One of the things I admired most about Tony was his eagerness to use new types of observations to improve the initial conditions of numerical weather prediction (NWP) models. He liked the challenge of using non-conventional data such as measurements of chemical constituents of the atmosphere or scatterometer observations over the oceans. Tony did not distinguish between “research” and “operational” satellite data—if a research mission was going to last for only three years, Tony wanted to use the research data operationally during the short period when they were available, even if the observations were not to be continued beyond the lifetime of the research mission. Not only could the research observations improve the operational forecasts, their impact in the model provided valuable information to the researchers about the characteristics and quality of the data.

The concept of radio occultation (RO) observations was very new to the meteorological community when the first RO soundings of Earth's atmosphere were made in 1995 in the successful proof-of-concept GPS-MET experiment. The RO observations produced by GPS-MET were not available in real time, their daily numbers were too small, and the experiment was terminated after two years for lack of operational funds; so research groups in Europe and the U.S. were doing limited, non-real-time experiments with RO observations until the joint Taiwan-U.S. mission COSMIC was launched in April 2006. However, the quality and potential of RO soundings had been demonstrated by GPS-MET and confirmed with launches in 2000 of CHAMP and SAC-C. Therefore, as the COSMIC program, which was designed to produce 2500 soundings per day in near-real-time developed in 2001-2005, several operational NWP centers, including NCEP and ECMWF, began preparing for success---developing the mathematical techniques and algorithms that could assimilate COSMIC data in their operational forecasts.

In my presentation I will outline how RO observations are assimilated in NWP models and show some examples of their impact. RO observations, because of their unique properties (high vertical resolution, accuracy and precision, and ability to “see” through clouds) complement infrared and microwave sounding systems, and therefore make a significant and cost-effective positive impact on numerical weather forecasts.