Validation of GOMOSEnvisat High Resolution Temperature Profiles HRTP Using Spectral Analysis

V.F. Sofieva, J. Vira, F. Dalaudier, A. Hauchecorne, and the GOMOS Team

Abstract The GOMOS (Global Ozone Monitoring by Occultation of Stars) instrument on board the Envisat satellite is equipped with two fast photometers operating at 1 kHz sampling frequency in blue and red wavelengths. The bi-chromatic scintillations recorded by the photometers allow the determination of refractive angle, which is proportional to the time delay between the photometer signals. The high resolution density and temperature profiles can be reconstructed (with a vertical resolution of about 200 m) from these data in the altitude range ~ 15-35 km. The validation of small-scale fluctuations in HRTP requires a very close collocation in time and space with high-quality data having comparable or better vertical resolution. Comparing spatial spectra of temperature profile fluctuations requires less strict collocation criteria. In this paper, we compared vertical wavenumber spectra of temperature fluctuations in HRTP and in collocated radiosonde data. We found that the vertical wavenumber spectra of HRTP and radiosonde temperature fluctuations are very similar in case of vertical occultations of bright stars. In case of oblique occultations or of dim stars, the HRTP fluctuations often have a larger spectral magnitude, despite of several good agreements. The spectral analysis has confirmed that the actual resolution of HRTP is 150-200 m.

1 HRTP: Measurement Principle and Retrievals

GOMOS on board Envisat is the first instrument performing synchronous scintillation measurements at two wavelengths. It is equipped with two fast photometers operating at blue (470-520 nm) and red (650-700 nm) wavelengths with a sampling frequency of 1 kHz (http://envisat.esa.int/instruments/gomos, Bertaux et al. 2004; Kyrola et al. 2004).

The retrieval of high-resolution temperature and density exploits chromatic refraction in the atmosphere. Due to dependence of air density (and, consequently,

Finnish Meteorological Institute, Earth Observation, Helsinki, Finland e-mail: [email protected]

Fig. 1 The schematic representation of chromatic refraction and the principle of refraction angle measurements by GOMOS

refractivity) on wavelength, the blue ray bends more than red one (Fig. 1), thus it is observed later by GOMOS.

The intensity structures (scintillation spikes), which are caused by atmospheric density fluctuations, are observed by both photometers with a time delay tB-tR (Fig. 2) proportional to the difference in refraction angle Aa = aB-aR. In the HRTP processing, the time delay is estimated as the position of the maximum of the cross-correlation function of photometer signals.

The difference in refractive angles Aa is transformed into the refractive angle aB at the reference wavelength. After that, the retrieval method is similar to the one used in radio occultation. Assuming local spherical symmetry of the atmosphere, the refractivity profile is retrieved from the refractive angle profile using the Abel transform. The refractivity is proportional to air density. The corresponding pressure

Fig. 2 Signals of GOMOS fast photometers (FP1, blue, and FP2, red). Time delay is clearly seen in the photometer data

profile is deduced by integrating the hydrostatic equation and, finally, the temperature profile is obtained from the state equation of a perfect gas. The details of the HRTP processing are described in Dalaudier et al. (2006).

At altitudes ~ 18-35 km, the estimated accuracy of HRTP is 1-2 K. The best accuracy is achieved for vertical (in orbital plane) occultations of bright stars. Below ~15 km, the quality of HRTP decreases due to low signal-to-noise ratio, broadening of scintillation peaks as a result of chromatic smoothing and the violation of the assumptions used in the retrievals (in particular, the weak scintillation assumption). The vertical resolution of HRTP profiles is ~200 m.

The HRTP profiles used in this validation were processed with the research software, which was developed within the scope of the ESA funded project "Algorithms for the estimation of high resolution temperature and density profiles from GOMOS measurements" aimed at optimization of HRTP retrievals. GOMOS data from September 2002 to January 2005 were used. Collocated sounding data available in the Envisat calibration/validation database at NILU (Norwegian Institute for Air Research) and in the SHADOZ (Southern Hemisphere Additional Ozoneson-des) database (Thompson et al. 2001) were used for the validation.

