INTRODUCTION 
When a satellite is out of range of a ground station any data acquired 
by on-board instruments must be recorded and then replayed when the ground 
station is in view. Clearly, the quantity of data that can be acquired in 
this way is Limited by the capacity of the tape recorder. In order to 
increase the coverage are, At is necessary to use data compression techni- 
ques 40 that imagery from a given area can be stored in a fewer number of 
bits. 
We have given particular consideration to an Ocean Colour Monitor (OCM) 
instrument which was originally scheduled to fly on ERS-1 but is now planned 
for a Later mission. 
2 THE RICE MACHINE 2 
The particular data compression technique we have studied is known as 
Rice Machine 2 (RM2). This has been described elsewhere (Rice 1985 ; McEvoy 
1981) but a brief description will be given here. 
Cascaded Hadamard transformations are applied to the image data, 
which {is processed in blocks of 64 square. Four Levels of Zhe transform are 
performed and the transformed data is entropy coded in blocks of 16. The 
Hadamard transformations produce unimodal data which 4s symmetrically 
distributed around zero. The basic principle is that the most commonly 
occurring values are given the shortest codes. This Type of compression 
which can be achieved depends on the randomness orn "activity" of the data. 
Using CZCS data from the North Sea we have obtained compression ratios of 
typically 2.8, 
Our objective has been a compression ratio of 4 and to obtain this 
it is necessary to reduce the bit quantisation of the transformed data. This 
naturally introduces errors in the data when it is reconstructed and we 
have found that the n.m.s evn is 0.7, compared Zo a maximum value of 255 
fon 8-bit data. 
3 COMPARISON WITH OTHER SYSTEM ERRORS 
  
This reconstruction error needs to be compared with other system 
errors. In particular : 
1) Detector noise. The detector themselves are noisy with a finite . 
signal-to-noise ratio. Fon the CICS, this is of the orden of 125 £o 1. 
7) Calibration errors. The data acquired by the instrument consists Of 
digital numbers which have to be converted Xo radiances through a Linear 
relationship. For the OCM, the absolute calibration is expected to be 
accurate to 2% and the interband calibration to 1$. 
3) Atmospheric correction. Before the imagery can be processed to calcutate 
chlorophyll concentrations, it must be corrected for atmospheric effects. 
At present, this is not a calculation which can be performed to any high 
degree of accuracy and there is considerable uncertainty concerning some 
of the critical parameters eg the aerosol optical thickness, 14 is unlikely 
that the atmospheric contriburion to the total radiance is known to better 
than 2%. 
  
  
  
  
  
  
  
  
  
   
  
  
  
  
   
   
   
  
  
  
  
  
  
  
   
  
  
  
  
   
  
  
  
  
  
  
   
  
  
  
  
  
  
  
  
  
  
  
  
     
By compari 
tively sma 
Amprovemen 
al 
We h 
concentu. 
experiment 
wher 
1 and 3 
The 
to sea-thu 
Data 
An the val 
proportion 
b but inde 
RANGANG (f 
terns. tuo 
directly p 
worst case 
This 
conceptu 
a 2$ enun 
approximat 
We 4 
nements. B 
sungace me 
tion withi 
to 50% (Ch 
5 COVE 
We h 
to ERS-1,