Forecasts of moist convective storms represent an important frontier for numerical weather prediction. A principal difficulty is to produce initial conditions through the assimilation of reflectivity and radial velocity observations from Doppler radars.

Recent results from assimilation experiments with an ensemble Kalman filter and both simulated and real radar observations will be presented. These experiments include assimilation of simulated observations from an idealized three-dimensional squall line, estimation of the environmental winds and sounding given radial velocity observations, assimilation of observations of zero reflectivity, and tests with real observations for the Arcadia supercell. Together, these experiments demonstrate the potential of the ensemble Kalman filter for convective-scale assimilation.