I’m posting what I noticed here, as there seems to be no mention at all in the ERA5-Land documentation of what seems to be a very large discrepency in ERA5-Land and ERA5 in South America and Africa prior to 1970.
ERA5-Land has central Africa and South America with several hot and arid episodes in 1951, 1961, and 1965, that are either non-existent or much less pronounced in the regular ERA5 dataset.
The documentation specifies that ERA5-Land "always uses forcing fields based on the final release of ERA5”. However, if one compares temperature, humidity, and precipiation in the Copernicus Interactive Atlas, the trends seem consistent after 1978 between ERA5 and ERA5-Land. However, prior to 1970, there are several spikes in the data in 1951, 1961, and 1965 in ERA5-Land but not in the regular ERA5.
The only way to explain these spikes in ERA5-Land data is that they came from a much older deprecated version of ERA5.P (preliminary version) back extension data which initially had these spikes, and not the final release that came after it. ERA5.P was said to be removed on August 15th 2023, but this data with the large spikes is still present in ERA5-Land to this day.
I would think it would be wise for the documentation to note that ERA5-Land data prior to 1978 is derived from the older deprecated ERA5.P (preliminary version) back extension data, as these spikes do not come from the final release of ERA5.
