Retrieval of atmospheric parameters from GOMOS data

Transcription

1 Atmos. Chem. Phys., 1, , 21 doi:1.5194/acp Author(s) 21. CC Attribution 3. License. Atmospheric Chemistry and Physics Retrieval of atmospheric parameters from GOMOS data E. Kyrölä 1, J. Tamminen 1, V. Sofieva 1, J. L. Bertaux 2, A. Hauchecorne 2, F. Dalaudier 2, D. Fussen 3, F. Vanhellemont 3, O. Fanton d Andon 4, G. Barrot 4, M. Guirlet 4, A. Mangin 4, L. Blanot 4, T. Fehr 5, L. Saavedra de Miguel 5, and R. Fraisse 6 1 Finnish Meteorological Institute, Earth Observation, Helsinki, Finland 2 Laboratoire Atmosphères, Milieux, Observations Spatiales, Université Versailles St-Quentin, CNRS-INSU, Verrières-le-Buisson, France 3 Institut d Aéronomie Spatiale de Belgique, Brussels, Belgium 4 ACRI-ST, Sophia Antipolis, France 5 European Space Research Institute (ESRIN), European Space Agency, Frascati, Italy 6 EADS-Astrium, Toulouse, France Received: 14 January 21 Published in Atmos. Chem. Phys. Discuss.: 19 April 21 Revised: 6 September 21 Accepted: 2 December 21 Published: 14 December 21 Abstract. The Global Ozone Monitoring by Occultation of Stars (GOMOS) instrument on board European Space Agency s ENVISAT satellite measures attenuation of stellar light in occultation geometry. Daytime measurements also record scattered solar light from atmosphere. The wavelength regions are ultraviolet-visible band nm and two infrared bands at nm and at nm. From UV-Visible and IR spectra vertical profiles of O 3, NO 2, NO 3, H 2 O, O 2 and aerosols can be retrieved. In addition re are two 1 khz photometers at blue nm and red nm. Photometer data are used to correct spectrometer measurements for scintillations and to retrieve high resolution temperature profiles as well as gravity wave and turbulence parameters. Measurements cover altitude region 5 15 km. Atmospherically valid data are obtained in 15 1 km. In this paper we present an overview of GOMOS retrieval algorithms for stellar occultation measurements. The low signal-to-noise ratio and refractive effects due to point source nature of stars have been important drivers in development of GOMOS retrieval algorithms. We present first Level 1b algorithms that are used to correct instrument related disturbances in spectrometer and photometer measurements The Level 2 algorithms deal with retrieval of vertical profiles of atmospheric gaseous constituents, aerosols and high resolution temperature. We divide presentation into correction for refractive effects, Correspondence to: E. Kyrölä (erkki.kyrola@fmi.fi) high resolution temperature retrieval and spectral/vertical inversion. The paper also includes discussion about GO- MOS algorithm development, expected improvements, access to GOMOS data and alternative retrieval approaches. 1 Introduction GOMOS (Global Ozone Monitoring by Occultation of Stars) is a spectrometer on board European Space Agency s Envisat satellite (see Bertaux et al., 1991, 2, 24, 21; Kyrölä et al., 24; ESA, 21, and handbooks/gomos/). GOMOS uses method of stellar occultations in monitoring of ozone and or trace gases in Earth s middle atmosphere (see 1). The wavelength regions are ultraviolet-visible band nm and two infrared bands at nm and at nm. The species that can be retrieved from se wavelength areas are O 3, NO 2, NO 3, H 2 O, O 2, neutral density and aerosols. The two photometers work at blue nm and red nm with a frequency of 1 khz. Photometer data are used to make scintillation correction for spectrometer data. From photometer data it is also possible to retrieve high resolution (resolution of 2 m) temperature profiles in range of 15 4 km using spatial separation of rays by chromatic refraction. Moreover, photometer data can be used for middle atmosphere turbulence studies. GOMOS measurements start at 15 km tangent altitude and may reach 5 km if not interrupted by clouds. Atmospherically valid data are obtained in 15 1 km. Published by Copernicus Publications on behalf of European Geosciences Union.

