Climatology and temporal evolution of the atmospheric semidiurnal tide in present‐day reanalyses

The solar semidiurnal atmospheric tide (S 2 ) was extracted from seven reanalysis data sets, including current data sets, such as CFSR (Climate Forecast System Reanalysis), MERRA (Modern‐Era Retrospective Analysis for Research and Applications), ERA‐Interim (ECMWF Reanalysis), and 20CR (Twentieth Century Reanalysis), and older frozen products, such as NCEP/NCAR (National Center for Environmental Prediction/National Center for Atmospheric Research), ERA‐40 (ECMWF Reanalysis), and JRA‐25 (Japanese 25 year Reanalysis). In this calculation, we emphasized the temporal variation of the tide. We also calculated the tidal error, which was sizable at high latitudes and over short averaging periods and large for 20CR at all latitudes. Because of the four standard daily samples, the interpolation scheme of van den Dool et al. (1997) was used when necessary. We found this method to be accurate for zonally averaged tides only. Comparing the climatology from the MERRA and CFSR S 2 with a recent empirical tide model showed that MERRA better represented the geographical structure of the tide, especially its phase. We found a bias in the phase in all of the reanalysis data sets except for MERRA. The temporal evolution of the tide was inconsistent between the different data sets, although similar seasonal variations were observed. The seasonal cycle was also better depicted in MERRA. The S 2 calculated from MERRA and satellite precipitation measurements from TRMM (Tropical Rainfall Measuring Mission) presented results that were inconsistent with the hypothesis in which rainfall latent heat release represents S 2 forcing and functions as a source of S 2 seasonal variability.