Based on the Brewer ozone spectrophotometer long-term (1993-2023) observations at Zhongshan Station, Antarctica, atmospheric total ozone column (TOC) of the Merra2 and ERA5 reanalysis are compared, evaluated and their trends are analyzed The results show that the reanalysis are generally in a good agreement with the ground-based data in the context of occurrence of the ‘ozone hole’ and the TOC seasonality The TOC bias (∆TOC (DU)) and relative difference (∆TOC(%)) on the daily mean scale are -2 0±9 6(1σ)DU and -0 6±4 3%(1σ) for Merra2 and -0 6±4 3%(1σ) for ERA5 respectively Both the probability distribution of ∆TOC(%)s exhibit each normal random processes and their large variations occurred at the end of March and during the ‘ozone hole’ period The reanalysis data were divided into two periods, 1993-2004 and 2005-2023, based on the changes of satellites to which the reanalysis data were assimilated, but the ∆TOC(%) values (including ERA5) during the ‘ozone hole’ increases with decreasing of the TOC in both periods, and the ∆TOC(%) values for Merra2/ERA5 were respectively 6 9%±4 6%(1σ)/4 6%±2 0%(1σ) and -0 4%~2 3%(1σ)/ 6 4%±3 1%(1σ) Whereas the corresponding averages for the non-ozone hole periods were respectively of only 0 3% ± 1 5% (1σ)/0 6% ± 1 4% (1σ) The ∆TOC (%) of Merra2 and ERA5 show an increasing trend with the solar zenith angle (SZA) during 1993 -2004, with each magnitude of 3% and 2%, while the opposite trend is observed from 2005 to 2023, with magnitude of -2% and 2% respectively for Merra2 and ERA5 Merra2 (ERA5) is systematically lower (higher) than the observed TOC after 2005 (2012), with ∆TOC (%) as low(high) as more than 6% Both the ERA5 and Brewer data show a clear TOC recovery trend during the last 30 years of the ‘ozone hole’ periods whereas neither of them is characterized by a clear trend during the last 31 years of the non-ozone hole periods Correspondingly, the data of Merra2 exhibits clear TOC deletion trends Both the reanalysis TOC data validated by Brewer's observation show their consistent recovery trends of TOC, and the recovering rate of ERA5 is 1 3 DU/10a The study suggests that raw reanalysis TOC data should be used with much caution before evaluating the long-term trends of the ozone layer, and the data from the ground-based observations, albeit the number is much lower than that of reanalysis outputs due to seasonal SZA or weather conditions, is critical for the reanalysis TOC validation and conclusions of TOC trend