Each dataset is made by two files: waveform (in HDF5 format) and metadata (in CSV format). All follow the structure of seisbench format. See here to learn more about the file structure.
We are hosting two copies of the dataset: one on Google Drive, another on UW-ESS server. All datasets are also available through SeisBench.
-
EH, BH, and HH channel (velocity)
-
EN (accelerometer)
- CSV (~93 MB): [GDrive]
Here are several ways to use the PNW dataset.
- Jupyter Notebook
A jupyter notebook is available to load and plot PNW dataset at here. Download and run it on a local machine to enable the interactive plotting (e.g., zoom in/out for checking the picks).
-
A notebook is available here on accessing the dataset with SeisBench APIs.
If you are more familiar with Google Colab, go to the link above. Note that interactive plotting is not available on Colab.
-
A micro version of the dataset, which contains 10 earthquake streams, 10 explosion streams, 10 sonic boom streams, 10 thunder streams, and 10 surface event streams. See data/microPNW.
-
A mini version of the dataset, which contains 500 earthquake streams, 500 explosion streams, 500 surface event streams, 126 sonic boom streams, and 94 thunder quake streams.
- waveform (640 MB): [Google Drive] | [UW-ESS]
- metadata (424 KB): [Google Drive] | [UW-ESS]
- A meso version with 10% of the full ComCat dataset (only earthquake + explosion).
- waveform (6.3 GB): [Google Drive] | [UW-ESS]
- metadata (4.7 MB): [Google Drive] | [UW-ESS]
Attribute | Description | Example |
---|---|---|
event_id | Event identifier | uw10564613 |
source_origin_time | Source origin time in UTC | 2002-10-03T01:56:49.530000Z |
source_latitude_deg | - | 48.553 |
source_longitude_deg | - | -122.52 |
source_type | - | earthquake |
source_type_pnsn_label | PNSN AQMS event type | eq |
source_depth_km | - | 14.907 |
source_magnitude_preferred | - | 2.1 |
source_magnitude_type_preferred | - | Md |
source_magnitude_uncertainty_preferred | - | 0.03 |
source_local/duration/hand_magnitude | Ml, Md, and Mh if available | 1.32 |
source_local/duration_magnitude_uncertainty | magnitude uncertainty if available | 0.15 |
source_depth_uncertainty_km | - | 1.69 |
source_horizontal_uncertainty_km | - | 0.694 |
station_network_code | FDSN network code | UW |
station_code | FDSN station code | GNW |
station_location_code | FDSN location code | 01 |
station_channel_code | FDSN channel code (first two digits) | BH |
station_latitude_deg | - | 47.5641 |
station_longitude_deg | - | -122.825 |
station_elevation_m | - | 220.0 |
trace_name | Bucket and array index | bucket1$0,:3:15001 |
trace_sampling_rate_hz | All traces resampled to 100 Hz | 100 |
trace_start_time | Trace start time in UTC | 2002-10-03T01:55:59.530000Z |
trace_P/S_arrival_sample | Closest sample index of arrival | 8097 |
trace_P/S_arrival_uncertainty_s | Picking uncertainty in second | 0.02 |
trace_P/S_onset | - | emergent |
trace_P_polarity | P-wave onset polarity | positive, negative, or undecidable |
trace_has_offset | Any visible offset in the trace | 1 |
trace_missing_channel | Number of missing channel of the trace | 2 |
trace_snr_db | SNR for each component | 6.135 |
Ni, Y., Hutko, A., Skene, F., Denolle, M., Malone, S., Bodin, P., Hartog, R., & Wright, A. (2023). Curated Pacific Northwest AI-ready Seismic Dataset. Seismica, 2(1). https://doi.org/10.26443/seismica.v2i1.368
BiBTex:
@article{ni2023pnw,
title={Curated Pacific Northwest AI-ready Seismic Dataset},
volume={2},
url={https://seismica.library.mcgill.ca/article/view/368},
number={1},
journal={Seismica},
author={Ni, Yiyu and Hutko, Alexander and Skene, Francesca and Denolle, Marine and Malone, Stephen and Bodin, Paul and Hartog, Renate and Wright, Amy},
year={2023},
month={05},
doi={10.26443/seismica.v2i1.368}
}
- [August 2023] Very few events (~15) in the ComCat dataset may have inconsistent
event_type_pnsn_label
andevent_type
. This issue comes from the outdated ComCat event metadata. Please prioritize PNSN label when such inconsistent occurs. - [June 2025] The
trace_start_time
field in the exotic metadata was delayed by 50 seconds. The metadata has now been corrected for all affected files.
If you find any issue in the dataset, please report through GitHub Issue or Email.