Differential magnetic fields have been monitored along the San Andreas fault and the Long Valley caldera since 1974. At each monitoring location, proton precession magnetometers sample total magnetic field intensity at a resolution of 0.1 nT or 0.25 nT. Every 10 min, data samples are transmitted via satellite telemetry to Menlo Park, CA for processing and analysis. The number of active magnetometer sites has varied during the past 21 years from 6 to 25, with 12 sites currently operational. We use this network to identify magnetic field changes generated by earthquake and volcanic processes. During the two decades of monitoring, five moderate earthquakes (M5.9 to M7.3) have occurred within 20 km of magnetometer sites located along the San Andreas fault and only one preseismic signal of 1.5 nT has been observed. During moderate earthquakes, coseismic magnetic signals, with amplitudes from 0.7 nT to 1.3 nT, have been identified for 3 of the 5 events. These observations are generally consistent with those calculated from simple seismomagnetic models of these earthquakes and near-fault coseismic magnetic field disturbances rarely exceed one nanotesla. These data are consistent with the concept of low shear stress and relatively uniform displacement of the San Andreas fault system as expected due to high pore fluid pressure on the fault. A systematic decrease of 0.8-1 nT/year in magnetic field has occurred in the Long Valley caldera since 1989. These magnetic field data are similar in form to observed geodetically measured displacements from inflation of the resurgent dome. A simple volcanomagnetic model involving pressure increase of 50 MPa/a at a depth of 7 km under the resurgent dome can replicate these magnetic field observations. This model is derived from the intrusion model that best fits the surface deformation data. ?? 1998 Elsevier Science B.V.