The Application of Radiosonde Observation Blacklisting Check to Variable Data Assimilation System

The radiosonde observations are very important for the numerical weather forecasting as they are more reliable data compared to other observations in the assimilation system. The radiosonde observation needs serious quality controlling before used in data assimilation, and the blacklisting check which uses monthly blacklist to delete the abnormal noise and bias data compared with background field is a significant step in observation quality monitoring. At present the radiosonde observations used in GRAPES (Global-Regional Assimilation and PrEdiction System) analysis system are scanned through preliminary checks such as climatological extreme check, consistency checks and background quality control in data assimilation system, but blacklisting check is not included. For the spurious high/low-pressure system structure of GRAPES analysis caused by the radiosonde observation in some region, the radiosonde observations are analyzed through comparing with NCEP analysis from June to August in 2007. Results show that the bias of radiosonde geopotential height observation and NCEP analysis is less than 50 gpm under 100 hPa. The bias is much bigger over 100 hPa and the radiosonde geopotential height is much smaller than NCEP analysis. Stations where the root mean square errors of the geopotential height between radiosonde observations and NCEP analysis for 500 hPa exceed 30 gpm are located in India, North Atlantic Ocean and Antarctic region. In those regions the unreliable observation ratios of some stations are above 20%, so they are listed in the blacklist. The geopotential height blacklist of radiosonde is provided by comparing observations with NCEP analysis and estimating the frequency of bad observation-quality station. The geopotential height blacklisting check of radiosonde is applied to GRAPES global 3D-var analysis system, and the cycle experiments of the analysis and forecast from 1 July to 21 July in 2009 are carried out. Two experiments are performed, Experiment 1 (control test) without blacklisting check and Experiment 2 with blacklisting check. The results indicate that the quality of geopotential height analysis of GRAPES is improved in India and Antarctic regions when considering geopotential height blacklist. The average root mean square errors of geopotential height GRAPES analysis and NCEP analysis for 500 hPa is reduced by 6 gpm in India Region and near 10 gpm in Antarctic Region after using blacklisting check in the 3D-var analysis system. In Antarctic region, the average biases and root mean square errors of geopotential height GRAPES analysis and NCEP analysis both decrease from 1000 hPa to high levels. The blacklisting check of radiosonde observation is a necessary step in the analysis system. The updated blacklist information will be used to assimilate the radiosonde observation more effectively.