A method for detecting transient signals in GPS position time-series: smoothing and principal component analysis

Ji, Kang Hyeun; Herring, T. A.

doi:10.1093/gji/ggt003

Cited by 42 publications

(22 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We analyze just one of the three components of horizontal strain at each strainmeter, which we refer to as ϵ LCV‐na , because this component of strain appears to have lower noise than the other two components. We isolate the component of interest in section S2 by first identifying two components with low sensitivity to atmospheric pressure variations (Wang et al, ; Roeloffs, ; Dragert & Wang, ; Hodgkinson et al, ; Hawthorne et al, ) and then using principal component analysis to isolate the lowest‐noise component (e.g., Bishop, ; Press, ; Rangelova et al, ; Kositsky & Avouac, ; Ji & Herring, ).…”

Section: Spectra From Borehole Strain Observationsmentioning

confidence: 93%

Observing and Modeling the Spectrum of a Slow Slip Event

Hawthorne

Bartlow

2018

JGR Solid Earth

View full text Add to dashboard Cite

We estimate and model the normalized moment rate power spectrum of large slow slip events in Cascadia. We estimate the spectrum using data from GPS-derived slip inversions, borehole strain records, and beamforming-based tremor amplitudes. The normalized power spectrum initially decreases with frequency but then may flatten at periods of 1 to 10 days before decaying as frequency −n m at higher frequencies, where n m is between 1.1 and 1.4 when estimated from tremor and between 0.4 and 1.5 when estimated from strain. We explore one way to understand the observed spectrum: by modeling a month-long slow slip event as the sum of a steady background moment rate and a population of subevents. The subevents represent the wide variety of observed slow earthquakes, ranging from 0.5-s-long tremor to 3-hr-long rapid tremor reversals. We parameterize the subevents' magnitude distribution and moment-duration scaling, and we examine how the subevent population determines the slow slip spectrum. There are not enough data to rigorously test the subevent model, but we show that the data are consistent with a single continuum of slow earthquakes whose moments scale linearly with their duration, as has been proposed previously. Ide et al. (2007) proposed that all these bursts of slip, from LFEs to RTRs to large slow slip events, are components of a single slow earthquake family. Most well-observed slow earthquakes have roughly the same moment to duration ratio, or average moment rate (e.g., Ide et al. A Single Slow Earthquake Family, With One Moment-Duration Scaling?

show abstract

Section: Spectra From Borehole Strain Observationsmentioning

confidence: 93%

Observing and Modeling the Spectrum of a Slow Slip Event

Hawthorne

Bartlow

2018

JGR Solid Earth

View full text Add to dashboard Cite

show abstract

“…A series of workshops were recently held on automated detection of transients to aid in regional earthquake hazard studies [Murray-Moraleda and Lohman, 2010]. Proposed methods of automated transient detection have included, but are not limited to, the network inversion filter [Segall and Matthews, 1997] and the subsequent network strain filter [Ohtani et al, 2010], principal component analysis [Dong et al, 2006;Ji and Herring, 2013], covariance descriptor analysis [Kedar et al, 2010], Gaussian wavelet transforms [Melbourne et al, 2005], template matching [Riel et al, 2014], and Multichannel Singular Spectrum Analysis [Walwer et al, 2016]. All of these methods have their individual strengths and weaknesses and aim to perform similar tasks.…”

Section: Introductionmentioning

confidence: 99%

Single‐station automated detection of transient deformation in GPS time series with the relative strength index: A case study of Cascadian slow slip

2016

View full text Add to dashboard Cite

The discovery of slow‐slip events over the past decades has changed our understanding of tectonic hazards and the earthquake cycle. Proper geodetic characterization of slow‐slip events is necessary for studies of regional interseismic, coseismic and postseismic deformation, and miscalculations can affect our understanding of the regional stress field and tectonic hazard. Because of the proliferation of GPS data over the last two decades, an automated algorithm is required to analyze the signals and model the deformation on a station by station basis. Using the relative strength index (RSI), a financial momentum oscillator, we test the ability to detect events of various sizes and durations. We first determine the statistics of the RSI under different noise conditions and then use this information as the basis for the automated transient detection algorithm by testing different synthetic signals. We then apply the technique to daily GPS displacement time series from 213 stations along the Cascadia subduction zone to form a record of transient deformation between 2005 and 2016. Our estimates of the spatial extent, duration, and propagation of major episodic tremor and slip (ETS) events are consistent with previous studies. We use the automated detections to remodel the displacement time series and obtain transient deformation rates over the past decade and discuss the tectonic implications. Finally, we analyze the correlation between transient detections and tremor showing good agreement between the two at slab depths commonly associated with ETS events.

show abstract

“…When there are data gaps in the time series, the iterative computations are required for extracting CMEs. Ji and Herring (2013) introduced a Kalman filtering approach of filling data gap and the uncertainties in PCA analysis. Based on the principle that a time series can be reproduced with its principal components, recently proposed a modified PCA for extracting CMEs from the GNSS time series with missing data and contaminated by uniform white noises.…”

Section: Introductionmentioning

confidence: 99%

Weighted spatiotemporal filtering using principal component analysis for analyzing regional GNSS position time series

Shen

2015

Acta Geod Geophys

View full text Add to dashboard Cite

Principal component analysis (PCA) is a powerful tool for extracting common mode errors from the position time series of a regional station network determined by global navigation satellite system (GNSS). It is implicitly based on the assumption that a time series dataset contains temporally uniform white noise. Since the position time series of a regional station network are not uniform and could have data gaps, this paper develops a PCA-based weighted spatiotemporal filtering (WSF) approach by taking into account the positioning formal error of daily solution and the data gaps in time series. The position time series of 27 GNSS stations of the Crust Motion Observation Network of China are analyzed to demonstrate the performance of WSF approach, and also compared with the modified PCA technique in Shen et al. (J Geod 88:1-12, 2014). It shows that the WSF approach outperforms the modified PCA at 21, 19 and 17 out of the total 27 stations for north, east and up components, respectively. The average formal standard deviation of unit weight derived from WSF and modified PCA are 2.12, 2.42, 5.88 and 2.21, 2.52, 6.05 for north, east and up components, respectively; the relative improvements are 4.1, 4.0 and 2.8 %. Moreover, two simulations of a network with 4 stations are processed to show the performance of WSF. The results show that WSF provides better results for all coordinate components of all stations when the local effects are small or negligible. For cases when the local effects becoming larger, the WSF performs better than the modified PCA from the statistical point of view. From the real and synthetic time series analysis results, it is reasonable to conclude that the positioning formal error of daily solution should be considered in spatiotemporal filtering.

show abstract

A method for detecting transient signals in GPS position time-series: smoothing and principal component analysis

Cited by 42 publications

References 46 publications

Observing and Modeling the Spectrum of a Slow Slip Event

Observing and Modeling the Spectrum of a Slow Slip Event

Single‐station automated detection of transient deformation in GPS time series with the relative strength index: A case study of Cascadian slow slip

Weighted spatiotemporal filtering using principal component analysis for analyzing regional GNSS position time series

Contact Info

Product

Resources

About