The well-known International Association of Geodesy (IAG) approach to the a
tmospheric geoid correction in connection with Stokes' integral formula lea
ds to a very significant bias, of the order of 3.2 m, if Stokes' integral i
s truncated to a limited region around the computation point. The derived t
runcation error can be used to correct old results. For future applications
a new strategy is recommended, where the total atmospheric geoid correctio
n is estimated as the sum of the direct and indirect effects. This strategy
implies computational gains as it avoids the correction of direct effect f
or each gravity observation, and it does not suffer from the truncation bia
s mentioned above. It can also easily be used to add the atmospheric correc
tion to old geoid estimates, where this correction was omitted. In contrast
to the terrain correction, it is shown that the atmospheric geoid correcti
on is mainly of order H of terrain elevation, while the term of order H-2 i
s within a few millimetres.