Atmospheric variables below model terrain

ERA5 reanalysis data on pressure levels at lower levels, eg at 1000 or 850 hPa, can show atmospheric variables that are below the model terrain. This happens in regions such as the Himalayas or Andes and is due to the interpolation of data from model to pressure levels  (the interpolation documentations is at
The data which is below the surface can be masked out using the surface pressure or the surface geopotential.
For more information about ERA5 levels see the ERA5 documentation.

The link you sent for the ERA5 documentation says the orography can be downloaded from the era5-single-levels dataset. 

I'm confused, because that dataset varies with time. I've tried, and I can indeed download "hourly fields" of geopotential at the surface (aka topography). Shouldn't the orography be an invariant field, like in ERA-Interim for instance?

Or does the topography in ERA5 really changes? 

I would be a good idea to clarify this point in the documentation.


the surface geopotential (orography in the CDS) is listed in Table 1: surface and single level parameters: invariants at the following link:

ERA5: data documentation - Copernicus Knowledge Base - ECMWF Confluence Wiki 

Please also have a look at the description for orography available on the overview page ERA5 hourly data on single levels from 1979 to present ( where it states that the parameter does not vary in time.



Hi Michela Giusti,

I am working with ERA5 reanalysis data on pressure levels and try to mask out atmospheric variables that are below the model terrain using the invariant geopotential from the era5-single-levels dataset.

However, I have realized that the geopotential in ERA5 reanalysis data on pressure levels varies over time.

Hence, I observe different values for the geopotential at different hours (and dates) for the same location (coordinates).

For instance, I observe the following values in my data:

level time geopotential longitude latitude
1 01.01.2020 10:00 458472 -100 40
1 01.01.2020 22:00 460877 -100 40
1 01.11.2020 10:00 458277 -100 40
1 01.11.2020 22:00 460718 -100 40
... ... ... ... ...

Shouldn't the geopotential be time-invariant within the same pressure level and coordinates?

How can I mask out atmospheric variables below the model terrain based on changing geopotentials in the ERA5 reanalysis data on pressure levels data?