||This paper discusses the design and implementation
of data compression applied to output from numerical
weather prediction models. Such techniques will
become essential when increased computational resources
lead to much increased data volumes at weather
forecast centers. The purpose of the data compression
is to reduce the data storage requirements as well
as transmission times and costs when serving remote
outstations and customers with direct model output.
Important issues to be considered are precision of the
decoded data, the encoding and decoding speed, the
complexity of the coding techniques, and the compatibility
with existing standards such as GRIB.
As more powerful computers become available
the precision of local weather forecasts can be improved
by using higher resolution in space and time
for weather simulations. For example, at the German
Weather Service (DWD), typically grids of 200 by 200
points are used for each of 20 model levels in the vertical.
It is anticipated to replace the current mainframe
computer in order to allow to increase the resolution
to 800 by 800 by 50. This will blow up the resulting
data file sizes by a factor of 40. Even though storage
devices are becoming less expensive there is obviously...