Novel Processing Flow for GPS data to help Suppress Non-geologic Phenomena

Global Positioning System (GPS) time series contain not only tectonic signatures, but also are affected byseasonal variations, atmospheric changes, drifts in satellite clocks, and changes in satellite orbit. Multipleapproaches are used to isolate tectonic signals. The most common data correction algorithms employedwhen isolating tectonic signals revolve around modeling seasonal changes and correcting for them. Here,we introduce a novel algorithm for isolating tectonic signals in GPS data based on observation of datacharacteristics from multiple regions in the world (Cascadia Subduction Zone, Alaska, and New Zealand).Raw displacement data along the Cascadia Subduction Zone from 2016 is shown in Figure 1a. Conven-tional processing involves detrending (Figure 1b) and other empirical corrections to account for seasonalchanges.However, we observe streaks along the station axis when all data are plotted together (Figure 1aand 1b). These streaks become even more prominent in the velocity time-series (Figure 1c) obtained bycomputing the first-order time derivative of the raw displacement data (Figure 1a). These streaks are likelynot geological and possibly arise from GPS positioning errors or atmospheric disturbances. GPS data fromthe Aleutian and Hikurangi Subduction Zones also show the same above-mentioned data characteristics.We apply a median filter along the station axis to suppress these streaks. The resulting GPS velocity(Figure 1d) shows effects of tectonic changes as is clear from the geodetic event (Episodic Tremor andSlip) in early 2016 in Figure 1d. As well-known from Episodic Tremor and Slip (ETS) studies, the mostprominent displacement/velocity changes are along the east-west direction.Analysis of the above processed data shows that there are not only spatially correlated trends in theeast-west and north-south direction during ETS but also up-down signals exhibit trends that are spatiallycorrelated. These observations will shed new light on ETS phenomena.

Figure: Three-component (a) raw GPS data and (b) detrended data from the Cascadia Subduction Zonerecorded in 2016. (c) GPS velocity data obtained using time-derivative of displacement data. (d) theresult of median filtering applied along the station axis.