SCIAMACHY OZONE COLUMN VALIDATION WITH MODELS AND ASSIMILATION

H. J. Eskes(1) and A. Dethof(2)

(1)Royal Netherlands Meteorological Institute, De Bilt,The Netherlands(2)European Centre for Medium-Range Weather Forecast, Reading, UK

ABSTRACT

In this paper we report the results of three validation stud-ies. First, a comparison of the recently processed SCIA-MACHY ”validation reference set” ozone columns (soft-ware version 5.01) with assimilated ozone fields basedon the new GOME ozone column retrieval of KNMI (TO-GOMI). Second, a direct comparison between the SCIA-MACHY ”validation reference set” and the scientific re-trieval of the KNMI (TOSOMI algorithm). Third, a moni-toring of the early 2004 SCIAMACHY total ozone meteoproduct with the ozone analyses of the ECMWF model.The SCIAMACHY operational ozone column product hasimproved compared to ACVE-1, but an upgrade of theprocessor is still urgently needed to include the latest al-gorithm developments inplemented for the GOME ozonecolumn retrieval. The TOSOMI scientifically retrieved O3

column of KNMI is a stable product with a small bias< 1% compared to GOME-TOGOMI and an overall biasof −1.5% compared to Brewer and Dobson. The SCIA-MACHY operational ozone data showed large problems inMarch and April 2003.

1. INTRODUCTION

Data assimilation complements validation with indepen-dent observations. The differences between the modelforecast and new observations that have not been usedin the assimilation (the observation minus forecast depar-tures, OMF) provide a wealth of information about themodel performance, the quality of the observations andthe retrieval approach. These departures form the coreof each data assimilation system: all new observation arefirst compared with a model prediction of this observa-tion. Based on the OMF departures and knowledge of theobservation and forecast error statistics, the model stateis updated in the assimilation step to account for the in-formation brought by the observations. With the forecastmodel one can construct a model predicted value for eachof the typically millions of satellite observations avail-able, and statements can be made with great statisticalconfidence. This is an advantage compared to validationstudies with ground stations and balloon and aircraft cam-paigns, where only typically between one and a few hun-dred collocations are available. The forecast, however, isitself based on earlier observations of the same, or simi-lar, satellite sensors. Therefore assimilation of one dataset alone can not fully determine the quality of the satel-lite observations, for which independent observations arecrucial.

In this contribution we will discuss the use of ozone dataassimilation models to learn about the quality of the SCIA-

Figure 1. Coverage of the Sciamachy validation referenceset, August-October 2002. Shown are the mean SCIA-MACHY total ozone values during this period, griddedon a 2x3 degree grid.

MACHY near-real time (NRT) total column ozone productand the results of the scientific DOAS ozone retrieval al-gorithm of KNMI, called TOSOMI. This paper discussesthree validation studies:(i) Monitoring of the SCIAMACHY operational ozone col-umn product of ESA and the TOSOMI scientific product ofKNMI for the SCIAMACHY validation reference set (SVRS).These two products are monitored with global analysedozone fields obtained from an assimilation of GOME ozonecolumns retrieved with the KNMI TOGOMI DOAS soft-ware. This for the period August-October 2002.(ii) A direct comparison between the two retrievals.(iii) A monitoring of the same two ozone retrievals withthe ECMWF IFS assimilation model for the period January-April 2004.

For data assimilation a continuous and global data setwithout large data gaps would be ideal. Unfortunatelythe availability of SCIAMACHY NRT ozone column dataprocessed with the latest version 5.01 of the processor hasbeen sparse (see, e.g. figure 1 for a geographical overviewof the states availability for validation). In this paper wewill focus on the period August - October 2002. Since thelimited coverage does not allow for an assimilation of thedata we have monitored the available NRT data againstassimilated GOME ozone fields.

