← Back to Gallery

Exploring the Impact of Lossy Compression on Passive Seismic Event Detection and Arrival Time Precision

New seismic sensing technologies such as distributed acoustic sensing (DAS) and low-cost accelerometer nodes are making it easier than ever before to continuously collect high-resolution, large-scale seismic data. Public seismology data archives are struggling to host the full datasets, research teams spend huge amounts of time and/or electricity to transfer the full datasets, and our ability to interactively visualize and rapidly understand these data is hindered by their large size. When data acquisition settings are selected appropriately (e.g. not oversampled in time or space), lossy compression techniques allow us to greatly reduce the size of the data without introducing the limitations of subsampling (e.g. array aliasing). However, lossy compression techniques are unable to exactly reconstruct our original array data at the same precision they were recorded at. While prior studies have characterized the errors introduced into the raw and reconstructed data for some types of lossy compression, we aim to understand the extent to which these errors propagate into meaningful changes in our characterization of seismic signals, including through template matching and event picking workflows. We quantitatively compare the changes in an open DAS dataset (the Bradys Hot Springs PoroTomo data) under varying levels of lossy compression using singular value decomposition (SVD) compression, wavelet-based compression, and zfp compression strategies. We show the quantitative tradeoffs in event catalogs resulting from template matching of data that has undergone substantial levels of compression. Further, we quantify the distribution of array-wide event pick time changes, which are shown to incur little to no average bias. We have released open-source, fully reproducible codes for comparing compression strategies for all workflows in this study to encourage others to use these tools in evaluating the suitability of lossy compression for sharing other datasets.

Session: Leveraging Cutting-Edge Cyberinfrastructure for Large Scale Data Analysis and Education - I

Type: Oral

Room: Tubughnenq’ 4

Date: 5/2/2024

Presentation Time: 05:15 PM (local time)

Presenting Author: Abdul Issah

Student Presenter: Yes

Additional Authors