Detecting and modelling transient crustal deformation using Sentinel-1 InSAR and Machine Learning

As the quality of satellite geodetic observations of surface deformation from Global Navigation Satellite Systems (GNSS) and Satellite Radar Interferometry (InSAR) have improved it has been increasingly clear that fault behaviour is not steady in time. On short time scales, slow earthquakes, often associated with seismic tremor, have been observed at numerous subduction zones around the world (Bürgmann, 2018), and some segments of strike-slip faults in the continents have also been shown to “creep” episodically in “slow” or “silent” earthquakes (Rousset et al., 2016) [Figure 1]. Significant changes in deformation rate have been observed on a decadal scale prior to major earthquakes (Mavrommatis et al., 2014). Following earthquakes, postseismic transient afterslip can last for up to a century (Ingleby and Wright, 2017). And on a longer time scale, faults have been observed to change their rates of slip over millennia (Cowie et al., 2017).

The student in this project will use the wealth of geodetic data from Sentinel-1 InSAR, processed by COMET scientists in Leeds, alongside archive data from satellites including ERS and Envisat, and any available GNSS (GPS) data, to investigate how widespread transient behaviour is on faults. They will use machine learning approaches to mine the data to identify deformation transients. This may include the use of supervised learning methods to classify faults, as well as tools from fields such as time series analysis, point processes, and anomaly detection. They will use estimates of Quaternary slip rates to understand how deformation rates vary over different timescales. They will use the results to build and test models of the earthquake deformation cycle, to understand the geological controls on deformation transients, and to explore the impact of transient behaviour on our understanding of seismic hazard. The results will have potential applications for monitoring different hazards, including volcanoes worldwide and landslides and sinkholes in the UK.

The project would suit a numerate student with a background in earth sciences, geology, or geophysics who is enthusiastic about problem solving and the use of EO and machine learning approaches. The student will be provided with training in state-of-the-art geodetic and machine learning methods and will have the opportunity to participate in field campaigns. The student will be part of the UK Natural Environmental Research Council’s Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics (COMET) and will be expected to interact with COMET students with different skills and backgrounds from across the UK.

Check https://eo-cdt.org/ for more information on funding and application process.

Note that the project is also co-supervised by Gordon Ross, University of Edinburgh https://www.maths.ed.ac.uk/school-of-mathematics/people/a-z?person=593