2 E. Kyrölä et al.: GOMOS retrieval algorithms E. Kyrölä et al.: GOMOS retrieval algorithms E. Kyrölä et al.: GOMOS retrieval algorithms 21 I ref (!) I occ (!) I ref (!) I occ (!) Measured signal (e) differen from z 2. If a s dens red wav The principle of stellar occultation measurement. The measurements above atmosphere are used to buildup up reference spectrum for star. The measurements through at- atmosphere are areused to calculate horizontal transmission spectra. These Theseserve as asa starting point for geophysical retrieval. The main light sources for GOMOS are stars that provide 1. sufficient The principle flux of in stellar spectral occultation range 25 1 measurement. nm where The measurements GOMOS above detectors atmosphere and star are tracking used to system build up work. reference spectrum Stars must be brighter for than star. The visual measurements magnitude through 4 and acceptable atmosphere are temperature used to calculate range of stars horizontal is 3 3 transmission K. spectra. Examples These serve of stellar as a starting spectra point measured for geophysical by GOMOS retrieval. are shown in 2. 1 The figure 5 shows clearly large differences between spectra from cool and hot stars. The point source character of stars and weakness of ir radiation make it necessary to consider several questions that are not relevant to more familiar solar occulta- 1 4 tion method (see, e.g. Chu et al., 1989; Glaccum et al., 1996). Light from a far away point source appears as parallel rays or equivalently 1 3 as plane waves when entering into atmosphere. In atmosphere rays are refracted from origi- 1 5 nal direction and deflection angle depends on wavelength and neutral density at tangent altitude. The originally parallel 1 2 light rays diverge after passage through atmosphere leading to a reduced Wavelength intensity (nm) (see 3). Moreover, fluctuations in neutral density distribution lead to inhomogeneous 2. Star spectra distribution measured ofby intensity, GOMOS. which The spectrum can be seen of as scintillations visually brightest in starradiation in sky, measurements. Sirius or α CMa, is shown by The black low line. signal-to-noise Sirius visual magnitude ratio means is 1.44 that we and must effective put additional effort on data retrieval from noisy data and we must 1 3 surface temperature (determined from spectral class) 11 K. The blue line shows spectrum for star β Cru with magnitude and what temperature additional 3 light K. sources, The red line likerepresents auroral and understand or spectrum natural of emissions, star ɛ Carmay withcompete magnitude with 1.86 and stellar temperature signal. As 41 evident K. Thefrom visual magnitudes 2, different determine stars provide flux at vastly 55 nmvary- F = spectra 3 F ref 1 m/2.5 4 and radiation. 5 6 fluxes. 7The 8 differences 9 between 1 by 1 2 2ing stars are most important Wavelength at (nm) edges of GOMOS spectral range. In UV-part cool stars radiate very weakly and this 2. isstar reflected spectra inmeasured poor retrieval by GOMOS. results forthe highspectrum altitude ozone. of In visually brightest water vapour star in band sky, Siriusnm, or αoncma, or is shown hand, by cool black stars line. aresirius relatively visual better magnitude than hot is stars 1.44but and brightness effective of surface temperature star cannot (determined be ignoredfrom as 2spectral shows. class) 11 K. The blue line shows spectrum for star β Cru with magnitude and temperature 3 K. The red line represents spectrumatmos. of Chem. star ɛ Phys., Car with 1, magnitude , and temperature 41 K. The visual magnitudes determine flux at 55 nm by F = F ref 1 m/2.5. Measured signal (e) Measured signal (e) Wavelength (nm) 2. Star spectra measured by GOMOS. The spectrum of visually brightest star in sky, Sirius or α CMa, is shown by black line. Sirius visual magnitude is 1.44 and effective surface temperature (determined from spectral class) 11 K. The blue line shows spectrum z 1 for star β Cru with magnitude and temperature 3 K. The red line represents z t spectrum of star ɛ Car 2 with magnitude 1.86 and 1 temperature 41 K. The visual magnitudes determine flux at 55 nm by t F = F 2 ref 1 m/ Chromatic refraction and its impact on detection times of different colours. A measurement z 1 at time t 1 receives blue light from tangent point z 1 and red light from tangent point z 2. If a satellite is assumed to move z t 2 clockwise along a circular 1 orbit, density fluctuations near tangent point z 2 are first observed at t red wavelengths at time t 1 and later at time t 2 at blue wavelengths Chromatic refraction and its impact on detection times of different colours. A measurement at time t 1 receives blue light from 3. Chromatic tangent point refraction z 1 and and its redimpact light from on detection tangenttimes pointof different z 2. If a colours. satellite is A assumed measurement to move at clockwise time t along 1 receives a circular blue orbit, light density fluctuations near tangent point z from tangent point z 1 and red light 2 are first observed at from tangent point red wavelengths at time t z 2. If a satellite is assumed 1 and later at time t to move clockwise 2 at blue wavelengths. along a circular orbit, density fluctuations near tangent point z 2 are first observed at red wavelengths Because a at half time ofteach 1 and orbit later at istime illuminated t 2 bluebywavelengths. sun, GOMOS detectors are illuminated by scattered solar light, which easily dominates stellar signal in daytime measurements. This is called background radiation. The twodimensional CCD detectors used by GOMOS make it, however, possible to estimate background radiation coming from sun and or extended sources and separate it from star signal. The former radiation is detected all over CCD whereas star radiation is localised. This estimation must be very accurate in order not to distort weak stellar signal. The separation also leaves noise to stellar signal. GOMOS occultation retrievals in daytime measurements suffer from inaccurate estimation of background and ir quality is presently low. The solar signal itself

