slightly underestimated. Some unphysical 25 outliers are also obvious in the MIPAS-E data which are connected to oscillations in the N 2 O and CH 4 profiles at lower altitudes.
For CH 4 comparisons, we obtain a statistically significant positive bias of 5 to 11% between MIPAS and FTIR lower-middle stratosphere partial columns, and a standard deviation of 4 to 7.5%, when the high variability period (winter-spring) for high latitude stations is not taken into account. For N 2 O comparisons, no statistically significant bias is seen between MIPAS and FTIR lower-middle stratosphere partial columns, and the 10 standard deviation is between of 2.5 to 6.8%, when the high variability period (winterspring) for high latitude stations is not taken into account. When the winter-spring period is included in the comparisons for the high latitude stations, we can reach standard deviations of 9 and 15%, for N 2 O and CH 4 respectively, probably due to collocation and horizontal smoothing errors.

Balloon measurements
The comparisons of MIPAS-E with balloon data of various types (remote sensing in emission or absorption, in situ) demonstrate the impact of remaining "oscillations". Reasonable agreement is however observed in the mid stratosphere between MIPAS-E and Balloon CH 4 and N 2 O. The MIPAS-E values in the very lower stratosphere present 20 a positive bias with respect to balloon measurements.

Aircraft measurements
General agreement is better at mid and high latitude than in the tropical region where a high deviation is observed by ASUR between 24 and 28 km. The CH 4

EGU
MIPAS-E v4.61 profiles present "oscillations" which are not observed in aircraft profiles in this UT/LS region, leading to relative differences which can reach ∼30% in this UT/LS altitude range, a region which is difficult for limb measurements from space. Around the 100 hPa level, MIPAS-E presents a positive bias with respect to correlative measurements as already noticed for other comparisons in the UT/LS

EGU
The individual used MIPAS data version 4.61 still exhibits in individual CH 4 and N 2 O profiles unphysical oscillations caused by the processing algorithm. As consequence, CH 4 and N 2 O values are sometimes uncorrelated; these specific pairs of values are recognized as outliers in the CH 4 /N 2 O correlation plots. Taking these problems into account the MIPAS CH 4 and N 2 O data are behaving as forecasted by the error estimation 5 (see Sect. 2.2.).
In order to investigate the causes of the non-physical oscillations in the CH 4 and N 2 O profiles retrieved with the ESA off-line processor, IFAC performed several tests using MIPAS-E scan #16 (lat. 46.4 • N) of orbit #2975 (24 September 2002), for which a correlative measurement by MIPAS balloon measurement is available. Retrievals 10 using different Occupation Matrices were performed. The results indicate that the N 2 O oscillations are reduced when more microwindows were used.
Other tests have been performed using of a temperature profile characterised by a better vertical resolution, but the oscillations are not significantly affected. The impact of the water vapor profile has been investigated by performing a retrieval using the H 2 O 15 profile derived from the coincident MIPAS balloon measurements. The impact on the CH 4 and N 2 O profile is negligible.
Additional tests have to be repeated for other scans for which other correlative measurements are available. The fact that N 2 O and CH 4 oscillations are correlated to each other could indicate the presence of a common systematic error. However, a sin-20 gle cause of the observed differences between MIPAS and correlative measurements could not be found.  SPIE Proceedings, 3501, 186-195, 1998. Höpfner, M., Oelhaf, H., Wetzel, G., et al.: Evidence of scattering of tropospheric radiation by PSCs in mid-IR limb emission spectra: MIPAS-B observations and KOPRA simulations,