Automatized Atmospheric Absorption Atlas (4A)
4A for Automatized Atmospheric Absorption Atlas is a fast and accurate line-by-line radiative transfer model particularly efficient in the infrared region of the spectrum.
Introduction
Calculating transmittances, jacobians, radiances and fluxes for a given input of atmospheric and surface conditions is known as forward modelling. Basically, both research and operational studies require fast, accurate and tractable forward models.
Applications of such forward models include :
- the simulation of high spectral resolution radiances and study of their sensitivity to surface and atmospheric variables to interface NWP models
- the selection of the best possible spectral intervals for the retrieval of atmospheric and surface parameters
- the investigation of the dynamic range of variation of radiances received at the satellite as a function of spectral and atmospheric variables
- the generation of “observations” to be used in simulation studies (retrievals, validation processes, direct assimilation, “Look-Up-Tables”,TIGR dataset, etc...)
- the validation of level1 or level2 of satellite data
- the modelling of the Earth radiation budget, simulations of atmospheric cooling rates and radiative forcing.
The 4A line-by-line model to calculate forward radiative transfer is an advanced version of the nominal line-by-line STRANSAC model. These forward models have become more and more accurate and efficient through the exploitation of new mathematical techniques, the availability of faster and faster computer systems, and, last but not least, the provision of better spectroscopic data ( GEISA database).
The 4A model has a long history of validation within the frame of the international radiative transfer community. Most of the validation results have been extensively discussed in a number of intercomparison exercises and in particular during the ITRA (Intercomparison of Transmittance and Radiance Algorithms) working groups - 1983, 1985, 1988, 1991 of the International Radiation Commission (see Chédin et al., 1988) and during the ICRCCM (Intercomparison of Radiation Codes in Climate Models ) campaigns (see F. Luther et al. 1988). More recently, launch of hyperspectral sounders (AIRS on Aqua platform and IASI on Metop platform) have led to more and more extensive validations, still within the frame of international campaigns or working groups.
4A/OP has been chosen by CNES as the official radiative transfer model for IASI level1 CAL/VAL and level1 operational processing.
The 4A (Automatized Atmospheric Absorption Atlas) : a brief description
Briefly, 4A is a compressed look-up-table of optical depths. The concept was described in Scott N.A. and Chédin A. (1981).
STRANSAC - the LMD genuine line-by-line and layer-by-layer model (Scott N.A., 1974; and later on Tournier et al., 1995) - is used to compute the atlases of optical depths :
- for 12 nominal atmospheres (12 temperature profiles, 7K distant)
- for each absorbing gas (reference concentration profiles)
- for a set of 43 (40 in the nominal version) pressure levels (between surface and top of the atmosphere; 43 is the current default value although not mandatory)
- at least, a fraction of half width spacing, thus varying with pressure : the so-called "spectral representation step"
4A processes 15 cm-1 blocks, each block being associated with thousands of matrices (number of spectral steps times the number of layers times the number of temperatures times the number of gases considered).
Each matrix is compressed in the wavenumber/layer/temperature space and in the wavenumber/layer/gas space and only the significant values are kept. Values (in transmittance units) too close to zero or too close to one are kept and treated separately in order to be reintroduced and taken into account in case of high values of viewing angles or absorber amounts.
The resulting compressed matrices are stored. To do a calculation, a compressed matrix is :
- uncompressed (in addition, "zero" and "one" are reintroduced)
- interpolated:
- to the correct temperature for each layer (interpolation between two adjacent temperature profiles)
- to the correct pressure levels (the nominal 43 pressure levels are not mandatory)
- scaled to the correct absorber amount and viewing angle (all gases may be considered as "variable" with respect to the altitude)
Optical depths from each individual gas are summed up and radiance calculation is performed. Starting from these high spectral resolution optical depths, transmittance profiles, Jacobian profiles (the latter two being optional), and brightness temperatures are generated using an appropriate "spectral integration step" (e.g. a fraction of the smallest encountered spectral representation step) combined with a relevant convolution step to take into account the various instrument functions. The description of the analytic computation of the Jacobians has been given in the mid 90’s (F. Chéruy et al, 1995). So far, the computation of temperature, absorbing gases, surface temperature, emissivity, jacobians is optionally performed.
The 4A model : Characteristics of the current version
During these last years, the model underwent important transformations, in relation with our in house applications (aerosols retrievals from AIRS and IASI, processing of the ACE-Scisat instrument for the retrieval of CO2 profiles, …), or in relation with applications within the frame of the CNES activities.
