Climate as Mean-Value of Variable Weather

Steven Goddard and Roger Pielke Sr discuss anew the basic question of the difference between computational simulation of climate vs weather.  

I have myself addressed this question in previous posts on short-time vs longtime accuracy with the following key aspects of turbulent (chaotic) flow developed in our book Computational Turbulent Incompressible Flow:

• point-values are short-time (not long-time) computable
• mean-values are long-time computable. 
  

Weather represents turbulent flow for which point-values may be computable over short-time (up to a week). Climate represents mean-values of weather and as such may be computable over long-time (years), if climate simulation is performed as weather simulation over long-time. 

In other words, if Global Circulation Models GCM reduce to weather simulation over long-time, it is possible that they may deliver some correct information as concerns climate. Current GCM do not seem to be of this form and their value is unclear.

Running a weather model over long time does not allow prediction of daily weather years ahead, but prediction of monthly averages may be possible. The basic reason is a certain cancellation property of turbulent flow, which can be expressed as "weather is variable" or "after rain comes sunshine".