3 E. Kyrölä et al.: GOMOS retrieval algorithms can be used for atmospheric retrievals in same way as done with OSIRIS/Odin (Llewellyn et al., 24) and SCIA- MACHY/ENVISAT (Bovensmann et al., 1999) instruments. Algorithms and processing for GOMOS scattered solar light retrievals are under development and y are not discussed in this paper (see Taha et al., 28, and Tukiainen et al., 28). GOMOS is a novel instrument concept in past re have been only a few attempts to exploit stellar occultations for studies of Earth s atmosphere. The first such measurements were made already 4 years ago (Hays and Roble, 1968; Roble and Hays, 1972). A more recent effort is UVISI instrument onboard MSX-satellite. It has carried out stellar occultations successfully during , providing samples of ozone profiles in stratosphere (Yee et al., 22). Most of GOMOS data processing algorithms are eir completely new or y have been tailored specifically for stellar occultations. The algorithms discussed here are used in ENVISAT GOMOS ground segment supported by ESA. The processing is divided into two levels. Level 1b deals with instrument related corrections, geolocation, and calculation of transmissions that are basis for Level 2 processing. Level 2 deals with retrieval of atmospheric parameters like constituent and temperature profiles. The algorithms have been developed in contract between ESA and ACRI-ST company in Sophia-Antipolis, France. The scientific institutes, Finnish Meteorological Institute, Service d Aéronomie and Institut d Aéronomie Spatiale de Belgique, form GO- MOS Expert Support Laboratory-coalition (ESL) that has been scientific partner of ACRI in algorithm development. The GOMOS Level 1b and Level 2 algorithms have been written down in great detail (presently in 827 pages) in GOMOS Detailed Processing Model-document (GOMOS ESL, 1998). The more scientific description of GO- MOS processing is GOMOS Algorithm Theoretical Basis document (ATBD, presently 74 pages) instruments/gomos/atbd/. The aim of this paper is to describe relatively concisely all methods that are used to process raw GOMOS data to atmospheric profiles of trace gases and temperature. The emphasis is in scientific justification of selected processing algorithms. The general GOMOS overview, validation, error analysis and scientific results are discussed in or papers of this special issue. The GOMOS data processing algorithms described in this paper are used by ESA operational processor version 5 (It corresponds to version 6 of ACRI s GOMOS processing prototype GOPR). This version deviates only slightly from earlier processor version 4. The differences to upcoming version 6 are discussed in Sect. 14. Section 2 presents what is known about physics involved in GOMOS measurements. This part concerns, refore, our understanding of GOMOS forward problem. We will concentrate on processes that are essential in understanding of GOMOS data retrieval algorithms. Section 3 summarises briefly properties of GOMOS instrument. Section 4 discusses all important geolocation for GOMOS measurements. Section 5 focuses on Level 1b processing where instrumental corrections are applied. In Sect. 6 we present briefly photometer Level 1b processing, which is very similar to spectrometer processing. Error estimation for transmissions are presented in Sect. 7. Introduction to Level 2 processing is given in Sect. 8. Refraction corrections that are essential for stellar occultation measurements are given in Sect. 9. The retrieval of high resolution temperature from photometer time delay is presented in Sect. 1. Sections deal with retrieval of atmospheric parameters from transmissions. In Sect. 14 we present main developments of GOMOS processing in past and expected future improvements. Section 15 gives basic information about access to GOMOS data and related documentation. Section 16 discusses alternative methods for GOMOS retrieval. 2 GOMOS measurement physics An occultation measurement can be thought as a measurement of atmospheric transmission between a source and instrument (for references and reviews for occultation method, see Elliot, 1979; Hays and Roble, 1968; Smith and Hunten, 199; Kyrölä et al., 1993; Bertaux et al., 21). Transmission cannot be measured directly but it needs to be calculated from measurements above and through atmosphere. As a first guess transmission can be modelled with well-known Beer-Lambert law T ext = e τ (1) where optical depth τ is given by τ(λ) = σ j (λ,t ( r(s)))ρ j ( r(s))ds. (2) j l Here ρ j s are constituent densities and σ j s are absorption or scattering cross-sections. Cross sections depend on wavelength and often also on temperature. The integration is made along light path joining instrument and source. The modelled transmission gives desired connection between measurements and geophysically interesting constituent profiles. In occultation measurements Beer-Lambert law alone is not able to explain observed transmissions. A parallel ray bundle from a star will be strongly disturbed by refractive effects in atmosphere. The following effects can be identified: 1. chromatic bending of rays, 2. refractive dilution, 3. scintillations. Atmos. Chem. Phys., 1, , 21

4 11884 E. Kyrölä et al.: GOMOS retrieval algorithms A nearly exponential decrease of atmospheric neutral density leads to a bending of light rays propagating tangentially with respect to atmosphere. The bending angle increases with a decreasing tangent altitude. Refraction transforms parallel incident rays into a diverging beam. This results in dilution of related intensity, see 3. The dependence of refractive index on wavelength leads also to a spatial separation of rays of different colours. As shown in 3, spatial separation of rays takes place not only inside atmosphere but separation increases in free propagation after atmosphere. Any instantaneous multi-wavelength measurement of a stellar spectrum through atmosphere is, refore, not connected to a unique ray connecting instrument and star. If we label rays by ir tangent heights, we see that at a given instant of time a measurement will be characterised by a range of tangent heights. If we want to attach only one tangent height we have to combine data from measurements at different observation times. The neutral air density distribution, origin of refraction, includes irregularities that are generated by internal atmospheric gravity waves and various processes generating turbulence. The characteristic scales range from a few kilometres down to a dissipation scale, which is typically smaller than 1 m. These air density irregularities lead to a variation in bending angle and refore also in transmitted intensity. Variations are called scintillations. Largest scale irregularities (associated with internal gravity waves) are strongly stretched in horizontal direction. The interaction of stellar light with se structures can be described in geometric optics approximation. It leads to random focusing and defocusing of light rays. For description of scintillations that are caused by small-scale irregularities generated by isotropic turbulence, diffraction ory needs to be used. For furr explanation, see Sofieva et al. (29). During daytime measurements GOMOS CCD s are illuminated by sunlight scattered in atmosphere one or more times. There may also appear so-called external stray light that