Processing of Doppler radar radial winds for data assimilation

Atmospheric wind field is 3-dimensional and variable over time and space. Doppler radar measures one component of the wind vector, ie. the radar radial component. For complete description of the wind field simultaneous measurements should be made with three Doppler radars. Usually only one radar is available and the availability of the other two radars is compencated by simplifying assumptions about the structure of the wind field. Widely used methods to retrieve vertical profiles of horizontal wind are VAD (Lhermitte and Atlas, 1961) and VVP (Browning and Wexler, 1968) methods. Both are based on linear wind field assumption. An alternative retrieval approach is to use a numerical model of the atmospheric dynamics as a constraint for resolving the full 3-dimensional wind field from the radial wind measurements of a single Doppler radar. Variational data assimilation technique provides the framework for appropriately relating the model variables to the observed variables, like radar radial winds. The benefit of direct use of radial winds is that observations include non-linear information of the wind field unlike wind profiles generated with VAD or VVP algorithms where the non-linear features are smoothed away. The drawback is that in the NWP (numerical weather prediction) models the implied background error covariance structures filter out small scale observational information. The amount of Doppler radial wind measurements is very large and some preprocessing prior to data assimilation must be made. One possibility, as discussed in this paper, is to generate spatial averages of the raw data. This article is organized as follows. Section 2 describes the principle of radar wind measurements and how the raw data can be preprocessed before introducing to an NWP model. Section 3 discusses the azimuthal behaviour of bias in radar