_______________________________________________________Proceedings of the Second Workshop on the Atmospheric Chemistry Validation of ENVISAT (ACVE-2)ESA-ESRIN, Frascati, Italy, 3-7 May 2004 (ESA SP-562, August 2004) ESC01HE

2. SCIAMACHY DATA SETS

2.1 ESA near-real time operational processor

The SCIAMACHY processor is described in the AlgorithmTheoretical Basis Document available on the ESA-ENVISATweb site [SCIAMACHY 2000]. The ozone column is re-trieved from the SCIAMACHY measurements based on theDOAS technique [Bovensmann 1999], and the current ver-sions of the software are based on the GOME data pro-cessor GDP v2.x. During the first ACVE meeting severalshortcomings have been identified with earlier versions3.51, 3.52, 3.53 of the near-real time (SCI NL) and meteo(SCI RV) total ozone products from SCIAMACHY [Lam-bert 2003; Eskes 2003b; Meirink 2003].

In early 2004, the SCIAMACHY NRT processor was up-graded to the newly operational version 5.01. The re-sults presented in this paper will be based on this newprocessor and on the latest operational data sets of early2004. In preparation of the ACVE-2 meeting ESA has pro-cessed about 1900 SCIAMACHY states and these stateswere made available to the validation teams in March2004. This was a subset of the 2591 states defined bythe validation teams, the so-called ”validation referencedata set”. The set of states is limited to the period fromJuly to November 2002 and does not allow for a study ofthe seasonal dependence of the data products.

2.2 TOSOMI Scientific Processor

The KNMI SCIAMACHY total ozone retrieval algorithm(TOSOMI, current version 0.31, November 2003) is an ap-plication of the TOGOMI algorithm to SCIAMACHY. Thenew GOME algorithm TOGOMI [Valks 2003] is based onthe total ozone DOAS (Differential Optical AbsorptionSpectroscopy) algorithm developed for the OMI instru-ment [Veefkind 2001]. With respect to total ozone col-umn retrieval using the DOAS method, the OMI, SCIA-MACHY and GOME instruments are very similar. Themain improvements of the new algorithm are:(i) treatment of the atmospheric temperature sensitivityby using effective ozone cross-sections calculated fromECMWF temperature profiles,(ii) improvements in the calculation of the air mass fac-tor, using the so-called empirical approach,(iii) using the Fast Retrieval Scheme for Clouds from theOxygen A-band (FRESCO) algorithm for the cloud cor-rection,(iv) a new treatment of Raman scattering in DOAS [DeHaan 2003]. This new formulation of DOAS explicitlyaccounts for the smearing of the solar Fraunhofer lines aswell as the atmospheric tracer absorption structures, and(v) air-mass factors based on semi-spherical polarization-dependent radiative transfer (KNMI DAK model).

The SCIAMACHY total ozone retrieval algorithm TOSOMIcombines a Sciamachy level-1 product reading modulewith the TOGOMI DOAS modules. The Fresco algorithm[Koelemeijer 2001] is applied to the Sciamachy spectra toobtain cloud fraction and cloud top height estimates. The

Figure 2. Latitude-dependent comparison of the meandifference betweenTOSOMI v0.31 ozone columns andground-based observations. Included are Brewer, Dob-son, Filter, SAOZ and DOAS measurements. Courtesy E.Brinksma, KNMI.

TOSOMI total ozone product is influenced by the qual-ity of the SCIAMACHY calibration procedure. No ad-ditional calibration corrections are applied to the ozonefit window (325-335nm), but a crude correction factorof 1.25 is applied to the radiances in the Fresco win-dow (758-772nm) because especially the cloud fractionretrieval is sensitive to the absolute reflectivity (status Oc-tober 2003).

The TOSOMI total ozone data is available in the form ofASCII data files (one file per day) from the TEMIS projectweb page [TEMIS].

