Multi-TASTE Phase F. Validation report

Transcription

1 Multi-TASTE Phase F Multi-Mission Technical Assistance to ESA Validation by Sounders, Spectrometers and Radiometers Validation report Delta-validation of SCIAMACHY SGP upgrade from V5.02 to V6.00 authors auteurs reference référence date of issue date d édition 27 August 2015 issue édition 1 revision révision 0 ESA contract Nr contrat ESA No /08/I-OL/CCN-1 D. Hubert, A. Keppens, J. Granville, F. Hendrick and J.-C. Lambert (BIRA-IASB); A. Van Gijsel (KNMI)

2 title titre Multi-TASTE Phase F Validation Report / Delta-validation of SCIAMACHY SGP upgrade from V5.02 to V6.00 reference référence date of issue date d édition 27 August 2015 issue édition 1 revision révision 0 status état Draft document type type de document Validation Report ESA contract Nr contrat ESA No /08/I-OL/CCN-1 prepared by préparé par D. Hubert, A. Keppens, J. Granville, F. Hendrick and J.-C. Lambert (BIRA-IASB); A. Van Gijsel (KNMI) document change record historique du document Issue Rev. Date Section Description of Change all Creation of this document all Input by authors 1 40

3 Executive summary This document has two objectives. The first is to report the data quality of the prototype SCIAMACHY Level-2 processor SGP 6.00, for a subset of the entire mission. And secondly, to compare the quality to that of preceding operational processors. The results presented hereafter are the first estimate of SGP 6.00 data quality and serve as direct input to the decision by the SCIAMACHY Quality Working Group whether the entire mission should be reprocessed with the new prototype, or not. The paragraphs below and Table 1 give an overview of our main observations and conclusions. The report contains detailed delta-validation analyses of Level-2 data by processors SGP 6.00, SGP 5.02 and (where applicable) SGP Several SGP Level-2 data products (nadir: vertical column of O3, NO2, BrO and CO; limb: vertical profile of O3) retrieved from a Diagnostic Data Set (DDS) of 5000 SCIAMACHY orbits, were evaluated through extensive comparison to correlative observations of reference collected from ground-based instruments certified for WMO s Global Atmosphere Watch and contributing networks like NDACC and SHADOZ (Dobson, Brewer, UV-visible instruments, FTIR spectrometers, ozonesonde, lidar and microwave radiometers). Table 1 presents an overview of our observations and conclusion. Overall, the data quality of SGP 6.00 data is similar to that of SGP 5.02, for all studied products. The new processor produces data of equal, if not slightly better quality (for a few products, in part of the atmosphere) than its predecessor. No unexpected changes were found for the overall bias and spread of the comparisons, and their dependence on geophysical parameters (latitude, altitude, season, year, solar zenith angle, cloud fraction, ). Perhaps the only worrisome result is the increase in drift (about 1.5 % over the mission lifetime) of the new O3 column data relative to correlative observations in the NH middle latitudes, most likely related to the changes in the IPF 8.01 Level-1 processor. For other products and quality indicators no degradation is seen relative to SGP 5.02, but also not a substantial improvement. This is mostly in-line with the expectation from the code changes made in the new Level-1 and Level-2 processors. Therefore we consider the current SGP 6.00 prototype sufficiently mature to reprocess the entire mission. 2 40

4 O3 limb profile 79 ozonesondes 12 stratospheric O3 lidars 3 microwave radiometers Representative global sample, except in mesosphere. As expected, V6 is very similar in quality to V5. It has slightly improved bias, short-term variability and estimates of random uncertainty in some regions of the atmosphere. Still, all major issues remain: AK-smoothing unreliable, large biases with altitude, latitude and SZA-dependence, degraded data quality in Arctic, large negative drift, inadequate auxiliary data for conversions. Reprocess, since at least no degradation relative to V5. However, many major issues remain and should be clearly documented in README file and fixed in later version. BrO nadir column 1 UV-visible spectrometer at Harestua (60 N, 11 E) Representative for one Arctic station, not a global perspective V6 is very similar in quality to V5. If not slightly better, with a less negative bias and possibly less outliers. Reprocess, since no degradation relative to V5. 12 FTIR CO nadir column Possible sampling issues, better precision expected for full mission. The relative bias of V6 is slightly reduced with respect to V5, of the order of 10 % for the Arctic, mid-north and midsouth regions. Large amount of outliers and negatives; the SGP CO product in general remains inadequate in both precision and accuracy. Reprocess, since there are indications of a reduced bias of V6 w.r.t. V5, while the spread remains similar. NO2 nadir column 19 UV-visible spectrometers Representative global sample, less so in Tropics Differences between the two SGP data versions are hardly noticeable and well below the detection limit of the groundbased measurements. Reprocess, since no degradation relative to V5. O3 nadir column 36 Dobsons 32 Brewers 17 UV-visible spectrometers Representative global sample Slight overestimation of up to 1.5% maximum. In the NH middle latitudes V6 shows a drift of about 1.5% over the mission lifetime. Dependence of bias on SZA beyond 80 (total ozone underestimation of up to 4%). Slight dependence on fractional cloud cover. Reprocess, since SGP 6.00 is an improvement w.r.t. SGP 5.02 regarding bias. Validation Report : SCIAMACHY SGP upgrade from V5.02 to V6.00 Table 1: Overview of the first delta-validation of SCIAMACHY prototype Level-1-to-2 processor SGP 6.00, and changes relative to the current operational processor SGP Correlative data sets Maturity validation V6 changes relative to V5 Other remarks Recommendation 3 40

5 Table of contents EXECUTIVE SUMMARY... 2 TABLE OF CONTENTS... 4 INTRODUCTION... 5 I SCIAMACHY DATA SETS... 5 I.1 DIAGNOSTIC DATA SET... 5 I.1.1 Optimization of sample size and coverage: 5011 orbits... 5 I.1.2 Impact of DDS on representativeness of validation results... 6 II DELTA-VALIDATION ANALYSES... 8 II.1 OZONE NADIR VERTICAL COLUMN... 8 II.1.1 Summary and conclusions II.2 NITROGEN DIOXIDE NADIR VERTICAL COLUMN II.2.1 Summary and conclusions II.3 CARBON MONOXIDE NADIR VERTICAL COLUMN II.3.1 Satellite and reference data II.3.2 CO column comparisons II.3.3 Summary table II.4 BROMINE MONOXIDE NADIR VERTICAL COLUMN II.5 OZONE LIMB PROFILE II.5.1 Algorithm changes in the V6 processor II.5.2 Comparison to ozonesonde and lidar (BIRA-IASB) II.5.3 Comparison to lidar and microwave radiometer (KNMI) II.5.4 Summary and conclusions III BIBLIOGRAPHY IV ACRONYMS AND ABBREVIATIONS V APPENDIX V.1 O3 NADIR COLUMN V.1.1 Dependence on solar zenith angle V.1.2 Dependence on cloud fraction V.2 O3 LIMB PROFILE V.2.1 Impact of smoothing V.2.2 Dependence on auxiliary data V.2.3 Dependence on solar zenith angle V.2.4 Dependence on month of year V.2.5 Dependence on scan angle

