¿ìè¶ÌÊÓƵ

Handling all the data produced by ERS-1 is causing concern. John Mason,
former director of Britain’s Meteorological Office, believes more thought
should have been put into the data processing at an earlier stage.

The colourful maps of different aspects of the Earth’s surface produced
by satellites require a large amount of processing. The data first received
are simply numerical data taken from sensor readings. ESA does some processing
on this to correct it for the position and attitude of the satellite and
then sends it to archiving centres in several ESA member countries.

Mason says that there is no funding in Britain for its archiving centre,
the Earth Observation Data Centre at Farnborough, to routinely process ERS-1
data into continuous maps of real physical properties such as temperature
and wind speed, or for combining the data with information from other sources.

But the amount of data produced by ERS-I will be dwarfed by the quantities
that will be sent down from NASA’s planned series of satellites in its Earth
Observing System. Peter Denning, a former director of NASA’s Research Institute
for Advanced Computer Science, says the agency will find it impossible to
keep all the data and make them available to researchers when the project
starts in 1998.

He says that the EOS satellites will send back more than a terabit (a
million million bits) of information a day. ‘If you imagine doing that for
10 years you have an immense storage and retrieval problem,’ says Denning.
He calculates the EOS system will fill 2500 compact discs of information
every day-9 million over 10 years.

NASA is planning a network of data centres for EOS. But even if the
agency decided to bear the cost of storing all the data, no communications
link would be capable of passing it on to researchers.

Denning says sensor technology is improving quicker than computing and
storage technology. So even if data retrieval gets 100 times faster over
the next 7 years, the amount of data transmitted by the satellites will
probably get 1000 times bigger.

‘It looks like a technological problem at first,’ says Denning, ‘but
it is a problem with our values as well.’ He thinks that as satellites get
better, scientists are going to have to sacrifice the principle of saving
all data. ‘Something will have to give.’

There are good reasons for saving all of the data. It is needed if someone
else wants to check a calculation, it costs a lot to collect and it may
contain important evidence that was overlooked first time round. But Denning
believes that in future researchers will have to rely on intelligent software
to select and discard information. ‘One possibility is to use intelligent
machines to sort through data and make meaningful statistical summaries.’

An example of the problem with this approach occurred with the discovery
of the ozone hole. For years, satellites flying over the Antarctic detected
dramatically low ozone levels, but software discarded them as errors. The
hole was finally discovered by a land-based sensor.

‘What gets left out may be important. I do not know how to handle that
one,’ says Denning. ‘But we make decisions all the time. We close off options
which later we may regret.’

William Bown