The TOSOMI algorithm has been validated against an ex-tensive set of ground-based observations from Brewer,Dobson, SAOZ, DOAS instruments (summarised in a re-port by E. Brinksma [2004] and discussed in this pro-ceedings [Lambert 2004]), see figure 2. TOSOMI ozonecolumns are, on average, 1.5% lower than the ground-based data, with a rms difference of 4.9% (all instru-ments) or 4.6% (Brewer and Dobson instruments). Theseresults are based on a collocation distance criterion of 600km. No clear dependence on geographical location wasfound. The comparison with FTIR measurements is dis-cussed by Kopp [2004].

3. MODELS AND ASSIMILATION

The NRT SCIAMACHY ozone column data and the scien-tific TOSOMI ozone columns have been monitored withthe KNMI ozone column assimilation scheme TM3DAMand the operational ozone analyses of ECMWF.

3.1 The tm3dam ozone data assimilation system

The assimilation model TM3DAM (TM3 Data Assimila-tion Model) is described in detail in Eskes et al. [2003a].The transport model is based on the second moment ad-vection scheme and has 44 vertical levels and is run with aresolution of 3x2 degree. The model is driven by the me-teorological analyses of the European Centre for Medium-Range Weather Forecasts. Stratospheric ozone chemistryis described by two parametrisations, one for gas-phasechemistry and one for heterogeneous ozone loss in thepolar regions. The assimilation uses a fast parameterisedKalman filter approach which provides ozone analyses

Figure 3. Latitude-dependent comparison of the meandifference betweenSCIAMACHY v5.01 ozone columns(upper panel) orTOSOMI (lower panel) and assimilatedGOME-TOGOMI. Shown are the bias (dashed line) andstandard deviation (solid line).

and detailed forecast error estimates. Both OMF and ob-servation-minus analysis data sets are routinely provided.

The model assimilates KNMI TOSOMI near-real time ozonecolumns derived from the SCIAMACHY measurements.The data base of ozone analyses and daily ozone forecastare made available via the TEMIS web site [TEMIS].

For the validation of the SCIAMACHY validation refer-ence set which has become available early 2004 we haveperformed a separate assimilation run for August-October2002. This is based on GOME ozone columns retrievedwith the new TOGOMI DOAS ozone column retrieval code[Valks 2003]. TOGOMI and TOSOMI are both based on theOMI-DOAS algorithm [Veefkind 2001]. They differ onlyin their specific application to GOME or SCIAMACHY mea-surements respectively.

3.2 ECMWF ozone analyses

Ozone is fully integrated into the ECMWF forecast modeland analysis system as an additional three-dimensionalmodel and analysis variable similar to humidity. The

Figure 4. Viewing angle dependence of the mean differ-ence betweenSCIAMACHY v5.01 ozone columns and as-similatedGOME-TOGOMI. Shown are the forward scan(crosses) and backscan (open symbols)OMF differences(in Dobson units). Negative angles: East. Positive an-gles: West. Shown are states with state identifiers 6 and7 (low- and midlatitudes).

forecast model includes a prognostic equation for the o-zone mass mixing ratio. Ozone chemistry is describedby a parametrisation which is an updated version of Car-iolle and Deque. Ozone data assimilation is performedin the operational ECMWF 4D-VAR assimilation systemand also in the 40-year reanalysis project (see [Dethof2002] for more details). The operational implementationis based on SBUV-2 ozone columns and stratospheric pro-files, GOME ozone columns (until June 2003), and, morerecently, on MIPAS NRT ozone profiles [Dethof 2004].ECMWF is actively participating in the validation activ-ities for ENVISAT.

4. SCIAMACHY VS ASSIMILATED GOME

Figure 1 shows the geographical distribution of the SVRS.In this section we present results of the comparison of theSCIA NL version 5.01 processed at these locations andthe TM3DAM assimilated total ozone analyses based onthe GOME-TOGOMI retrievals. For comparison we alsoshow the TOSOMI ozone columns retrieved for the sameSVRS.

In figure 3 we show the latitude dependence of the com-parison. In the Northern Hemisphere over Europe thereare a reasonable amount of reference states available formeaningful conclusions.

