A 66-station GPS network spanning central Greece, first observed in 1989, has been occupied fully on three occasions: June 1989, October 1991 and May 1993. Subsets of this network bounding the Gulf of Korinthos have also been occupied in June 1995, October 1995, May 1996 and September/October 1997. The first three occupations were processed using a fiducial GPS methodology, whereas later surveys were processed using CODE precise orbits. Combination of data from different surveys to yield smooth site velocities requires global network translations at each epoch to compensate for errors in the realization of the reference frame. This method provides a posteriori estimates of the relative coordinate errors and reference frame noise. Only one earthquake, the 1995 June 15 Egion event, has caused significant local coseismic displacement, and its effects on the interseismic velocity field are removed using an elastic dislocation model.
We constrain the orientation of the 100 yr triangulation–GPS velocity estimates of Davies et al. (1997) using 14 sites common to the two networks. The goodness of fit of this transformation indicates that the short-term and 100 yr geodetic estimates of deformation are highly compatible. We infer that short-term geodetic studies are capable of determining longer-term deformation rates provided that transient, local effects can be modelled. From the combined velocity field, we estimate principal strains and rigid-body rotation rates at points on a regular grid using data from neighbouring sites. Strain rates are high within the Gulf of Korinthos and much lower elsewhere. The extension rate across the Gulf of Korinthos increases from east to west. Comparison of the extension rate with historical and recent rates of seismic release of strain reveals significant medium-term seismic hazard in the western Gulf of Korinthos, and may also indicate long-term aseismic strain.