Document not found! Please try again

Approach to the Cross-Validation of MIPAS and ... - Semantic Scholar

2 downloads 0 Views 124KB Size Report
The approach to cross-validate vertical profiles of temperature ... terminology (Rodgers 1976), the improved guess xi+1 of the solution is cal- culated from the ...
Approach to the Cross-Validation of MIPAS and CHAMP Temperature and Water Vapour Profiles Gabriele Petra Stiller, Tilman Steck, Mathias Milz, Thomas von Clarmann, Udo Grabowski, and Herbert Fischer Institut f¨ ur Meteorologie und Klimaforschung (IMK), Forschungszentrum Karlsruhe GmbH / Universit¨ at Karlsruhe, Postfach 3640, 76021 Karlsruhe, Germany [email protected]

Summary. The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) is a space-borne limb viewing mid-infrared high-resolution spectrometer launched on ENVISAT into polar orbit on 1 March 2002. Water vapour and temperature profiles are retrievable from spectral data in an altitude range from approximately 5 km up to the mesosphere. Complementing the operational analysis of ESA, a non-operational MIPAS level 2 processor has been developed at IMK to derive geophysical data from the MIPAS observations. The IMK MIPAS level 2 data processing scheme supports a multitude of regularization approaches. Full retrieval diagnostics in terms of retrieval covariance matrices, averaging kernel matrices etc. are available. Vertical profiles of temperature and humidity have been identified to be candidate quantities for comparison with corresponding CHAMP data as part of commissioning phase as well as long-term cross-validation activities. The accuracy of MIPAS water vapour and temperature measurements is investigated by test retrievals based on simulated data. The approach to cross-validate vertical profiles of temperature and humidity from CHAMP and MIPAS, respectively, is based on the χ2 of the difference of MIPAS and CHAMP retrieved profiles under consideration of error correlations, averaging kernels, and spatial and temporal mis-matching of the observations. Key words: Remote sensing, atmospheric observation, MIPAS, water vapour

1 Introduction On 1 March 2002, the high-resolution mid-IR FTS limb emission sounder MIPAS (Michelson Interferometer for Passive Atmospheric Sounding) has successfully been brought into space on board of ENVISAT. ENVISAT is a sun-synchronous polar orbiter with 98.55◦ inclination in approximately 800 km altitude; subsequent orbits will provide global coverage about every 3 days with 14.4 orbits per day. MIPAS is a limb-viewing Fourier transform infrared (FTIR) emission spectrometer with 0.035 cm−1 spectral resolution (unapodised), covering the mid infrared from 685 cm−1 to 2410 cm−1 (14.6 – 4.15 µm). Tangent heights from about 5 to 150 km are supported. The

552

Gabriele Stiller et al.

field of view (FOV) of MIPAS corresponds to a width of 30 km in horizontal direction and approximately 3 km at the tangent point in vertical direction. The standard limb observation grid covers 6 to 68 km with a step width of 3 km. The sampling rate is about 500 km along-track and about 2800 km across-track at the equator. The standard observations cover the complete globe for day, night, or twilight conditions from the upper troposphere to the mesosphere.

2 IMK level-2 data processing Besides the operational data processing under ESA responsibility, nonoperational science-oriented data analysis will be performed at IMK within which vertical profiles of temperature, water vapour, and some 30 other species will be retrieved. The IMK level-2 data analysis (i.e. retrieval of atmospheric parameters) is based on the operational ESA level-1 data (i.e. calibrated and geo-located radiance spectra). As part of extensive validation efforts during commissioning phase as well as long-term activities, intercomparison and cross-validation of the vertical profiles of temperature and humidity as derived from MIPAS and CHAMP for the altitude range covered by both instruments is planned to be done. These activities will cover data amounts of several thousand limb sequences per year, an amount equivalent to several days of global MIPAS data. 2.1 Retrieval strategy Inferring vertical profiles of atmospheric parameters from a limb sequence of infrared emission spectra involves the inverse solution of the radiative transfer equation. In order to account for nonlinearities, the linear estimation of the solution is embedded in a Newtonian iteration scheme. Following Rodgers’ ˆ i+1 of the solution is calterminology (Rodgers 1976), the improved guess x ˆ i as culated from the current guess x −1 ˆ i+1 = x ˆ i + (KTi S−1 x y Ki + R + λI)

