Saroja Polavarapu
The recent launches of ENVISAT and AURA represent a significant increase
in the number of measurements of the middle atmosphere, especially of
chemical species. At the same time, weather forecast models are raising
their lids to include the mesosphere, while middle atmosphere climate
models are developing data assimilation systems to provide chemical
climate analyses and to more directly confront the models with
measurements. For all these purposes it is necessary to extend
assimilation schemes into the upper stratosphere and mesosphere.
However, this extension is not as simple as it might seem. The
mesosphere is a region of strong variability and short time scales, and
thus large forecast errors. As a result, assimilation schemes that rely
on forecast error statistics to transfer information in space and across
variables are very sensitive to the specification of these statistics.
In addition, assumptions normally made with operational systems such as
unbiased forecast and observation errors come into question. Thus, the
mere presence of a mesosphere in an assimilation model forces a
re-evaluation of the data assimilation problem. This talk will discuss
some of the challenges of middle atmosphere data assimilation,
illustrated by experiences with the Canadian Middle Atmosphere Model.