INFLUENCE OF SCINTILLATION ON QUALITY OF OZONE MONITORING BY GOMOS

Transcription

1 INFLUENCE OF SCINILLAION ON QUALIY OF OZONE MONIORING BY GOMOS V.F. Sofieva (), E. Kyrölä (), F. Dalaudier (2), V. K (3), A.S. Gurvich (3), d the GOMOS team (4) () Finnish Meteorological Institute, P.O. Box 3, FIN-000, Helsinki, Finld, viktoria.sofieva@fmi.fi (2) Service d Aeronomie, BP 3, 937, Verrieres-le-Buisson CEDEX, Frce (3) A.M. Oboukhov Institute of ospheric Physics, Pyzhevsky 3, Moow, Russia (4) FMI, Finld; Service d Aeronomie, Frce; BIRA, Belgium, ACRI-S, Frce, ESA, Netherlds d Italy ABSRAC Point-like nature of stars leads to significt disturbces of light beams passed through the osphere by ractive effects. hese effects should be taken into account in retrievals from stellar occultation measurements. We consider the influence of intillation on stellar occultation measurements d on quality of ozone retrievals from these measurements, based on experience of the GOMOS (Global Ozone Monitoring by Occultation of Stars) instrument on board the Envisat satellite. In the GOMOS retrieval, the intillation effect is corrected using intillation measurements by the fast photometer. We present qutitative estimates of the current intillation correction quality d of the impact of intillation on ozone retrievals by GOMOS. he alysis has shown that the present intillation correction efficiently removes the distortion of trsmission spectra caused by isotropic intillations. he impact of errors of dilution d isotropic intillation correction on quality of ozone monitoring is negligible. However, the current intillation correction is not able to remove the wavelength-dependent distortion of trsmission spectra caused by isotropic intillations, which c be present in oblique (off orbital ple) occultations. his distortion may result in error of ozone retrievals of % at altitudes - km. his contribution to the error budget is significt for bright stars. he advced inversion methods that c minimize the influence of intillation correction error are also diussed.. INRODUCION he stellar occultations measurements have a set of beneficial features, such as self-calibration measurement principle, global coverage, good vertical resolution, wide altitude rge of measurements: from the troposphere to the thermosphere. However, using stars impose certain requirements on instruments d retrievals. Since stars are point sources of quite lowintensity light, special instruments are needed for recording stellar spectra. he stellar spectra observed through the Earth osphere are not only attenuated by absorption d attering (this phenomenon is used for reconstruction of the chemical composition), but also modified by ractive effects. Almost exponential decrease of the ospheric air density leads to bending of rays coming from a star, the lower tgent altitude, the larger bending. Refraction in the osphere trsforms parallel incident rays into diverging beams, thus resulting in dilution of the registered intensity, the effect known as ractive attenuation or ractive dilution. he dependence of ospheric ractivity on wavelength leads to a spatial separation of rays of different color in the osphere, the effect known as chromatic raction. he chromatic raction d ractive dilution are related to a smooth dependence of ractive index on wavelength d altitude. However, the air density d, as a consequence, ospheric ractivity, always has a fluctuations caused by ospheric processes, such as internal gravity waves (IGW), turbulence, different kinds of ospheric instabilities. he interaction of light waves with ractivity irregularities result in intillation, i.e., fluctuations in the measured intensity of stellar light. If the stellar light passed through the osphere is recorded with high-frequency devices, the measured intensity fluctuations may exceed the me value by several hundred percent. he upper pel in Fig. shows example of intillation measurements by the GOMOS fast photometer operating with the sampling frequency of khz. he rms of relative fluctuations of intensity recorded by the photometer rapidly grows with decreasing altitude until it saturates at values ~ at altitude of ~ km. Ozone d other trace gases are retrieved from UV-VIS spectrometer measurements, which have significtly lower sampling rate, 2 Hz, in GOMOS. However, the photometers signal averaged down to 2 Hz still exhibit fluctuations (blue line in Figure, top), thus showing the possible modulation of the spectrometer signals caused by intillation. he depth of this modulation exceeds the instrumental noise, especially for very bright stars. his intillation noise is a nuie for ozone retrieval, d it should be corrected as much as possible before starting the inversion procedure. In the GOMOS inversion [], the intillation effect is corrected using intillation measurements by the fast photometer. In this paper, we give deription of the Proc. Envisat Symposium 07, Montreux, Switzerld April 07 (ESA SP-636, July 07)