6 Introduction This document reports on the delta-validation of the upgrade of SCIAMACHY operational Level-2 processor SGP from version 5.02 to version Several SGP Level-2 data products (nadir data products: vertical column of O3, NO2, BrO and CO; limb data products: vertical profile of O3) retrieved from a Diagnostic Data Set (DDS) of 5000 SCIAMACHY orbits, were evaluated through extensive comparison to correlative observations of reference collected from ground-based instruments certified for WMO s Global Atmosphere Watch and contributing networks like NDACC and SHADOZ (Dobson, Brewer, UV-visible instruments, FTIR spectrometers, ozonesonde, lidar and microwave radiometers). This work was carried out by independent validation teams at BIRA-IASB and KNMI, in the frame of the Multi-TASTE Phase-F project. The main objective of this report is to provide feedback for the decision by the SCIAMACHY Quality Working Group and ESA to reprocess the entire mission with version 6.00 of the SGP level-1-to-2 data processor. We therefore focus on delta-validation analyses of the ground-based comparisons for subsequent SCIAMACHY data versions (SGP 6.00, SGP 5.02 and, where applicable, SGP 3.01) and for identical SCIAMACHY measurements. The Diagnostic Data Set plays a central role in all studies, representing about 10% of the entire mission. The SCIAMACHY data sets are described in Section I, together with a brief discussion of the representativeness of the results presented in this report. Section II contains concise evaluations of the geophysical data quality of each Level-2 product, differentiated by altitude and latitude where sensible and feasible. Specific attention is given to the evolution of SGP 6.00 relative to earlier SCIAMACHY data releases, and to the verification that the newest version meets expected improvements. Each section concludes by a summary table. I SCIAMACHY data sets The quality of the Level-2 data products retrieved by prototype processor SGP 6.00 is the main subject of this report. It is compared to that of two earlier operational offline processors SGP 6.00/Y (prototype) : diagnostic data set (DDS), Aug 2002 Apr 2012, retrieved from IPF 8.01/Y Level-1 data. SGP 5.02/W (operational): complete data set (subsetted to DDS), Aug 2002 Apr 2012, retrieved from IPF 7.04/W Level-1 data. SGP 3.01/R (operational): complete data set (subsetted to DDS), Aug 2002 Jan The switch off of SGP 3.01 in the operational ground segment occurred in Jan Therefore, it is not possible to compare all three SCIAMACHY versions in the last two years of the mission. It is important to note that the different Level-2 data sets were retrieved from different Level-1 data sets. Changes in the Level-1 processor may therefore impact Level-2 data quality. At some instances in this report a shorter notation for the processor versions is used: V3, V5 and V6. I.1 Diagnostic data set I.1.1 Optimization of sample size and coverage: 5011 orbits All following V6 prototype studies are based on a subset of the SCIAMACHY mission, processed by the Level-2 V6 prototype. The validation teams optimized a list of 5011 orbits specifically for deltavalidation studies, covering various SCIAMACHY data products, atmospheric states, latitudes, time 5 40

7 periods, and considering the availability of complementary correlative instrument types. The optimization algorithm and the expected performance of the Diagnostic Data Set (DDS) were presented at Progress Meeting 1 (20 Oct 2014) and subsequent Teleconference (20 Nov 2014, see Noël et al., 2014). Figure 1 below shows the co-location statistics expected for different choices of the orbit sample size. In the end, an orbit sample size of 5011 orbits was found to be optimal in terms of validation capabilities and processing time constraints (Minutes of Meeting: Noël et al., 2014). The resulting DDS represents about 10% of the entire SCIAMACHY mission. The latitude-time coverage of the colocation of each SCIAMACHY data set with different types of ground-based measurements is shown in Figure 2 for the 5011 orbit list. The final DDS orbit list was circulated to the SQWG on 25 Nov Figure 1: Co-location statistics (Y-axis) as a function of DDS orbit sample size (X-axis) for the different validation analyses (see legend). Left: total number of co-location pairs. Right: fraction of selected co-location pairs relative to the total possible number of co-location pairs. An orbit sample size of 5011 orbits was found to be optimal in terms of validation capabilities and processing time constraints. I.1.2 Impact of DDS on representativeness of validation results Figure 1 and Figure 2 show the appropriateness of the DDS for most validation analyses. It is not expected that the O3 column, NO2 column and O3 profile comparison results will change significantly when the full mission is processed. The DDS-based V6 validation results presented hereafter should therefore be representative of the full mission data set, on the global scale and in the long-term. The DDS-based BrO validation results should be robust compared to the full mission results, but it should be noted that only one station is included in this study and therefore it cannot provide a global view on the BrO data quality. The DDS contains 80% of the (space, time) co-locations for the SCIAMACHY CO versus FTIR analysis. However, the final delta-validation analysis does not include the temporal co-location condition, since the analysis is based on the difference in monthly mean of SCIAMACHY overpasses over the station (as recommended by the QWG, Section II.3.1). Therefore, ten times more data will be available once the entire mission is processed, which should lead to a factor 10 improvement in the precision of the yearly averaged bias estimates. 6 40

8 Figure 2: Latitude-time cross-section of the co-location of SCIAMACHY Level-2 data with ground-based instruments (different marker types), for each of the studied data products (different panels). 7 40

9 II Delta-validation analyses The following sections contain concise reports of the initial results of the comparison of SCIAMACHY V6 prototype data (and earlier V5/V3 data) to ground-based measurements. Each section concludes by a table summarizing the data quality in terms of bias and comparison spread 1, and their dependence on latitude, altitude and other parameters where relevant. Clear statements are made about the changes of V6 data relative to earlier SCIAMACHY data, and whether these represent an improvement or degradation relative to independent correlative measurements. II.1 Ozone nadir vertical column Method: The SCIAMACHY ozone column Diagnostic Data Set reprocessed for the delta validation of SGP upgrade from v5.02 to 6.00 has been compared to correlative data sets collected from groundbased Brewer and Dobson spectrophotometers and DOAS UV-visible spectrometers performing network operation in the framework of WMO s Global Atmosphere Watch (GAW) and its contributing network NDACC. The comparison methodology is described basically in Lambert et al. [1999,2000], with updates in Balis et al. [2007]. Correlative data undergo first quality checks and filtering procedures, e.g., only Brewer and Dobson direct sun data are considered, and among those obtained by single monochromator instruments, only data at solar elevation higher than 15 are kept to avoid data affected by the well-known air mass dependence. Co-location between SCIAMACHY and groundbased observations is then defined according to geographical criteria depending on the type of validation data: for Brewer and Dobson direct sun measurements there is a maximum distance of 150 km between the SCIAMACHY footprint centre and the station; for UV-visible zenith-sky observations, the footprint of the SCIAMACHY field-of-view must intercept the ground-based twilight air mass estimated with a ray tracing model; in all cases observations must take place the same day. Uncertainties: The study reported by Van Roozendael et al. [1998] has shown that mutual agreements between Dobson, Brewer and UV-visible data can reach the percent level when major sources of discrepancy are properly accounted for. For Dobson instruments, temperature dependence of the ozone absorption coefficients used in the retrievals can account for a seasonal variation in the error of ±0.9% in the Alps and ±1.7% at Sodankylä (Finland, 67.4N), and for a systematic errors of up to 4% Bernhard et al. [2005]. The effect can dramatically increase in extremely cold conditions (winter polar vortex). The effect is smaller for Brewer instruments thanks to their use of wavelength pairs with reduced ozone absorption dependence on the temperature. Dobson and Brewer instruments might also suffer from long-term drift associated with calibration changes. Assuming that the Dobson and Brewer instruments are well-calibrated, and that data are filtered to avoid air mass dependence, values of 1% on the Brewer and 2% on the Dobson total ozone data can be considered as indicative of the systematic component of the network data uncertainty. For zenith-sky UV-visible spectrometers measuring ozone in the Chappuis band, the use of DOAS minimizes uncertainties associated with the temperature dependence of the absorption cross sections [Burkholder and Talukdar, 1994]. Major uncertainties on the vertical column are associated mainly with the AMF based conversion from spectrally derived slant columns to vertical columns. Real-time data can be based e.g. on a single-profile standard AMF calculated at 60N in winter and at sea level [SAOZ standard AMF, Sarkissian et al., 1995]. Seasonal changes of the ozone profile and scattering geometry are then responsible for a systematic bias of about 5 6% amplitude at 67N, falling to 3 4% at 44N. This is in addition to the ±1% scatter that might result from short-term fluctuations. The use of the standard SAOZ AMF, e.g., also introduces an average meridian 1 Comparison spread includes a combination of the precision of satellite and ground-based measurements and random uncertainties of metrological origin (i.e. differences in smoothing and sampling). 8 40