The operational retrieval, SCIAMACHY NRT, show a biaswhich, on average is smaller than 1% compared to theassimilated GOME-TOGOMI results. This result is in con-trast to the large negative bias reported during the firstACVE of December 2002, and demonstrates an improve-ment over the previous versions of the operational ozonecolumn product.

The bottom figure shows that GOME-TOGOMI and SCIA-MACHY-TOSOMI retrievals have a very small average rel-ative bias (< 1%). Since the underlying algorithms areidentical, this result shows that calibration aspects (e.g.radiometric calibration) do not substantially influence theozone column DOAS retrieval in the 325-335 nm window.The fit residuals of the TOSOMI and TOGOMI retrievalsare both small, of the order of 0.5%. It is important tonote that the monthly-mean latitude-dependent bias be-tween the model forecasts and the TOGOMI data that isassimilated is generally < 1%, and the global averagebias is negligible. Latitude-longitude plots of the biasseem to show a better consistence (less scatter) for TO-SOMI than for SCIAMACHY NRT when compared with theassimilated model fields.

In the Southern Hemisphere it is more difficult to drawconclusions because of a more sparse distributon of thereference states. For instance the positive bias observedbetween 30-40S is related to a single state west of SouthAfrica, and is not statistically meaningful. The compari-son is further complicated by strong gradients related tothe ozone hole, the anomalous splitting of this ozone holein 2002, the large solar zenith angles and possibly biasesintroduced by the model (due to values transported fromthe unobserved dark winter pole). More data is definitelyneeded to arrive at conclusions.

Figure 4 shows the OMF difference for the SCIAMACHYoperational ozone columns. This shows an increase inozone from East to West viewing pixels of about 2%.On top of this there is a difference between backscan andforward scan pixels: the backscan is higher for the Eastpixels and lower for the West pixels. Similar plots weremade for the TOSOMI which shows very little viewing-angle dependence. However, the backscan - forward scanasymmetry is also observed in the TOSOMI data set. Apossible reason for this is a small error in the geolocationparameters in the level-1 files.

Figure 5 shows that the operational and TOSOMI retrievalsboth have a significant cloud-fraction dependence. Forthe SCIAMACHY NRT product the ozone amount increasesby 2-3% when the cloud fraction increase from 0 to 1.The TOSOMI retrieval increase by 1-1.5% from 0 to 0.9cloud fraction. Cloud fraction 1 is exceptional, with 1.5to 2% higher ozone and a high population compared toOCRA. This suggests that at least part of the cloud frac-tion dependence is related to peculiarities of the FRESCOscheme (see also the paper by Fournier [2004]).

5. SCIAMACHY NRT VS TOSOMI

For the direct comparison of the operational SCIAMACHYozone product and the TOSOMI scientific product we haveapplied the TOSOMI algorithm to the updated level-1 dataof the SCIAMACHY validation reference set (SVRS) thatbecame available early 2004. The two level-2 productswere compared on a 1 to 1 basis.

Figure 6 shows the direct comparison of the ozone valuesfor the northern hemisphere, tropics and southern hemi-

Figure 5. Cloud-fraction dependence of the mean differ-ence betweenSCIAMACHY v5.01 ozone columns (upperpanel) orTOSOMI (lower panel) and assimilatedGOME.Solid line: Observation-minus-forecast inDU. Dashedline: Number of measurements with a certain cloud frac-tion for OCRA (upper panel) orFRESCO (lower panel).

sphere mid-latitudes (top panel). There is a good corre-spondence between the retrievals, with a mean agreementbetter than 15 DU for all ozone values (mean bias 0.5%)and a standard deviation of 6 DU (about 2%).

The South Pole shows a different behaviour, with largedifferences for the larger ozone values. Above 350 DUthe two retrievals start to deviate. The SCIA NL productreaches higher ozone levels of around 400 DU, whereasthe higher ozone values of TOSOMI reach up to 500 DU.We note that these high values are exceptional and arerelated to the split ozone hole conditions. These valuesoccur close to the pole with high solar zenith angles.

