Phased-array radar (PAR) technology offers the flexibility of sampling the storm and clear-air regions with different update times. As such, the radial velocity from clear-air regions, typically with lower signal-to-noise ratio, can be measured more accurately. In this work, observing system simulation experiments (OSSEs) are conducted to explore the potential value of assimilating clear-air radial velocity observations to improve numerical prediction of supercell thunderstorms. Synthetic PAR observations of a splitting supercell are assimilated at different lifecycle stages using an ensemble Kalman filter. Results show that assimilating environmental clear-air radial velocity can reduce wind errors in the near-storm environment and within the precipitation region. Improvements in the forecast are seen at different stages, especially for the forecast after 30 min. After assimilating clear-air radial velocity observations, the probabilities of updraft helicity and precipitation within the corresponding swaths of the truth simulation increase up to 30–40%. Additional diagnostics suggest that the more accurate track forecast, stronger vertical motion, and better-maintained supercell can be attributed to the better analysis and prediction of the mean environmental winds and linear and nonlinear dynamic forces. Consequently, assimilating clear-air radial velocity produces accurate storm structure (rotating updrafts), updraft size and storm track, and improves the surface accumulated precipitation forecast. The performance of forecasts with a higher frequency of assimilating clear air radial velocity does not show systematic improvement. These results highlight the potential of assimilating clear-air radial velocity observations to improve numerical weather prediction (NWP) forecasts of supercell thunderstorms.