10 dependence of 3% at 67N to +2.8% at the tropics. In recent years the SAOZ sub-network of the NDACC UV-VIS network has undergone a reprocessing at all stations with the new version 3 of the algorithm, which includes now climatological air mass factors for the conversion of ozone slant column data into ozone vertical column data, plus several other improvements. The use of climatological air mass factors instead of the standard air mass factor changes DOAS/SAOZ total ozone data by at least one percent in the Southern middle latitudes. According to recent work by Hendrick et al., a total uncertainty of 4.7%, of which a systematic bias lower than 2.5%, can be used as indicative values of the uncertainty of UV-visible total ozone measurements. Ground-based validation of SCIAMACHY ozone column data retrieved at DLR with SGP 3.01 and 5.02 was reported in detail in the Multi-TASTE final report (January 2012), which provided estimates of SCIAMACHY uncertainties as well as dependences on solar zenith angle, time, latitude, and cloud parameters. While ground-based validation results of SCIAMACHY SGP 5.02 based on correlative Brewer and Dobson measurements remain unchanged, ground-based validation results based on the SAOZ sub-network of the NDACC UV-VIS network are slightly modified by the current reprocessing of all SAOZ data. In this section we compare new SGP6.00 validation results to the corresponding update of SGP 5.02 validation, based thus on reprocessed UV-visible data. Validation results: Both SGP 5.02 and SGP 6.00 ozone column DDSs are generally mutually consistent and also consistent with GAW/NDACC ground-based ozone column data records. On a statistical basis, Figure 3 presents from pole to pole the mean relative difference between the two version of SCIAMACHY data and the Brewer, Dobson and UV-visible data. At most stations the mean difference between the two SGP data sets ranges to within 0.2% to 0.6%, SGP 6.00 providing always lower ozone column values than SGP At a few stations only this relative difference between the two processors rises up to 0.8%. The result of this general decrease in total ozone values is a better global mean agreement between SCIAMACHY SGP 6.00 and ground-based networks: the 1-2% overestimation of ground-based data reported for SGP 5.02 (with the DDS as well as previously with 2002/2003/2004/2006 data) decreases now with SGP 6.00 to a slighter overestimation of maximum %, with some stations reporting even an underestimation (estimates based on the DDS only). At most stations the relative difference is now comparable with the level of uncertainty attributable to the ground-based data themselves. It is interesting to note that the uncertainty reported in the SCIAMACHY data files (about ±1%) is also comparable to the uncertainty attributable to the groundbased data. 9 40

11 Figure 3: 10-year mean value (bullet) and 1sigma spread (error bar) of the percent relative difference between SCIAMACHY SGP Diagnostic data Set and ground-based network total ozone data, as a function of latitude. SGP 5.02 results are depicted in green and new SGP 6.00 results in red. Upper graph: SGP vs. Brewer network; middle graph: SGP vs. Dobson network; lower graph: SGP vs. DOAS UV-visible network. The shaded area represents the indicative uncertainty of 1%, 2% and 2.5% on the Brewer, Dobson and UV-visible ozone column measurement, respectively. Green and red square contours represent the median value of the uncertainty on SCIAMACHY total ozone data as reported in SGP 5.02 (green) and 6.00 (red) data files, respectively

12 With the earlier version SGP 3.01 a negative drift in SCIAMACHY ozone column values had been noticed at numerous but not all stations, resulting of a transitory decrease in total ozone values in the first couple of years, followed by a more stable behaviour after An expected effect of the recent SCIAMACHY level-1 data improvements is a change of this drift. Figure 4 and Figure 5 show the longterm evolution of the relative difference between SCIAMACHY and Brewer data: in the middle latitudes of the Northern Hemisphere, the region where the network offers the best geographical and temporal sampling, and also at all latitudes of the Northern Hemisphere. From SGP 5.02 to 6.00 the relative stability after 2003 has degraded now to a slight drift, of about 1.5% over the full lifetime of SCIAMACHY. Figure 6 show similar results with the Dobson instruments in the Northern hemisphere. The latter exhibit the seasonality of about ±2.5% expected from the temperature dependence of the ozone absorption coefficients, nevertheless the same drift of SGP and the absence of drift in SGP5.2 data - as noticed with respect to the Brewer network is also visible when SGP6.00 data are compared to the Dobson network. Figure 4: 10-year time series of the monthly mean value (triangle) and 1sigma spread (error bar) of the percent relative difference between SCIAMACHY SGP Diagnostic data Set and Brewer network total ozone data averaged in the 30 N to 60 N latitude zone. For clarity, SGP 6.00 results depicted separately in upper graph and SGP 5.02 results in lower graph

13 Figure 5: Same as Figure 4 but now with respect to all Brewer instruments of the Northern Hemisphere. A permanent caveat of the application of single-wavelength-amf DOAS in the nm spectral range, thus within an optically thick atmosphere due to strong ozone absorptions, is the dependence of the retrieved ozone value on the solar zenith angle (SZA) associated with the measurement. This dependence usually shows up in the form of an artificial decrease of ozone values at large SZA (typically beyond 80 ). Differences between previous SGP 3.01/5.02 data versions and ground-based networks, based on 4 full years of SCIAMACHY data, were shown to be affected by such a SZA dependence, with an artificial decrease of total ozone data by about 4% at SZA beyond 80. Figure 20 in Appendix, based on comparisons of the SCIAMACHY DDS with correlative Brewer network data in two latitude zones, confirms that the SZA dependence of SGP data has not changed from SGP 5.02 to Clouds are also an important source of uncertainty for total ozone retrieval from nadir ultraviolet measurements. The SGP data files contain the fractional cloud cover of the SCIAMACHY ground pixel, the cloud top pressure, and the cloud optical depth. Examples of the dependence of SCIAMACHY ozone data on the cloud fraction are given in Figure 21 (see Appendix) in two latitude zones. At individual stations this dependence usually is small, within 1-4%, with in general the best agreement at low cloud fractions and a variable agreement at high cloud fractions

14 Figure 6: Same as Figure 4 and Figure 5 but now with respect to all Dobson instruments of the Northern Hemisphere. II.1.1 Summary and conclusions Table 2: Summary table of the nadir vertical ozone column validation results. SGP 6.00/Y Bias (DU) Slight overestimation of up to 1.5% maximum, but within the ground-based and SCIAMACHY uncertainty bars at most stations. SGP 6.00 is an improvement w.r.t. SGP Spread (%) No changes w.r.t. to previous SGP versions. Spread dominated by a combination of measurement uncertainties and atmospheric noise (mismatch uncertainties) Other Drift: in the NH middle latitudes SGP 6.00 shows a drift of about 1.5% over the mission lifetime, while SGP 5.02 was affected by a transitory drift only in the first years of the mission. Dependence of bias on SZA beyond 80 (total ozone underestimation of up to 4%). Slight dependence on fractional cloud cover

