A New Method to Evaluate Reanalyses Using Synoptic Patterns: An Example Application in the Ross Sea/Ross Ice Shelf Region

• Published in: Earth and Space Science
• Main Authors: Cairns LH, McDonald, Adrian
• Format: Article in Journal/Newspaper
• Language: English
• Published: American Geophysical Union (AGU) 2021
• Subjects: Fields of Research::37 - Earth sciences::3701 - Atmospheric sciences::370108 - Meteorology; Fields of Research::37 - Earth sciences::3701 - Atmospheric sciences::370105 - Atmospheric dynamics; Ross Ice Shelf; Ross Sea; Antarc*; Antarctica; Ice Shelf
• Online Access: https://hdl.handle.net/10092/102367; https://doi.org/10.1029/2019EA000794

We compare the consistency between eight reanalyses: CERA20C, ERA5, ERA-Interim, ERA20C, NCEP-DOE, MERRA2, JRA55, and 20CRV2c. This comparison uses daily surface winds near Antarctica to classify synoptic patterns using the self-organizing map technique. The relative frequency of occurrence (RFO) of these patterns are very similar during the satellite era in each reanalysis. The three most common patterns are the same in each reanalysis and changes between the reanalyses only display a 12% relative variation. Examination of the RFOs over time highlights that the CERA20C and 20CRV2c reanalyses display large changes previous to 1957. These changes are likely connected to model relaxation toward their climatology because of a lack of observational constraints. Primarily, we introduce the entropy coefficient (U) which quantifies the consistency between reanalyses in their representation of synoptic patterns. Examination of U shows current reanalyses (ERA5, ERA-Interim, JRA55, and MERRA2) are highly consistent in the satellite era likely due to good observational coverage. However, centennial reanalyses (CERA20C, ERA20C, and 20CRV2c) show two upward step changes in consistency as measured by U at around 1957 and to a lesser extent 1979. Low values of U before 1957 suggest that centennial reanalyses are of limited use before this date, but may be useful after 1957 in this region. We also show that the entropy coefficient displays an inverse relationship with ensemble spread metrics of individual reanalyses. We conclude that the entropy coefficient provides a powerful quantification of the influence of changes in observation density on reanalysis quality in data sparse regions.