2 intillation correction that is applied in GOMOS processing d diuss its quality d limitations. eliminated from the ospheric trsmission data. Main steps of the intillation-dilution correction deribed in [2] are shown in Fig. 2. measured used in retrievals o be estimated d removed Estimated using ray tracing through the ECMWF air density field ˆ ( λ) = dilu dα ( λ ) + L dp = ext( λ, ( λ, = ( dilu in Estimated using intillation measurements by fast photometers I t ˆ ( ) in ( = I Estimated intillation-dilution term ˆ = ˆ ˆ in( dt t dilu t averaging over t =0.5 s integration time of spectrometers Figure. op: intillation measurements by the GOMOS red photometer in occultation of Sirius R02833/S00 (latitude S), the intillation averaged to 2 Hz (blue line), d the smooth signal obtain from the intillation data using filtration by the Hning window with the cut-off ale 3 km (dashed gray line). Bottom: rms of relative fluctuations of intensity We combine theoretical estimates, experimental results d simulation in order to obtain qutitative estimates of the impact of intillation. We restricted ourselves to consideration of the influence on retrieval of ozone only, as it is the main target of the GOMOS mission. 2. GOMOS MEASUREMENS AND SCINILLAION CORRECION In occultation measurements, the spectrometer measures stellar light passing through the osphere continuously (in case of GOMOS, with the sampling frequency of 2 H as star sets behind the Earth limb. he ospheric trsmission spectra are obtained by dividing the spectra measured at different tgent altitudes by the erence spectrum, measured above the osphere. hese trsmission spectra contain spectral signatures of absorption d attering in the osphere, which are also modified by ractive effects. In order to eliminate the ractive perturbations of trsmission spectra, it is assumed that absorption d raction affect the ospheric trsmission spectra independently, so it c be expressed as a product []: ( λ, = ext, ( ) where ext is the trsmittce due to absorption d attering, d represents the combined effect of raction d intillation. In the GOMOS retrieval, the component due to ractive effects is estimated d Figure 2. Scintillation-dilution correction implemented in GOMOS retrievals. wo main hypotheses are used in the GOMOS intillation correction. First, it assumed that all highfrequency fluctuations in the photometer signal (except for noise) are due to intillations, while fluctuations due to structure in the vertical profiles of absorbing constituents affect only the low-frequency component of the signal. Second, it is assumed that light rays of different color come through the same ractive structures. his hypothesis is always satisfied in vertical occultations (in orbital ple), but may be violated in oblique occultations if isotropic turbulence is well developed. Validity of these assumptions is diussed in [2, 3], d will be considered further in Section 4 of this paper. 3. QUALIY OF ANISOROPIC SCINILLAION CORRECION Under the assumption of strong isotropy of air density irregularities d provided that the me raction is perfectly known d intillations are weak, we c expect that the intillation correction deribed above almost perfectly eliminates the intillation-dilution component from the measured trsmission spectra. he main error of the isotropic intillation correction comes from impossibility of clear separation of dilution d intillation terms. Other error sources are noise in the photometer data d the fact that the photometer records not monochromatic intillations but averaged over the wavelength bd of the optical filter. At altitudes below 25- km, the weak intillation assumption is violated, that also results in reduced accuracy of intillation correction. In order to estimate the best quality of the intillation correction, the noise-free signal of the red photometer (isotropic intillations) was simulated with the intillation model. he realistic me ractivity

3 profile from ECMWF alysis data is used. he trsmission due to absorption d attering was simulated with LIMBO [4]. For simulation of the Level b spectrometer data we used the following approach. First, monochromatic intillations at khz sampling frequency corresponding to the wavelengths of each pixel were simulated (with the absorption effect included), d then the signal was integrated down to 2 Hz sampling frequency of the spectrometer rue d estimated trsmittces Quality of isotropic intillation correction 0 nm 0 nm ( true - est )/ true % Relative error of trsmittce estimate true 0 nm true 0 nm dilution correction, 0 nm dilution correction, 0 nm dilution & intillation corrrection, 0 nm dilution & intillation correction, 0 nm dilution correction, 0 nm -2 dilution correction, 0 nm dilution&intillation correction, 0 nm dilution& intillation correction, 0 nm Figure 3. op: true (solid lines) d estimated trsmittces due to absorption d attering. Bottom: relative error of the trsmittce estimates, with d without intillation correction Fig. 3 illustrates the quality d usefulness of the GOMOS intillation correction: most of the modulation caused by intillation, which