Measurement errors need to be specified so that radar observations can be properly assimilated for numerical weather prediction. There are two related aspects to this problem: (1) errors in the original measurements within each radar pulse volume, and (2) representativeness of the radar data estimates used in the assimilation process. For radial velocities, the first error source depends on the strength of the return signal and the spread or width of the Doppler velocity spectrum. Spectral width in turn depends mainly on reflectivity and velocity gradients within and across the pulse volume and velocity turbulence within the pulse volume. Estimation of these errors is complicated by the fact that the components needed for reliable error estimation are themselves only measured and, therefore, have inherent uncertainties.

We will review some of the basic error sources and their specification. We will also present a simplified approach to the specification of the spatial distribution of Doppler velocity error that includes not only random measurement error, but also includes a measure of representativeness when velocity data are gridded in preparation for the assimilation process.