In order to speed up production, ERA5 is being produced by several parallel experiments, each covering a different period, which are then appended together to create the final product. The disadvantage of this approach, is that there can be discontinuities in the final product at the transition points between the different experiments. Here, we consider the transition at the end of 2009 and the beginning of 2010. Each experiment begins with a warm up of a year, typically, which does not become part of the final product, but does overlap with the experiment for the preceding period. This overlap facilitates a comparison of the two experiments to see how well they have converged to a common solution.

Fig. 1 shows the ERA5 zonal mean temperature (contours) and difference between the final product and the warm up (colours), for three months of the overlap, October to December 2009. In the troposphere and lower stratosphere, between 1000 hPa and 10 hPa, the differences are small (less than 0.2 K). From the mid-stratosphere to the stratopause (from 10 hPa to 1 hPa), the differences are larger, but generally below 2 K, except near the equator at about 1 hPa where the final product is more than 2K warmer than the warm up. In the mesosphere (from 1 hPa to 0.01 hPa), the final product is generally more than 2 K colder than the warm up near the equator, and its more than 5 K colder at about 0.05 hPa. In addition, this colder region spreads into the northern hemisphere near 0.5 hPa, and the southern hemisphere above 0.05 hPa. Differences of more than 2 K, but of different signs, are also apparent in both polar regions above about 0.3 hPa.

A similar, but not identical, comparison for the transition from 1988 to 1989 in ERA-Interim (Fig. 2) reveals that in the troposphere and stratosphere (from 1000 hPa to about 1 hPa), the differences have, if anything, a slightly larger magnitude than those in the 2009/2010 ERA5 transition, and ERA-Interim exhibits transition differences lower down in the atmosphere. However, in the lower to mid-mesosphere (from about 1 hPa to 0.1 hPa, with the latter being the top of the ERA-Interim domain), the differences in the ERA-Interim transition are generally smaller than in the ERA5 transition, having magnitudes less than 2 K. The spread in temperature, for September - November 2009, in the ERA5 EDA 10 member ensemble (Fig. 3), which gives some indication of the uncertainty, is generally larger than the ERA5 transition differences in the troposphere and stratosphere, particularly at lower latitudes. The largest values of the spread are located in the mid to upper stratosphere, at lower latitudes, where values exceed 0.8 K. This region merges into a second area of relatively large uncertainty, in the northern mid-latitudes of the mid-mesosphere, where values exceed 0.7 K. In general, the spread in the mesosphere is smaller than the ERA5 transition differences.

Fig. 4 shows the ERA5 zonal mean eastward wind (contours) and difference between the final product and the warm up (colours). In the troposphere and lower stratosphere, between 1000 hPa and 10 hPa, the only region with transition differences greater than 0.2 ms^-1 is a latitudinally confined region near the equator, above about 50 hPa, where the differences of alternating sign are no larger than about 1 ms^-1. From the mid-stratosphere to the stratopause (from 10 hPa to 1 hPa), the equatorial transition differences are slightly larger, with a larger latitudinal extent. However, in the mesosphere (from 1 hPa to 0.01 hPa), the differences are much larger, with peak values greater than 50 ms^-1 near the equator at about 0.1 hPa and values greater than 2 ms^-1 are widespread. These equatorial mesospheric differences in eastward wind, and possibly temperature too, are related to the spurious jet that this cycle of the IFS (CY41R2) suffers from, particularly in the transition seasons.

The equatorial transition differences in eastward wind in ERA-Interim in the lower stratosphere (Fig. 5) are slightly smaller than those in ERA5, though above in the mid to upper stratosphere, the differences are somewhat larger than those in ERA5. In the mesosphere, the transition differences in ERA-Interim are much smaller than those in ERA5. The spread in the eastward wind in the ERA5 EDA 10 member ensemble (Fig. 6) is generally larger than the ERA5 transition differences in the troposphere and stratosphere, except near the equator. In the mesosphere, the spread is smaller than the ERA5 transition differences.

The ERA5 transition differences for specific humidity (Fig. 7) are generally below 5 %, apart from at low latitudes in the mesosphere above 0.05 hPa, where differences exceed 10 %. The transition differences for ozone (Fig. 8) are generally below 5 % in the troposphere and stratosphere, except for a few small regions, the most marked being the upper troposphere/lower stratosphere (UTLS) of the tropics and southern hemisphere. However, the ozone differences are large in the mesosphere in the polar regions, where differences exceed 20 % or even 50 %.

In summary, the ERA5 transition differences for 2009/2010 in the troposphere and stratosphere are generally smaller than transition differences in ERA-Interim and are generally smaller than the ensemble spread. However, in the mesosphere the transition differences are much larger, exceed those in ERA-Interim and are larger than the ensemble spread. The large differences in the mesosphere are particularly apparent in the polar and equatorial regions. For the latter region, the differences in eastward wind, and possibly temperature too, are related to the spurious equatorial mesospheric jet that this cycle of the IFS (CY41R2) suffers from, particularly in the transition seasons. For more information on this spurious equatorial mesospheric jet, see Section 7 in the ECMWF Technical Memorandum No. 824 Report on Stratosphere Task, Force by T.G. Shepherd, I. Polichtchouk, R.J. Hogan and A.J. Simmons.

Related articles