originates from sun illuminated parts of satellite or atmosphere not directly in FOV of GOMOS. As already mentioned, scattered sunlight can be used for atmospheric retrieval (not discussed in this paper) but at same time this part is unwished background radiance that needs to be removed from star related occultation signal before retrieval based on Eq. (1). Background radiation can also be generated by auroral and atmospheric emissions. Clouds are usually factor that terminates an occultation. Clouds also affect GOMOS retrievals in some cases. Noctilucent clouds (NLC) reside high in atmosphere (about 82 km). NLCs are optically (also vertically) thin and star tracking is not interrupted. NLCs are easily detected during daytime from increased scattered sunlight in photometers and spectrometers (see Pérot et al., 21, in this special issue). The background radiance removal, however, does not work correctly for very thin sources of radiation and this produces erroneous transmission values. During nighttime background removal is not used but contribution to transmission is too small to be detected in individual occultations. 3 GOMOS instrument GOMOS is a medium resolution spectrometer (see 4). Spectrometers A1 and A2 cover UV-visible wavelength region nm with 1416 pixels with spectral width.31 nm. The spectral resolution is.8 nm. The spectrometer B1 covers nm with nm wide pixels. The spectral resolution is.13 nm. The spectrometers B2 covers nm with 5.52 nm wide pixels. The spectral resolution is.13 nm. The two photometers work at blue nm and red nm wavelengths at a frequency of 1 khz (see Popescu and Paulsen, 1999). The high sensitivity requirement (stars are weak sources of light) down to 25 nm was a significant design driver and led to an all reflective optical system design for UVIS part of spectrum and to a functional pupil separation between UVIS and IR spectral regions. Due to requirement of operating with very faint stars (down to magnitude 4) sensitivity requirement to instrument is very high. Consequently, a large telescope (3 cm 2 cm aperture) had to be used to collect sufficient signal. Detectors with high quantum efficiency and very low noise had to be developed to achieve required signal-to-noise ratios. GOMOS includes a slit to limit background radiation (mainly solar radiation scattered from atmosphere) entering to detectors. The GOMOS star tracker is sensitive inside range 6 1 nm. The star tracker needs to keep star image in centre of slit with accuracy of one pixel. The pointing stability is better than 4 microradians. The star tracker is able to follow a star down to 5 2 km. The altitude limit depends on qualities of a star, state of atmosphere (e.g., clouds), and limb illumination condition. GOMOS observes only setting stars. Acquisition of a setting star is much easier than a rising star (due to refraction) and getting both modes was impossible due to demanding visibility requirements in Envisat platform design. 4 Geolocation The accurate geolocation is of central importance for all satellite instruments. This is especially important for limb viewing instruments where accuracy of vertical geolocation affects directly accuracy of vertical profiles. For GOMOS stars provide excellent means to get accurate vertical knowledge of measurements. Each GOMOS measurement of atmospheric transmission is geolocated. Several different factors contribute to this: orbit of satellite, shape of Earth, state of atmosphere, and stellar co-ordinates. In absence of refraction, Atmos. Chem. Phys., 1, , 21

5 E. Kyrölä et al.: GOMOS retrieval algorithms E. Kyrölä et al.: GOMOS retrieval algorithms nm star tracker 1 Hz photometers nm 65-7 nm grating grating R=637 km 21 km 88 km 3262 km 1 km 3 km 15 km 2 Hz nm nm nm spectrometers MOS instrument optical 4. layout. GOMOS instrument optical layout. R=637 km GOMOS line-of-sight (LOS) is defined by direction of star and location of Envisat. Geometry and scales of a GOMOS measurement are shown in 5. The LOS varies from 28 to 22 below satellite s horizontal plane. As a result, for a 28 LOS, a positioning error L along track (along 3262 km orbit) will produce an error z for 21 km altitude of LOS tangent height 88 km z = Lcos28.5 L (3) 1 km The GOMOS geolocation 3 km requires, refore, a good orbit determination (about 5 m accuracy) and accurate timing of measurements. The direction to star must take 15 km into account proper motion of star, parallax effect (small), and aberration of light (Karttunen et al., 27). The aberration depends on time of year ( change in orbital velocity of Earth during year) giving maximum aberration angle of 1 4 radian. This angle, projected on limb, would yield an error of about 3 m in altitude if not accounted for. The ENVISAT orbit information can be obtained from orbital data in GOMOS Level data. Depending on time delay of processing with respect to measurement, re may be also more accurate orbit data available. The shape of Earth is approximated by well-known model WGS84. For near real time processing, made within 3 h after measurements, atmospheric profile is computed from ECMWF 24 h forecast (made day before) for levels up to 1 hpa (about 45 km) and completed by climatological model MSIS9 model above 1 hpa. The final processing is using ECMWF analysis up to 1 hpa. In both cases, a continuous density profile is built up, satisfying hydrostatic law. Because of atmospheric refraction, light from a star to GOMOS does not follow a straight line. A full ray tracing gth scales of GOMOS measurements. The satellite (Ende is assumed to be 8 km. 6. Tangent point movement during occultation. The blue line represents a short occultation of 44 s and red line a long occultation of 164 s. The refraction effect is to decrease vertical velocity of LOS for altitudes below 4 km Length Length scales scales of of GOMOS GOMOS measurements. measurements. The The satellite satellite (Envisat) altitude is assumed to be 8 km. (Envisat) altitude is assumed to be 8 km. calculation is performed to compute path of stellar light. The refracted path is determined from equation ( d n dr ) = n. (4) ds ds The satellite position and stellar direction provide boundary conditions. The source term is produced by gradient of refractive index n. The refraction is computed at several points along path, taking into account atmospheric model and index of refraction which varies with wavelength according to Edlen s law ( n = λ λ 2 Here wavelength λ is expressed in micrometers. The overall multiplicative factor 1/1.62 is introduced to account for deviation from ideal gas law in pressure range relevant to stratosphere. The approximation applied in this calculation is that re is no contribution from density gradients or than ones in plane determined by GOMOS, star and Earth s centre. Besides refraction, an important factor for occultation geometry is angle of occultation with orbital plane. We characterise obliquity of occultation by angle θ between motion of line of sight with respect to atmosphere and direction of Earth s centre. Figure 6 shows tangent altitude change in a short occultation (obliquity near zero) and in a long occultation (large obliquity). Figure 7 shows how vertical sampling distance differs in occultations of 6. Figures 8 and 9 show how GOMOS measurements probe atmosphere in short and long occultations. The occultations are same used in Figs We have shown movement of tangent point in latitude-longitude values ) (5) 6. T line repre occultatio velocity o Atmos. Chem. Phys., 1, , 21