15 II.2 Nitrogen dioxide nadir vertical column Method: The SCIAMACHY nitrogen dioxide column Diagnostic Data Set reprocessed for the delta validation of SGP upgrade from v5.02 to 6.00 has been compared to correlative data sets collected from ground-based DOAS UV-visible zenith-sky spectrometers performing network operation in the framework of WMO s Global Atmosphere Watch (GAW) contributing network NDACC. The comparison methodology is described extensively in the Multi-TASTE final report. Correlative data undergo first quality checks and filtering procedures, e.g., enhanced NO 2 column values occurring simultaneously with O 4 and H 2 O column enhancements are filtered out since most likely affected by uncorrected multiple scattering within dense clouds or snow showers (Pfeilsticker et al., GRL 1998). Co-location between SCIAMACHY and NDACC zenith-sky observations is selected where the footprint of the SCIAMACHY field-of-view intercepts the ground-based air mass at sunrise as estimated with a ray tracing model; in all cases observations must take place the same day. To ensure similarity of the satellite and ground-based vertical sensitivity, SCIAMACHY measurements including tropospheric pollution are rejected through a cloud-based filtering approach since zenith-sky air masses are sensitive mainly on the stratospheric column. A photochemical correction based on the SZA difference between SCIAMACHY and twilight data is applied to account for NO 2 diurnal cycle effects. According to NDACC intercomparison campaign results, the uncertainty on ground-based retrievals should be around 10% [Vandaele et al., 2005; Roscoe et al., 2010]. Validation results: Details of the ground-based validation of SCIAMACHY nitrogen dioxide column data (retrieved at DLR with earlier SGP 3.01 and 5.02, with data available in 2002, 2003, 2004 and 2006) was reported in the Multi-TASTE final report, which provides estimates of SCIAMACHY uncertainties as well as their dependence on solar zenith angle, season, and latitude. Here, Figure 7 shows from pole to pole the 10-year mean agreement between SCIAMACHY SGP 5.02 and 6.00 DDS NO 2 column data and NDACC data. Differences between the latest two SGP data versions are hardly noticeable and well below the detection limit of the ground-based measurements. At stations free of tropospheric pollution and where the diurnal cycle can be accounted for accurately, that is, where direct comparisons between satellite nadir and ground-based zenith-sky measurements provide the most quantitative results, the median agreement ranges to within ± molec.cm -2. A few molec.cm -2 of agreement is equivalent to a few percent up to about 10%, that is, within the error bar of the satellite and ground-based retrievals [e.g. Vandaele et al., 2005; Roscoe et al., 2010]. There are two stations exhibiting a non-negligible systematic difference of molec.cm -2. After verification it turns out that the NDACC instruments at those two stations report too high values by about 20% in summer and are probably using NO2 cross-sections at inappropriate temperature. There is also a bias of about molec.cm -2 between comparison results in the middle latitudes of the Southern Hemisphere and of the Northern Hemisphere, already noticed in previous studies and reports. Regarding the spread of the absolute difference between SCIAMACHY and NDACC data, values of a few molec.cm -2 are themselves within the error bar of the measurements and of the validation method. Enhanced spread at NDACC stations surrounded by pollution sources visible by the satellites like all Northern middle latitude sites (Europe and Japan) and polar sites where the diurnal cycle is less predictable in spring and winter, is attributable partly to the difference in vertical sensitivity and/or to residual diurnal cycle effects

16 Figure 7: 10-year mean value (bullet) and 1sigma spread (error bar) of the absolute difference between SCIAMACHY SGP Diagnostic data Set and NDACC/UV-visible network NO2 column data, as a function of latitude. SGP 5.02 results are depicted in green and new SGP 6.00 results in red. The shaded area represents the indicative uncertainty of the NDACC/UV-visible NO2 column measurement. Green and red square contours represent the median value of the uncertainty on SCIAMACHY NO2 column data as reported in SGP 5.02 (green) and 6.00 (red) data files, respectively. II.2.1 Summary and conclusions Table 3: Summary table of the nadir vertical nitrogen dioxide column validation results. SGP 6.00/Y Arctic 30N-60N 30N-30S 30S-60S Antarctic Bias (10 14 molec.cm -2 ) ±3 +4 ±3-5 ±3 Spread (10 14 molec.cm -2 ) Other Apparent bias between NH and SH data, maybe due to difference in sensitivity to tropospheric pollution and/or to residual diurnal cycle effects between SCIAMACHY and NDACC/UV-visible twilight data II.3 Carbon monoxide nadir vertical column II.3.1 Satellite and reference data The CO nadir column data selection and screening for both the SCIAMACHY satellite instrument and the ground-based FTIR reference data have been extensively described in the Multi-TASTE Phase F midterm report (Hubert et al., 2014). Due to the unrealistically large variability of SCIAMACHY CO column data, users are advised to average SCIAMACHY data at least on monthly scales (whereby negative averages are omitted from further analysis). However, a prerequisite for the meaningful use of monthly means is that the data subsets (SCIAMACHY data versions 5.02 and 6.00, and NDACC data) offer a similar sample of the atmosphere. If this is not the case, comparison sampling errors will add to the targeted discrepancies between measurements. As the DDS of SGP version 6.00 has only about 10 % of the data coverage of the SGP version 5.02 full dataset, the delta-validation analysis is based on the relative difference of monthly means for measurement pairs within 300 km of the groundbased FTIR instrument locations and for the same subset of SCIAMACHY pixels for both algorithm versions (applying a 0.1 s tolerance). When constructing monthly weighted averages of the ground-based FTIR measurements and of the overlapping SCIAMACHY data subsets, the amount of monthly co-locations is strongly reduced. It has been decided to exclude stations with less than 10 overlapping monthly means from further analysis, which is additionally motivated by the later need for yearly statistics (as in Figure 9). Out of the 17 FTIR stations that regularly provide the NDACC DHF with CO column data, the 12 stations that are 15 40

17 eventually used for the comparative analysis are indicated by blue dots in Figure 8. This distribution reveals that there is a relative over-concentration of measurements in the Arctic and Northern middle latitudes (30-60 N), which has to be taken into account when evaluating the comparison statistics. Figure 8: Locations of the 12 ground-based FTIR stations that have been used in the SCIAMACHY CO (delta-) validation (blue). Five NDACC stations dropped out of the delta-validation because of a lack of overlapping data for both processor versions (red). Green lines mark the edges of the latitude bands considered in Table 4. II.3.2 CO column comparisons The methodology for determining CO column comparison statistics on both monthly and yearly scales has also been outlined in the Multi-TASTE Phase F midterm report (Hubert et al., 2014). Yearly mean differences and spreads for SCIAMACHY SGP versions 5.02 and 6.00 are shown in Figure 9 for all 12 FTIR stations under consideration. The comparison results are summarized in Table 4, differentiated over five latitude bands. The major observations are the following: - SCIAMACHY CO column data show a large amount of outliers and negatives (even for monthly means) for both SGP versions. Negative monthly means are omitted from the comparative analysis. - Due to the large variability of the SCIAMACHY CO column data, even on monthly scales, no seasonal cycle or trend is to be observed, nor any meridian dependence. - The relative bias of SCIAMACHY SGP V6 CO columns with respect to ground-based FTIR stations can be considered to be slightly reduced with respect to V5, i.e. of the order of 10 % for the Arctic, mid-north and mid-south regions. At Arrival Heights (the only Antarctic station) the bias is even significantly reduced by 63 %. Yet one remarkable bias increase occurs at Izaña (the only equatorial station) from 25 to 55 % (see fourth column in Table 4). - The average (yearly) comparison spread amounts about 20 to 35 % for all latitude bands and for both SGP versions under study (see fifth column in Table 4). - The SCIAMACHY SGP CO product in general remains inadequate in both precision and accuracy

