Channel Compression of Trace Gas Remote Sounder by Expanding the Weighting Function with Empirical Orthogonal Functions
25 Aug 2004-Journal of the Meteorological Society of Japan (Meteorological Society of Japan)-Vol. 82, Iss: 4, pp 1081-1093
TL;DR: A method is proposed to compress the multi-channel sounder data for global greenhouse gases monitoring into a small number of hypothetical channels loosing a negligible information content that the original channel has, showing that the 240 channels of original radiance data can be condensed to 5 channels or less of hypothetical radiances, with loosing negligible information that the originally data has.
Abstract: In this paper a method is proposed to compress the multi-channel sounder data for global greenhouse gases monitoring into a small number of hypothetical channels loosing a negligible information content that the original channel has. Recent Earth observation satellites provide radiance data of thousands of channels at one observation point, piling up gigabyte of data per day. Meanwhile it has recently been recognized that the atmospheric field analysis for the numerical weather prediction produces a better result, when it uses the radiance data of satellite sounder rather than such the retrieved physical parameters, as the temperature and humidity. The difficulty of using radiance data of large number channels is that it uses too much resources in the assimilation loop of three, or four-dimensional variational formulation. In the present work we propose a method to compress a large number of original channels to a small number of hypothetical channels, whose weighting functions are the eigenvectors made from the matrix of original weighting functions for various atmospheric and observational conditions. The original weighting functions of an arbitrary spectrum concerned are, in turn, expanded with the eigenvectors obtained as above. Using this relation, the radiances of hypothetical channels are obtained by a linear combination of originally observed radiances. The error covariance matrix of the hypothetical channels is also obtained from measurement error of original channels. Since, in the present method of hypothetical radiance, much more channels of original spectrum can be incorporated in the analysis, almost all the information that the original spectrum has is maintained. An example of the application is shown for the simulated radiance data of a remote sounder of the atmospheric greenhouse gases, which observes the solar radiation reflected in the Sun glint region of the water surface, with a high spectral resolution. The result showed that the 240 channels of original radiance data can be condensed to 5 channels or less of hypothetical radiances, with loosing negligible information that the original data has.
Citations
More filters
TL;DR: In this paper, the authors describe a method of simultaneously retrieving atmospheric temperature, moisture, and cloud properties using all available IASI channels without sacrificing computational speed using an empirical orthogonal function (EOF) transformation.
Abstract: . The Infrared Atmospheric Sounding Interferometer (IASI) is an ultra-spectral satellite sensor with 8461 spectral channels. IASI spectra contain high information content on atmospheric, cloud, and surface properties. The instrument presents a challenge for using thousands of spectral channels in a physical retrieval system or in a Numerical Weather Prediction (NWP) data assimilation system. In this paper we describe a method of simultaneously retrieving atmospheric temperature, moisture, and cloud properties using all available IASI channels without sacrificing computational speed. The essence of the method is to convert the IASI channel radiance spectra into super-channels by an Empirical Orthogonal Function (EOF) transformation. Studies show that about 100 super-channels are adequate to capture the information content of the radiance spectra. A Principal Component-based Radiative Transfer Model (PCRTM) is used to calculate both the super-channel magnitudes and derivatives with respect to atmospheric profiles and other properties. A physical retrieval algorithm then performs an inversion of atmospheric, cloud, and surface properties in the super channel domain directly therefore both reducing the computational need and preserving the information content of the IASI measurements. While no large-scale validation has been performed on any retrieval methodology presented in this paper, comparisons of the retrieved atmospheric profiles, sea surface temperatures, and surface emissivities with co-located ground- and aircraft-based measurements over four days in Spring 2007 over the South-Central United States indicate excellent agreement.
67 citations
Cites methods from "Channel Compression of Trace Gas Re..."
...Aoki (2004, 2005) described a method of compressing high resolution infrared spectra into a few hypothetical channels using a regression matrix and EOFs derived from weighting functions....
[...]
TL;DR: A novel radiative transfer model and a physical retrieval algorithm are applied to hyperspectral data taken during the European Aqua Atmospheric Thermodynamics Experiment (EAQUATE) campaign to retrieve atmospheric temperature, moisture and ozone profiles and surface properties such as surface skin temperature and surface emissivity.
