Noise filtering in a variational wind field retrieval from Doppler radar

Two different concepts for noise filtering in a DOPPLER radar wind field retrieval are compared. The first technique is a smoothing on the raw radar data. The second method uses a smoothness constraint within a variational technique in which a model wind field is adjusted to radar data by minimizing a cost function. The wind field retrieval algorithm is described by PROTAT and ZAWADZKI(1999) and is further extended here by adding a topography constraint for its application in mountainous terrain. The experiments are carried out with analytic wind fields artificially affected by noise. In a first step only the horizontal wind is retrieved. In a second step the continuity equation serves to calculate the three-dimensional wind field. Power spectra and an error analysis of the retrieved wind fields as well as the performance of the algorithm lead to the following results. An efficient noise filtering by a mere data smoothing is accompanied by a lack of accuracy in the retrieved wind components. The smoothness constraint eliminates the noise more efficiently than smoothing of raw data while producing an accurate wind field. For the retrieval of the vertical velocity wind field, the smoothness constraint is indispensable.