Precise measurements of local magnetic fields have been obtained with a differentially connected array of seven synchronized proton magnetometers located along 60 km of the locked-to-creeping transition region of the San Andreas fault at Parkfield, California, since 1976. The M 6.0 Parkfield earthquake on 28 September 2004, occurred within this array and generated coseismic magnetic field changes of between 0.2 and 0.5 nT at five sites in the network. No preseismic magnetic field changes exceeding background noise levels are apparent in the magnetic data during the month, week, and days before the earthquake (or expected in light of the absence of measurable precursive deformation, seismicity, or pore pressure changes). Observations of electric and magnetic fields from 0.01 to 20 Hz are also made at one site near the end of the earthquake rupture and corrected for common-mode signals from the ionosphere/magnetosphere using a second site some 115 km to the northwest along the fault. These magnetic data show no indications of unusual noise before the earthquake in the ulf band (0.01–20 Hz) as suggested may have preceded the 1989 ML 7.1 Loma Prieta earthquake. Nor do we see electric field changes similar to those suggested to occur before earthquakes of this magnitude from data in Greece. Uniform and variable slip piezomagnetic models of the earthquake, derived from strain, displacement, and seismic data, generate magnetic field perturbations that are consistent with those observed by the magnetometer array. A higher rate of longer-term magnetic field change, consistent with increased loading in the region, is apparent since 1993. This accompanied an increased rate of secular shear strain observed on a two-color EDM network and a small network of borehole tensor strainmeters and increased seismicity dominated by three M 4.5–5 earthquakes roughly a year apart in 1992, 1993, and 1994. Models incorporating all of these data indicate increased slip at depth in the region, and this may have played a role in the final occurrence of the 28 September 2004 M 6.0 Parkfield earthquake. The absence of electric and magnetic field precursors for this, and other earthquakes with M 5–7.3 elsewhere in the San Andreas fault system, indicates useful prediction of damaging earthquakes seems unlikely using these electromagnetic data.

You do not currently have access to this article.