6 11886 E. Kyrölä et al.: GOMOS retrieval algorithms - E. Kyrölä et al.: GOMOS retrieval algorithms 6. Tangent point movement during occultation. The blue 8. Geographical coverage a short GOMOS occultation.the 6. Tangent point movement during occultation. The blue line represents a short occultation of 44 s and red line a long 8. duration Geographical of occultation coverage is 44 sof and a short obliquity GOMOS 3.9. The occultation.the red line, lineoccultation representsofa164 short s. The occultation refraction effect of 44isstoand decrease red vertical line duration a longhardly of discernible occultation in centre is 44of s andmain figure, obliquity gives3.9. latitude- discernible valuesinof centre tangentof points main below figure, 1 km. gives The inserted latitude- The red line, occultation velocity of of 164 LOS s. The for altitudes refraction below effect 4 km. is to decrease vertical hardlylongitude velocity of LOS for altitudes below 4 km. longitude figurevalues shows ofzoom tangent of tangent points point below location. 1The km. blue The lines inserted show latitude-longitude projections of ray tracing paths figurebelow shows 1 km. zoom of tangent point location. The blue lines show latitude-longitude projections of ray tracing paths below 1 km. definition of illumination flag of GOMOS measurements. Stellar occultations are naturally working best when re is a full dark limb condition (illumination flag = ). In bright limb conditions scattered solar light easily dominates stellar signal and retrieval is difficult. The full dark limit of Table 1 constrains strongly amount of scientifically usable data and refore more relaxed limits are usually applied without evident corruption of data. Almost all scientific work so far has been using only dark limb measurements and dark limb is usually defined by requiring solar zenith angle at tangent point to be larger than 17. This is a less restrictive constraint than full dark case in Table 1, which requires solar zenith angle at tangent point to be larger than 11 and solar zenith angle at satellite Tangent point vertical sampling distance as aas function a function of ofposition to be larger than Geographical coverage. The geophysical validation of a long GOMOS occultation. T altitude. The The blue line represents a ashort short occultation of 44of s and 44 s andof measurements with different illumination conditions have duration of occultation is 164 s and obliquity The red red line line a along occultation of of s. s. The The off-orbital off-orbital occultation occultationbeen presented in Meijer et al. (24) and van Gijsel et al. (red) samples atmosphere with a better vertical resolution. The line gives latitude-longitude values of tangent points bel (red) samples atmosphere with a better vertical resolution. The (21). temporal sampling is in both cases nominal.5 s. 1 km. The blue lines show latitude-longitude values of temporal sampling is in both cases nominal.5 s. ray tracing paths below 1 km. The occultation starts from 5 Level right1b: lower spectrometer side and moves data processing to centre of figure. and latitude-longitude values of ray tracing paths. From geometry we know that points around tangent point 5.1 Wavelength assignment contribute strongest to horizontal columns. In short occultations tangent point is almost stationary in latitudelongitude coordinates. In long occultations movement calibration auxiliary product and spectral shifts identified The nominal wavelength assignment is carried out using of tangent point can be more than 1 in latitude and longitude. assignment, corresponding to a perfect star tracking, is pro- using data from star tracker. The nominal wavelength An important part of geolocation is to determine vided by spectral assignment of one CCD column and by illumination condition of measurement. Table 1 provides spectral dispersion law of spectrometers determined Atmos. Chem. Phys., 1, , 21

7 E. Kyrölä et al.: GOMOS retrieval algorithms Table 1. Illumination flag for GOMOS products. SZA is solar zenith angle. Flag Term Requirement full dark Not in bright limb nor twilight nor straylight conditions. 1 bright SZA at tangent point <97 for at least one measurement with altitude at tangent point <5 km. 2 twilight Not in bright limb and SZA at tangent point <11 for at least one measurement with altitude at tangent point <1 km. of nd on he he e, e- ed es hs 9. Geographical coverage of a long GOMOS occultation. The 9. Geographical coverage of long GOMOS occultation. The duration of occultation is 164 s and obliquity The red duration line gives of occultation latitude-longitude is 164 values s and of obliquity tangent points below The red line1 gives km. The latitude-longitude blue lines show values latitude-longitude of tangent values points of below 1 km. raythe tracing blue paths lines below show 1 km. latitude-longitude The occultation starts values from of right ray tracing lower side paths andbelow moves 1 to km. centre Theofoccultation figure. starts from right lower side and moves to centre of figure. in on-ground characterisation. Spectral shifts due to vibrations and imperfect tracking are estimated by pointing data history from SATU (Star Acquisition Tracking Unit). These shifts are used to spectrally assign each CCD column during each spectrometer measurement. The SATU output provides shift in horizontal (spectral) and vertical (spatial) directions. Mean SATU output data are computed at a frequency of 2 Hz and it is used to compute physical and spectral shifts of star signal on spectrometer CCD arrays. The mean spatial shift (2 Hz) expressed in fraction of spectrometer B2 CCD lines (or pixels) is plotted in Bad pixels, cosmic rays and modulation signal Bad CCD pixels are specified in bad pixels list coming from GOMOS calibration database. The initial list is defined by on-ground characterisation of GOMOS. This list is regularly updated by GOMOS in-flight calibration process (Uniformity monitoring mode). There is no on-line detection of bad pixels but cosmic rays correction processing may prevent use of unexpected bad pixels. Once bad pixels are identified, two complementary actions are performed: (1) bad pixels are flagged and are ruled out from any furr processing. (2) New values in ADU (Analog to Digital Unit) are interpolated from neighbouring pixels by a median filter 1. but Shifts y from are SATU not used forin whole furr occultation processing. in Y- (spatial) direction. The CCD The detectors unit of aremispointing sensitive tois high fraction energyofparticles vertical sizearriving of from B2 spectrometer Earth s radiation pixel. The belts, size interplanetary of one GOMOS space B2 spectrometer or interstellar pixel medium. is Theyµm may (spatial spectral). also be generated by radioactive decay of impurities in spacecraft and GOMOS 3 straylight Not in bright limb and SZA of ENVISAT <12 for at least one measurement. 