The impact of supplemental rawinsonde observations on numerical weather prediction investigated in an OSSE

While most rawinsondes are launched once or twice daily at 0000 and/or 1200 UTC ('on-hours'), supplemental rawinsondes are sometimes launched at 0600 and/or 1800 UTC ('off-hours') in an effort to improve the skill of weather forecasts, such as for Hurricanes Irene (2011) and Sandy (2012). In this study, the National Aeronautics and Space Administration Global Modeling and Assimilation Office (NASA/GMAO) observing system simulation experiment (OSSE) framework is used to investigate the impact of these supplemental rawinsondes on analysis quality and forecast skill. The GMAO OSSE employs the European Centre for Medium-Range Weather Forecasts (ECMWF) Nature Run at T511/91L resolution with the Gridoint Statistical Interpolation (GSI) data assimilation system and Global Earth Observing System Model, version 5 (GEOS-5) forecast model. A moist adjoint of the GEOS-5 shows that at off-hour cycle times, the global observing network is not able to fully compensate for the lack of rawinsonde observations in comparison to on-hour cycle times. When supplemental rawinsondes at off-hours are added to the global rawinsonde network, there is a small but statistically significant improvement in the Northern Hemisphere analysis quality at all cycle times. This improvement is retained into the medium-range forecast, with 1 to 3% improvement in root-mean-square error of wind and temperature at the 24 and 48 hour forecast times, and slight improvement in 5-day anomaly correlations. Over the Atlantic Ocean basin, downstream advection of information from North America results in statistically significant improvements of short-term forecast skill.>