Figure 7 shows the comparison as a function of latitude.Between 50S and 50N the bias between the two productsis within 4%. For larger latitudes (and solar zenith angles)the difference becomes larger, especially in the southernhemisphere.

Figure 8 looks at the intermediate DOAS result, namelythe slant column. This is directly related to the depth of

Figure 6. Sciamachy v5.01 operational ozone columnversus theTOSOMI ozone column, for latitudes North of60S (top) and South of 60S (bottom). Solid line: mean.Dotted line: mean± standard deviation. Dashed: mini-mum, maximum.

the absorption features and is related to the amount ofozone along the path the light tarvels through the atmo-sphere. The figure shows striking differences. Two im-portant aspects to explain the differences are the differenttreatment of rotational Raman scattering [de Haan 2003]which causes TOSOMI to be higher than SCIAMACHY NRTby 2-3% (low latitudes) to 7-9% (high latitudes). A sec-ond important aspects is the difference in cross-sectionsused (SCIAMACHY-FM versus GOME-FM) which causesTOSOMI to be lower than SCIAMACHY NRT by about 6%.Especially the large impact of the rotational Raman ef-fect demonstrates the need for an urgent upgrade of theSCIAMACHY processor to a DOAS scheme that accountsfor this.

6. OZONE MONITORING AT ECMWF

ENVISAT retrieval products are routinely monitored and/orassimilated with the ECMWF operational model. Theseinclude:(i) ESAs SCIAMACHY operational total column ozone me-teo product, monitored at ECMWF since February 2003,and(ii) KNMIs TOSOMI NRT total column ozone data, moni-tored at ECMWF since 21 March 2004.

Figure 7. Difference betweenSCIAMACHY NRT (v5.01)and TOSOMI ozone columns (in %) as a function of thelatitude. Different symbols correspond to different stateid’s (2-7).

Figure 8. Difference betweenSCIAMACHY NRT (v5.01)and TOSOMI slant columns (in %) as a function of thelatitude. The symbols correspond to state id’s 7 (nearequator), 6, 5, 4, 3, and 2 (high solar zenith angle).

Figure 9 shows the dependence of the SCIAMACHY me-teo product with time. The figure demonstrates problemsin March 2004, which are related to inadequate leakagecurrent calibration and a wrong initialisation file. A fur-ther change happens around 28 March and from this timethere is a worse agreement with ECMWF ozone columns.

Figures 10 and 11 shows the ozone measurements and theOMF departures for April. Clearly the operational prod-uct is unrealisticly low. The two jumps around 11 and25 March in the SCIAMACHY METEO product are not ob-served in the TOSOMI product. Also observed from thefigure is a data gap between about 90-150E: the data fromftp-pde.envisat.esa.in is not available in NRT, and is toolate for analysis. A similar problem occurs with the TO-SOMI product: although all orbits are available, they donot arrive in NRT.

Figure 11 demonstrates the large negative bias of around60 DU for the SCIAMACHY METEO product. For TOSOMI

Figure 9. SCIAMACHY meteo product ozone columns asa function of time, from January 1 to April 18 2004, andlatitude, in DU.

the mean departure with the ECMWF model is about 20DU. At least part of this 20 DU will be related to theECMWF analyses, which is currently assimilating only alimited set of ozone data, consisting of SBUV columnsand stratospheric profiles. Both ENVISAT MIPAS ozoneprofiles and GOME ozone column data are no longer avail-able in this month (on a global scale).

The problems with the meteo products have in the meantime been identified and corrected.

