LOS Time Series
We have compared the integrated InSAR/GNSS LOS time series with the
GNSS-only time series and find both excellent and poor agreements (Table
S2). Note the two time-series will not match exactly as our approach is
essentially trusting GNSS at longer wavelengths while InSAR at short
wavelengths. Fig. 4b shows several examples where there is excellent
agreement between the InSAR and GNSS time-series. The typical deviation
of GNSS daily solutions from its moving average is around 3-4 mm, while
the integrated InSAR deviates only slightly larger, around 3-6 mm (Fig.
4a). The vertical component could largely reproduce this deviation and
can be considered as the major source of discrepancy. However, there are
a number of cases where the integrated InSAR GNSS time-series fail to
match the GNSS solutions (Fig. 4c). In most cases this is a direct
result of anomalous displacements or artifacts in the GNSS displacement
time series. An extreme example is station p800 in Los Angeles showing
systematic effects due to the growth of thick vegetation nearly
concealing the GNSS antenna and causing a significant drift in the north
direction over a period of about 2 years (overlapping with the 4.5 years
of our time series) until the vegetation is cleared in May 2018 (add
figure in supplement). Station CRHS also in Los Angeles and affected by
vegetation and whose data after 2006 are suspect and should not be used
(personal communication, Chris Walls, UNAVCO NOTA engineer) – the
station was subsequently abandoned. Station CUHS in the Cuyama Valley is
subject to significant subsidence of about 30 mm/yr with a seasonal
signature resulting in non-tectonic horizontal displacements with
amplitudes up to 30 mm. Station CJMG in the San Gabriel Mountains has
annual horizontal artifacts starting in 2017 with an amplitude of 30 mm
(peak to peak). Another example is station P789 on the San Andreas fault
in the transition zone between locked and creeping section is
experiencing uplift since 2013 of about 2.5 mm/yr but not fully
sufficient to explain its misfit with InSAR. For the purpose of this
study these GNSS data were not excluded. Rather, we allowed for
discrepancies to exist through the integration, especially when a single
station shows large mismatch, because the GNSS correction model, applied
to each interferogram, is smoothed with a robust filter that down weighs
the anomalous misfits. Thus, in cases a GNSS station starts to behave
anomalously, the InSAR time-series remain largely unaffected (Fig. 4c).
In further studies, we recommend a more conservative approach in
choosing stations for the GNSS/InSAR integration. More assessment and
results are available in Table S2 and Fig S1 in the supplements.