Inter-comparison of extra-tropical cyclone activity in nine reanalysis datasets

This study inter-compares extratropical cyclone activity in the following nine reanalysis datasets: the ERA-20C Reanalysis (ERA20C), the Twentieth Century Reanalysis, version 2c (20CR), the Japanese 55-year Reanalysis (JRA55), the Modern Era Retrospective-analysis for Research and Applications (MERRA), the NCEP Climate Forecast System Reanalysis (CFSR), the ERA-Interim Reanalysis (ERAint), the ERA40 Reanalysis, the NCEP–NCAR Reanalysis (NCEP1), and the NCEP-DOE Reanalysis (NCEP2). The inter-comparison is based on cyclones identified using an objective cyclone tracking algorithm. In general, reanalyses of higher horizontal resolutions show higher cyclone counts, with MERRA and 20CR showing the highest and lowest mean counts of all-cyclones, respectively. However, MERRA shows the highest mean intensity (i.e., geostrophic winds) of all-cyclones, and CFSR the lowest, although MERRA and CFSR share a similar horizontal resolution. MERRA is most different from the other datasets, showing many more cyclones of shallow-medium core pressures and much higher counts of cyclones of strong intensity than the others, while CFSR shows many more cyclones of moderate intensity than the others. MERRA cyclones tend to have weaker surface winds but stronger geostrophic winds than the corresponding CFSR cyclones. The track-to-track agreement between the datasets is better for moderate-deep cyclones than for shallow ones, better in the NH than in the SH, and better in winter than in summer in both hemispheres. There is more similarity in temporal trends and variability than in specific cyclone counts and intensity, and more similarity in deep-cyclone (core pressure ≤ 980 hPa) statistics than in all-cyclone statistics. In particular, all the four datasets that cover the period from 1958 to 2010 agree well in terms of trend direction and interannual variability in hemispheric counts of deep-cyclones, showing a general increase in both hemispheres over the past half century, although the magnitude of increase varies notably from dataset to dataset. The agreement in trends of deep-cyclone counts is generally better in winter than in summer, and better in the NH than in the SH, with nearly perfect agreement for the counts of NH winter deep-cyclones. However, the nine datasets do not agree well in terms of trend and interannual variability in the mean intensity of deep cyclones, especially in summer and in SH winter. The temporal homogeneity of cyclone statistics in each dataset was also analyzed. The results show that ERAint, NCEP2, MERRA, ERA40, and CFSR are homogeneous for the NH, and that ERAint and NCEP2 are also homogeneous for the SH. However, large inhomogeneities were found in the other datasets, especially in the earlier period. Most of the identified inhomogeneities coincide with changes in the quantity and/or types of assimilated observations. These inhomogeneities contribute notably to the differences between the datasets, which are larger in the earlier period than in the recent decades. Better trend agreements between these datasets are seen after the inhomogeneities are accounted for. It is critically important to identify and account for temporal inhomogeneities when using these datasets to analyze trends.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.