18 17 40

19 18 40

20 Figure 9: Time series of yearly relative differences (left) and spreads (standard errors on the mean, right) between SCIAMACHY CO total column data and NDACC ground-based FTIR data at 12 stations (rows, sorted north to south). The number of comparisons equals the number of months for which both SCIAMACHY and the NDACC FTIR instrument provide CO columns. II.3.3 Summary table Table 4: Summary of SCIAMACHY CO column validation outcome divided into 5 latitude bands. Subsequent columns provide the number of stations, the number of monthly means, the bias shift from SGP V5 to SGP V6, and the comparison spread (which is similar for V5 and V6) in each band. # monthly Bias Bias Spread Latitude band # stations means SGP V5 - FTIR SGP V6 - FTIR SGP V5 & V6 Arctic (60N-90N) % -6% 19% Mid-north (30N-60N) % 36% 24% Tropics (30N-30S) % 55% 20% Mid-south (30S-60S) % 29% 37% Antarctic (60S-90S) % 2% 35% II.4 Bromine monoxide nadir vertical column The nadir total columns of BrO were compared to ground-based (GB) UV-visible zenith-sky measurements at Harestua, Norway (60 N, 11 E). The validation methodology was presented in Hendrick et al. (2009) and applied to full mission SGP 5.02 data and the DDS SGP 6.00 data. This analysis is therefore not a delta-validation in the strict sense, and the differences observed between the data versions may be partially explained by differences in sampling. In order to properly take into account the relative contributions of both the stratosphere and troposphere to the total columns, the SCIAMACHY total vertical columns were recalculated on a daily basis by dividing the slant columns given in the SGP data files by total AMFs derived from stratospheric and tropospheric vertical profiles retrieved from GB UV-visible zenith-sky observations (see Hendrick et al. (2007) for more details)

21 Table 5: Validation methodology for nadir BrO total columns. Correlative data UV-visible zenith sky instrument at Harestua (60N, 11E) Screening SCIAMACHY data: CF<40% Co-location SCIA-GND <300km from station, temporal window: all coincident mornings are selected Photochemical correction Ground sunrise columns converted to satellite overpass SZA (Hendrick et al., 2007 & 2009) Figure 10 shows time series of BrO total columns for both SCIAMACHY processors and ground-based observations around Harestua. The annual cycle in BrO is well reproduce by both SGP versions. The V5 full mission data set has clear outliers in the early years of the mission, which is not seen for V6, perhaps since those pixels are not in the DDS. Figure 11 presents absolute difference time series, with mean and standard deviation per year. Overall, SGP 6.00 produces slightly more BrO leading to a small reduction in bias compared to that of its predecessor (-12.2% versus -12.8%), a change which is not significant. Also the spread is very similar, except in the years 2003, 2004 and 2007, during which V5 produces a few severely outlying pixels. Nevertheless, it is quite clear that the data quality of the V6 prototype is it least as good as that of the V5 processor at the Arctic station of Harestua. Quality at other locations may differ however, and will require further study. Figure 10: Time series of the BrO total columns around Harestua (60N, 11E) by SCIAMACHY SGP 5.02, SGP 6.00 and UV-visible zenith-sky observations. Figure 11: As in left, but for the absolute differences between SCIAMACHY and ground-based data. Solid and dashed lines indicate annual mean and standard deviation respectively

22 Table 6: First estimates of data quality of SCIAMACHY SGP 6.00 nadir total BrO column based on comparisons to UV-visible zenith-sky observations at Harestua (60 N, 11 E). SGP GB SGP GB Absolute difference Bias (x10 12 molec/cm 2 ) Standard deviation Relative difference Bias (%) Standard deviation Remark V5 standard deviation affected by large outliers in and II.5 Ozone limb profile Two separate comparison analyses were performed for SCIAMACHY ozone limb profile data. The first analysis focuses on the UT/LS region and the stratosphere on a global scale, through comparisons to network observations by ozonesonde and stratospheric O3 lidar instruments. The second analysis focuses on the upper stratosphere and mesosphere for two mid-latitude ground stations where observations are available from a stratospheric O3 lidar and a microwave radiometer. The results of both analyses are consistent in the overlapping altitude and latitude region. II.5.1 Algorithm changes in the V6 processor Opposed to SGP V3 and V5 which reported a priori information in the mesosphere, SGP V6 actually retrieves ozone. This requires the combination of the radiances from all four measurement states per scan sequence. Hence, where the V3 and V5 processors retrieved up to four ozone profiles per scan sequence, the V6 processor provides at most one. As a result, the horizontal coverage of the new product is four times smaller (see bottom row Figure 12) while the vertical coverage is larger, now extending the retrieved information from the stratopause well into the mesosphere. One of the focuses is therefore on the data quality of V6 in the upper stratosphere and mesosphere, where previously only a priori profile information was provided in the V3/V5 data files. In addition, we also show results of the V3 & V5 data quality for each scan angle, since any scan angle dependence may lead to differences relative to V

23 Figure 12: Spatial sampling of SCIAMACHY limb ozone profiles co-locating with ozonesonde observations at the RMI ground station Uccle (51 N, 4 E). Colours differentiate between SCIAMACHY scan angle, from left (pink) to right (dark red) relative to flight direction. Top panels show all data available for V3, V5 and V6; bottom panels show the data subsetted to the same V3-V5-V6 measurements. Only one V6 profile is retrieved from the four states per scan sequence. Black contour lines indicate 500km and 100km distance from the station location (large black marker). II.5.2 Comparison to ozonesonde and lidar (BIRA-IASB) METHODOLOGY This section compares SCIAMACHY V3-V5-V6 data to observations by ozonesonde and stratospheric lidar networks contributing to GAW, NDACC and SHADOZ, sampling a variety of atmospheric regions and states from the Arctic to the Antarctic and from the ground up to the upper stratosphere. Table 7 summarizes the adopted methodology which is identical to that of earlier Multi-TASTE reports and was recently published by Hubert et al. (2015). It is important to note that this is a true deltavalidation analysis, in the sense that only SCIAMACHY-ground co-locations were considered where all three processors produced an ozone profile for the same scan sequence. For the period, for which no V3 data exists, only matching SCIAMACHY-ground pairs of V5-V6 were considered. As a result, differences in quality between V3 and later version can be partially caused by differences in temporal coverage. In practice, this turns out to be of minor importance. Table 7: Validation methodology for vertical ozone profiles. Correlative data NDACC/GAW/SHADOZ ozonesonde; NDACC stratospheric O3 lidar Screening SCIAMACHY data: none, since no QWG recommendation in README; Correlative data: standard procedure (Hubert et al., AMTD 2015) Co-location SCIA-GND <500km from station, < ±12h Delta-validation matching Identical SCIA scans (SCIA time within 40sec); : V3-V5-V6; : V5-V6 Comparison grid Vertical grid of SCIAMACHY (differs for V3 and V5/V6) Coordinate system Number density on altitude levels Vertical smoothing Correlative data not smoothed with SCIAMACHY AK, only regridded (see text) Quality indicators Statistical properties of distribution ΔO3 = (O3 SCIA - O3 GND ) / O3 GND Other Reported SGP 6.00 error estimates divided by x8 prior to analyses (ad-hoc correction suggested by DLR for bug in processor) 22 40

