4-D Var in high-resolution versions of the UM

Mark Dixon, Ross Bannister

4D-VAR has produced very successful results in large-scale forecast systems, and so it is natural to ask whether similar benefits can be obtained in a high-resolution forecasting context. However, performing 4DVAR at high resolution poses a number of additional challenges which are not present at large-scale. At small scales, trajectories can be highly non-linear, owing both to the dynamics and micro-physical processes, as well as to the strong interaction between them. This project will investigate ways of coping with such non-linearity in the 4D-VAR system. Another salient research topic is related to the specification of background error covariances. Traditionally, covariances have been constructed using approximations based on large-scale balance relationships. At small-scales these forms of balance break down and it is essential that they are replaced with alternatives.

You are visitor number [an error occurred while processing this directive] to this page.


Page navigation