Abstract: The conventional observing system serves as an indispensable constraint to the atmospheric reanalysis. Surface observations were available from ships, drifting buoys, land stations, and airports. In situ measurements of upper-air were available from radiosonde, pilot balloon, aircraft, and wind pro file. Data collection and merging of different data source, quality control, and data selection are always important for global atmospheric reanalysis. This article describes the input conventional datasets used to produce the global atmospheric reanalysis project of CMA (CRA-40, 1979-2018), and discusses the processing method of conventional data before being assimilated. During the development stages of CRA-40, conventional data from 15 data sources are integrated, quality controlled and evaluated. Compared with NCEP Climate Forecast System Reanalysis (CFSR), the CRA-40 assimilates more Chinese surface and radiosonde data. The speci fic humidity of TAMDAR and ACARS reports are assimilated. Observations that are expected to have a negative impact on the reanalysis are rejected. These observations include surface pressure observations over high terrain, radiosonde observations below the model surface, humidity above 300 hPa, and other data of blacklists. Blacklists provide the identi fiers of station, aircraft, ship and buoy with systematical quality problem in different periods. Blacklists information of CRA-40 come from the statistic results of the departure between observation and ERAInterim reanalysis, and the historical blacklists provided by CFSR and ECMWF. Blacklists of CFSR and ECMWF are accessed before being using in CRA-40. Parts of research results are of guiding signi ficance for improving the historical and real-time data processing and application. Quality controlled historical conventional observations are assimilated in CRA -40. Some of conventional observations are serviced by China Integrated Meteorological Information Service System (CIMISS).