New estimates of time-dependent noise and velocity uncertainties in GPS position time series

Abstract:

GPS station velocities inferred from position time series have become vital for many subfields of geophysics. However there is still no consensus on how large GPS velocity errors truly are. This is especially true for long-duration time series. The reason is that long-period, time-dependent noise in GPS time series is difficult to detect. Random-walk noise process is especially difficult to estimate, but has a large effect on velocity uncertainty.

Previously we developed a Network Noise Estimator that provides more precise estimates of time-dependent noise than traditional, methods of noise estimation for single time series. The method uses Kalman filtering to determine maximum likelihood estimates of noise parameters for networks of stations. Simulations show that this method allows us to estimate more precise velocity uncertainty in GPS position data within networks of stations.

In this study we present estimates of GPS velocity uncertainties from many dozens of stations and compare them to the traditionally estimated values. We largely focus on the central region of the North American continent. Our preliminary estimates of velocity uncertainty are 0.5-0.6 mm/yr for 10 years of data, which is twice the level found with traditional methods. We show that small unmodeled linear trends do not affect the noise estimates, so that some minor tectonic signal should not affect our results. We provide confidence intervals on the noise parameters and explore correlations between flicker-noise and random-walk estimates. Additionally, we determine typical noise estimates for groups of stations with similar monumentation. These estimates could be used to determine velocity uncertainty for any future use of the data from stations with similar monument types.