Effects Of ATMS And CrIS Data Assimilation On Weather Forecasts Over Tibetan Plateau

The effects of ATMS and CrIS data assimilation on weather forecasts over Tibetan Plateau was investigated by using NOAA's Gridpoint Statistical Interpolation(GSI) data assimilation system and NCAR's Advanced Research Weather Research and Forecasting (ARW-WRF) regional model. The experiment was designed as 4 parts: (1) a control experiment (CTRL) and three data assimilation (DA) experiments with different data sets, including (2) conventional data only (CONV), (3) a combination of conventional and ATMS satellite data (ATMS), and (4) a combination of conventional and CrIS satellite data. The forecast abilities of the temperature (T)and the relative humidity (RH) at 2-meter height, and the wind speed (WS) at 10-meter height above the earth surface in January and July 2015 of DA were evaluated for the model. Those variables in different vertical layers over the terrain at different elevations were analyzed to improve the forecast as well. The results showed that the improvements of the three DA experiments were not the same. Both F24H(first 24-hour) and L24H (last 24-hour) forecasts of 10-m winds (WS) in January and the 2-m RH forecasts in July could be modified by assimilating ATMS over higher-elevation region, while 2-m T prognosis could be rectified over lower-elevation region. In CRIS case, a good performance is showed over higher-elevation region for F24H 2-m T prediction in July. Meanwhile, CRIS can also improve the prediction accuracy of 10-m WS over higher-elevation region in both January and July, to a certain extent. Considering the vertical stratification, the CRIS DA gave a negative contribution in all vertical layers while ATMS DA had different forecast accuracy in different vertical layers for different variables. The forecast error in T was typically caused by the systematic error which was controlled largely by the physical representation within the model. In contrast, the inaccuracies in the RH and in WS forecasts are dominated more by nonsystematic errors, derived from the random inadequacies of the initial conditions. The overall improvement in experiment with the ATMS DA was better than that in case of CRIS DA. After the assimilation, the wind field forecasts were much improved as compared to the forecasts of the temperature and the humidity field.Besides, precipitation prediction over the Tibetan Plateau in July 2015 was also evaluated. The results showed that the location of monthly mean of precipitation belt is shifted northward in the simulations and shows an orographic bias described as an overestimation in the upwind of the mountains and an underestimation in the south of the rainbelt. The rain shadow mainly influenced prediction of the quantity of precipitation although the main rainfall pattern was well simulated. For the F24H and L24H accumulated daily precipitation, in the model outputs, the amount of precipitation was generally overestimated, but it was underestimated during the heavy rainfall periods of 3-5, 13-16, and 22-25 July. The observed water vapor transport from the southeastern Tibetan Plateau was stronger than in the model simulations,which induced inaccuracies in the forecast of heavy rain on 3-5 July. The data assimilation experiments, particularly the ATMS assimilation, were closer to the observations for the heavy rainfall process than the control. Overall, the satellite data assimilation can enhance the WRF-ARW model's ability to predict the spatial and temporal pattern of precipitation in July 2015 although the model capability exists a significant limitation in the complex terrain area.