Automatic detection of slow slip earthquake events

The below case study shares some of the technical details and outcomes of the scientific and HPC-focused programming support provided to a research project through NeSI’s Consultancy Service.

This service supports projects across a range of domains, with an aim to lift researchers’ productivity, efficiency, and skills in research computing. If you are interested to learn more or apply for Consultancy support, visit our Consultancy Service page.

Research background

A slow slip event (SSE) is a type of recurring slow earthquakes, part of which plays an important role in releasing strain in subduction zones and have been found to precede large natural earthquakes. Over the past few decades, a wealth of GPS stations (over 15,000) have been deployed to detect SSEs worldwide. The huge amount of daily GPS time series makes it impossible to virtually inspect SSEs, therefore demanding a robust automatic detection method. 

Project challenges

University of Otago PhD candidate Yiming Ma has developed an R code that detects slow slip events, a type of slow earthquakes. The code relies on the IDetect package to identify the change points in GPS time series (jumps in value or derivatives). These points are challenging to detect because of the noise in the signals, which mask the jumps in signal values and derivatives. One function (ID_cplm) was found to account for 92 percent of the execution time. The authors of IDetect are actively working on another version of the code with a more efficient implementation of this function. In the meantime NeSI wanted to help Yiming run 100s of white noise scenarios in the most efficient way. 

What was done

NeSI Research Software Engineers parallelised the outer loop by breaking the iterations into multiple HPC jobs through parallelisation.

Main outcomes

Parallelisation led to a 20x reduction of the turn-around time for 100 iterations. Apart from improving the productivity of the researcher, this leads to a more efficient use of the Mahuika cluster by spreading the workload into small jobs, which are easier to fit in the queue than jobs that require a lot of resources and large wall clock times. 

Researcher feedback

"I wouldn’t have been able to conduct such large-scale simulations without the help of the NeSI support. It’s really an excellent and productive journey working with the NeSI support." 

- Yiming Ma, PhD Candidate, Department of Mathematics & Statistics, University of Otago 

Do you have a research project that could benefit from working with NeSI research software engineers or our data engineer? Learn more about what kind of support they can offer and get in touch by emailing support@nesi.org.nz.