Global climate models do not easily downscale for regional predictions | EurekAlert! Science NewsThe problem has been well known for some time, but I guess this is putting some more specific details into it. Also, it makes it clear what nonsense it was for the head of the CSIRO to suggest we could afford to move on from climate modelling work.
Zhang and Michael Mann, Distinguished professor of atmospheric
science and director, Earth System Science Center, were concerned that
the direct use of climate model output at local or even regional scales
could produce inaccurate information. They focused on two key climate
variables, temperature and precipitation.
They found that projections of temperature changes with global
climate models became increasingly uncertain at scales below roughly 600
horizontal miles, a distance equivalent to the combined widths of
Pennsylvania, Ohio and Indiana. While climate models might provide
useful information about the overall warming expected for, say, the
Midwest, predicting the difference between the warming of Indianapolis
and Pittsburgh might prove futile.
Regional changes in precipitation were even more challenging to
predict, with estimates becoming highly uncertain at scales below
roughly 1200 miles, equivalent to the combined width of all the states
from the Atlantic Ocean through New Jersey across Nebraska. The
difference between changing rainfall totals in Philadelphia and Omaha
due to global warming, for example, would be difficult to assess. The
researchers report the results of their study in the August issue of Advances in Atmospheric Sciences.