4 twilight + straylight Not in bright limb and both conditions for twilight and for straylight are verified. materials. A cosmic ray will produce an additional charge in a single pixel or in several pixels if angle of incidence with respect to surface of CCD is small. This phenomenon is seen in all space-flying CCDs. When full CCD image is available, standard way to identify cosmic hits is to pass a median filter on image and to compare content of each pixel to median value of window centred on this pixel. For GOMOS, we do not get a full CCD image. Rar, we have several blocks of lines electronically binned toger (three per CCD) at each exposure of.5 s. Therefore, a pseudo-image is reconstituted for each measuring band of CCDs: top line of image being first measurement in band, second line being second measurement, and so on for a whole occultation. We assume here that a pixel has only a very small chance to be hit by a cosmic ray during two consecutive measurements. This abnormal CCD response is eliminated by a median filtering in both spectral and temporal directions. The content of each pixel is compared to median value of filtering window. A cosmic hit is detected if this difference is greater than a given threshold. This threshold value depends of gain index of CCD. Threshold values for four different gain index are read in Level 1b processing configuration product. The value of corresponding sample is replaced by computed median value. It is flagged and not used furr in processing. In bright limb condition cosmic ray processing is disabled. After launch it was discovered that electronic chain offset level shows a small variation. This modulation has a reproducible pattern in successive samples and an amplitude of.5 to 1.5 ADU (changing with specific CCD A1 or A2). If measurement is not towards bright limb, a Atmos. Chem. Phys., 1, , 21

8 11888 E. Kyrölä et al.: GOMOS retrieval algorithms The ine, derted ines aths 11. Temporal dark charge evolution of a single binned CCD column. Expected long sensor warming and seasonal rmal variations 11. are Temporal perfectly visible. dark charge Sharpevolution variations of of a single dark charge binned level CCD column. (e.g. around Expected orbits 85, long18 sensor 5, 2 warming ) areand due to seasonal occurrence rmal ofvari- ations new hot are perfectly pixels. visible. Sharp variations of dark charge level 1. Shifts from SATU for whole occultation in Y- (spatial) direction. The unit of mispointing is fraction of vertical 1. Shifts from SATU for whole occultation in Y- (spatial) direction. The unit of mispointing is fraction of vertical new hot pixels. (e.g. around orbits 85, 185, 2) are due to occurrence of size of B2 spectrometer pixel. The size of one GOMOS B2 size spectrometer of B2 pixel spectrometer is pixel. µm (spatial The size spectral). of one GOMOS B2 5.4 Dark charge spectrometer pixel is µm (spatial spectral). During Envisat flight a majority of CCD pixels have correction of this modulation is performed for spectrometer A1 and A2 (after cosmic ray correction). For spec- degradation of CCD s has been main negative sur- become hot, with a higher and variable dark charge. This trometer B1 and B2 modulation is not corrected. There is prise in GOMOS mission. The precipitation of protons is a slow change in amplitude of this modulation, which is reason for this continuous increase of dark current. monitored. In absence of light, level of dark charge is much higher than on ground and it is highly variable from one 5.3 Conversion into electrons pixel to next (so-called DCNU, Dark Charge Non Uniformity). Also, its average level for one pixel may change, A conversion of ADU values into number of electrons is performed in each spatial band of spectrometers A and B including also a non-linearity correction. This conversion needs value of CCD gain (in e/adu) for each electronic chain and for each programmable gains (4 gains per spectrometer). The non-linearity correction of electronic chain is performed in this processing step. with a time scale that can vary from a few seconds to several hours. This phenomenon is called RTS (Random Telegraphic Signal). Therefore, a new observing strategy was set up. At each orbit, when Envisat is near equator in deep night, GOMOS is pointed toward a dark area in sky, collecting refore only dark charge on all CCDs. These measurements are used to correct all or occultations in same D LIN = (D d )f nlin (D d ) (6) orbit. Figure 11 shows evolution of dark charge for individual CCD-column and 12 increase of mean N RO = D LIN G (7) dark charge with time for spectrometer A2. Here D is spectrometer pixel value in ADU (after elimination of saturated, bad and cosmic ray pixels), f nlin is non-linearity factor, G is electronic gain factor, d is off-set and N RO is number of electrons collected by pixel. The non-linearity factor can be determined by using so-called linearity monitoring mode of GOMOS where integration time can be specified between.25 to 1 s. Originally it was measured on ground in instrument characterisation process. 5.5 Mirror reflectivity and vigneting The plane mirror of SFA (Steering Front Assembly) is oriented 12. Mean in such Darka Charge way that for spectrometer line of sight A2. isthe reflected blackin curve shows a fixed seasonal direction, variations along due to telescope position axis. ofthe sun reflectivity and thus to ofchange mirror of varies temperature with wavelength of CCD. and The redangle curveofhas in-beecidence, bywhich normalising 25. black The incidence curve at aangle reference depends temperature on obtained of 5elevation C. Therefore, and azimuth redangles. curve shows Since that incidence an increase angle of Dark Charge variesindependent by a few degrees of during temperature one occultation, still exists. it is important to correct this effect. Orwise transmission ratio of two star spectra measured with different incidence angles would be biased. A look-up-table has been built from Sirius Fi era Fi no is red Atmos. Chem. Phys., 1, , 21