24 AK-BASED SMOOTHING INTRODUCES VERTICAL OSCILLATIONS IN COMPARISONS As a mandatory step before the comparison itself, the high-resolution ozonesonde and lidar profile data are smoothed vertically to the lower resolution of SCIAMACHY. One widely used technique to do so is to convolve the high-resolution data with the vertical averaging kernels (AKs) and a priori profile associated with the SCIAMACHY retrieval (Rodgers formula). In theory, smoothing high-resolution data to the lower resolution SCIAMACHY data using its AKs and a priori is supposed to remove the comparison error due to the differences in vertical sensitivity. The underlying assumption is obviously that the vertical averaging kernels do describe accurately how the SCIAMACHY system smoothes the atmospheric profile. But here it appears, like in Figure 13, that the relative difference ΔO3 degrades significantly when the correlative profile data are smoothed with the SCIAMACHY vertical averaging kernels. The bias (top row) exhibits vertical oscillations between neighbouring grid levels of up to 5% in amplitude, over the entire stratosphere, at all latitudes. The comparison spread (bottom row) exhibits similar vertical oscillations, but solely in the polar middle stratosphere. The oscillating behaviour is also seen with V5, but not with V3. Such oscillations are not seen when AK-independent techniques of smoothing are applied, e.g., using the regridding approach of Calisesi et al. with implicit smoothing (Figure 14 and Figure 15), or when synthetic low-pass filters are used as a smoothing function (e.g. triangular, box-shaped and Gaussian functions, see Figure and Figure 23 in Appendix V.2). The strong degradation of the ground-based comparison results when SCIAMACHY AKs are used suggests issues with those AKs, the way they are calculated, and maybe the underlying forward and inverse models. As a result, in the rest of the report only comparisons will be shown for unsmoothed correlative data. It must be remembered that the vertical regridding process inevitably introduces some smoothing

25 Figure 13: Impact of smoothing correlative profiles with SCIAMACHY vertical averaging kernels. (Top) Vertical structure of median bias of SCIAMACHY relative to AK-smoothed NDACC/GAW/SHADOZ ozonesonde (solid lines) and NDACC lidar (dashed lines) network observations in different latitude bands (panels left to right). Colours differentiate between SCIAMACHY Level-2 processors (see legend) and represent V3-V5-V6 matched data in the Diagnostic Data Set. The shaded area indicates the statistical uncertainty (1σ) on the bias. Positive bias values indicate where SCIAMACHY overestimates ozone relative to correlative measurements. (Bottom) Same as top row, but for the half-width of 68% inter-percentile of the relative differences, equivalent to 1σ of a Gaussian distribution. CHANGES BETWEEN V6 AND EARLIER VERSIONS The main results are shown in Figure 14 (median bias), Figure 15 (comparison spread) and Figure 16 (reported ex-ante random error), in five latitude bands and vertically resolved. Generally, no major differences are observed in data quality between V6 (blue) and V5 (red) in the UT/LS and stratosphere. This is consistent with what was expected based on the implementation changes in the Level-2 V6 processor (Section II.5.1). However, in some parts of the atmosphere the comparison results have changed somewhat, which generally represent a small improvement in data quality relative to V5. These are detailed below. Firstly, the comparison spread of V6 is 2-5% smaller than that of its predecessors in the upper stratosphere and 1-2% smaller than V5 below 30km. The extension of the altitude range of the ozone retrieval hence seems to improve the short-term variability substantially in the upper stratosphere. At lower altitudes, the improvement is marginal. Secondly, in the Arctic the V6 short-term variability clearly improves over V5, with 5-10% smaller observed comparison spreads. Finally, V6 ozone values are on average 3-7% lower than V5 in the lower and middle stratosphere, except in the Arctic. While this improves the bias relative to correlative data over 30N-90S, that is not the case at Northern midlatitudes. The change in bias is therefore not an unequivocal improvement. In addition, the ad-hoc correction factor 1/8 for the reported V6 ex-ante uncertainties brings the estimates in line with those of its predecessors. Vertical oscillations (V6 and V5) are seen in the middle stratosphere, which are, again, likely caused by the finer retrieval grid since not seen for V3. The ~5% reduction in V6 uncertainty over earlier versions in the upper stratosphere concurs with the observed reduction in comparison spread mentioned in previous paragraph

26 In the following section we show that there is some dependence of V5 data quality on scan angle. The reported location of V6 profiles are closest to the -1 scan angle states of V3/V5. The V3/V5 data quality of the -1 state is within 1-2% of that of all V3/V5 states taken together (Figure 30 and Figure 31). Hence, differences between V6 and earlier data versions are not expected to originate from the small scan angle dependence of SCIAMACHY data quality. Figure 14: As in Figure 13, for the median bias when correlative profiles are only regridded, not smoothed. Apparent inconsistencies between ozonesonde and lidar-derived results are a result of differences in temporal sampling of the co-locations. Figure 15: As in Figure 14, but for the half-width of 68% inter-percentile of the relative differences, when correlative profiles are only regridded, not smoothed. Figure 16: As in Figure 14, but for the reported uncertainty estimates in the data files (1σ) of SCIAMACHY (thick lines) and correlative data (thin lines). Correlative profiles are only regridded, not smoothed. The SCIAMACHY V6 uncertainties were reduced by a factor 8, as an ad-hoc correction for a bug in the prototype. REMAINING MAJOR QUALITY ISSUES V6 ozone data quality represents only a minor improvement relative that of V5. None of the major quality issues observed for SGP 5.02 are truly addressed by the new processor. The issue with the averaging kernels was already mentioned above. Other issues are summarized below. (To avoid clutter in the main report, all relevant figures are included in Appendix V.2) 25 40