2 Using Spectral Analysis for HRTP Validation: Motivation

The beneficial feature of HRTP is a good vertical resolution. In order to validate small-scale fluctuations in HRTP that are assumed to be generated by internal gravity waves (IGW), correlative measurements should provide high-quality temperature in the same altitude range with a similar or better resolution.

Let us discuss the data collocation criteria, i.e., a space-time window, where the temperature field does not change much and where we can expect similarity of temperature profiles, including their small-scale fluctuations. In the stratosphere, the characteristic ratio of horizontal and vertical scales is typically equal to the ratio of maximal and minimal intrinsic frequencies of the gravity waves field (N/f is typically larger than 100, N is the Brunt-Vaisala frequency and f is the Coriolis parameter). Since the vertical resolution of HRTP is ~200 m, the horizontal separation of collocated measurements should ideally be less than 20 km. HRTP is most sensitive to gravity waves with small vertical and large horizontal wavelengths, i.e., waves of low intrinsic frequencies (Fritts and Alexander 2003). Since the characteristic time for evolution of such gravity waves is few hours, the time difference between collocated profiles should not exceed 2-3 h.

The criteria for data selection derived above are very strict. In the altitude range of HRTP, only radiosonde data satisfy the vertical resolution requirements. It is important to note that the time separation results in additional spatial separation in the atmosphere caused by advection of air masses. For example, provided a very moderate wind speed of 20 m/s for altitudes 15-30 km, air parcels probed at the same geographical location with a time interval of 30 min will be separated by 36 km in the atmosphere. Relatively long measurement time of temperature profiles by radiosondes during balloon flights (it takes ~1 hour for balloon to raise from 10-30 km) has a similar effect. If a satellite (nearly instantaneous) measurement is perfectly collocated in time and space with a radiosonde measurement at 10 km, then they will be separated in the atmosphere up to several tens of kilometers at 30 km due to displacement by wind.

It is nearly impossible to find data satisfying these very strict collocation criteria (all our searches were not successful so far). According to the analysis of collocated radiosonde data (Sofieva et al. 2008), difference in temperature rapidly grows with increasing separation distance. Rms of temperature differences for profiles having ~200 m vertical resolution is ~0.4 K for 40 km separation, ~0.7 K for 80 km separation, and it is ~ 1-1.5 K for separations 200-1000 km.

If the separation of measurements in the atmosphere exceeds 20-30 km, the small scale structures of temperature profiles should not coincide. However, we can expect similar spectral properties of the temperature field at locations not far from each other (e.g., less than 500 km) during some time period (few hours). The spectral analysis of collocated radiosonde profiles (Sofieva et al. 2008) has confirmed this hypothesis: the spectra of temperature fluctuations look similar, even for profiles significantly separated in time and space (a few hundreds of kilometers, several hours). The values of rms of profile fluctuations are found to be very close to each other (the smaller distance, the smaller difference in rms of fluctuations). For radiosonde profiles separated in the stratosphere by 300-600 km, rms of temperature fluctuations are within ± 40% interval in majority of cases (Sofieva et al. 2008).

The spectral analysis approach allows using data collocated in a wider space-time window, thus more data are suitable for validation. Furthermore, the validation of the HRTP spectra is of high importance, as one of the possible HRTP applications is the study of internal gravity wave activity in the stratosphere.

The vertical resolution of HRTP is expected to be ~200 m (worse than of radiosonde profiles but better than that from radio-occultation measurements). This resolution is sufficient for probing vertical spectra of gravity waves. Comparison of vertical spectra of temperature fluctuations in HRTP and collocated radiosonde profiles allows experimental estimation of actual HRTP vertical resolution: it corresponds to a cut-off of vertical spectra at large wavenumbers.

