Fault creep and interseismic loading in the northern San Francisco Bay region

Collaborators: Li-Chieh Lin (UCR), Ryan Rivera (UCR), Simran Sangha (JPL), Marin Govorcin (JPL), Mike Floyd (MIT)

Past collaborators: Lizhen Jin, Jerlyn Swiatlowski

InSAR evidence for creep on the Rodgers Creek fault between Healdsburg and Santa Rosa

Shallow fault creep is an unusual behaviour of some faults in California and a select few other places in the world. Rather than remaining locked at the surface for long periods and releasing strain in earthquakes (the 'classic' stick- slip behaviour seen on most faults), creeping faults move slowly over time, and without producing large earthquakes. To understand if there are geological controls on the occurrence of creep, we are searching using persistent scatterer/small baseline subset InSAR, focusing on the faults of the northern San Francisco Bay region, where creep has been observed in several locations, including on the Rodgers Creek fault, for the first time.

In order to 'ground truth' our InSAR observations, as well as measure the loading of faults between earthquakes ('interseismic loading'), we have been making extensive campaign GPS measurements across the region since 2008. Our preliminary results imply that more slip may be occurring on the Rodgers Creek-Maacama and Green Valley- Bartlett Springs fault systems to the east and less on the main plate boundary fault, the San Andreas, in the west than current USGS models suggest.

Example publication: Funning et al., 2007, GRL

The InSAR Centroid Moment Tensor (ICMT) project

Collaborators: Karlee Rivera (UCR), Ana Ferreira (UCL)

Past collaborators: Nader Shakibay Senobari (now at UCR Computer Science and Engineering), Jennifer Weston, Michael Frietsch (now at KIT) 

InSAR, through showing us in great detail where the ground was displaced in earthquakes, can give us very accurate locations of earthquake centroids (the center of moment release for each earthquake). We find in general that centroid locations obtained by long-period seismology (e.g. from the Columbia University 'Global Centroid Moment Tensor', or GCMT, project), disagree with these accurate locations, often by tens of kilometers. Since seismologically- derived earthquake locations depend on an Earth model (a model of how seismic wavespeeds vary within the Earth), our hypothesis is that any error in earthquake location can be attributed to an error in that Earth model.

Our project involves two parallel efforts. Firstly, we have been compiling a catalogue of published earthquake source models derived from InSAR data, in a format similar to the GCMT project; we call this the 'ICMT catalogue'. Second, we have been testing a variety of high resolution Earth models to investigate which features in those models may be required to correctly locate earthquakes seismologically – and if they cannot be located accurately using those models, what biases in seismic wave travel time need to be included to do so.

Example publications: Weston et al., 2011, JGR; Ferreira et al., 2011, JGR; Weston et al., 2012, Tectonophysics

Difference in location between GCMT and ICMT catalogues for earthquakes in South America

Estimates of event duration from the Matrix Profile cross correlation (MPCC). Each P-wave onset results in a sharp rise in MPCC, which continues to the end of the event coda. We are exploring the use of this duration estimate to estimate event magnitude. 

Seismic data mining using the Matrix Profile

Collaborators: Norma Contreras (UCR), Nader Shakibay Senobari (UCR), Eamonn Keogh (UCR),  Philip Brisk (UCR), Roby Douilly (UCR), Zach Zimmerman (now at Google)

The Matrix Profile (MP) is an efficient and highly parallelizable method for discovering repeating patterns in time series data. It works very well at identifying earthquakes in seismic waveform data, and can handle extremely long time series (we have tested it on 1 billion data points - 579 days of data from a single station component at 20 Hz). It has the advantage over template matching algorithms in that it does not require a priori templates.

We currently have NSF funding to explore developing applications of the MP for seismology, and refining the algorithms that compute it.

Example publications: Zhu et al., 2016, IEEE ICDM; Zimmerman et al., 2019, ACM SoCC

Detailed studies of earthquakes using InSAR

Past collaborators: Christos Kyriakopoulos (now at U. Memphis), Tisha Irwin (now at Saddleback College), Chad Severson (now at Idaho Power)

The high spatial resolution of InSAR allows us to image earthquake slip and fault geometry in fine detail. We use this capability to investigate specific earthquake phenomena in detail, such as the relationship between fault geometry and fault slip, the control that data distribution has on fault location, or the effect of surface topography on surface displacement. In addition, we are currently investigating how combining InSAR studies with complementary techniques, such as body wave seismology or dynamic earthquake rupture modelling, can lend us greater insights into the earthquake rupture process, and ultimately, into the underlying physics and physical properties controlling it.

Example publication: Funning et al., 2014, GJI

Faulting and deformation in southern California

Collaborators: Scott Marshall (Appalachian State), Eileen Evans (CSU Northridge)

GPS velocities, and the deformation strain rates that they imply, show us the 'forcing' that drives faulting in tectonically active areas. We combine GPS strain rate tensors with boundary element models incorporating realistic fault geometry in an attempt to estimate fault slip rates in specific regions of southern California. In one of these regions, the Ventura Basin, where some researchers have proposed evidence for infrequent but large thrust faulting earthquakes, our compilations of GPS data show the expected horizontal contraction, but very little of the expected vertical motion that would be associated with that shortening. We have also been collecting campaign GPS data in the Inland Empire area to densify the GPS velocity field in the vicinity of the San Jacinto fault, the second most dangerous fault in the region, and the closest to Riverside.

Example publication: Marshall et al., 2013, JGR

Deformation caused by extraction and injection at The Geysers, California

Collaborators: Mike Floyd (MIT)

Past collaborators: Rachel Terry (now at UNAVCO)

Some of the largest deformation signals we see in California are associated with human activity. The Geysers, located in northern California, around 100 miles north of the Bay Area, is the largest active producing geothermal field in the world. The large volumes of steam extracted from the ground, when insufficiently replaced with injected fluids or rainwater, led to a decline in steam pressures through the 1990s, accompanied by rapid subsidence that was measured by both GPS and InSAR. Efforts to mitigate these effects in the early 2000s, through the construction of pipelines to inject wastewater from local cities into the geothermal reservoir, resulted in reduced subsidence rates that we detected via a series of GPS campaigns from 2008 to 2011. 

Injection at the field is ongoing, in an attempt to enhance geothermal production. We are currently monitoring the effects of this through three continuous GPS sites that we installed at The Geysers in 2012 and 2013. Current indications are that the northwestern part of the field, a region of net injection at present, is uplifting as a result.

Example publication: Floyd and Funning, 2013, GRC Transactions