Abstract: The objective of the paper is to apply a novel radiative transfer model and a physical retrieval algorithm to hyperspectral data taken during the European Aqua Atmospheric Thermodynamics Experiment (EAQUATE) campaign. A principal-component-based radiative transfer model (PCRTM) is used to calculate projection coefficients of the radiance spectrum onto a set of predefined empirical orthogonal functions (EOFs) and associated derivatives with respect to the state vector. Instead of fitting channel radiances, the physical retrieval algorithm iteratively fits the principal component (PC) scores or the EOF projection coefficients of the observed radiance spectrum using the PCRTM as its forward model. Since the EOFs are orthonormal to each other, only a few PC scores are needed to capture the information content of the radiance spectrum, therefore reducing the computational time needed for running both the forward model and the inversion. This paper demonstrates the application of such a physical algorithm for retrieving atmospheric temperature, moisture and ozone profiles, and surface properties such as surface skin temperature and surface emissivity. The results have been compared with those obtained with a NAST-I channel-based physical retrieval algorithm and with those obtained from collocated radiosonde and LIDAR measurements. Copyright © 2007 Royal Meteorological Society
31 citations
01 Dec 2006
TL;DR: In this paper, a fast radiative transfer model based on empirical orthogonal functions (EOF) is proposed for the simulation of sensors with different characteristics and in different spectral ranges from the solar to the infrared.
Abstract: Remote sensing with the new generation of highly spectrally resolving instruments like the Atmospheric Research Interferometer Evaluation System (ARIES) or the assimilation of highly resolved spectra from satellites into Numerical Weather Prediction (NWP) systems requires radiative transfer computations that deliver results essentially instantaneous. This paper reports on the development of such a new fast radiative transfer model. The model is based on an Empirical Orthogonal Functions (EOF) technique. The model can be used for the simulation of sensors with different characteristics and in different spectral ranges from the solar to the infrared. For the purpose of airborne remote sensing, the fast model has been designed to work on any altitude and for slant paths whilst looking down or up. The fast model works for situations with diverse temperature and humidity profiles to an accuracy of better than 0.01K for most of the instrument channels. The EOF fast model works for clear-sky atmospheres and is applicable to atmospheres with scattering layers of aerosols or clouds. The fast model is trained with a large set of diverse atmospheric training profiles. In forward calculations corresponding high resolution spectra are obtained. An EOF analysis is performed on these spectra and only the leading EOF are retained (data compression). When the fast model is applied to a new independent profile, only the weights of the EOF need to be calculated (=predicted). Monochromatic radiances at suitable frequencies are used as predictors. The frequency selection is done by a cluster algorithm, which sorts frequencies with similar characteristics into clusters.
27 citations
Cites background from "Channel Compression of Trace Gas Re..."
...An alternative is discussed in (Aoki, 2004 and Aoki, 2005), where the EOF analysis is done for spectra of weighing functions (the derivatives of radiances with respect to layer temperatures or water vapour amounts, also referred to as Jacobians)....
[...]
TL;DR: In this article, the authors examined the performance of this method for the simulated radiance data of high spectral resolution in the thermal infrared region that is used for the remote measurements of temperature and humidity vertical profiles.
Abstract: A method has recently been shown by Aoki (2004) to compress the number of channels of trace gas remote sounder preserving almost all the information content that the original data has. In this method the weighting function of the original channels is expanded with empirical orthogonal functions (EOFs), and a set of hypothetical radiances, whose weighting functions are the EOFs, are used for the analysis. It has been shown that the radiance data of 240 original channels, which is obtained by the above trace gas sounder, can be compressed to 3 or less channels of hypothetical radiances, loosing negligible information content. In the present paper, the performance of this method is examined for the simulated radiance data of high spectral resolution in the thermal infrared region that is used for the remote measurements of temperature and humidity vertical profiles. The number of hypothetical channels that is required for preserving the information content of original data has been examined for two different spectral resolutions, 0.1 and 0.3 cm - 1 . The present study showed that the radiance data of 1200 channels of high spectral resolution 0.1 cm - 1 is compressed to less than 23 hypothetical channels with negligible loss of information content in the case of wavenumber region 640-760 cm - 1 , which is used for the temperature soundings. For the low resolution 0.3 cm - 1 , the number of hypothetical channels needed becomes smaller than this. In the case of the humidity sounding from the region 1300-1600 cm - 1 with the spectral resolution 0.1 cm - 1 , 3000 channels are compressed to less than 20 hypothetical channels.
