The atmospheric correction procedure for the ATSR-2 solar reflective channels, specifically
for landsurfaces, is intended for application within an operational environment to provide
atmospherically corrected surface reflectance image products. The overall correction procedure
has been separated into two discrete components: first the derivation of the atmospheric code
for rapid computation of the pixel surface reflectance from the sensor apparent reflectance for
changing atmospheric conditions and, second, incorporation of the code into an atmospheric
correction procedure. The present paper deals with the first of these components and outlines
the rationalization of the 5S code to provide a rapid execution 5S pseudo-code for atmospheric
correction.
Fast atmospheric correction procedures which integrate the results from radiative transfer codes
such as LOWTRAN or 5S have been outlined by Richter (1990) and Telliet (1992). Richter
used LOWTRAN to produce a catalogue of atmospheric correction functions for TM bands 1-5
and 7; Telliet constructed look-up tables of the inversion coefficients / 1 (\,), B (L,)and spherical
albedo S(\,) for AVHRR data using a modified 5S code, for a series of solar zenith, view
zenith and relative azimuth angles. The present solution differs from that of Telliet in that
the computations of the inversion coefficients and spherical albedo for specific illumination
and viewing geometries, and for changes in the aerosol optical depth, are included within the
operational procedure. The pre-computation phase is restricted to those atmospheric parameters
which are independent of the atmospheric path length.
Regarding the second component of the study, the atmospheric correction procedure for
land-surfaces is intended for application without the requirement for in-situ atmospheric data:
the correction will be derived using standard atmospheric models, characteristic aerosol models,
and the apparent reflectance data in the four ATSR-2 solar reflecting channels at the two
observation angles. It is anticipated that an iterative procedure will be employed within the
atmospheric correction in which the pseudo-code is deployed to calculate the surface reflectances
assuming a range of aerosol optical depths. The appropriate surface reflectances and atmospheric
optical properties will be selected to be consistent with some pre-selected reflectance criteria.
The use of this type of iterative procedure intensifies the use of the atmospheric code and the
requirement for a computationally efficient operating code.
It is also pertinent to note that an important feature of this atmospheric modelling is that by
assuming standard atmospheric profiles and aerosol models, the variability in the atmospheric
optical properties can be related solely to changes in the aerosol optical depth at a reference
wavelength x 71 (550 nm). Therefore, in operation, the success of the atmospheric correction
procedure relates to the ability to solve for a single parameter.
The 5S computer code (Tanre et al ., 1986) was developed to allow estimation of the solar
radiation backscattered by the Earth-surface-atmosphere system, as it is observed by a satellite
sensor. This signal depends on the surface reflectance and the perturbation by two atmospheric
processes: gaseous absorption and scattering by molecules and aerosols. Given a Lambertian
surface reflectance, the apparent reflectance at the sensor is calculated from separate estimates
of the gaseous transmission and the atmospheric functions: intrinsic atmospheric reflectance,
direct and diffuse transmittances and spherical albedo. Inhomogeniety of the surface reflectance
may also be accounted for approximately by estimating the environmental effects using an
environmental weighting function (Tanre et al., 1990).
The draw-back to the direct use of the 5S code within the correction procedure is the time
required for repeated calculations for large volumes of data, especially if the calculations are
to be iterated for each location (agglomeration of pixels). The 5S pseudo-code has therefore
been developed to improve the execution speed of the computations using the 5S model.
The computational procedure of the pseudo-code is modified from 5S to reduce the run time.
A significant computational load is eliminated by pre-calculating the Rayleigh optical depth
and gaseous absorbing parameters for each standard atmospheric model; also pre-calculated are,
the aerosol optical parameters: phase function P(9, A) (0= phase angle), single scattering albedo
oo 0 (A,), asymmetry factor g(A.,), and extinction coefficient for each characteristic aerosol type.
These parameters are required for each spectral bandpass. The pre-calculated parameters are