Back in the late 80's I had some college rooommates who did this for a living. After moderate earthquakes in southern California, groups of geophysics graduate students would be sent to the channel islands off the coast with huge old clunky GPS receivers. They would align the GPS receiver over a benchmark and camp out for 3 days collecting data. Similar groups would do the same thing all over SoCal. Combining the data makes for a super differential GPS data set. As each receiver is at a known location (well sort of, it is what they are there to determine), each has the accuracy of a single diff. GPS receiver. However, what the scientists cared about was not the aboslute positions of the receivers, but their relative positions. As I recall, 0.5cm resoultion was routinely achieved event back then. I'm sure todays systems are automated, and remotely read out. Today's grad students won't have stories about being buzzed by navy jets or herds of ferrel cats.