However, comparing spectra of temperature fluctuations in HRTP and collocated radiosonde profiles should be performed with care, because the vertical wavenumber spectra of temperature fluctuations in the ground-based (HRTP) and in GW intrinsic reference frame (balloon measurements) can differ as a result of the wind-shifting effect (e.g., Eckermann 1995). Gardner and Gardner (1993) estimated influence of horizontal winds on vertical wavenumber spectra. They found that modifications in vertical wavenumber spectra caused by background winds are negligible if:

(i) ratio of horizontal wind speed to balloon ascend speed is much smaller than the anisotropy coefficient n (in Gardner and Gardner (1993), n ^ 22 is taken). This means that the time required to ascend through a temperature irregularity of characteristic vertical dimension is short compared to the time required to advect the balloon horizontally through the irregularity with the characteristic horizontal size.

(ii) the time required to ascend through a temperature irregularity of characteristic vertical dimension is short compared to the lifetime of the irregularity.

The second condition can be considered as always satisfied, as balloons ascend with the mean velocity of 4-5 m/s. The first condition can be violated only in case of very strong horizontal winds.

For HRTP, the influence of obliquity of occultation on the distortion of vertical wavenumber can be estimated in a similar way. The difference between spectra obtained using instantaneous vertical and oblique profiles is expected to be small, if obliquity a of occultation satisfies tan a<<n (a is the angle between the local vertical and the direction of star motion). This condition is satisfied for all occultations considered in the present work.

3 Validation Results 3.1 Data Selection

We selected temperature profiles from PTU (pressure, temperature, humidity) and ozone soundings at high and mid latitudes available in NILU database, located at < 300 km distance and having < 4 h time difference with the corresponding GOMOS measurements. The location of stations and sonde characteristics are collected in Table 1. 63 profiles were found satisfying the collocation criterion. However, only 27 sounding profiles were covering a significant part the HRTP altitude range, i.e., altitudes ~ 18-30 km. Most of the selected GOMOS occultations are in "stray light" illumination conditions (i.e., not in full dark) and oblique (off orbital plane). Vertical occultations of bright stars (|a|<5°), where the best accuracy of HRTP is expected, were not presented within this selected set. In order to have possibility to estimate HRTP quality in "the best" occultations, we found 12 vertical occultations of bright stars collocated with SHADOZ soundings. However, the spatial separation of these occultations and SHADOZ soundings is large, 300-600 km, as well as time difference, ~12 h. The information about the

Table 1 Location of sounding stations at mid and high latitudes and some characteristics of soundings

Station

Location

Sonde type and typical vertical resolution

Legionowo

52.40°N, 20.97°E

50 m (PTU)

Uccle

50.8°N, 4.35°E

75 m (PTU and ozone sondes)

Jokioinen

52.40°N, 20.97°E

10 m (PTU), 50 m (ozone sondes)

Sodankyla

67.37°N, 26.63°E

10 m (PTU and ozone sondes)

Ny-Âlesund

78.92° N, 11.93° E

50 m (ozone sonde)

Scorebysund

70.50°N, 22.00°W

50 m (ozone sonde)

Marambio

64.2° S, 56.7°W

50 m (ozone sonde)

Dumont D'Urville

66.67° S, 140.01°E

90 m (ozone sonde)

Fig. 3 GOMOS HRTP from occultation of star S018 and collocated radiosonde temperature profile at Jokioinen on September 25, 2002. Distance between profiles, time difference and obliquity of occultation a (the angle between the local vertical and the direction of star motion in the phase screen) are specified in the figure header

Fig. 3 GOMOS HRTP from occultation of star S018 and collocated radiosonde temperature profile at Jokioinen on September 25, 2002. Distance between profiles, time difference and obliquity of occultation a (the angle between the local vertical and the direction of star motion in the phase screen) are specified in the figure header

HRTP meas.err.

-sonde

HRTP

220 225 Temperature, K

HRTP meas.err.

-sonde

HRTP

220 225 Temperature, K

location of SHADOZ sounding stations and sonde characteristics can be found at http://croc.gsfc.nasa.gov/shadoz/and in Thompson et al. (2001).