7. CONCLUSIONS AND PERSPECTIVES

The main conclusions for the three inter-comparisons pre-sented in this contribution are:

Sciamachy vs assimilatedGOME-TOGOMI

The reprocessed SCIAMACHY operational data set, andthe TOSOMI scientific product have been compared withassimilated GOME TOGOMI ozone fields. This was donefor the validation reference set, for the period August-October 2002. The main results are:• The overall bias between (assimilated) GOME TOGOMI,SCIAMACHY-operational and TOSOMI is small, < 1%.This is a clear improvement compared to the status of theoperational products reported at the first ACVE.• The TOSOMI product is about 1.5% lower than Brewerand Dobson measurements.• The standard deviation of TOSOMI and SCIAMACHYNRT minus assimilated GOME is about 3%. This is agood result, demonstrating that the noise in both prod-ucts is low.• A viewing angle dependence was detected in the NRTproduct of about 2% difference between East and West.•A cloud fraction dependent bias is found in both the TO-SOMI and SCIA NL products. There is a 2 - 2.5% ozonedifference between cloud fraction 0 and cloud fraction 1.For the TOSOMI product the pixels with cloud fraction =1 are anomalous, which may be related to the FRESCOcloud retrieval scheme.• The validation reference set is too limited for validationbased on data assimilation: full coverage is needed formore detailed studies.

Figure 10. SCIAMACHY meteo product (top) andTO-SOMI (bottom) ozone columns in April 2004, in DU.

SCIAMACHY NRT v5.01 versusTOSOMI

The SCIAMACHY NRT level-2 ozone column product wascompared on a one-to-one basis with the TOSOMI scien-tific product for the validation reference set. The mainfindings are:• There is a reasonable comparison between SCIAMACHYNRT and TOSOMI ozone columns for low- and mid-lat-itudes: a latitude dependent bias less than 4% is foundfor the latitude range [ 50S, 50N ], and the overall bias isonly about 0.5%. This result is quite surprising, given thelarge differences between the two algorithms.• We find considerable differences at high latitudes, es-pecially south of 60S: a bias up to 5% (NH), 10% (SH)is found between the two retrievals. Note also the specialconditions and anomalously high ozone values during the2002 vortex splitting.• The differences in slant column retrievals reflect differ-ences in the treatment of rotational Raman scattering andthe different choice of cross sections. These two aspectshave a large effect on the retrieval of opposite sign.• The geolocation information (angles) provided in thelevel-1 product needs to be checked again (especially theviewing zenith angle). This is illustrated, for instance, bythe forward scan / backscan a-symmetry.

Sciamachy ozone monitoring atECMWF

The operational SCIAMACHY meteo product and the KNMITOSOMI ozone columns have been monitored with theECMWF model. The main results are:• The ESA product is normally relatively stable, but therewere large jumps observed in ozone (order 40 DU) inMarch and April 2004.• A negative bias (−30 DU) is found in the compari-son between the SCIAMACHY meteo data and the ECMWF

Figure 11. Observation minusECMWF forecast depar-tures for theSCIAMACHY meteo product (top) and fortheTOSOMI product (bottom). Scale in %.

model ozone. This bias got worse after 28 March 2004(values of 60 DU in global mean).• The TOSOMI product agrees better with the ECMWFozone analyses than the SCIAMACHY NRT product (−20DU difference).• A data gap exists between about 90-150E. Data fromftp-pde.envisat.esa.in is not available in NRT and comestoo late for analysis.• Geolocation information (e.g. FOV, SZA) is still not in-cluded in the SCIAMACHY meteo data product, despiterepeated requests.• The stability and NRT delivery of the SCIAMACHY me-teo product is currently not good enough for an assimila-tion of the data at ECMWF.

To conclude, we find the the SCIAMACHY operationaltotal ozone product has improved compared to the sta-tus in 2002 and 2003. Biases are of the order 0-4% forthe tropics and mid-latitudes. Larger differences, up to10%, are found at high latitudes or solar zenith angles.The noise on the product is low. Recently developedDOAS ozone column retrieval algorithms (including TO-SOMI/TOGOMI) demonstrate the need for an upgrade ofthe SCIAMACHY operational processor. The KNMI TO-SOMI product is available for the period January 2003 -present. It is a stable product which has shown an in-sensitivity towards level-1 product updates and shows agood agreement with the GOME data set processed withthe TOGOMI algorithm. The monitoring with the ECMWFmodel has revealed problems with the ESA product dur-

ing March and April 2004. There is a data gap in thenear-real-time data and viewing geometry information isstill missing in the meteo product.

