A two-season impact study of four satellite data types and rawinsonde data in the NCEP Global Data Assimilation System

Extended-length observing system experiments (OSEs) during two seasons are used to quantify the contributions made to forecast quality by conventional rawinsonde data and four types of remotely sensed satellite data. The impact is measured by comparing the analysis and forecast results from an assimilation forecast system using all data types with those excluding a particular observing system. The impact of the particular observing system is assessed by comparing the forecast results over extended periods. For these observing system experiments, forecast results are compared through 168 h for periods covering more than a month during both the summer and winter seasons of each hemisphere. The assimilation - forecast system used for these experiments is the National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) and the Global Forecast System (GFS). The case studies chosen consist of periods during January-February 2003 and August-September 2003. During these periods, a T254L64 layer version of NCEP's global spectral model was used. The control run utilized all data types routinely assimilated in the GDAS. The experimental runs individually denied data from the Advanced Microwave Sounding Unit (AMSU), the High-Resolution Infrared Radiation Sounder (HIRS), geostationary satellite atmospheric motion vectors (GEO winds), in situ rawinsondes (raobs), and surface winds derived from the Quick Scatterometer (QuikSCAT). Differences between the control and denial experiment forecasts are accumulated over the two 45-day periods and are analyzed to demonstrate the impact of these data types. Anomaly correlations (ACs), forecast impacts (FIs), and hurricane track forecasts are evaluated for all experimental runs during both seasons. The anomaly correlations used the standard NCEP software suite and are partitioned into subsections covering the polar caps (60 degrees-90 degrees) and midlatitudes (20 degrees-80 degrees) of each hemisphere and the tropical region (20 degrees N-20 degrees S). Anomaly correlations of geopotential heights are shown at several pressure levels in the polar regions and midlatitudes. The root-mean-square error (RMSE) for 850- and 200-hPa wind vector differences are shown for the tropical region. The geographical distributions of forecast impacts on geopotential heights are also examined. The influence these data types have on tropical cyclone track forecasts are shown for both the Atlantic and Pacific basins and again are computed using standard algorithms developed and maintained at NCEP. The results demonstrate a positive impact from all data types with AMSU and rawinsonde data providing the largest anomaly correlation improvements in all zonal regions examined. Smaller forecast improvements are noticed from each of the other data types. In the Atlantic basin, each of the four satellite data types provides nearly equal improvement to the tropical cyclone track forecasts; however, GEO winds provide the largest improvement to track forecasts in the Pacific basin.