The variational assimilation of measurements from a vertically pointing, ground-based 12-channel microwave profiling radiometer is tested with the MM5 3dvar and 4dvar data assimilation systems. The Radiometrics TP/WVP-3000 radiometer measures brightness temperature in 7 microwave frequencies from 51 to 59 GHz on the flank of the 60 GHz oxygen resonance line and 5 frequencies from 22 to 30 GHz associated with the water vapor resonance peak at 22.2 GHz. Retrieved profiles of temperature, water vapor, and cloud liquid water are obtained through neural net inversions of the brightness temperatures, where the neural net is trained using radiosonde soundings and corresponding forward modeled brightness temperatures. An observation operator has also been developed for the direct assimilation of the measured brightness temperatures, using the forward model and its adjoint.

The impact of assimilation of the retrieved profiles, and direct assimilation of the brightness temperatures, on analyses and subsequent forecasts have been tested using the MM5 mesoscale model and its associated 3dvar and 4dvar assimilation system. Results will be presented from case studies using profilers located at the ARM CART site in Oklahoma, and the NCAR Foothills campus in Boulder, Colorado.