Classically, atmospheric corrections for terrestrial gravimetry are computed from the local air pressure record and a conventional admittance factor of \(-3.0\, \mathrm {nm}/\mathrm {s}^{2}/\mathrm {hPa}\) is usually adopted to derive gravity effects. The reference level is based on a standard atmosphere in agreement with the Resolution \(\mathrm {N}^\circ 1\) of the International Association of Geodesy (IAG) of 2023, which defines the International Terrestrial Gravity Reference System (ITGRS). Roughly 90% of the atmospheric contributions to gravity variations are covered by this approach. To account also for the spatial distribution of air masses around a given station, the Atmospheric attraction computation service (Atmacs*) was established that relies on numerical weather models of the German Weather Service (Deutscher Wetterdienst – DWD). However, while Atmacs provides accurate time-variations of atmospheric corrections, the long-term stability is currently limited by occasional model improvements at DWD. Such model updates do not only include improvements in the spatial and/or vertical resolution but also modifications of the associated model orography, and may thus cause discontinuities in the time series of atmospheric corrections which may rise the tens of \(\mathrm {nm}/\mathrm {s}^{2}\) level and are typically altitude-dependent. In the present study, we aim to solve this problem by referring the time series to a common reference ensuring the compatibility with the conventions adopted for the ITGRS, and we show the advantage of using Atmacs for correcting absolute gravity observations.