For the stations used in this study, with heights up to 500 m, no relation was found between station height and larger residuals when using the Hopfield model. The increase in the residuals with respect to VMF1 seems to be strongly correlated with regions of large variability in the pressure and temperature fields and only moderately with station height. Although in this study stations with heights up to 500 m are used, preliminary results indicate that a similar accuracy level (up to 1–3 mm, depending on the variability of the pressure and temperature fields) can be obtained for stations with heights up to about 1000 m. Using VMF1 grids, with 2.0 2.5 spacing, ZHD can be estimated with 2–4 mm accuracy in parts of the world in which the atmospheric pressure has small to moderate variability, provided the Hopfield height reduction is used. The Berg reduction introduces strong seasonal signals in most stations located in regions where the pressure and temperature have a strong annual variation. Therefore, when using the VMF1 it is strongly recommended the use of the Hopfield or an equivalent type of height reduction. Even using a proper height reduction, some areas of the globe still exist where the VMF1 lead to a determination of ZHD with errors that, in some stations, exceed 4 mm (1r). For some of these stations, the errors present on VMF1-derived ZHD have clear seasonal patterns, which exceed the centimetre level during some periods of the year. These errors will fully propagate into derived products such as the wet tropospheric corrections for coastal altimetry, thus inducing erroneous seasonal signals which can have an impact, for example, in altimeter derived sea level variability.