Yannick Tremolet, ECMWF
Data Assimilation with Overlapping Windows
Abstract: Several studies in recent years have shown that there are theoretical benefits to the use of a weak-constraint formulation and a long assimilation window in 4D-Var. A central aspect of the long window weak-constraint formulation is that subsequent assimilation widows overlap. We will show that this overlapping window also works for strong constraint 4D-Var with 12 or 24 hour windows, with examples from surface pressure only re-analyses and operational NWP configurations. Using overlapping windows is one way to resolve a contradiction in operational NWP, which on one hand requires the latest arriving observations essential for the quality of the forecast to be included in the analysis---implying a short data cut-off---and on the other hand requires later arriving observations essential for the quality of the background and subsequent forecasts---implying a long data cut-off. A potential issue with overlapping windows is the correlation of background and observation errors if observations are re-used in subsequent analysis, and we will discuss strategies to avoid this correlation and compare performance to see which gives best result in practice. Finally we will discuss further possible developments, including quasi-continuous overlapping windows.