What kind of ‘geographic’ coordinates are used for ICON output?

Chapter 2.1.3 of the ICON Tutorial [1] describes the mapping of geodetic (‘geographic’) to spherical coordinates for being used as input to ICON. While geodetic / geographic coordinates are typically defined by the WGS84 ellipsoid, ICON makes use of a spherical approximation to render the internal calculations more efficient.

The ExtPar tool, used for ICON input, interprets WGS84 latitudes as geocentric latitudes without further mathematical transformation.

In contrast to this, ICON output variables like CLAT, CLON, or HHL are defined as ‘geographical’ latitude, longitude or geometrical height above MSL [2].

Does this mean, that these variables thus refer to the WGS84 ellipsoid, making a subsequent and explicit geocentric to geodetic transformation unnecessary?

Does this also mean, that ”zero topography heights”, as defined by HHL for model half levels, can be thought of layered shells surrounding the internal WGS84 ellipsoid?

Thanks in advance!

References:
[1] F. Prill, D. Reinert, et.al. - “ICON Tutorial 2025”
[2] D. Reinert, F. Prill, et.al. - “DWD Database Reference for the Global and Regional ICON and ICON-EPS Forecasting System”, Version 2.5.6

You are right, the ICON model uses a spherical approximation of the Earth but it does not perform an explicit transformation of its input data to the WGS‑84 ellipsoid. The ExtPar tool simply re‑interprets the coordinates: WGS‑84 latitudes are taken directly and applied to the sphere without any conversion.
This approach also influences the handling of the output. If an explicit transformation to the WGS‑84 ellipsoid were applied, the atmospheric flow computed using the ExtPar data would be shifted from its original position and you would inadvertently introduce spatial errors. Therefore, the variables CLAT and CLON should instead be interpreted as geographical latitude and longitude (therefore the database manual is right here). In practice, you can plot CLAT/CLON directly on any standard map projection without performing an additional datum conversion; just remember that the underlying dynamics were calculated on a sphere rather than on the true ellipsoidal Earth.