27 Most importantly, the SCIAMACHY bias shows more complex patterns and dependences than many other limb/occultation ozone profilers (Hubert et al., 2015). It exhibits clear dependences on latitude, altitude, solar zenith angle and time at decadal scale. - Most affected seems the middle and upper stratospheric part of the profiles, with a large positive bias in the Southern hemisphere and Tropics, which changes sign around Northern mid-latitudes to become negative in the Arctic. - Also, the bias changes over the course of the mission. It decreases by up to 8% per decade above 30km (>4σ significance), and increases below 30km by up to 3% per decade (almost 2σ significance). - Furthermore, the solar zenith angle may influence the agreement between SCIAMACHY and ground-based data in the upper stratosphere, above 30km (Figure 26 and Figure 27). Here, large SZA coincide with larger variability (globally) and higher ozone values compared to small SZA. The more positive bias at high SZA is especially clear at mid-northern lidar and ozonesonde sites. Since SZA and time of year are correlated at each location, the data quality may equally depend on month and season. It is expected that metrological uncertainties increase at large SZA, due to increasing mismatch between the probed air masses. Whether or not this is fully explaining the observed dependences of bias and comparison spread is a matter of further research. - Another (minor) dependence is that on scan angle, where the difference in bias between East and West states are up to 4% at the peak in O3 number density, around 25km (Figure 30). The variability is not affected (Figure 31). SCIAMACHY data quality degrades strongly in the Arctic. - The bias (Figure 26) and comparison spread (Figure 27) depend strongly on SZA (or season), with much higher values in both indicators around Boreal winter (high SZA). - In addition, there is a sudden increase in bias and spread at the O3 number density peak (around 25km) compared to other altitudes. - The scan angle dependence is also much more pronounced than at other latitudes, with states facing the Pole (i.e. negative values, or West) having higher variability in MS and US (Figure 31) and more positive bias around 25km and in US (Figure 30). The data quality in Antarctica is in line with that of other sounders. - Nevertheless, the SCIAMACHY bias shows a clear dependence with scan angle. States facing the Pole tend to be more positively biased relative to ozonesonde and lidar. The short-term variability is much less dependent on scan angle compared to that in the Arctic. - The apparent small positive bump in bias around 22km is caused by extremely low ozone values (which are overestimated by SCIAMACHY) during the Antarctic ozone hole season. At other times of the year the bias is close to 0%. Not directly related to the retrieval, but still important for many applications is the quality of the auxiliary information provided in the SCIAMACHY data. Pressure and temperature profiles are required for the conversion of SCIAMACHY s native (altitude, number density) profile representation to e.g. pressure or VMR. Figure 24 clearly shows that the bias changes by more than 5% in large parts of the atmosphere when such conversions are done. The observed spreads are almost not affected (Figure 25). It should be noted that many of the major SGP 5.02 quality issues (positive bias in US which has N-S asymmetry, degraded quality in Arctic and negative drift in MS) were also observed for the scientific Level-2 processor developed by IUP Bremen (V2.9). This indicates that the Level-2 issues may have their origin in the Level-1 data

28 II.5.3 Comparison to lidar and microwave radiometer (KNMI) This section shows the results of the comparison between SCIAMACHY and collocated lidar & microwave radiometer ozone profiles at Lauder (45.0 S) and at Mauna Loa (19.5 N). Note that these double-collocation datasets are somewhat sparse (451 collocations at Lauder and 231 at Mauna Loa). Furthermore, the collocations are sought closest in time to the SCIAMACHY observation and thus the two ground-based observations are not necessarily simultaneous. In fact, for Mauna Loa, less than 2% of the used lidar and microwave radiometer data are acquired within 3 hours of each other. 28% is within 6 hours, nearly 39% within 9 hours and over 44% is within 12 hours. For Lauder, less than 3% of the used collocated lidar and microwave radiometer data is within 3 hours, 5% within 6 hours, 25% within 9 hours and 54% within 12 hours. Figure 17 shows the comparison results when both a lidar observation and a microwave radiometer observation were paired with a single SCIAMACHY ozone profile at Mauna Loa. The left panel shows the ozone number density (mean and standard deviation per instrument), the middle panel shows various percentiles of relative differences (with respect to the ground-based observation) and the right panel shows median/mean differences with standard deviations/errors, all as a function of altitude. The upper parts correspond to the comparison with microwave radiometer data (e.g. blue lines for the middle panel). The altitude region where both lidar data and microwave radiometer data overlap, is that between 30 and 45 km. There is a few percent difference (<5%) at some altitudes. The spread is very similar around 30 km, but at the highest overlapping altitudes (close to 45 km), the spread is larger when comparing with lidar. At these altitudes natural variation (due to differences in time of observation) is also larger, and lidar data quality reduces. Overall we see a mostly positive bias of the SCIAMACHY data, with exceptions around 20 km and 60 km

29 Figure 17: Comparison between SCIAMACHY v6.00 limb ozone profiles and ground-based observations at Mauna Loa (19.5 N). Left panel: Average ozone profiles (thick lines) with standard deviations (thin lines) in number density as a function of altitude for SCIAMACHY (red and purple) with lidar (dark blue) and microwave radiometer (light blue) ozone concentrations. Note that above 50 km, a log-scale is used. Middle panel: from left to right are shown the 2.5, 16, 50 (=median), 84 and 97.5 percentiles of the relative differences with respect to the ground-based instrument of choice. Comparisons with lidar are in black/grey and comparisons with the microwave radiometer are in light blue. The number of collocations is given on the right side of the right axis (microwave in light blue). The total number of collocations is 231 for both sets. Right panel: mean plus/minus one standard deviation and plus/minus two standard error and median relative differences as a function of altitude. Medians are shown as the green (dark green for microwave radiometer comparison) lines, the thick black lines correspond to the means. The thin blue lines are used for the standard errors/deviations in the microwave radiometer comparison. Figure 18: As Figure 17, but for Lauder (45.0 S). Figure 18 then shows similar results for the comparison at Lauder. Also here we see a positive bias in the SCIAMACHY data which is somewhat more pronounced. Agreement between the comparison 28 40

30 with microwave and lidar is very good up to about 42 km. Above they deviate, which is possibly due to the reducing quality of the lidar data at these altitudes. We stress the good agreement between the lidar results of this analysis and the one in previous section. Given the stronger deviations above 60 km, it is advisable to carry out a comparison with other data sources to verify the origins of the observed deviations. Figure 19 finally shows the comparison between microwave radiometer and SCIAMACHY ozone profiles at Payerne in the northern hemispheric mid-latitudes. Here we observe a bias in the SCIAMACHY data that is close to zero at 30 km (where the comparison starts) which almost linearly increases to nearly 20% close to 47 km and then steadily decreases to -50% at 70 km, passing 0% around 53 km. In this case, the spread is largest when the bias reaches the largest positive value. The Payerne microwave results in the stratosphere agree with the 30N-60N lidar results reported in previous section. Figure 19: Comparison between Payerne microwave radiometer and SCIAMACHY version 6 ozone profiles. II.5.4 Summary and conclusions A first detailed comparison of the SCIAMACHY SGP 6.00 Diagnostic Data Set shows that the limb ozone data quality is very similar to that of the previous offline processor SGP A few minor improvements relative to V5 are seen in some parts of the atmosphere (less variability in upper stratosphere, reduced positive bias in middle and lower stratosphere in Tropics and Southern hemisphere). Nonetheless, none of the major data quality issues of SGP 5.02 have been resolved. Such a progress was also not expected, since not the target of the current development cycle. Most of the issues are likely caused by deficiencies at Level-1. Improvement in a future processor hopefully makes the SCIAMACHY limb ozone data set more competitive with that of other limb/occultation sounders. A first comparison with microwave radiometer ozone profiles at Lauder (southern hemisphere midlatitudes) and Mauna Loa (northern tropics) indicates a mostly positive bias in the SCIAMACHY mesospheric ozone profiles. For the overlapping altitude region (between 30 and 45 km) with collocated lidar observations, the agreement between the comparison of SCIAMACHY ozone profiles 29 40

