solar zenith angle of the block of pixels being 
processed. 
Unlike channels 1 and 2 , which receive energy 
from the sun reflected from the earth-atmosphere 
complex, channels 4 and 5 are sensitive to the 
radiation emitted by the earth at long infra-red 
wavelengths. The equivalent brightness 
temperatures of channels 3, 4 and 5 are 
calculated by first applying the calibration 
coefficients provided by the AVHRR and then 
inverting the Planck function. The intensity of 
the response in channel 3 is partly due to 
reflectance of solar energy and partly due to the 
radiation of the earth in that portion of the 
spectrum. The portion due to reflectance is 
estimated by using the channel 4 brightness 
temperature. 
ATMOSPHERIC CORRECTION 
Because the atmosphere does not scatter and 
absorb all wavelengths equally, the radiation 
reaching the sensor has a spectrum different from 
that which would have been received in the 
absence of an atmosphere. The radiance measured 
by the orbital instrument is therefore only 
partially dependent on the reflectance of the 
target (Tanr6 e t al. 1986). As a result of 
atmospheric scatter a substantial portion of the 
radiation reaching the sensor is contributed by 
reflection of the sun's radiation within the 
atmosphere. The target is illuminated not only 
by direct sunlight, but also by skylight, and a 
non-negligible part of the energy reaching the 
sensor has been scattered more than once in its 
double passage through the atmosphere. 
The algorithms used in the pre-processing 
software will as far as possible estimate surface 
geophysical parameters (bi-directional 
reflectance factors of channels 1 and 2 and 
brightness temperatures of the surface) 
independently of atmospheric properties. 
The physics of the radiative transfer through the 
atmosphere are well documented, and form the 
basis of several models providing atmospheric 
correction. However, most such models involve 
recursion to account for multiple scattering or 
ray- or photon-tracing and are too cumbersome to 
incorporate into the operational preprocessing of 
AVHRR data, whose main advantage is its high 
temporal frequency, and hence the potential 
timeliness of the data. 
Most of the effects of the atmosphere can 
nevertheless be corrected operationally, and to 
this aim the LOA, under the Tecnodata contract, 
is to optimise its "Simulation of the Satellite 
Signal in the Solar Spectrum", or "5S" model 
(Tanre et a l . 1986) for incorporation in the pre 
processing software. 
The model, as its name suggests, was originally 
designed to use ground and atmospheric 
characteristics to model the signal received at 
the satellite, and not to derive ground 
characteristics from the satellite signal. 
However, it is based on a set of analytic 
expressions of the optical properties of the 
atmosphere, each of which is reversible, given 
the inherent symmetry and reversibility of the 
passage of light through any transparent medium. 
For the purposes of atmospheric correction of 
AVHRR data, the model has therefore been re 
organised to derive the reflectance of the 
surface from 
satellite. 
the signal received at 
the 
The original 5S model is described in detail 
(with FORTRAN listings of the code) in Tanre e t 
al. ( 1 986 ) . 
Environment 
correction 
and 
nature of the atmospheric 
Water vapour. Channel 2 suffers markedly 
from absorption by water vapour. This leads in 
particular to considerable sensitivity of the 
NOVI to the water vapour content of the 
atmosphere. In order to correct for this effect 
it is necessary to know or to estimate the water 
vapour content at the location at which the 
radiometric data are being collected. The 
European Centre for Medium Range Weather 
Forecasts at Reading have archives (dating from 
1/1/1980) of data derived from a model of 
atmospheric conditions which provide an estimate 
of relative humidity and air temperature at 
midday at 7 barometric altitudes for cells in a 
2.5°x2.5° grid. The European Community is 
covered by about 200 of these cells. At present 
it is not known whether the JRC will have access 
to these data in real-time for inclusion in the 
pre-processing chain, or whether it will be 
necessary to assume seasonal and regional mean 
conditions in water vapour content. 
Aerosols. The software will not attempt to 
correct the data for atmospheric aerosols, since 
on the scale of Western Europe, aerosol loadings 
vary locally and unpredictably with time. 
Insufficient data exist on their distribution and 
particle size for any kind of routine mapping, 
and it is therefore not possible to eliminate 
their effect on a continental scale and on an 
operational basis using present techniques. This 
is to be regretted since for channel 1 and 2 of 
the AVHRR the contribution of aerosols to noise 
in the signal is often more important than that 
of any other atmospheric source. 
Surface characteristics. The 5S model used to 
derive the algorithm for the atmospheric 
correction assumes that the surface is spatially 
non-uniform, but that it is a Lambertian 
reflector. This is clearly incorrect, but we 
lack the necessary knowledge of the bi 
directional reflectance of surfaces in Europe at 
the AVHRR scale. We hope that in the mid to long 
term we will possess this information as one of 
the results of a major collaborative effort 
between the JRC and the University College 
London. 
Look-up tables. The Joint Research Centre 
requires an algorithm that functions efficiently 
and rapidly to remove the major calculable 
effects of the atmosphere on channel 1 and 
channel 2 data. The need for a rapid algorithm 
implies the use of a formula coupled with look-up 
tables that are pre-loaded with the results of 
any calculation that can be carried out 
beforehand. Thus, for example, the software will 
set up sun angle and view angle tables which it 
can index when the geometry of the orbit is 
known . 
access 
Auxiliary files are set up 
to current estimates of water 
for rapid 
vapour and 
oxygen concentrations, 
181 
sags 
1Ü 
;SP