COLA Report 24
Predictability and Error Growth in a Coupled Ocean-Atmosphere
Model
J. Shukla and Ben P. Kirtman
March 1996
Abstract
Error growth rates in a dynamical coupled ocean-atmosphere prediction
system are calculated by examining forecasts made from a large number of
initial conditions based on consecutive months. The error is defined as
the difference between consecutive forecasts valid at the same time.
The error growth calculations have been made using the forecasts from
the original Zebiak-Cane model in which initial conditions for coupled
predictions were obtained by forcing the ocean model with observed wind
stress. The error growth calculations are repeated for revised forecasts
using the same model, but with the initial conditions obtained by
combining the wind stress from observations and the coupled model
simulation. The revised initialization scheme gives greater hindcast.
It is found that the growth rate of the initial errors for forecasts
from the revised initialization scheme is entirely due to reduction of
the intial error. These results are encouraging for more complex
prediction systems, where it is anticipated that reductions in initial
errors via data assimilation techniques will lead to improvements in
forecast skill.
Complete copies of this report are available from:
Center for Ocean-Land-Atmosphere Studies
4041 Powder Mill Road, Suite 302
Calverton, MD 20705-3106
(301) 595-7000
(301) 595-9793 Fax
last update: March 26, 1996
comments to: www@grads.iges.org