is well observed in the curves corresponding to the dilution correction only, is eliminated by the applied intillation correction. he residual (non-corrected) intillation modulation is below % for altitudes above ~ km altitude rge, in the ozone layer. o estimate average quality of the isotropic intillation correction, we carried out Monte Carlo simulations (00 runs) of the intillation correction deribed above, with different intillation realizations. he relative error of the isotropic intillation correction (i.e., the error in estimated trsmittces) is shown in Fig. 4. It demonstrates that the GOMOS intillation correction efficiently eliminates modulation caused by isotropic intillations: the remaining error is below %. he residual error of correction of isotropic intillation (dilution correction error) δ, leads to perturbation of the estimated trsmittces due to absorption d attering: relative error, % Figure 4. Relative error in estimated trsmittces caused by error of isotropic intillation correction: results of Monte Carlo simulations ext ext ˆ δ = = ext ext (2) ˆ δ + Because δ does not depend on wavelength in case of isotropic intillations, the erroneous correction of dilution d isotropic intillation does not modify the spectral shape of the trsmittce due to absorption d attering, but only chges its value (equivalently, the optical depth values are shifted by a const. Such modulation of the trsmittce spectra leads to synchronous fluctuations in all retrieved profiles. he simulations have shown that the sensitivity of ozone retrieval to the spectrally flat perturbation of trsmittce spectra is negligible, 2% trsmittce perturbation results in only 0.00% perturbation in ozone line density. 4. ISOROPIC SCINILLAIONS AND HEIR IMPAC ON OZONE RERIEVAL In the presence of isotropic small-ale air density irregularities, the main assumption of the GOMOS intillation correction that light rays of different colors come through the same ractive structures - c be violated (Fig. 5). In reality, we always observe a mixture of isotropic d isotropic intillations. In vertical occultations, intillations of both types are correlated. In oblique occultations, the isotropic intillations become uncorrelated when the separation of the colored rays becomes larger th ~max(l K,ρ F ) (here ρ F is the Fresnel ale d l K is the Kolmogorov s ale). he GOMOS intillation correction is not able to remove not correlated fluctuations. Furthermore, correcting isotropic intillations as though they were isotropic introduces additional error.

4 vertical occultation α oblique occultation Figure 5. Colored lines: ray trajectories in the phase reen for vertical d oblique occultations. Grey ovals hematically show isotropic irregularities of air density d gray circles denote isotropic irregularities. he effect of non-corrected isotropic intillations is well seen in residuals R(λ)= ext (λ) mod (λ) (the difference between measured d modeled trsmittces) in case of oblique occultations. he residuals have oillating features, which are observed clearly in case of bright stars (Fig.6). statistically independent fluctuations in measured intensity, provided intillations are weak. is δi δ he additional term, ε =, ext is I represents the intillation correction error that is present in GOMOS measurements in case of oblique occultations in turbulent osphere. It is assumed to be a Gaussi rdom variable with the zero me d the covarice matrix C mod. he covarice matrix C mod was estimated using the statistical alysis of residuals in sequential oblique occultations of Sirius. he alysis of amplitude d correlation of residual fluctuations has shown that their dependence on wavelength, on obliquity d on altitude is in a good agreement with the predictions by the theory of isotropic intillations, generated by locally isotropic turbulence. his has allowed the parameterization of the intillation error, which depends mainly on one parameter ch sinα ξ =, the ratio of the chromatic distce to ρf the Fresnel ale (here ch is the vertical chromatic shift d α is the gle between the direction of the apparent star motion d the local vertical at the ray perigee poin. Figure 6. Residuals in occultation R058/S00, obliquity ~75. Left: color plot; right: aled residuals (by the factor ) at selected altitudes. Assuming that (i) the dilution estimate is error-free, (ii) isotropic component of the intillation is estimated with the error + ˆ δ, d (iii) the δ : = modulation due to isotropic intillation c be is is δi presented in the form = +, the estimated is I trsmittce due to absorption d attering c be approximated as: is is extdiluinin δi δ (3) ˆ = + ext ext ext is ˆ dilu ˆ in δ I + in diluin in he factorization of the trsmission due to intillation is in inin in (3) approximates the assumption that isotropic d isotropic irregularities generate m=0 m= σ /σ noise Figure 7 op: Cross-correlation function of the spectrometer pixels at km, obliquity ~. Bottom: altitude dependence of ratio, for the stars of magnitude 0 d 2 d the effective temperature = 000 K.