An example of collocated HRTP and sounding profiles is shown in Fig. 3. Two error estimates are shown for HRTP: measurement error (shaded area) and the total error estimate, which includes the measurement error and the upper limit initialization error (dashed lines). ECMWF temperature profile at the occultation location is also shown.

3.2 Results of Spectra Comparison

For computing power spectral density, the profiles were interpolated to a common equidistant 30 m altitude grid. For detection of temperature fluctuations, the smooth component is obtained using Hanning filtering with the cut-off scale 3 km.

The power spectral density of relative fluctuations of temperature &T/T is estimated using the method of averaged periodogram. In our analysis, we focus on wave numbers lower than 0.01 cy/m, as the high-frequency part of spectra is influenced by aliasing. For the spectral analysis, we used profiles in the altitude range 20-30 km.

Figure 4 shows several spectra of relative temperature fluctuations in the collocated HRTP and sounding profiles at mid and high latitudes (occultations of different types, mostly oblique). The "model" lines correspond to the model of the saturated gravity waves, which predicts kz~3 shape of the vertical wavenumber spectrum V5T/T of the relative temperature fluctuations:

10-3 10-2 10-3 10-2 10-3 10-2 vertical wavenumber, cy/m

Fig. 4 Spectra of relative temperature fluctuations in sounding profiles and in HRTP, collocations at high and mid latitudes o

10-3 10-2 10-3 10-2 10-3 10-2 vertical wavenumber, cy/m

Fig. 4 Spectra of relative temperature fluctuations in sounding profiles and in HRTP, collocations at high and mid latitudes

N4 3

g where kz is the vertical wave number, g is the acceleration of gravity, and A«0.1 is an experimental constant (Smith et al. 1987). They are presented only for reference, as the current analysis is not aimed at checking the hypothesis on universality of GW vertical spectra. It is known that departures from the spectrum (1) are rather common in the stratosphere (Eckermann 1995; Fritts and Alexander 2003).

The rms of temperature fluctuations (computed as a sample standard deviation) and information about GOMOS occultations and sonde measurements are also specified in Fig. 4.

The spectral density of HRTP fluctuations shown in Fig. 4 is usually larger than that of soundings fluctuations. The horizontal wind velocity is not more than about 10 times the balloon ascent velocity in all the considered collocations. Thus the difference of HRTP and radiosonde wavenumber spectra is unlikely explained by the influence of background wind estimated in Gardner and Gardner (1993) and discussed in Sect. 2 of this chapter. We have not found any clear dependence of spectra discrepancy on magnitude of horizontal winds. All occultations selected at high and mid latitudes are either of dim stars (thus the measurement noise is significant) or/and oblique. In oblique occultations, isotropic scintillations caused by turbulence disturb the correlation between photometer signals. Thus accuracy of HRTP retrievals, which are based on finding the correlation between photometer signals, decreases in oblique occultations. Therefore, disagreement in spectral characteristic for these occultations is not very surprising. The situation drastically changes for vertical occultations of bright stars (Fig. 5, these occultations are collocated with SHADOZ soundings in tropics). The vertical wavenumber spectra are very similar for HRTP and sondes, as well as the rms of fluctuations.

Figure 6 shows the scatter plot of rms of temperature fluctuations in HRTP and radiosonde profiles, for these two datasets. Figure 6 summarizes observations described above. For vertical occultations of bright stars, rms of HRTP fluctuations

