Notes on characteristics of radar radial winds
Kirsti SALONEN
Finnish Meteorological Institute, Finland
Abstract
Doppler radar measures one component of the wind vector, that is, the radar radial component. With constant range and elevation the radar radial wind has a form of a sine as a function of azimuth angle, if an assumption of a linear wind field is made. The amplitude of the sine defines the wind speed and and the phase of the sine defines the wind direction.
Variational data assimilation provides a framework to interpret and to use the radar radial wind data in Numerical Weather Prediction (NWP) models. This consists of modelling the radar measurement with the available NWP model variables. The amount of raw measurements is however enormous, and some preprocessing is needed before using the radial wind observations in data assimilation. One solution is to calculate spatial averages, so called superobservations (SOs), from the raw data. SOs are smoother than the raw observations and better represent the model resolution. Another, but not as good possibility is to use data thinning, i.e. to use raw observations with sparse resolution.
An important aspect when defining the quality of radar radial wind observations, raw observations or SOs, is the possible azimuth dependence of verification scores, especially the azimuth dependece of bias. Calculating bias over the whole measurement area may result near zero and indicate that the data is unbiased. Still, there might be systematic difference between observed and modelled wind direction or wind speed. Studying the bias as a function of azimuth angle reveals those differences immediately. If observations and model are unbiased, bias as a function of azimuth angle is zero. If observations, or model, are not unbiased, the bias will also have a form of sine as a function of azimuth angle. The amplitude of the sine defines the maximum bias in the data and the phase difference to the reference wind direction defines the bias in wind direction.
In this presentation verification results will be presented from a set of experiments with one month radar radial wind data.