ISOLATING LONG-TERM TRANSIENTS IN CASCADIA GPS DATA USING A SEASONAL-TREND DECOMPOSITION PROCEDURE

The absence of long-term slow slip events (SSEs) in Cascadia is enigmatic on account of the diverse group of subduction zone systems that do experience long-term SSEs, which includes Japan, Alaska, New Zealand and Mexico. At these subduction zones, long-term SSEs serve as an important mode of stress and strain release and influence how strain is accumulated spatially and temporally. The conditions that encourage long-term slow slip are not well established due to the variability in thermal parameter and plate dip amongst subduction zones that host long-term events. The Cascadia Subduction Zone likely has the capacity to host long-term SSEs, and the lack of such events motivates further exploration of the observational data. In order to search for the existence of long-duration transients in surface displacements, we examine Cascadia GPS time series from PANGA and PBO to determine whether or not Cascadia has hosted a long-term slow slip event in the past ~20 years. A careful review of the time series does not reveal any large-scale multi-year transients. In order to more clearly recognize possible small amplitude long-term SSEs in Cascadia, the GPS time series are reduced with two separate methods. The first method involves manually removing (1) continental water loading terms, (2) transient displacements of known short-term SSEs, and (3) common mode signals that span the network. The second method utilizes a seasonal-trend decomposition procedure based on loess (STL) to filter and decompose the time series into seasonal, trend and remainder components. We find that the STL method successfully partitions seasonal, common mode, and short-term SSE signals into the seasonal and remainder components, thus producing a trend component that encapsulates long-term motion at GPS stations. A comparison of the manually cleaned data with the trend components from STL shows that both methods produce similar surface displacement time series products. Therefore, we believe that STL proves to be a reliable and efficient alternative to manual cleaning of GPS data when investigating long-term trends. Manual inspection of both of these products also reveals intriguing long-term changes in the longitudinal component of several GPS stations in central Cascadia, although the origin of these changes remains unclear.