-sonde -HRTP---model

R03875/S012 At = 14.5 h, As = 385.8 km,a = 5° a . = 0.99 K, au„ „ = 0.95 K

R03875/S012 At = 14.5 h, As = 385.8 km,a = 5° a . = 0.99 K, au„ „ = 0.95 K

R09300/S010 At = 10.8 h, As = 358.1 km, a = -4° a . = 0.83 K, au„ „= 1.49 K

R09300/S010 At = 10.8 h, As = 358.1 km, a = -4° a . = 0.83 K, au„ „= 1.49 K

R11891/S009 At = 15.1 h, As = 432.9 km, a = 6° a „=1.19 K, au„TO = 1.75 K

R11891/S009 At = 15.1 h, As = 432.9 km, a = 6° a „=1.19 K, au„TO = 1.75 K

sonde HRTP

R14310/S020, a = 2° At = 10.6 h,As = 441.6 km sonde

HRTP

R08535/S029 At = 12.9 h, As = 106.9 km, a= 3° a „=1.29 K, au„ „= 1.29 K

R14297/S012 At = 12.8 h,As = 550.1 km,a = 2° a „ = 0.72 K, a„„TO= 1.26 K

R14297/S012 At = 12.8 h,As = 550.1 km,a = 2° a „ = 0.72 K, a„„TO= 1.26 K

10-2 10-3 10-' vertical wavenumber, cy/m

R14180/S012 At = 11.8 h,As = 616.2 km, a= -6° a „= 0.97 K,au__ = 1.33 K sonde HRTP

R14180/S012 At = 11.8 h,As = 616.2 km, a= -6° a „= 0.97 K,au__ = 1.33 K sonde HRTP

10-3

Fig. 5 Spectra of relative temperature fluctuations in SHADOZ sounding profiles and in HRTP

sonde HRTP

S 10

sonde

HRTP

O high and mid latitudes, oblique occultations O high and mid latitudes, vertical occultations of bright stars * tropics, vertical occultations of bright stars

O high and mid latitudes, oblique occultations O high and mid latitudes, vertical occultations of bright stars * tropics, vertical occultations of bright stars

Fig. 6 Rms of temperature fluctuations of HRTP profiles versus that of sounding profiles. Solid line: y=x, dashed lines: y=1.2x and y=(1/1.2) x; dotted lines: y=1.5x and y=(1/1.5)x

Fig. 6 Rms of temperature fluctuations of HRTP profiles versus that of sounding profiles. Solid line: y=x, dashed lines: y=1.2x and y=(1/1.2) x; dotted lines: y=1.5x and y=(1/1.5)x is close to that in the collocated radiosonde temperature profiles. The scattering of the data is very similar to that observed in collocated radiosonde data (Sofieva et al. 2008). In case of oblique occultations or dim stars, fluctuations in HRTP are larger than in radiosonde temperature profiles, but several good agreements were also observed.

A clear cut-off corresponding to scales ~ 150-200 m is observed in HRTP spectra (Fig. 5). This confirms that the actual HRTP vertical resolution is 150-200 m.

4 Conclusions and Outlook

The validation of small-scale structures in high-resolution profiles is the problem of high complexity, because the collocated data with the same vertical resolution should be available nearly at the same time and location. In this work, we based and described the spectral analysis approach to validation of high-resolution profiles, which requires less strict collocation criteria.

The application of this method for validation of high-resolution temperature profiles retrieved from bi-chromatic stellar scintillation measurements by GOMOS fast photometers has shown that HRTP fluctuations are realistic (in terms of their 1D vertical spectra) in vertical occultations of bright stars. In case of oblique occultations or of dim stars, the HRTP fluctuations often have a larger spectral magnitude, despite of several good agreements. The spectral analysis has confirmed that the actual resolution of HRTP is 150-200 m.

The large class of occultations - close to vertical occultations of medium-brightness stars (with visual magnitude 1 < m < 2.5) is not presented in the considered data sets. We expect that the accuracy of HRTP profile retrievals from such occultations is close to that with bright stars, as moderate noise does not affect the computation of cross-correlation function and thus the accuracy of time delay reconstruction. The validation of these occultations is the subject of future work.

In addition, it is possible also to use all available radiosonde data world-wide and find suitable collocations with the GOMOS occultations for spectral validation of HRTP. This will possibly give a larger statistics and a better coverage of different types of occultations.

