Interpretation of Global Forecast Model "Flipflops"
Two (00 UTC 8 and 9 April 2002) ensemble forecasts separated
by 24 hours, valid 12 UTC 16 April 2002, for 5700-m(?) 500-hPa height
contour. The operational MRF T170 forecast is in bold yellow. Graphic
courtesy of Stephen Jascourt, UCAR/COMET
Abstract
All forecasters are familiar with the radical changes from run to run
that seem to occur with medium-range (and sometimes even short-range) forecasts
in the Global Forecast Model. Do such 'flipflops' mean that there has been
a fundamental change in the model forecasts?
In these cases, we will describe two recent model "flipflops" in a pair of
time-adjacent MRF runs, and show how ensemble forecasts shed light on what is actually
occurring in the MRF.
Learning objectives
In this case study, the main lessons are:
- Model flipflops usually result from forecast uncertainty, rather than
radical changes in the initial data, because the data assimilation system
gives high weight to the first guess (i.e. the short-term forecast at analysis time).
- As a result, individual operational model forecasts often will look like one
of the ensemble members from an older forecast valid at the same time, and almost
always fall within the range of forecasts given by the prior ensemble run.
- The ensemble mean and spread is a better measure of fundamental run-
to-run model forecast change than consecutive operational forecasts.
Credits
by Dr. Bill Bua
Thanks to:
Bernard Meisner, SOO/SR (for bringing attention to the April cases)
Stephen Jascourt (for images, discussions, and review)
Steve Tracton (for discussions and review)
Pete Manousos (for discussions and review)