Recently it was realized that in the past ERA5 has potentially assimilated incorrect snow observations. This has occurred because of problems in the quality control of these observations. Although this could have had local negative effects on the ERA5 snow product and other related surface parameters, a scan through the ERA5 archive did not clearly show many affected cases prior to 2021. However, a deeper investigation needs to be performed and will be reported in this page.
However, in September and October 2021 the problem in ERA5T (the preliminary version of ERA5 made available about 5 days after real time) became noticeable in central Asia, due to the assimilation of incorrect snow depth observations for some stations in this region. This resulted in erroneously large values of snow depth and soil moisture, and to a lesser extent, erroneous values of 2m temperature and 2m dewpoint temperature. The stations which were identified have been omitted from assimilation in ERA5T from 15 November 2021 onward.
It was decided that this issue needed correction prior to the final, quality-checked, release of ERA5. As a result, for the months September, October, November and at least a part of December 2021, the final ERA5 product is (and will be) different from ERA5T. Although the differences are mostly limited to central Asia and mainly to surface parameters (in particular snow depth, soil moisture and - to a lesser extent - 2m temperature and 2m dewpoint temperature), all the resulting reanalysis fields do differ slightly over the whole globe but should be within their range of uncertainty (which is estimated by the ensemble spread and which can be large for some parameters). On the Climate Data Store (CDS) disks, the initial (ERA5T) fields have been/will be overwritten, i.e., for these months, access to the original CDS disk ERA5T product will not be possible after it has been overwritten.
2. A data corruption example
2.1. Description of the issue in ERA5T
The impact of the assimilation of incorrect snow observations can be clearly seen in the maps of the monthly mean anomalies in soil moisture and snow depth for September and October 2021 (Figure 1). The maps highlight several "bull's eye" patterns mostly located in central Asia (primarly Kazakhstan). The most prominent of these patterns is located north of the Caspian Sea and appeared in the hydrological summary of the monthly C3S Climate Bulletin.
The "bull's eyes" in the anomaly maps are centered around stations which - as our investigation later revealed - started reporting spurious positive snow depth data at the beginning of September despite near-surface temperature well above 0ºC. The circular pattern of the anomalies around the stations highlight the effect of the two-dimensional optimal interpolation (2D-OI) scheme used in the Land Data Assimilation System (LDAS) of ERA5. Further information on the LDAS can be found in Hersbach et al. (2020). The presence of snow on the ground in the ERA5T model led to surface cooling, which is how the problem also affected the 2m air temperature and 2m dewpoint. However, as the 2m temperatures at the stations' locations remained well above 0ºC, the snow generated by the surface analysis rapidly melted, resulting in a rapid increase in soil moisture in ERA5T during the course of September and October.
|Figure 1. Monthly mean anomalies in soil moisture (layer 1) and snow depth for September and October 2021 based on data from ERA5T. The maps highlight the impact of the assimilation of incorrect snow observations from stations from central Asia (northern Caspian Sea region and eastern Kazakhstan). These stations are denoted with a '+' symbol in the maps. For soil moisture anomalies, the same mask (in darker grey) as the one used for the Climate Bulletin is applied over desert areas.
2.2. Corrected data from the final ERA5 product
In the final ERA5 product that will replace ERA5T for the months of September through December 2021, the stations that were identified as reporting incorrect snow depth data were excluded from the surface analysis, and the results are shown in Figure 2 .
|Figure 2. Monthly mean anomalies in soil moisture (layer 1) and snow depth for September 2021 based on the final ERA5 product. The final anomalies for October 2021 will be added when the data become available.
- Hersbach, H., et al.: The ERA5 global reanalysis, Q. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803, 2020.