The past fifteen years has seen the development of a variety of new techniques to estimate the three-dimensional vector wind field from single-Doppler radar data (e.g., Tuttle and Foote 1990; Sun et al. 1991; Sun and Crook 1994; Qiu and Xu 1992; Shapiro et al 1995; Laroche and Zawadzki 1994, 1995; Xu et al 1994a,b, 1995; Zhang and Gal-Chen 1996, Gao et al. 2001, Lazarus et al. 2001, Weygandt et al. 2002a). These Single-Doppler Velocity Retrievals (SDVRs) can be used as stand-alone algorithms for hazard-warning or nowcasting, or as a means for numerical model initialization or data assimilation (e.g., Wilson et al. 1998; Weygandt et al. 2002b). Although the various SDVRs have successfully retrieved cross-beam winds in a number of simulated and real radar datasets, solution non-uniqueness and other difficulties are persistent problems in some circumstances. Our presentation describes the use of rapid-scan radar data to alleviate this problem for the case of a fast-moving thunderstorm outflow observed in northern Oklahoma on 20 June 2000. A series of retrieval experiments were performed with rapid-scan data (~ 1 min scan period) of the gust front gathered by two Doppler-on-Wheels (DOW) research radars. For each experiment, data from one radar were supplied to the retrieval, while data from both radars were used to construct a dual-Doppler wind analysis to verify the retrieved azimuthal wind component. The retrieval algorithm used in these experiments is a simple hybrid 4DVAR method that combines the Lagrangian framework of Laroche and Zawadzki (1994, 1995) with the method of Xu et al. (1995) in which a bulk source term is retrieved as part of the solution. In our latest version of the retrieval we make use of two constraints: (i) the radial component of the equation of motion and (ii) a background term obtained from a linear wind analysis similar to VVP. This latter constraint is only applied in regions where the cost-function associated with the linear wind analysis is less than a threshold value. Key experiments focussing on optimal time and space resolutions for the retrieval and the optimal length of the assimilation window will be reported.