ACKNOWLEDGEMENTS

HE acknowledges Ankie Piters for a careful reding of themanuscript. The work presented includes contributionsfrom several colleagues, in particular Jan Fokke Meirink,Ronald van der A en Ellen Brinksma. HE further ac-knowledges financial support by the Netherlands Agencyfor Aerospace Programmes (NIVR).

REFERENCES

Bovensmann, H., et al., SCIAMACHY: Mission Objec-tives and Measurement Modes, J. Atm. Sci.,56, 127-150, 1999.

Brinksma, E., Validation of SCIAMACHY TOSOMIozone columns with groundbased data, draft report,KNMI, De Bilt, The Netherlands, 2004.

Dethof, A., and E. Holm, Ozone assimilation at ECMWF,Q. J. R. Meteorol. Soc., in press, 2004.

Dethof, A., and E. Holm, Ozone in ERA40: 1991-1996,ECMWF Technical Memorandum no. 377, August2002.

Eskes, H. J., van Velthoven, P. F. J., Valks, P. J. M.and Kelder, H. M., Assimilation of GOME total ozonesatellite observations in a three-dimensional tracertransport model, Q. J. R. Meteorol. Soc.,129, 1663-1681, 2003a.

Eskes, Henk, Jan Fokke Meirink, and Ankie Piters,Comparison of SCIAMACHY near-real-time ozonecolumns with GOME assimilated ozone, Proceedingsof the Envisat Validation Workshop, 9-13 December2002, ESRIN, Frascati, Italy, 2003b.

Fournier, N., et al., SCIAMACHY cloud product valida-tion, this proceedings, 2004.

de Haan, J.F., Accounting for Raman Scattering inDOAS, SN-OMIE-KNMI-409, KNMI, 2003.

Koelemeijer, R.B.A., P. Stammes, J.W. Hovenier, and J.F.de Haan, A fast method for retrieval of cloud parame-ters using oxygen A-band measurements from GOME,J. Geophys. Res.,106, 3475-3490, 2001.

Kopp, G., et al., Validation of SCIAMACHY ozone col-umn densities and profiles using ground-based FTIRand Millimeter Wave measurements, this proceedings.

Lambert, J. C., et al.,SCIAMACHY ozone column valida-tion, this proceedings, 2004.

Lambert, J. C., et al.,, Proceedings of the Envisat Vali-dation Workshop, 9-13 December 2002, ESRIN, Fras-cati, Italy, 2003.

Meirink et al.,Verification of SCIAMACHY near-real timeand meteo level-2 products: DOAS UV-Vis major tracegases, clouds, aerosols, and geolocation, Proceedingsof the Envisat Validation Workshop, 9-13 December2002, ESRIN, Frascati, Italy, 2003.

SCIAMACHY Validation Site, http://www.sciamachy-validation.org/.

SCIAMACHY Level 1c to 2 Off-line Process-ing Algorithm Theoretical Basis Document,R. Spurr and W. Balzer, DLR, December2000, available at http://atmos.af.op.dlr.de/cgi-bin/home.cgi?page=projdocs.

TEMIS project web ite, http://www.temis.nl/.

Valks, P., and R. van Oss, TOGOMI Algorithm Theoreti-cal Basis Document, KNMI, November 2003.

Veefkind, J.P. and J.F. de Haan, OMI Algorithm Theo-retical Basis Document, Barthia, P.K (ed), Volume II- Chapter 3, DOAS Total Ozone Algorithm, ATBD-OMI-02, Version 1.0, September 2001.