7 citations
20 Jan 2005
TL;DR: In this paper, the authors examined the information content of the vertical profile of trace gas, which the spectrum of the reflected solar radiation has, and showed that the crrelation rapidly decreases with the decrease of the range of the data that generates the EOFs.
Abstract: A method has recently been developed to compress the number of channels of trace gas remote sounder preserving almost all the information content that the original data has 1 . In this method, the weighting function of the original channels is expanded with empirical orthogonal functions (EOFs), and a set of hypothetical radiances, whose weighting functions are the EOFs, are used for the analysis. It has been shown that the radiance data of 240 of original channels of CO 2 spectrum at around 6207 cm -1 can be compressed to about 3 channels of hypothetical radiances with loosing negligible information content. This means that the information content of the vertical profile of trace gas, which the spectrum of the reflected solar radiation has, is not so much. In the present paper, the information content of the vertical profile is examined for typical two types of spectra of CO 2 and CH 4 absorption bands at around 1.6 and 2.1 µ regions. Another issue of this paper is the correlation between the measurement error of the hypothetical channels. Since the hypothetical radiance is generated by the linear combination of radiances of the original channels, it could have the correlation between the radiance errors in hypothetical channels. It is shown that the crrelation rapidly decreases with the decrease of the range of the data that generates the EOFs. Keywords: Channel compression, EOFs, hypothetical radiance, information content, correlation of error
3 citations
References
More filters
TL;DR: In this article, the authors present a review of the methods which may be used to estimate the state of the atmosphere, i.e., the distribution of temperature and composition, from measurements of emitted thermal radiation such as are made by remote sounding instruments on satellites.
Abstract: This paper reviews the methods which may be used to estimate the state of the atmosphere, i.e., the distribution of temperature and composition, from measurements of emitted thermal radiation such as are made by remote sounding instruments on satellites. The principles of estimation theory are applied to a linear version of the problem, and it is shown that many of the apparently different methods to be found in the literature are particular cases of the same general method. As an aid to understanding, the optimum linear solution is described in terms of the geometry of n dimensions, with n = 3 for illustration. In generalizing the approach to the nonlinear problem there are two stages: (1) finding any member of the infinite family of possible solutions, which may be done by any convenient iterative method, and (2) finding the optimum solution by satisfying appropriate constraints.
1,483 citations
TL;DR: In this paper, an approximation to 4D-Var, namely the incremental approach, is considered and is shown to produce the same result at the end of the assimilation window as an extended Kalman filter in which no approximations are made in the assimilating model but in which instead a simplitied evolution of the forecast error is introduced.
Abstract: SUMMARY An order of magnitude reduction in the cost of fourdimensional variational assimilation (4D-Var) is required before operational implementation is possible. Reconditioning is considered and, although it offers a signi6cant reduction in cost, it seems that it is unlikely to provide a reduction as large as an order of magnitude. An approximation to 4D-Var, namely the incremental approach, is then considered and is shown to produce the same result at the end of the assimilation window as an extended Kalman filter in which no approximations are made in the assimilating model but in which instead a simplitied evolution of the forecast error is introduced. This approach provides the flexibility for a cost-benefit trade-off of 4D-Var to be made. The development of variational four-dimensional assimilation (4D-Var) from the stage of being a theoretical possibility to being a practical reality is progressing at a rapid pace. The first results of four-dimensional variational assimilation using real observations were provided by Thbpaut et al. (1993b) using an adiabatic primitive-equation model at truncations "21 and T42. More recently Andersson ef al. (1994) used 4D-Var with a T63 model to assimilate remotely-sensed data such as infrared and microwave TOVS radiance measurements, while Thdpaut et d. (1993a) used 4D-Var with the same model to assimilate normalized radar backscatter cross-section measurements from the ERS-1 scatterometer.
