Two versions of a two-dimensional approach to Doppler radar velocity de-aliasing were developed and tested on various datasets. One version used fully recursive segmentation of the data and the technique of spatial computer vision to de-alias isolated areas of velocity, whereas the other version utilized spatially constrained segmentation, objective analysis and a nonmeteorological knowledge base for de-aliasing homogenous velocity regions. The utility of the NEXRAD VAD algorithm-derived windfield information was evaluated. A recursive three-dimensional de-aliasing approach was also developed in an attempt to reduce the need for auxiliary windfield information.
These approaches were tested on two- (and three) dimensional data in both R, Θ, (Φ) and X, Y, (Z) coordinates and found to produce nearly identical results. Spatial filtering of the data proved to be a nearly preprocessing step and an elective filter was developed which preserves valid velocity information while reducing computational requirements. The importance of removing ground clutter was found to increase as the unambiguous velocity of the radar is reduced and a new scheme was developed and included in the de-aliasing process.