Improved Quantitative Precipitation Forecasts by MHS Radiance Assimilation with a New Quality Control Algorithm

A detailed analysis of a degradation caused by AMSU-B/MHS data assimilation for a real case points to a potential problem of the MHS quality control algorithm associated with cloud detection in the National Centers for Environmental Prediction (NCEP) Gridpoint Statistical Interpolation (GSI) analysis system. The GSI cloud detection based on differences of two MHS window channels between observations and model simulations failed to identify cloudy points when a phase error is present in water vapor and cloud hydrometeor background fields. A new MHS cloud detection algorithm is developed based on a statistical relationship between MHS data and GOES imager channel at 10.7 . The modified MHS quality control leads to a marked improvement of AMSU-B/MHS data assimilation, leading to a positive impact on quantitative precipitation forecasts (QPFs) instead of degradation. The temporal evolution of 3-h accumulative rainfall distributions compared favorably with that of multi-sensor NCEP observations and GOES-12 imager observations. The precipitation threat scores are increased by more than 50% after 3-6 hours of model forecasts for thresholds exceeding 1 mm. This study highlights the importance of quality control for satellite data applications in NWP.