Bob Noel writes:

btw - back in 1984 I attended a computing symposium at Goddard
Space Flight Center (if that's now the official name, sorry, but it was
at Goddard). One talk of interest was the computing power needed
for weather forecasting. The supercomputers used for the models
in 1984 required 48 hours of wallclock time to render a 24 hour
forecast. That forecast was only a bit better than 50% (?) correct.
They were estimating the need for several orders of magnitude more
CPU power as well as more efficient models. Well, we certainly have
much more CPU power available now...

But we also need much more accurate and extensive data, which we do not have.

The computing power and data increase more than exponentially with the length
of the forecast required, because of the chaotic nature of weather. That's
why weather forecasting has barely made any progress at all, even as computing
power has increased by many orders of magnitude.

--
Transpose mxsmanic and gmail to reach me by e-mail.