1,473 citations
TL;DR: The 3D-Var model as discussed by the authors uses a spherical-harmonic expansion, much as the ECMWF optimal interpolation (OI) scheme used an expansion of Bessel functions.
Abstract: In the first of this set of three papers, the formulation of the European Centre for Medium-Range Weather Forecasts (ECMWF) implementation of 3D-Var is described. In the second, the specification of the structure function is presented, and the last is devoted to the results of the extensive numerical experimentation programme which was conducted. The 3D-Var formulation uses a spherical-harmonic expansion, much as the ECMWF optimal interpolation (OI) scheme used an expansion of Bessel functions. This formulation is introduced using a convolution algebra over the sphere expressed directly in spectral space. It is shown that all features of the OI statistical model can be implemented within 3D-Var. Furthermore, a non-separable statistical model is described. In the present formulation, geostrophy is accounted for through a Hough-modes separation of the gravity and Rossby components of the analysis increments. As in OI, the tropical analysis remains essentially non-divergent and with a weak mass-wind coupling. The observations used, as well as their specified statistics of errors, are presented, together with some implementation details. In the light of the results, 3D-Var was implemented operationally at the end of January 1996.
652 citations
TL;DR: 1DVAR scheme is developed at the European Centre for Medium-range Weather Forecasts as a method for extracting information from TIROS Operational Vertical Sounder radiances for use in the operational data-assimilation system and has demonstrated consistent positive impacts on forecast skill in the northern hemisphere.
Abstract: In recent years difficulties have been experienced in exploiting satellite sounding data in numerical weather prediction (NWP) in the form of independently retrieved temperature and humidity profiles. Attention has now focused on methods through which the information in the radiance measurements may be assimilated more directly into the NWP system%.
A scheme known as ‘one-dimensional variational analysis’ (1DVAR) has been developed at the European Centre for Medium-range Weather Forecasts as a method for extracting information from TIROS Operational Vertical Sounder radiances for use in the operational data-assimilation system. The 1DVAR scheme is based on variational principles applied to the analysis of the atmospheric profile at a single location, using a forecast profile and its error covariance as constraints. The details of the scheme are presented. Errors in 1DVAR products are correlated with those of the short-range forecast which serves as a background for the subsequent three-dimensional analysis. Methods for addressing this aspect of the assimilation problem are discussed.
The characteristics of 1DVAR products and their impact on the analysis are described. A series of forecast impact experiments has been conducted and has demonstrated consistent positive impacts on forecast skill in the northern hemisphere.
259 citations
TL;DR: The direct use of TOVS (TIROS Operational Vertical Sounder) cloud-cleared radiances in a three/four-dimensional variational data assimilation scheme is described and an impact on the tropical wind field from the use of the humidity-sensitive TOVS channels is shown.
Abstract: The direct use of TOVS (TIROS Operational Vertical Sounder) cloud-cleared radiances in a three/four-dimensional variational data assimilation scheme is described. This scheme uses a fast radiative transfer model and its adjoint. Radiances are used together with all the other observational data. Global spectral fields of mass, wind and humidity are analysed simultaneously under certain mass/wind balance constraints which control the degree to which gravity waves enter into the analysis. In this way the need for a subsequent initialization is avoided. The scheme thus combines retrieval, analysis and initialization in one step and makes it possible to achieve a more optimal combination of the information contained in the radiances, the conventional data and the background (a six-hour forecast).
At spectral truncation T63, a global three-dimensional variational analysis (3D-Var) of TOVS radiances and conventional data is compared with the ECMWF operational Optimum Interpolation scheme, which uses TOVS radiance information in the form of profiles of temperature and humidity, retrieved using a one-dimensional variational method. The results of 3D-Var are in good agreement with the operational analysis at the same resolution.
In an application of the four-dimensional scheme (4D-Var), data covering a period of 24 hours were used simultaneously. In 4D-Var consistency in time is ensured through the evolution of the forecast model and its adjoint. Using the adiabatic version of the ECMWF forecast model (spectral resolution T42) we show that 4D-Var is able to extract additional information from the dynamics of the model. In particular we show an impact on the tropical wind field from the use of the humidity-sensitive TOVS channels.
209 citations