ˆ i )), (KTi S−1 xi )) + R(xa − x y (y − F(ˆ

(1)

ˆ of the ith iteration, Sy is the covariwhere Ki is the Jacobian matrix ∂y/∂ x ance matrix of the measurement error, and R is a regularization term. The damping term λI shall improve convergence in nonlinear cases and consists of the scalar Marquardt parameter λ and unity matrix I (Marquardt 1963; Levenberg 1944). F is the forward radiative transfer model which provides ˆ i , and y contains the measured simulated measurements as a function of x spectral radiances; xa is the a priori information the solution shall be constrained to.

MIPAS and CHAMP Cross-Validation

553

Fig. 1. Left: retrieval results (solid lines) and true profiles (dashed lines) of test retrievals, for H2 O (black) and temperature (grey). Middle: relative difference to true H2 O-profile (solid) and estimated noise error (dotted). Right: as middle but absolute difference of T-profiles. The spectra have been perturbed by detector noise as specified for MIPAS.

Atmospheric state parameters are retrieved on a fixed, i.e. tangent heightindependent, altitude grid finer than the tangent altitude spacing. Typical gridwidth is 1 km. Retrieval of quantities where independent measurements are available, such as line of sight, are constrained by optimal estimation (Rodgers, 1976). Retrieval of quantities where only climatological a priori knowledge is available are constrained by Tikhonov’s 1st and 2nd derivative operators (Tikhonov 1963). Further details of the IMK retrieval strategy are published in Clarmann et al. (2001). In Fig. 1 examples for both, a simulated water vapour and a simulated temperature retrieval are given. The simulated measurement spectra are representative for midlatitudinal atmospheric conditions with a tropopause altitude of 11 km and a hygropause altitude of 16 km. The estimated error due to measurement noise is in the order of 10 % for H2 O, which is confirmed by the test retrieval. For the temperature, the estimated random error is in the order of 1 K for the complete stratospheric as well as the upper tropospheric region. Except for the middle and lower troposphere the temperature retrieval results show good agreement with the reference. The vertical reso-

554

Gabriele Stiller et al.

lution is high enough to reproduce the tropopause and the hygropause with good accuracy. Test retrievals based on more realistic scenarios including all relevant error sources in the simulated reference spectra confirm these estimates of achievable accuracies. 2.2 Retrieval Diagnostics Besides the retrieved vertical profiles, retrieval diagnostics to evaluate the data quality and further properties of the data products are available. Here we focus on the ones relevant to validation. The averaging kernel matrix A is given as ˆ ∂x −1 T −1 (2) A= = (KT S−1 K Sy K = Gy K, y K + R) ∂x ˆ /∂y is the gain matrix. A provides the dependence of the where Gy = ∂ x ˆ on the true profile x. The width of the maxima in the rows retrieved profile x of A gives an estimate of the vertical resolution (Carli 1998; Steck 2001). The covariance matrix Sx of the retrieval, which includes the error component due to measurement noise is given as Snoise = Gy Sy GTy

(3)

The systematic or forward model parameter covariance matrix Sf representing the error components due to uncertain model parameters represented by their covariance matrix Sb is given as Sf = Gy Kb Sb KTb GTy .

(4)

3 Cross validation of different measurements For the intercomparison of retrieved quantities from two instruments, the quantities have to be made comparable, e.g. by interpolation on a common grid. Then, by application of the formulation of Rodgers and Connor (1999) and Rodgers (2000) the mean expected difference between the two retrieved ˆ 1 and x ˆ 2 from different instruments can be expressed in terms of profiles x covariances Sδ as Sδ = (A1 − A2 )T Sc (A1 − A2 ) + Sx1 + Sx2 ,

(5)

where Sc is the climatological covariance of a comparison ensemble, which is needed to estimate the effect of different smoothing characteristics on the retrievals. Spatial-temporal mismatch can be considered by adding an additional term Smm to Sδ . Finally the χ2 of the difference between the retrieved profiles is ˆ 2 )T (Sδ + Smm )−1 (ˆ ˆ 2 ). χ2 = (ˆ x1 − x x1 − x (6) This quantity indicates if the two retrieved profiles are in agreement or not: The χ2 expectation value for interconsistency measurements is equal to the number of degrees of freedom.

