Annual report on GFS GCM performance

A mailing list I receive from UCAR brought news of an addition to the GFS forecast archive “New Dataset: NCEP GFS 0.25 Degree Global Forecast Grids Historical Archive” and links to a web page.

On browsing an excellent general interest snippet appeared

Page 16/17, performance over time for a few parameters of various GCM and with the observation on the GFS useful horizon “Increase is about one day per decade”.

Review of GFS Forecast Skills in 2013
Fanglin Yang
IMSG – Environmental Modeling Center
National Centers for Environmental Prediction

Link is on the performance review page, 2013 rev. is PDF (2.7MB)

In the review the improvements over the years are shown and also a good deal of comparison with competitive GCM so this helps gauge the general situation.

Fanglin has been candid or at least as far as is sensible in publication which is a refreshing change after reading so many works of spin and in contrast to the dismal works of climatic “GCM” which are a totally different animal.

The computing facilities and code has of course been ramping up over the years, gridding has become finer so this looks like slow progress. I think we are seeing the intractable nature of the problem, against ultimately chaos. Any major improvement is going to need a breakthrough, an unlikely development.

In addition there have been changes in weather observations used as input parameters.

I’m sure all of us who live or visit weather wild parts of the world wish them luck, a job for life.