5 he covarice matrix C mod has off-diagonal elements, which deribe the correlation between spectrometer pixels. An example of the correlation function of the spectrometer pixels B(λ, λ 2 ) at km is shown in Fig. 7 (top). he covarice matrix of the trsmission errors, C tot, c be presented as a sum of two matrices (provided that errors are Gaussi) C tot =C noise +C mod, ( 4) where the diagonal matrix C noise corresponds to the measurement noise, while the non-diagonal matrix C mod corresponds to modelling error. he altitude dependence of the ratio of intillation d noise stdard deviations σ r = is shown in Fig.7 (bottom). For very bright σ noise stars, the turbulence error c be twice larger th the instrumental noise. he error of the line densities reconstruction in the spectral inversion [5] c be estimated via Gaussi error propagation: C N = Σ CtotΣ, ( 5) where C N is the covarice matrix of the line density errors, Σ is the matrix of cross section d C tot is the covarice matrix of the total error (4). he impact of the non-corrected isotropic intillations on the ozone retrievals quality is illustrated in Fig.8, which compares the line density error estimates for the perfect intillation correction (corresponding to vertical occultations) with the error estimates (5) (corresponding to non-corrected isotropic intillations in oblique occultations). he turbulence error results in additional error of 0.5- % in line density reconstruction. altitude (km) 70 m=0, vertical m=0, oblique m=2, vertical m=2, oblique line density error (%) Figure 8 Relative error of the ozone line density retrievals, solid lines: oblique occultations (error due to non-corrected isotropic intillations is included), dashed lines: vertical occultations (the measurement error contains instrumental noise only), for stars of magnitudes 0 d 2 d of effective temperature 000K. Obliquity α=75 is used in the calculations. Although the absolute value of the intillation correction error is relatively small, the isotropic intillation has significt contribution to the error budget in case of bright stars. he spectral inversion is followed by the vertical inversion aimed at reconstruction of local densities of ozone, NO 2, NO 3 d aerosols [6]. he ikhonov-type regularization is applied in the vertical inversion for its stabilization. It is formulated in the grid-independent way [6, 7] so that the actual resolution of the retrieved profiles, which takes into account the smoothing properties by inversion, is independent of the retrieval grid. he regularization parameter depends also on sampling resolution, which c be significtly better in oblique occultations. As a result, we apply more smoothing in oblique occultations, which are affected by isotropic intillations. altitude (km) 70 m=0, α=0 m=0, α=75 m=2, α=0 m=2, α= local density error (%) Figure 9: As Fig.8, but for the ozone local density. he vertical inversion has been performed without regularization. he different sampling resolution in vertical d oblique occultations is taken into account. altitude (km) 70 m=0, α=0 m=0, α=75 m=2, α=0 m=2, α= local density error (%) Figure 0: As Fig.9, but the grid-independent regularization is applied in the vertical inversion. Fig. 9 shows the errors of ozone local density for oblique d vertical occultations provided the vertical inversion is performed without regularization. If the regularization is not applied, the incomplete intillation

6 correction results in ~- 2.5% error in ozone local density retrieval at altitudes - km. If the grid-independent regularization applied (Fig.0), the intillation correction error is still visible in oblique occultations of very bright stars at ~- km, but it is ~0.5% (against ~2% if the regularization is not applied). For typical stars (lines corresponding to visual magnitude 2, m=2 in Fig.0) d significtly oblique occultations, the accuracy of ozone local density retrievals c be very similar in oblique d vertical occultations (compare blue curves in Fig.0). Since the sampling resolution is twice denser