The phased-array radar (PAR) of the National Weather Radar Testbed (NWRT) in Norman, Oklahoma represents a paradigm shift for weather radar observations. It can adaptively scan multiple regions of interest and provide rapidly-updated weather observations. Through beam multiplexing, increased measurement accuracy can be achieved without increasing scan time. Alternatively, at the same measurement accuracy, more independent samples can be collected within a given time, allowing for, e.g., effective spatial over-sampling. Since the NWRT PAR radar has wider beams (on average about 2°) than the operational WSR-88D radar (about 1°), spatial resolution becomes low at far ranges. In this study, we examine the impact of spatial over-sampling on the analysis of thunderstorms, when simulated radar observations are assimilated using the ensemble Kalman filter (EnKF) method. The truth simulation and data assimilation are carried out at up to 500 m horizontal resolution and the ARPS EnKF system is upgraded to allow for the assimilation of radar data on the radials. Different over-sampling rates, defined as the ratio between beam width and the sampling increment in azimuth or elevation, will be examined, for storms at different distances from the radar, and the impact and optimal scanning configuration will be determined based on the experiment results.