9 E. Kyrölä et al.: GOMOS retrieval algorithms observations with various azimuth and incidence angles. The variation of reflectivity is assumed to be a linear function of angle of incidence. At extreme range of azimuth angles (negative angles), plane mirror is vignetting (partially masking) a part of collecting area for IR spectrometer B. The degraded performance of IR spectrometer at se azimuth angles (nominally from 1 to 5 ) is taken into account by a vignetting correction. 5.6 Stray light and bright limb background Photometer flux The internal stray light is light coming from region of FOV (Field Of View) limited by slit but whose photons do not fall on ir nominal position on CCD. It is mainly generated by grating as scattered light and it is most manifest as light from star collected in background bands. At present time, no correction of internal stray light has been implemented. The external stray light is light that makes its way through slit, though its origin is outside nominal FOV of slit. There are mainly two sources of external stray light. One is solar light scattered by some Envisat hardware into baffle and optics of GOMOS. The or is coming from sun-illuminated limb or nadir. The external stray light of solar origin does not go through ozone layer and refore contains some UV even at maximum ozone absorption wavelengths (near 255 nm). At present time, no correction of external stray light has been implemented. Still, geometrical computations are performed to characterise illumination conditions of GOMOS, Envisat, and tangent point of occultation at limb at a reference altitude ( nominal value is 5 km). The illumination condition is given in flag stored in Level 1b product. This flag may be also used as a guide for stray light contamination. In bright limb occultations background radiation falling on central band needs to be estimated and subtracted from measured central signal in order to get star signal alone. This estimate is also used for photometers. The estimates of upper and lower background are stored in GOMOS Limb product. Before estimation, a flat field correction is applied to upper and lower bands. The sensitivity of each pixel is retrieved from a PRNU map (Pixel-to-pixel Response Non- Uniformity). The PRNU values of all pixels have been determined in-flight. The processing software includes two different methods for estimating background contribution inside stellar band: (1) A linear interpolation between upper and lower bands. (2) An exponential interpolation (this is method currently activated in processing). The estimate for central background is obtained as N B C = N NST C Q C (8) 12. Mean Dark Charge for spectrometer A2. The black curve shows seasonal variations due to position of sun and thus to changeof of temperature of CCD. The red curve has been obtainedby by normalising black curve at a reference temperature of of 5 C. C. Therefore, red curve shows that an increase of Dark Charge independentof of temperature still exists. where NC NST is background contribution for central band from method 1 or 2 and Q C is PRNU averaged for central band. Once central background NC B is estimated, it is subtracted from central band to get stellar signal. This is not done in full dark limb conditions for spectrometer A. In full dark conditions background is usually negligible and subtracting it would add noise of dark charge of two background bands. There are, however, some atmospheric emissions lines that can be detected in GOMOS nighttime spectra. Examples are famous auroral and airglow (green) line at nm and oxygen line around 63 nm. In GOMOS Level 2 processing spectral range nm (pixels ) covering oxygen line has been flagged and this area is not used in retrieval. For spectrometer B, background is not negligible even during night. There are O 2 emissions around 76 nm and OH emissions near 94 nm. On day side, limb brightness is large. It saturates spectrometers usually at altitudes lower than 26 km so that background correction is no longer feasible. When any of upper, central or lower band is saturated, corresponding column is flagged and not used in data processing. 5.7 Star signal computation The stellar signal created by background subtraction needs to be corrected for flat-fielding in order to get a valid estimation of star spectrum. The nominal location of star signal on CCD, SATU output data, photometer engineering data and PNRU are used in flat-fielding. 14 normal occ beg is 3.9. red and Atmos. Chem. Phys., 1, , 21

10 24 E. Kyrölä et al.: GOMOS retrieval algorithms 1189 E. Kyrölä et al.: GOMOS retrieval algorithms The flat-fielding is most important for spectrometers B1 and B2. During one spectrometer integration time, image of star spectrum is moving on CCD due to refraction and residual pointing errors. Therefore several pixels with varying responses contribute to measured signal. The signal at band level will depend on exact position on which star spectrum is formed. This position is known from SATU data, at 1 Hz sampling rate. Coupled with a refraction model and with instrument point spread function, SATU information allows to estimate current effective PRNU of CCD pixels of central band, which contains star spectrum image. The scintillations measured from photometers data allow to take into account fast brightness variations of star spectrum, for each position of star spectrum at 1 Hz within a sample time (.5 s). 11. Temporal dark charge evolution of a single binned CCD column. If star Expected imagelong in sensor slit warming deviates andtoo seasonal muchrmal from variations are central perfectly position, visible. Sharp slitvariations may block of a part darkof charge sig- level nominal nal. (e.g. Because around orbits position 85, 185, of 2) star are image due to can beoccurrence estimatedof from new hotstar pixels. tracker, a look-up table can be used to recalibrate signal as N S = ˆN S τ slit (9) where slit transmission is calculated blocking factor. Figure 13 shows an example of occulted star spectra (central band) at several tangent altitudes. The Level 1b processing for spectrometers is finalised (see, however, error analysis in Sect. 7) by calculating so-called full transmission T = N S N REF (1) The full transmission is computed as ratio of estimated star spectrum to reference spectrum of current occultation. All spectra used in this processing must be spectrally aligned before computation. The wavelength assignment of star