Towards the assimilation of precipitation data from ground-based radars in ECMWF's 4D-Var system

The feasibility and potential benefits of the variational assimilation of ground-based radar rain rates in ECMWF's global 4D-Var system have recently been investigated.

As a first attempt, the two-step '1D+4D-Var' approach initially chosen by ECMWF to assimilate satellite cloudy/rainy microwave radiances operationally has been applied to assimilate hourly rain rates from the NCEP Stage IV combined radar-gauge analyses. In this method, rain data are first assimilated through a 1D-Var procedure to create pseudo-observations of total column water vapour, which are then assimilated in 4D-Var. More recently, the direct assimilation of NCEP Stage IV data in 4D-Var has also been tested.

Results from experiments using both methods will be presented and the pros and cons of '1D+4D-Var' versus direct 4D-Var will be discussed. The main issues, requirements and limitations of the variational approach will be summarized. These include uncertainties in observation and model error statistics, the effect of non-linearities, the screening of observations, the proper choice of observable (logarithmic transform) and the issue of non-rainy model background situations.

In addition, results from denial experiments in which radar observations were assimilated in the absence of all other competing moisture observations (from surface stations, radio-soundings or satellites) over the USA will demonstrate that radar observations alone can have a positive impact on the quality of 4D-Var analyses and subsequent forecasts.

Finally, the prospects and requirements for the operational assimilation of radar data at ECMWF will be discussed.