The Study On Variational Quality Control Of Non-Gaussian Distribution Observation Error Data For GRAPES-3DVAR

Quality control of observations directly affects the analysis quality of numerical prediction data assimilation. After discussing the observation error of non-Gaussian error distribution models. Based on the "Gaussian plus flat" distribution model of observation error, this paper developed the variational quality control scheme for the 3D-Var of assimilation and forecast system in regional GRAPES, furthermore analyzed and verified its applicability and effectiveness. Meanwhile, the heavy rainfall in southern China was choosed as a case for assimilating and forecasting analysis by using the Global Forecast System (GFS) data as the background field, and the conventional observation data including TEMP, SYNOP, SHIPS, AIREP, SATOB and COSMIC satellite retrievals data, so also do batch tests of a total of 31 days on August 2013. The results showed that variational quality control reasonably adjust observation weight according to different quality of observation, and significantly improve the analysis increment field and analysis field of geopotential height, pressure, wind, specific humidity, especially it has a more positive effects on heavy rainfall areas. The forecast quality are farther ameliorated for the precipitation area, precipitation intensity and center position of precipitation, the ability to forecast of the heavy rainfall, the big rainstorms and other precipitation larger level reflect the better effect particularly. And it also had certain effect on the typhoon track. Therefore, variational quality control plays an important role in assimilating and forecasting analysis of the mesoscale and microscale severe weather processes.At the same time, the initial startup, key parameters of variational quality control scheme were discussed, and discussed the parameterization scheme in detailed. Four groups of tests were designed for analyzing and comparing. The experimental results show that, compared to the other scheme, the test of parameterization configuration of the third scheme processing and adjusting more reasonable for observation weight, analysis incremental. The analysis field, the precipitation field of the precipitation are improved more significantly. The correctness that according to the different elements of the different observation types using different parameters configuration was verified, too.