Assimilation Application of AMDAR Data to the Operational NWP System of Beijing

AMDAR data has been widely proved valuable in improving accuracy of numerical weather forecast. But assimilation application of AMDAR data in operational NWP model is seldom accessed in China. BJ RUC, the operational rapid update and cycling NWP system of Beijing Meteorological Bureau, is used to evaluate the impacts of assimilating AMDAR data on forecast performance. Durative experiments with and without AMDAR data are conducted from 15 Jul to 15 Aug, 2007. The average RMSE difference and average precipitation scores difference of AMDAR and NOAMDAR experiments are compared to discuss the impacts of AMDAR on short range numerical weather forecast.Analysis shows that the assimilation of aircraft data has remarkable positive impacts on the first 9 hours forecast especially on upper levels. Forecasts of wind on the levels of 925 hPa to 250 hPa and temperature on the levels of 850 hPa to 400 hPa are improved. The assimilation of AMDAR data contributes more in rapid updated cycle mode than in cold start mode. It has positive impacts on the threshold score of accumulated precipitation forecast of 0—12 hours, 0—24 hours, 12—18 hours, and 18—24 hours. It also improves the overestimating in the NOAMDAR experiments. The increment analysis shows parallel characteristics with the horizontal distribution of AMDAR data and the vertical improvement characteristic of the 3 hours forecast. The comparability indicates the improvement of wind and temperature in first guess field made by assimilating AMDAR data. The relative observation errors of AMDAR against Beijing radiosonde are discussed and proved reliable. It suggests that determining AMDAR observation errors properly in assimilation can help make better use of AMDAR data.Experiments on AMDAR data assimilating application in the operational BJ RUC system show that AMDAR assimilating has important positive impacts on improving the upper temperature, wind and precipitation forecast. Reasonable amendment of observation errors will bring more efficient utilization of AMDAR data, so priority is assigned to adjusting the AMDAR observation errors in BJ RUC.