4A is maintained at LMD. This includes introducing the newly derived parameters for spectroscopy, for line-coupling or for continua, for aerosols and CFCs, ..., as soon as they have been validated. The current version is referred to as 4A/OP 2009.
Items | 2009 Version | 2006 Version |
Spectroscopic database | GEISA 2009 | GEISA 2003 |
H2O continuum | Clough et al., 1989 | Clough et al., 1989 |
N2 continuum | Lafferty et al, 1996 | Lafferty et al, 1996 |
O2 continuum | Thibault et al,1997 | Thibault et al,1997 |
Partition Function | Gamache et al., 2000 | Gamache et al., 2000 |
Line coupling | Niro et al, 2004 | Rodrigues et al, 1998 |
χ correction of Voigt profiles | Perrin et al, 1989 | Perrin et al, 1989 |
pressure shift for H2O | yes | no |
Spectral dependence of the surface emissivity | yes | no |
Simulation of Scattering effects of aerosols | yes, based on DISORT K. Stamnes et al, 1988 |
no |
Implementation of the limb viewing geometry | yes | no |
Number of nominal layers / Top of the atmosphere | 43 / 0.0026 hPa | 40 / 0.05 hPa |
Adjustment of gas concentration profiles in the Atlas computations | yes | N.R. |
Illustration of some 4A model outputs
4A can be used for a wide variety of atmospheric and surface conditions. Since spectra are computed at a high spectral resolution (5.10-4 cm-1), 4A keeps specificities of a line-by-line model. As a consequence, any kind of apparatus function may be used at the time of the convolution. So, 4A is and has been used for many different applications including low and high resolution spectra of the Earth atmosphere (HIRS, AVHRR, METEOSAT, AIRS, IASI, MODIS, SEVIRI, IIR ...) and spectra of planetary atmospheres (e.g., within the frame of the VOYAGER mission to giant planets).
The good quality of 4A and of the IASI spectra (spectral resolution and signal-to-noise ratio) together with the good description of the atmosphere ( ARSA database) makes it possible to analyze errors along the spectra as signatures of insufficient spectroscopy or missing or uncorrectly represented minor absorbers.
Figure below presents "4A-simulated versus IASI-observed" brightness temperature spectra (blue, right y-axis) and residuals spectra (red, left y-axis) as well as their associated standard deviations (green) (in°K) for a set of 764 IASI/ ARSA 50km*3hours space*time window collocations (from July 2007 to December 2009, clear sea/night cases, tropical situations). From top to bottom: IASI bands I, II, and III and, for each of these band, spectra, residuals and standard deviations. (Click the figure to enlarge it).
The figure below shows an example of the variation of the 4A Jacobians versus pressure with respect to perturbations of respectively 1K and 10% on the temperature profile (left) and CO2 concentration (right) for the mean tropical atmosphere of the TIGR dataset . (Click the figure to enlarge it).
Figure below presents an example of 4A-simulated transmittances (green) versus ACE-FTS observed transmittances (red) spectra (top) and residuals spectrum (bottom) for a 11.612 km tangent height observation, in the 2450-2550cm-1 range. (Click the figure to enlarge it).
An illustration of the 4A/DISORT model is given below. It shows the spectral dependence – between 650 and 900 cm-1 or 15.4 µm - 11 µm -, of the sensitivity of IASI channels to an aerosols layer of IR optical depth 1.5 (at 10 µm), at an altitude of 2400 m (solid black line and right ordinate). Superimposed are spectra using an atmosphere increased by 20% of water vapour (magenta) or ozone (cyan), and, for comparison, 20%, 10%, and 5% of the aerosol contribution (all results on the left ordinate). Black dots (•) indicate channels selected for aerosol retrieval in this spectral region. Example below is for a mean tropical atmosphere). (Click the figure to enlarge it).
4A/OP, the operational version of 4A
In accordance with the convention signed between CNES, LMD/CNRS, NOVELTIS is currently in charge of the consolidation and the distribution of 4A : such a version is referred to as 4A/OP. The 4A/OP 2009 version is presently distributed.
The code is written in standard Fortran 90 under UNIX or Linux operating systems, to be ran on different platforms. A Graphical User Interface (GUI) is implemented. More information on this operational version, is given in the NOVELTIS 4A/OP website.
Previous page: Tools
Next page: Improved Initialization Inversion system (3I)