Bruce Ingleby, ECMWF

Forecast and analysis sensitivity to in situ observations

Abstract:

The sensitivity of forecast and analysis fields to surface, aircraft and radiosonde observations is examined to look at the effects of observation density and frequency, specified observation errors and the location within the time window (12 hour ECMWF 4D-Var). It is already known that forecasts are more sensitive to observations near the end of the time window - because they are more able to constrain growing error modes. Recent work indicates that analysis fields are more sensitive to observations near the start of the window - because they project more onto decaying modes. In an NWP context the quality of the forecast fields is more relevant than that of the analysis fields (the two are not independent, but it is quite possible to make the analysis 'draw closer' to observations without improving forecast quality). Some of the results (and related OSEs) are for upper tropospheric radiosonde humidity and new sources of aircraft humidity.

File available here