The usability of ECMWF's forecasts for the determination of diurnal and semidiurnal mass variations in the atmosphere-ocean system due to atmospheric pressure tides is examined and contrasted to corresponding variabilities deduced from ECMWF's analyses. While the diurnal pressure tide and the oceanic response simulated with a baroclinic ocean model are well resolved from both analyses and forecasts, the semidiurnal tide can be recovered from 3 hourly forecasts only. In terms of rms values of geoid height anomalies, forecast errors cause 0.18 mm, different wind representations 0.09 mm, and the doubled temporal resolution 0.20 mm of deviations between forecasts and analyses. Since atmospheric tides are highly variable, a time-invariant harmonic approach might not meet high precision requirements as for the GRACE mission and for high-resolution Earth rotation parameters. Considering these forecast errors, forecasts allow to account for atmospheric variability and corresponding oceanic responses down to semidiurnal timescales, dispensing with any additional model of atmospheric tides.