Research Observational Seismology
I am interested in general in tectonics and fault mechanics, currently I study these areas through observational seismology. Within the field of observational seismology I endeavour to create complete catalogues of seismicity (within an area) to enable studies of spatial and temporal changes in seismicity. To generate these complete catalogues we need to have an objective method of detecting and analysing earthquakes over large time-periods, without the inherent subjectivity that normal obervatory practices have. To this end I am also interested in applying big-data type methods for data mining, clustering and visualisation to seismic data.
Low-frequency earthquakes on the Alpine Fault
Following on from the discovery of tremor on the deep extent of the Alpine Fault by Wech et al., (2012), we extracted high amplitude periods from the tremor to use a templates to scan for low-frequency earthquakes. In this way we provided the first documentation of 14 low-frequency earthquake (LFE) families. These LFE families locate near the inferred (from GPS, seismic reflection and magnetotelluric studies) deep extent of the Alpine Fault, within a region of low quality factor (high attenuation), which we hypothesise is a region of high fluid pressure. We calculated magnitudes for all the events in our catalogue and demonstrated an exponential frequency-magnitude relationship. We also saw increases in LFE detection rates following large regional earthquakes which may be due to triggering. These results were published as Chamberlain et al., (2014).
Summary cartoon of LFEs in light of other studies
We are now working on extending the catalogue using a longer continuous seismic dataset, and more templates in an effort to generate a spatially and temporally continuous catalogue. This has been difficult as many previously used methods have either failed, due to high noise levels and low inter-station coherance, or are computationally inefficient for large datasets. We are developing a new method at the moment… to be continued.
Deep-Fault Drilling Project
Alpine Fault from space - from the DFDP wiki
I was lucky to be involved in the second phase of the deep-fault drilling project (DFDP-2), where we tried to drill into the Alpine Fault in Whataroa. My role in the project was characterising the seismicity in the Whataroa Valley prior to drilling, and co-leading the real-time seismic monitoring around the drill-site. We have used the EQcorrscan package to detect (and an as yet, non-distributed correlation based picking routine) and locate near-repeating seismicity near the drill-site. Currently this work is unpublished, as we refine locations using a variety of velocity models.
During and before drilling we operated a 24-7 real-time seismic monitoring operation using open-source (RTQuake) software to detect and generate initial locations and magnitudes. To check and refine picks and locations we operated in an international team of seismologists, making use of different time-zones to check detections within 30 minutes of them being made.
Tasman glacier speed-up
Tasman Glacier in the central Southern Alps from seismic site LABE on De La Beche ridge.
The Tasman Glacier is a large temperate glacier (containing 29% of New Zealand’s
perenial ice) near Mt. Cook in the central Southern Alps.
Horgan et al., (2015) documented speed-up events using geodetic observations on the glacier, following large rainfall events. We have a seismic site as part of the SAMBA network within 1km of the glacier (above photo taken from the site, LABE), which has increased amplitudes during large speed-up events. How this seismic signal relates to the speed-up events is a topic of current research within our group at Victoria University of Wellington.
Southern Alps Micro-earthquake Borehole Array
Less a research interest, more the tool to achieve the above interests; the Southern Alps Micro-earthquake Borehole Array (SAMBA) is a network of 13 seismometers deployed in the central Southern Alps to monitor micro-seismicity on and near the Alpine Fault. The SAMBA deployment began in late 2008 under the Royal Society of New Zealand, Marsden funded project: “Putting a stethoscope on the Alpine Fault”. The initial deployment and data analysis formed was completed by Dr. Carolin Boese during here PhD at Victoria University of Wellington. The network has since been maintained and added to during my PhD tenure.
Map of the SAMBA network taken from Chamberlain et al. (2014)
SAMBA is a network of short-period sensors deployed mostly in shallow (post-hole) boreholes, with three deeper borehole sites, and four short-period surface sensors to extend the network over the LFE and tremor source region.
Thanks to the amazing situation of the network we have some fun fieldtrips to collect the data every six months. Ideally data would be telemetered from the sites to reduce costs (helicopter bills rack up), but a lack of sight-lines and no cell-phone reception mean that this is currently very difficult.
Calum at the Solution Ranges (SOLU) SAMBA site in the central Southern Alps, overlooking the Landsborough Valley
The network produces some amazing data in a relatively seismically quiet region of New Zealand. There are very few earthquakes above M 4 within the network, and many noise sources (e.g. helicopters, storms, rockfalls, glacier motion, ocean), which make earthquake detection and analysis difficult. As such having a borehole deployment, with the reduction in noise that these bring, is essential to study this crucial section of the on-land plate boundary in New Zealand.
Developing multi-parallel open-source software for earthquake detection
To achieve any of the large-scale (long-duration datasets in spatial and temporal resolution) catalogues in an objective (because catalogues should be measurably complete, rather than subjectively complete according to which analyst was working a particular day, or whether that analyst was hungover) manor, we need to process all the data in the same way. One way to do this is using the matched-filter technique. However this requires a-priori knowledge of the earthquake source. I am working on a method of extracting earthquake information from the continuous data using a matched-filter approach, without this prior knowledge (it’s already in development on EQcorrscan).
To allow this method to work we first need an efficient matched-filter routine. I have been developing (and continue to develop) network-based matched-filter routines and analysis functions as part of the EQcorrscan package. These routines can be used on small or large computers, for large datasets they work best on big machines. I have been fortunate enough to obtain time on the PAN cluster through the NeSI project. This has allowed me to use over 600 templates, scanned through the full 6.5 year dataset, with compute clock times of less than 10.5 hours (less than 4 hours if I could use all the resources!).
The first low-frequency earthquake catalogue for the Alpine Fault
After Aaron Wech and Carolin Boese found tremor using the SAMBA network (Wech et al. 2012) I set about looking for low-frequency earthquakes (LFEs) within the tremor. I found a small set of LFEs by manual inspection, and, alongside David Shelly of the USGS, applied a matched-filter cross-correlation routine to find more, similar, LFEs. We documented 14 LFE clusters or families, which repeated throughout the 36 month period we studied. The catalogue we generated contained over 8000 discrete LFEs, with higher detection rates during tremor periods, and following large regional earthquakes. You can read more in the paper we published in Geochemisty, Geophysics, Geosystems: Chamberlain et al. 2014.
Stress and strain rates near the San Andreas Fault
Working with Nicolas Houlie during my masters project, we compared stress and strain rates in the vicinity of the San Andreas - Chamberlain et al., (2014).