Rapid resolution of crustal motion at short ranges with the global positioning system

On the basis of the analysis of Global Positioning System (GPS) data collected at the U.S. Geological Survey Kennedy Ranch House alinement array near Parkfield, California, in November 1990 and February 1991, we assess the temporal resolution of crustal deformation for GPS receiver systems operating as strain meters and provide guidelines for achieving millimeter-level precision with short-range, kinematic-type GPS measurements. Site occupation of about one-half hour is sufficient for day-to-day repeatability of submillimeter precision in relative position during good satellite geometry. Signal multipath effects which dominate dispersive baseline errors are essentially averaged out and need not be modeled. However, as occupation times decrease, baseline solutions are increasingly sensitive to multipath effects which together with nondispersive receiver measurement noise, are of millimeter to centimeter magnitude in epoch-by-epoch solutions. On repeatedly monitored baselines we can account for both of these noise sources separately. The mostly random receiver noise is effectively suppressed by low-pass filters with corner frequencies of about 20 mHz. The multipath noise is filtered using the fact that it essentially repeats from day to day. Residual errors in filtered 1-s epoch-by-epoch solutions are below 2 mm during periods of good satellite geometry. Hence for a continuously monitored baseline, actual deformation in excess of this threshold can be detected with a time lag comparable to the group delay of the low-pass filter, i.e., typically within a minute. In terms of strain along a typical short baseline of 1 km sampled once per second we can detect 1–2 μs in l min with multipath modeling. Without multipath modeling, it takes 10 min to resolve the same magnitude of deformation. In 30 min we can detect 0.1 μs independent of multipath modeling. These results have important implications for kinematic-type GPS surveys. For millimeter-level precision without impractical multipath modeling it is necessary to occupy a site during good satellite geometry for a period longer than that of the multipath signature. Our experience in both benign and difficult multipath environments indicates that approximately 10 min are sufficient. The measurements at the Kennedy Ranch House alinement array suggest that this segment of the San Andreas fault appears to be virtually locked at the surface during the 3-month period between the two surveys.