Presenting DASPack at IAGA/IASPEI 2025
I’m in Lisbon for the IAGA / IASPEI Joint Scientific Meeting, presenting our work on DASPack, a high-performance, open-source compression library purpose-built for distributed acoustic sensing (DAS) data.
This post includes the poster and a photo of the presentation.
Why compression matters for DAS
DAS enables dense, long-range, and high-rate measurements along optical fibers, powering applications in geophysics, infrastructure monitoring, and environmental sensing. But the data "tsunami" is real: large deployments already produce hundreds of terabytes. Storage, transfer, and real-time processing become bottlenecks without domain-aware compression.
What is DASPack?
DASPack targets DAS signals directly and supports both lossless and controlled lossy compression. The pipeline combines wavelet transforms, linear predictive coding, and entropy coding to maximize compression while preserving signal fidelity in the regimes that matter.
- High-performance, designed for DAS
- Lossless or controlled lossy (max absolute error bound)
- Real-time throughput on modern hardware
Results at a glance
Below is a simplified snapshot from our evaluation across diverse datasets. In practice we observe strong, consistent gains while keeping DAS-relevant features intact.
| Mode | Compression | Fidelity |
|---|---|---|
| Lossless | ≈ 3× | Bit-perfect |
| Lossy | ≈ 6× | Near-perfect |
| Lossy (aggressive) | ≈ 10× | Acceptable for some tasks |
Throughput reaches ~100–200 MB/s single-threaded and ~750 MB/s on 8 threads, making real-time compression feasible even for high-rate acquisitions.
Learn more
If you’re working with DAS and dealing with storage/transfer at scale, we’d love your feedback. The preprint includes methodology and a broader benchmark suite, and the repository contains CLI and library usage examples.
Thanks to collaborators and the community for the insightful discussions in Lisbon. Lots of great ideas for next steps.