channel calibration techniques for channel 1 and 
channel 2 of each AVHRR since TIROS-N. The 
review will use a single terminology and notation 
irrespective of the terminology and notation used 
in the original articles. For each technique 
presented it will give values of all 
coefficients found by the researchers. 
The work was started, at SCRIPPS (University of 
California), after signature of contract on 6 
June 1989. 
Outline of method 
Selection of targets Several target types were 
selected at various locations in the world. The 
principal characteristics of these targets are 
that they can be reliably identified in a series 
of images and that their radiometric properties 
are known or are likely to remain stable over 
time. Within these categories several study 
sites were selected on the basis of (1) 
availability of suitable data for the site and 
(2) sufficiently large spatial extent of a 
radiometrically homogeneous and stable area. 
The sites chosen for study using 1-km (HRPT or 
LAC) imagery collected at SCRIPPS include desert 
(White Sands in New Mexico, Edwards Base in the 
Mohave Desert in California, the Yuma desert 
which straddles the frontiers between Arizona, 
California and Sonora in Mexico), snow fields 
(high in the Sierra Nevadas), cloud tops . r e gio n s 
in the edges of sun glint over the ocean, and 
well to the north-west of such regions for high- 
view-angle, high-sun-angle data over the ocean. 
Almost all of the signal in these parts of the 
images comes from atmospheric molecular 
scattering. 
Other sites are examined with 3x5-km (GAC) data: 
desert sites in the Sahara include two bright 
Lambertian sites (located roughly at 17N 7E and 
25N 1 1E) , one darker non-Lambertian site which is 
thought to be lightly vegetated (near 19N 14E), 
and three further sites (18N 8W, 28N, 36W and 31N 
OW) whose directional reflectance characteristics 
are not known at the time of writing; and as for 
the 1-km data, cloud tops (mainly in the inter 
tropical convergence zone over West Africa), sun 
glint and atmospheric molecular scattering. 
Preliminary results 
At the deadline for acceptation of papers for 
this conference the contract had not ended. In 
this paper it is therefore only possible to 
report the initial results, based on the first 
few months of the work. The oral presentation 
will be based on the full results which will by 
then be available. 
In order to establish that the methods used 
provide reasonable results, data from the AVHRR 
on board NOAA-9, the best-known of the AVHRRs, 
were given priority. In this way it was possible 
to compare the data with those given in the 
literature. 
The various targets are used to calibrate (1) the 
visible channel (C h1) on its own (desert targets 
and atmospheric molecular scattering) or to 
provide inter-calibration coefficients between 
channels (sun glint, cloud and snow targets). 
Preliminary results (based on 6 measurements 
starting from late 1 985 to late 1 988) for the 
AVHRR on board NOAA-9 show an almost negligible 
drift (of the order of -0.1 NC.year 1 ) in the 
offset of the zero ( N C Q ) , corresponding to the 
deep-space values, for both channels. This 
result confirms that reported by Kaufman and 
Holben. From the data of the same period on the 
desert sites, a loss of sensitivity in channel 1 
of roughly 6% per year (-0.06±0.01 NC.year 1 ) is 
detected. This drift is also in close accord 
with figures published in the literature (Staylor 
1989, Kaufman and Holben). 
At the time of writing, no data were available on 
the water vapour content of the atmosphere, so 
that the desert sites could not be used to 
calibrate channel 2. 
Inter-channel calibration on sun glint targets 
shows an apparent steady decrease in (about 
0.5 yr ^) from launch to the start of 1986, and 
an increase whose rate is of the same order of 
magnitude from 1986 to 1989. The behaviour of 
the inter-calibration coefficient on snow-field 
targets is similar. More results are needed 
before this intriguing observation can be 
explained satisfactorily. 
Selection and treatment of AVHRR data Each 
site is identified in several images of AVHRR 
data collected shortly after launch and every six 
months thereafter until the end of the 
operational life of each instrument. 
Data are carefully selected to avoid any in which 
there are apparent problems (such as unwanted 
cloud or haze). Rather than working pixel-by 
pixel, solar and view angles are calculated for 
the centre of 32x32 pixel blocks. Depending on 
the target, pixels are selected for inclusion in 
the study on the basis of local radiometric 
characteristics. Thus for example cloud-free and 
aerosol-free pixels from molecular scattering 
targets are characterised by high Ch1/Ch2 ratios, 
while "good" glitter pixels are characterised by 
low Ch1/Ch2 ratios. 
In order to eliminate the effect of variations in 
illumination, all the data were converted to 
reflectance values, taking into account the 
equivalent solar constant for the two channels, 
the solar zenith angle and the sun-earth 
d i stance. 
No results are yet available for the other 
targets or for the AVHRRs on N 0A A 7 and 11. 
Conclusions of initial study 
The methods used give results that are well in 
keeping with those available in the literature. 
This encourages us to continue with the work on 
the other satellites using the same techniques 
and the same targets, in the knowledge that the 
final data set will provide us with good in-orbit 
calibration coefficients. 
Pre-processing algorithm 
The appropriate calibration coefficients for 
channels 1 and 2 for a given date are calculated 
by interpolating between known values using a 
least-squares best fit. For the thermal channels 
the values given by NOAA are pre-processed into a 
suitable look-up table making it possible to 
retrieve values rapidly from any one of the 
possible 1023 levels. They are applied in the 
standard manner to the raw data and are then 
converted to reflectance using the earth-sun 
distance, the mean solar irradiance and the mean