spectrum is re-sampled over nominal spectral grid before computation of transmission. A reference spectrum of star is measured at each occultation by averaging 1 measurements obtained above 12 km. Examples of Level 1b transmissions can be found in Mean Dark Charge for spectrometer A2. The black curve in Level 2 processing. shows seasonal variations due to position of sun and thus to change of temperature of CCD. The red curve has been obtained by normalising black curve at a reference temperature 6 Photometer data processing of 5 C. Therefore, red curve shows that an increase of Dark Charge independent of temperature still exists. The processing applied to fast photometer data is to a large extent same as done for spectrometers. Specific steps for photometers are unfolding and corrections for spikes. The background subtraction uses estimated central background of spectrometers as an input to remove background contribution for photometers. Because of 12 bits capacity of on-board adder included in photometer electronic chain, only 12 least Measured signal (e) x km 9.3 km 8.1 km 69.9 km 6.4 km 5.3 km 4.2 km 3.1 km 2.2 km 12.2 km Wavelength (nm) 13. Sirius spectrum measured through atmosphereat at several tangent altitudes. significant bits are written in GOMOS packets. In case of overflow, analysis of temporal series usually allows to rebuild original signal (unfolding algorithm). The current unfolding algorithm compares local variation of signal to a threshold. However, in some circumstances (e.g. Sir- 6 ius occultations with strong atmospheric scintillations), real variability of signal flux can exceed this threshold, leading 5 to false overflow detection. Photometer flux An incorrect unfolding of photometer signal has a negative4 impact on scintillation/dilution correction of transmission spectra (Level 2 processing) and thus on quality of species retrieval. Discontinuities of NO 2 3 column density profiles have been observed at altitudes of false reset corrections. The analysis of processed occultations 2 leads us to definitely avoid applying unfolding process for occultations of Sirius or in case of full dark limb1 condition where overflow is assumed to be only occasional. The photometer data are measured with a sampling rate of khz and each sample corresponds Time (ms) to a time integration of light intensity during one millisecond (1 ms). For each spectrometer 14. Blue and measurement red photometer (.5signals. s), reboth is consequently signals have been 5 photometer normalised bysamples dividingfor by each mean. photometer. In horizontal However, axis timelast is sample occultation of eachtime series measured of 5 from is obtained altitudewith of 26a km time (35integra- tion beginning of.9 ms of (instead occultation). of 1 ms), The thus obliquity producing of a occultation downward s since spike. is 3.9. Some This instrumental peaks show clearly effect istime corrected delay effect by dividing between red and blue photometers. value of last sample in each series of 5 by.9. An example of photometer data is shown in 14. The figure shows time delay effect explained in 3. The red light from a given structure is detected first when using setting occultations. The vertical separation chromatic lag varies (exponentially) from 1 m at 45 km to1 m at 3 km to 6 m at 2 km and 3 m at 1 km The related time delay is obtained by dividing by vertical velocity of ENVISAT (less than Atmos. Chem. Phys., 1, , 21

11 E. Kyrölä et al.: GOMOS retrieval algorithms k curve thus to as been erature f Dark Photometer flux Time (ms) 14. Blue and red photometer signals. Both signals have been normalised by dividing by mean. In horizontal axis timeis is occultation time measured from altitude of 26 km (35 s since beginning of occultation). The obliquity of occultation is 3.9. Some peaks show clearly time delay effect between red and blue photometers. 3.4 km/s). The time delay is n from 3 ms at 3 km and 88 ms at 1 km. 7 Error analysis All Level 1b products are accompanied by error estimates. These estimates are calculated assuming that re is no modelling errors and assuming a normal Gaussian error statistics. The pixel data are assumed to be independent on each or. We ignore all correlations between pixels. In following main error contributions for main products are reviewed. The error on any signal coming from a CCD is N = N + 2 dc +R2 ro + G2 12 (11) The first component under square-root is photon noise, second ( dc ) is noise associated with dark charge removal, third one (Rro) is read-out noise and last one is quantisation noise. The error on background has been estimated by b = B est (12) where B est is estimate of background signal in central band. The error on estimated stellar signal is N = N + 2 dc +R2 ro + G sl + 2 b (13) where sl is error associated to subtraction of stray light, and b is error associated to subtraction of background. At present, sl is set to zero because stray light correction is not activated. The error on reference spectrum is given by N ref = 1 p Nj 2 (14) p j=1 where p=1 is number of measurements averaged. The error on transmission is ( N ) 2 ( ) 2 Nref T = T + (15) N N ref The error of photometer signal is dominated by shot noise, and estimate is taken as square root of signal. All se error estimation formulas have been validated by simulations and by in-flight statistical analysis of measurements. 8 Introduction to GOMOS geophysical retrieval The GOMOS Level 1b products that are used in GOMOS Level 2 processing are: transmission data (T obs ); photometer data (I obs pho ); geolocation data (l(λ,t)); a priori atmospheric data (ρ air ( r),t( r)). Main data are transmission spectra at different tangent heights. Photometric data from two fast photometers are used to correct transmissions from scintillation effects. Photometric data are also used to retrieve a high resolution temperature profiles of atmosphere. Geolocation and a priori atmospheric data are necessary information in dealing with refractive effects and at initialising stage of retrieval. In GOMOS Level 2 processing data set processed is measurements from one star occultation at a time. Even if some occultations would probe same atmospheric region at successive orbits se occultations are treated separately. Therefore, possibility to carry out some kind of atmospheric tomography by GOMOS has not been considered in ESA Level 2 processing. In present GOMOS ground processing limb spectra from Level 1b are not used for atmospheric retrieval but this possibility is under development. The geophysical retrieval strategy in Level 2 data processing assumes that a measured transmission can be taken as a product of two transmissions T obs = T ref T ext (16) Atmos. Chem. Phys., 1, , 21