The problem
explored in this thesis is the response of an idealized GCM to El
Niño-like and how to causally link the model's response to
its forcing. Once the forcing is determined, the model can then be
corrected toward some designed goal, e.g. an improved simulation of El
Niño. Since this approach implicitly includes the effects of
transients, it provides a better understanding of how transients affect
the overall climate response to SST anomalies and their importance in
better overall climate simulations.
The first experiment
places a Gaussian cooling anomaly in the central Pacific centered at
180W and the Equator. The anomaly itself extends from 140E to 140W and
20S to 20N. The GCM is integrated forward in time for 1000 days with
this cooling anomaly to generate a climate anomaly. Then, the adjoint
is integrated with this anomaly introduced at each time step and run
backward in time to determine the sensitivity to that particular
climate pattern. The sensitivity generated by the adjoint is centered
roughly on the date line and has an overall character quite similar to
the imposed Gaussian cooling anomaly.
In the second
experiment, the GCM was integrated forward in time for 1000 days to
determine the errors associated with the model as measured by
departures from the DJF climate from the NCEP reanalysis. The model was
then run backwards using the adjoint to determine the forcing that
would cancel this climate anomaly. The model can then be adjusted to
account for the error and run once again to examine if an improvement
in the model climate is obtained. The thermal forcing increments
associated with this iteration process seek to improve the climate by
adjusting the tropical/extratropical heating contrasts.
The results obtained
appear to provide a useful way to correct climate models, and future
research will focus on implementing this approach in a full physics GCM.
The University of Wisconsin-Milwaukee, 1999
Under the Supervision of Dr. Kyle Swanson