An evaluation of the Southern Annular Mode in the Twentieth Century Reanalysis

We explore whether improvements to the Twentieth Century Reanalysis (20CR) in the most recent version (v3) have improved representation of the Southern Annular Mode (SAM).  The negative SAM index bias in the first half of the 20th century compared to instrumental SAM reconstructions in previous 20CR versions (due to a systematic high latitude high pressure bias) is still present. It is reduced in summer and autumn, but not in winter and spring.  Correlations between reanalysis and reconstructed SAM indices through the full series in all seasons do show improvements in v3 compared to previous versions.   A reduction in SAM index ensemble spread is evident during periods with higher numbers of assimilated observations, in particular in summer and autumn.  Analysis of the spatial distribution of assimilated observations shows clear improvement in years/periods with greater numbers of ships observations in the Southern Ocean and Antarctic observations (e.g. early 20th century Antarctic expeditions).  However it is not until the advent of greater numbers of ships observations in the Southern Ocean and regular data from Antarctic meteorological stations in the late 1940s that there are enough high latitude observations to realistically constrain the reanalysis.   Enhancements have been enough to improve how the reanalysis follows the observations temporally, highlighting the benefit of data rescue, but due to the bias, we recommend that the 20CR should still not be used for the analysis of long-term SAM trends, and caution should be exerted when using SLP data from the high latitude Southern Hemisphere from all 20CR versions prior to 1957.