Observing System Experiments Using the AFGL Four-Dimensional Data Assimilation System

Abstract

Five observing system experiments have been performed with the Air Force Geophysics Laboratory (AFGL) global data assimilation system (GDAS). Each experiment includes one week of data assimilation during each of the two special observing periods of the First GARP Global Experiment (FGGE) of 1979, and three 4-day forecasts. The first experiment (STATSAT) provides a baseline for comparison. It uses all the available level II-b FGGE data, except surface observations. Comparisons with the National Meteorological Center (NMC) level III-A analyses show that the AFGL GDAS performs adequately. In the second experiment (NOSAT) all satellite data are removed from the dataset, showing the sensitivity of the AFGL GDAS to satellite data. As expected the impact on the analysis is small in the Northern Hemisphere, but large in the Southern Hemisphere. In the third experiment (GLASAT), the effect of a physical satellite retrieval method is tested by using the temperature profiles provided by NASA Goddard laboratory for Atmospheres (GLA). The differences between the GLASAT analyses and the STATSAT analyses are generally fairly small. Comparisons of forecasts made from the two analyses to radiosonde data give no clear indication whether physical or statistical retrievals lead to better forecasts for the AFGL system. The fourth experiment (NOCON) was designed to test the degradation of the analyses when all conventional data are denied and only satellite data are used. Differences between this and the STATSAT experiment appear first over continental areas, but propagate quickly over the oceans, especially in winter. The differences keep growing throughout the assimilation period. The forecasts from this experiment, especially the later forecasts, have little skill beyond 2 or 3 days. Finally a fifth experiment (NOCOR) examines the effect of denying conventional data only over a particular region (Europe and USSR). The effect is much smaller than in the NOCON experiment, even in the region where data were denied. Nevertheless the differences keep increasing during the assimilation period and some propagation of the differences across the Pacific and into North America is observed, especially in winter. The forecast errors, initially largest in the region of data denial, spread over the whole hemisphere in 3 or 4 days.