for α=75, the gridindependent regularization applies more smoothing in the oblique occultation th in the vertical one, thus removing measurement noise d the intillation correction error. Note that local density errors are smaller in oblique occultations at all altitudes outside the turbulent altitude region ~- km, for the same reason. hese estimates of the impact of isotropic intillation on quality of ozone retrieval were obtained with the aid of the GOMOS data alysis, d theore they are close to reality. Figs. 8 d 0 show typical values of ozone retrieval errors induced by the incomplete intillation correction. he magnitude of the turbulence error depends on ratio of isotropic d isotropic intillation power, which, in turn, depends on ospheric stability. he shape of the error correlation function depends on obliquity of occultation that also affects the ozone retrieval error. he parameterization of the intillation correction error that uses the chromatic separation of rays enables qutitative characterization of the intillation correction error for different obliquities of occultations. he obtained parameterization of the modelling error c be directly used in the inversion thus providing correct error estimates. However, the non-diagonal covarice matrix of the modelling errors reduces significtly the numerical efficiency of the GOMOS spectral inversion. he implementation d assessment of this method c be the subject of future work. 5. SUMMARY We have presented qutitative estimates of the current intillation correction quality d of the impact of intillation on ozone retrievals by GOMOS. he following main conclusions c be drawn from this study:. he present intillation correction efficiently removes the modulation of the trsmittce spectra caused by isotropic intillations. 2. he impact of errors of dilution d isotropic intillation correction on quality of ozone monitoring is negligible. 3. he current intillation correction is not able to remove the wavelength-dependent modulation of trsmission spectra caused by isotropic intillations, which is present in oblique occultations. his modulation may result in error of ozone line density retrievals of 0.5-% at altitudes - km. his contribution to the error budget is significt for bright stars. 4. he grid-independent regularization of ikhonov type implemented in the GOMOS vertical inversion significtly reduces the retrieval error. By applying more smoothing in oblique occultations, which are affected by incomplete intillation correction, it makes the retrieval accuracy in oblique occultations less th -.5 % worse th in vertical occultations of the same star. In case of significtly oblique occultations d not very bright stars, the accuracy of ozone retrieval is very similar in the oblique d vertical occultations at altitudes - km. 6. REFERENCES [] GOMOS ESL, Algorithm heoretical Basis Document, version 2.0, 07, ABD_v2.pdf [2] Dalaudier F., V. K, d A. S. Gurvich: Chromatic raction with global ozone monitoring by occultation of stars. I. Deription d intillation correction, Applied Opt.,, , 0 [3] K V., F. Dalaudier, d A. S. Gurvich: Chromatic raction with global ozone monitoring by occultation of stars. II. Statistical properties of intillations, Applied Opt.,, , 0 [4] Kyrölä, E., J amminen, L. Oikarinen, E. Sihvola, P. Verronen, d G. W. Leppelmeier: LIMBO- Limb d occultation measurement simulator, in ESAMS99, Europe Symposium on Atmospheric Measurements from Space, vol.wpp-6, pp , ESA, Noordwijk, 999 [5] Kyrölä, E. et al. (993), Inverse theory for occultation measurements.. Spectral inversion, J. Geophys. Res., Vol. 98, No. D4, pp [6] Sofieva, V. F., J. amminen, H. Haario, E. Kyrölä d M. Lehtinen. Ozone profile smoothness as a priori information in the inversion of limb measurements, Annales Geophysicae, 04, Vol. 22, No. 0, pp [7] amminen, J., E. Kyrölä d V. Sofieva, Does a priori information improve occultation measurements? in Occultations for Probing Atmosphere d Climate, edited by G. Kirchengast, U. Foelshe d A. Steiner, Springer Verlag, 04, pp