Apparent stability of GPS monumentation from short-baseline time series

Long-running, short (<< 1 km) GPS baselines offer an insight into the accuracy budget of geophysical estimates inferred from GPS coordinate time series. In this paper, we report on analysis of 10 short baselines using data spanning several years and examine deviations from a constant baseline length. Annual signals with amplitude >0.5 mm are evident in various coordinate components at 6 of the 10 sites, with amplitudes exceeding 2.5 mm on two baselines. These signals are largely invariant to elevation cutoff angle. Multipath and phase center model errors are shown to probably have negligible effects at most sites at annual periods. They are therefore likely real monument motions, although linear thermal expansion is negligible at most sites. Linear trends >0.25 mm yr(-1) are evident on four baselines in at least one coordinate component. Subdaily signals are present at all sites at similar to K1 and its higher harmonics. Time-correlated noise, possibly in the form of flicker noise, dominates despite many common signals being eliminated as a result of the small baseline lengths, although it is around 1 order of magnitude smaller than for single site PPP solutions. We constrain monument random walk motion to be no more than 0.5 mm yr(-1/2). If these baseline motion results were representative of the similar to 300 currently active IGS sites, 180 would have annual signals >0.5 mm in at least one coordinate component, 120 would have linear rates >0.25 mm yr(-1), and almost all sites would have subdaily (K1, K2) signals of >0.1 mm, each solely due to local site effects.