MIPAS and CHAMP Cross-Validation

555

4 Future application to MIPAS and CHAMP CHAMP currently provides temperature profiles in the troposphere up to the middle stratosphere, but water vapour profiles in the troposphere only. While the planned observation scenarios for MIPAS allow coverage of the troposphere down to about 5 km, retrievability of water vapour still needs to be confirmed with real MIPAS data. Studies based on simulated data show that, in the cloud-free case, meaningful water vapour volume mixing ratios can be retrieved down to about 11 km for the tropics; for midlatitude observations, the full altitude range observed can be exploited. Retrievability of mid-tropospheric water vapour depends strongly on the presence of (thin) clouds and aerosol layers. Limitations in the altitude range where meaningful MIPAS retrievals can be performed could result in a reduced set of observations available for crossvalidation. Crucial for the intercomparison is the availability of data sets with sufficiently small temporal and spatial mis-match. What sufficiently small is, however, currently is under investigation. If it should turn out that there are not sufficient observations available to perform any significant intercomparison ensemble mean data have to be compared instead of individual profiles. In this case the temporal and local scatter are expected to average out while biases remain visible. Once successfully cross-validated, both observation systems will benefit from their consistency and can provide synergistic information. Tropopause and tropospheric water vapor information from CHAMP could be used as initial guess or a priori information for the MIPAS retrievals, or temperature information from MIPAS could be used by CHAMP within the water vapour retrieval. Finally, the retrieved profiles from both observing systems could be merged to provide consistently increased altitude coverage.

5 Conclusion and outlook The IMK level 2 retrieval processor provides good results and useful retrieval diagnostics. The retrieval errors for H2 O and temperature in the stratospheric and upper tropospheric regions are in the order of 10 % and 1 K, respectively for MIPAS. Especially during the commissioning phase the cross validation with CHAMP will provide insight in the data quality and data characteristics of both instruments.

6 Acknowledgements The main part of this work has been done under BMBF contract no. 01 SF9953/8. T.S. and M.M. have been supported by the EU project AMIL2DA (EVG1-CT-1999-00015) and BMBF contract no. 07 ATF43.

556

Gabriele Stiller et al.

References 1. Carli B, Ridolfi M, Raspollini P, Dinelli BM, Dudhia A, Echle G (1998) Study of the retrieval of atmospheric trace gas profiles from infrared spectra. Technical report, European Space Agency, Final Report of ESA Contract 12055/96/NL/CN 2. Levenberg A (1944) A method for the solution of certain non–linear problems in least squares. Quart Appl Math, 2, 164–168 3. Marquardt DW (1963) An algorithm for least–squares estimation of nonlinear parameters. J Soc Indust Appl Math, 11(2), 431–441 4. Rodgers CD (1976) Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation. Rev Geophys Space Phys, 14(4), 609–624 5. Rodgers CD, Connor BJ (1999) Intercomparison of remote sounding instruments. In: Optical Remote Sensing of the Atmosphere, OSA Technical Digest, pages 46–48, Optical Society of America, Washington DC, 1999 6. Rodgers CD (2000) Inverse Methods for Atmospheric Sounding: Theory and Practice. World Scientific, Series on Atmospheric, Oceanic and Planetary Physics, FW Taylor, ed, Vol 2 7. Steck T, von Clarmann T (2001) Constrained profile retrieval applied to the observation mode of the Michelson interferometer for passive atmospheric sounding. Appl Opt, 40(21), 3559–3571 8. Tikhonov A (1963) On the solution of incorrectly stated problems and method of regularization. Dokl Akad Nauk SSSR, 151:501 9. von Clarmann T, Fischer H, Funke B, Glatthor N, Grabowski U, H¨ opfner M, Kiefer M, Mart´ın-Torres FJ, Milz M, Stiller GP (2001) MIPAS interactive semi– operational level–2 data processing. In: Smith WL, Timofeyev Yu M (eds) IRS 2000 Current Problems in Atmospheric Radiation, pages 785–788. A Deepak Publishing, Hampton, Va, USA

Suggest Documents