Project Info

detecting events in lossy-compressed seismic data

Eileen Martin

Project Goals and Description:

New technologies have enabled geophysicists to collect orders of magnitude more seismic data thanks to new sensor technologies, but a major challenge is that they have no public data archive that can support data volumes of this size. They need to take advantage of lossy compression methods that approximately preserve all information about the data in many fewer bytes (e.g. aiming to reduce data volumes by 10x to 100x). However, if the data preserved are only approximate, scientists want to understand how differences from the raw data may propagate into differences in the final analysis of these data. For example, can weak seismic events (small earthquakes or other small events) still be detected in the compressed data? Specifically, the student will theoretically and computationally investigate a type of compression called zfp, which can be applied to 2D and 3D arrays of scientific data. The student, in collaboration with Eileen Martin (faculty mentor) and Hafiz Issah (grad student mentor), will apply this compression method to data acquired by fiber optic cables repurposed as seismic arrays via a technology known as distributed acoustic sensing (DAS). They will develop mathematical models of the compression process applied to data governed by the wave equation. They will perform testing of seismic event detection algorithms such as STA/LTA or FAST to quantify differences in results obtained with the original data versus the compressed data. The student will integrate this compression scheme into the under-development community DAS Data Analysis Ecosystem (DASDAE) software.

More Information:

Grand Challenge: Engineer the tools of scientific discovery.
Lindstrom, Peter. "Fixed-rate compressed floating-point arrays." IEEE transactions on visualization and computer graphics 20.12 (2014): 2674-2683. Diffenderfer, James, et al. "Error analysis of zfp compression for floating-point data." SIAM Journal on Scientific Computing 41.3 (2019): A1867-A1898. Trainor-Guitton, Whitney, et al. "Distributed Acoustic Sensing and Machine Learning Hone Seismic Listening." Eos, 103. March 2022. Trnkoczy, Amadej. "Understanding and parameter setting of STA/LTA trigger algorithm." New Manual of Seismological Observatory Practice (NMSOP). Deutsches GeoForschungsZentrum GFZ, 2009. 1-20. Yoon, Clara E., et al. "Earthquake detection through computationally efficient similarity search." Science advances 1.11 (2015): e1501057. Distributed Acoustic Sensing Data Analysis Ecosystem (DASDAE), with documentation and community codes at

Primary Contacts:

Eileen Martin, | Hafiz Issah,

Student Preparation


Required: linear algebra, numerical methods, partial differential equations, ability to do programming in one of C, C++ or Python Preferred: background in signal processing or image processing




numerical error analysis, signal processing, data compression and information theory, statistical error analysis in scientific workflows, scientific software engineering skills (unit testing, software validation, scalability testing), communication and visualization skills


once-per-week half-hour one-on-one meeting with Eileen Martin, half-hour weekly group meeting with the Martin Group, occasional meetings with graduate student Hafiz Issah (AMS), discussions once a semester about goals and next steps (career paths, graduate school and fellowship applications)


Share This