Excessive amplitude of HRTP fluctuations, which was detected in this study in case of dim stars or oblique occultations, can potentially be reduced by applying regularization in retrievals. The HRTP profiles presented in this study were processed with the minimum of a priori information used in the inversion. The inversion method based on statistical optimization (Bayesian approach) has also been developed within the scope of the project "Algorithms for the estimation of high resolution temperature and density profiles from GOMOS measurements". Validation of retrievals with the regularization applied will also be the subject of future work.

Acknowledgements This work has been performed within the scope of the ESA funded project "Algorithms for the estimation of high resolution temperature and density profiles from GOMOS measurements" (Contract number 17895/04/I-LG). The authors thank ESA and the GOMOS team for the GOMOS data. The authors greatly appreciate the use of correlative data from the ENVISAT Cal/Val database at NILU, and thank the principal investigators of the radiosondes and SHADOZ sondes. The work of V.F. Sofieva was supported by the Academy of Finland (postdoctoral researcher project).

References

Bertaux JL, Hauchecorne A, Dalaudier F, Cot C, Kyrola E, Fussen D, Tamminen J, Leppelmeier GW, Sofieva V, Hassinen S, Fanton d'Andon O, Barrot G, Mangin A, Theodore B, Guirlet M, Korablev O, Snoeij P, Koopman R, Fraisse R (2004) First results on GOMOS/Envisat. Adv Space Res 33:1029-1035, doi:10.1016/j.asr.2003.09.037 Dalaudier F, Sofieva V, Hauchecorne A, Kyrola E, Blanot L, Guirlet M, Retscher C, Zehner C (2006) High-resolution density and temperature profiling in the stratosphere using bi-chromatic scintillation measurements by GOMOS. Proceedings of the First Atmospheric Science Conference, European Space Agency, ISBN 92-9092-939-1-ISSN 1609-042X Eckermann SD (1995) Effect of background winds on vertical wavenumber spectra of atmospheric gravity waves. J Geophys Res 100(D7):14097-14112 Fritts DC, Alexander MJ (2003) Gravity wave dynamics and effects in the middle atmosphere. Rev

Geophys 41(1):1003, doi:10.1029/2001RG00106 Gardner C, Gardner N (1993) Measurement distortion in aircraft, space shuttle, and balloon observations of atmospheric density and temperature perturbation spectra. J Geophys Res 98(D1):1023-1033

Kyrölä E, Tamminen J, Leppelmeier GW, Sofieva V, Hassinen S, Bertaux JL, Hauchecorne A, Dalaudier F, Cot C, Korablev O, Fanton d'Andon O, Barrot G, Mangin A, Theodore B, Guirlet M, Etanchaud F, Snoeij P, Koopman R, Saavedra L, Fraisse R, Fussen D, Vanhellemont F (2004) GOMOS on Envisat: An overview. Adv Space Res 33:1020-1028, doi:10.1016/S0273-1177(03)00590-8

Smith SA, Fritts DC, VanZandt TE (1987) Evidence of a saturation spectrum of atmospheric gravity waves. J Atmos Sci 44(10):1404-1410

Sofieva VF, Dalaudier F, Kivi R, Kyro E (2008) On the variability of temperature profiles in the stratosphere: Implications for validation. Geophys Res Lett 35(L23808), doi:10.1029/ 2008GL035539

Thompson AM, Witte JC, McPeters RD, Oltmans SJ, Schmidlin FJ, Logan JA, Fujiwara M, Kirchhoff VWJH, Posny F, Coetzee GJR, Hoegger B, Kawakami S, Ogawa T, Johnson BJ, Vomel H, Labow G (2001) Southern Hemisphere Additional Ozonesondes (SHADOZ) 1998-2000 tropical ozone climatology 1. Comparison with Total Ozone Mapping Spectrometer (TOMS) and ground-based measurements. J Geophys Res 108(D2):8238, doi:10.1029/2001JD000967

Renewable Energy Eco Friendly

Renewable Energy Eco Friendly

Renewable energy is energy that is generated from sunlight, rain, tides, geothermal heat and wind. These sources are naturally and constantly replenished, which is why they are deemed as renewable.

Get My Free Ebook


Post a comment