31 with microwave radiometer or with lidar data yields consistent results. For Payerne (northern midlatitudes), we observe SCIAMACHY limb ozone being positively biased below ~53 km and negatively above. Table 8: First estimates of data quality of SCIAMACHY SGP 6.00 limb ozone profile based on comparisons to ozonesonde, stratospheric ozone lidar and microwave radiometers. Values represent the range of the quality indicator values across each (latitude, altitude) bin. SGP 6.00/Y 60N-90N 30-60N 30N-30S 30S-60S 60S-90S MEDIAN BIAS 15-20km % % % % -8 +8% 20-30km % -5 +2% 0 +15% % 0 +5% 30-45km % 0 +12% % % % 45-70km* % % % Other Major issues remain : AK-smoothing introduces vertical oscillations, very positive bias in US which differs NH/SH, strong SZA and altitude-dependence in Arctic, global dependence on SZA, scan angle dependence at O3 peak, negative decadal drift in MS, auxiliary data impact bias considerably. COMPARISON SPREAD (1σ) 15-20km 17-22% 12-40% 20-40% 12-40% >40% 20-30km 15-22% 7-12% 6-20% 8-12% 8-40% 30-45km 18-45% 8-27% 7-20% 8-40% 14-45% 45-70km** >35% >20% >35% Other Major issues remain : AK-smoothing introduces vertical oscillations, strong SZA and altitude-dependence in Arctic, increases during Antarctic O3 hole season, global dependence on SZA especially in US. *Based on one station per latitude bin. **Typically increasing towards mesopause. III Bibliography Hendrick, F., et al.: Retrieval of stratospheric and tropospheric BrO profiles and columns using groundbased zenith-sky DOAS observations at Harestua, 60 N, Atmos. Chem. Phys., 7, , doi: /acp , Hendrick, F. et al.: Multi-year comparison of stratospheric BrO vertical profiles retrieved from SCIAMACHY limb and ground-based UV-visible measurements, Atmos. Meas. Tech., 1, , doi: /amt , Hubert, D., N. Kalb, J.-C. Lambert et al.: Multi-TASTE Final Report (October 2008 October 2011), TN- BIRA-IASB-MultiTASTE-FR, MultiTASTE-FR-iss1revC-Oct2012.pdf, Hubert, D., A. Keppens, J. Granville, F. Hendrick and J.-C. Lambert: Multi-TASTE Phase F Progress Report #1 / October 2013 March 2014, Project Progress Report, TN-BIRA-IASB-MultiTASTE- Phase-F-PR1, 37pp., 17 Apr Hubert, D. et al.: Ground-based assessment of the bias and long-term stability of fourteen limb and occultation ozone profile data records, Atmos. Meas. Tech. Discuss., 8, , doi: /amtd , Noël, S., et al.: SCIAMACHY Quality Working Group (SQWG-3), Minutes of Status Telecon 20 Nov 2014, Dec

32 SCIAMACHY Quality Working Group: Readme file for SCIAMACHY Level 2 version 5.02 products, ENVI- GSOP-EOGD-QD , issue 1.2, 2P_README.pdf, IV Acronyms and abbreviations AK Averaging Kernel AMF Air Mass Factor BIRA-IASB Belgisch Instituut voor Ruimte-Aëronomie / Institut d aéronomie spatiale de Belgique (Belgian Institute for Space Aeronomy) CNRS/LATMOS Centre national de la recherche scientifique / Laboratoire "Atmosphères, milieux, observations spatiales" DDS Diagnostic Data Set DHF NDACC Data Host Facility DIAL Differential Absorption Lidar DLR Deutsches Zentrum für Luft- und Raumfahrt (German Aerospace Center) DOAS Differential Optical Absorption Spectroscopy Envisat ESA s Environmental Satellite ESA European Space Agency / Agence spatiale européenne EVDC Envisat Validation Data Centre FTIR Fourier Transform Infrared Spectroscopy GAW WMO s Global Atmospheric Watch HDF Hierarchical Data Format IUP Institut für Umweltphysik, University of Bremen KNMI Koninklijk Nederlands Meteorologisch Instituut Lidar Light Detection and Ranging LS Lower Stratosphere MS Middle Stratosphere Multi-TASTE Technical Assistance to Multi-mission Validation by Sounders, Spectrometers and Radiometers MWR Millimetre Wave Radiometer NDACC Network for the Detection of Atmospheric Composition Change (formerly NDSC) NILU Norwegian Institute for Air Research QWG Quality Working Group SAOZ Système d'analyse par Observation Zénithale SCD Slant Column Density SCIAMACHY Scanning Imaging Absorption spectrometer for Atmospheric CHartographY SCIAVALIG SCIAMACHY VALidation and Interpretation Group SGP SCIAMACHY Ground Processor SHADOZ Southern Hemisphere ADditional Ozonesondes SZA Solar Zenith Angle TASTE Technical Assistance to Envisat validation by Soundings, Spectrometers and Radiometers US Upper Stratosphere UT Upper Troposphere UV Ultra Violet UV-VIS Ultra Violet - Visible VCD Vertical Column Density VMR Volume Mixing Ratio WMO World Meteorological Organization 31 40

33 V Appendix V.1 O3 nadir column V.1.1 Dependence on solar zenith angle Figure 20: 10-year mean value (bullet) and 1sigma spread (error bar) of the percent relative difference between SCIAMACHY SGP Diagnostic data Set and ground-based network total ozone data, as a function of SCIAMACHY solar zenith angle (SZA). SGP 5.02 results are depicted in green and new SGP 6.00 results in red. Upper graph: SGP vs. Brewer network at stations from 60 N to 90 N; lower graph: SGP vs. Brewer network at stations from 30 N to 60 N

34 V.1.2 Dependence on cloud fraction Figure 21: 10-year mean value (bullet) and 1sigma spread (error bar) of the percent relative difference between SCIAMACHY SGP Diagnostic data Set and ground-based network total ozone data, as a function of SCIAMACHY fractional cloud cover. SGP 5.02 results are depicted in green and new SGP 6.00 results in red. Upper graph: SGP vs. Brewer network from 60 N to 90 N; lower graph: SGP vs. Brewer network from 30 N to 60 N

35 Validation Report : SCIAMACHY SGP upgrade from V5.02 to V6.00 V.2 V.2.1 O3 limb profile Impact of smoothing SCIAMACHY SGP AK Implicit (regridding) Gaussian Triangle Square Figure 22: Vertical and meridian structure of the median bias of SCIAMACHY limb ozone relative to correlative data which were smoothed using various low-pass filters (different rows). Solid lines indicate ozonesonde comparisons, dashed lines lidar comparisons 34 40

36 AK None Gaussian Triangle Box Figure 23: As above, but for the observed spread in the relative differences

37 V.2.2 Dependence on auxiliary data altitude, number density altitude, VMR pressure, number density pressure, VMR Figure 24: Vertical and meridian structure of the median bias of SCIAMACHY limb ozone relative to unsmoothed ozonesonde (solid) and lidar (dashed) data in different ozone profile representations

38 altitude, number density altitude, VMR pressure, number density pressure, VMR Figure 25: As above, but for the observed spread in the relative differences

39 V.2.3 Dependence on solar zenith angle Figure 26: Dependence on solar zenith angle of median bias of SCIAMACHY SGP 6.00 relative to lidar (top) and ozonesonde (bottom). From left to right: 60N-90N, 30N-60N, 20N-30S, 30S-60S, 60S-90S. Figure 27: As above, but for the observed spread in the relative differences

40 V.2.4 Dependence on month of year Figure 28: Dependence on month of median bias of SCIAMACHY SGP 6.00 relative to lidar (top) and ozonesonde (bottom). From left to right: 60N-90N, 30N-60N, 20N-30S, 30S-60S, 60S-90S. Figure 29: As above, but for the observed spread in the relative differences

41 V.2.5 Dependence on scan angle Figure 30: Dependence on scan angle of median bias of SCIAMACHY SGP 5.02 relative to lidar (top) and ozonesonde (bottom). Positive scan angles lie to the left ( East ) of the central viewing axis, negative to the right ( West ). Remark: the SGP 6.00 processor produces only one profile from four measured states in the scan sequence, its geolocation generally located closest to that of the -1 profile of SGP From left to right: 60N- 90N, 30N-60N, 20N-30S, 30S-60S, 60S-90S. Figure 31: As above, but for the observed spread in the relative differences