Characterising flow dependent background errors in data assimilation
Sarah Dance
Department of Meteorology, University of Reading
Abstract:
Background error statistics are a key component of operational data
assimilation systems. Unfortunately, they are also a component that is not
well known. Thus, operational centres tend to use simple, tuned
climatological estimates. It is widely believed that employing more
appropriate, flow dependent error covariance matrices will lead to improved
analyses. In this context, we take a wide definition for flow dependence, to
include not only time and state dependence, but also climatological variations
in flow character that are dependent on physical space e.g. the balances at
the tropics are different to those in the midlatitudes. In this talk we will
describe the background error statistics used at the UK Met Office. We will
consider research into improved characterisation of the boundary layer, and
look at alternative ways of incorporating state dependent information into the
error covariances.