A forward operator for Global Positioning System (GPS) slant total delay (STD) data and its adjoint were implemented into the Mesoscale Model version 5 (MM5) 4DVAR system to investigate its impact on quantitative precipitation forecasting (QPF). An operational forecast system was set up providing two forecasts per day, one driven by ECMWF forecasts only and the other additionally by 4DVAR of GPS STD data.

The investigation of statistics for August 2007 demonstrated a positive impact on the representation of the water-vapour field and the diurnal cycle of precipitation in southwest Germany. The spread of observation-minus-model departures was strongly reduced during the first 6 h of the simulations. As compared to the control simulation, the averaged diurnal cycle of precipitation in the 4DVAR integration was closer to observations in spite of the limitations of the 4DVAR system. Especially promising is the almost complete removal of the spin-up at the beginning of the simulation.

This study not only demonstrates the potential of water vapour data assimilation for nowcasting and short-range QPF. It also suggests that improved GPS retrievals and extended networks are important to further improve the forecast performance and demonstrates the potential to apply observation operators for model verification.