Global Positioning System (GPS) surveys were used to determine coseismic displacements within the epicentral region of the 1989 Loma Prieta earthquake near Santa Cruz. The California Highway Department (Caltrans) conducted the preearthquake survey in February and March 1989 using single-frequency GPS receivers. In March 1990, we reoccupied eight of the Caltrans stations using dual-frequency receivers. Relative displacements were computed by differencing the stations coordinates determined from these two surveys. Displacements were determined relative to sites remote from the earthquake by incorporating GPS measurements from Loma Prieta (station LP1) to the Very Long Baseline Interferometry (VLBI) site at a Fort Ord. Examination of repeated measurements in the single frequency preearthquake survey indicates precison is ~10 mm in the horizontal, although it may be as low as ~40 mm in the weakly connected part of the network. The precision in the vertical is generally 30--40 mm. In contrast, the largest observed coseismic displacements is 410 mm horizontal and 341 mm vertical. We use a quasi-Newton method to estimate the parameters of the uniform rectangular dislocation source that best fits the observed displacments.

The best fitting dislocation dips 68¿ southwest, extending from 3 to 16 km depth, strikes N66¿W, and has 1.8 m of dip slip and 1.2 m of strike slip. The best fitting dislocation overlaps the aftershocks but is rotated 15¿ counterclockwise from the trend of the aftershock zone. We find that the data can be fited nearly as well with a dislocation that is constrained to lie in the plane of the aftershocks. In this case the fault extends from 5 to 15 km depth. The constrained model explains 96% of the data. Both models have a moment of M0=2.9¿1019 N m. We also use the quasi-Newton method to estimate rectangular dislocation models from the leveling data of Marshall et al. (1991), and the trilateration and GPS data of Lisowski et al. (1990). In contrast to the previous studies, we find that both of these data sets are independently best fit with dislocations embedded within the aftershock zone. ¿American Geophysical Union 1993