Datasetpublic

erzanugroho/seismicid-earthquake-forecast-dataset

SeismicID Earthquake Forecast Dataset Dataset for SeismicID, Indonesian earthquake risk forecasting and visualization. Live app: https://seismicid.erzanugroho.xyzSource code: https://github.com/erzanugroho/SeismicID Important notice This dataset and model outputs are experimental research artifacts. They are not earthquake early warning, official hazard advice, or replacement for BMKG / official authorities. Files file rows description… See the full description on the dataset page: https://huggingface.co/datasets/erzanugroho/seismicid-earthquake-forecast-dataset.

sourceHugging Facecc-by-4.0updated 4mo agoView on Hugging Face

0likes33downloads

Dataset Card

SeismicID Earthquake Forecast Dataset

Dataset for SeismicID, Indonesian earthquake risk forecasting and visualization.

Live app: https://seismicid.erzanugroho.xyz Source code: https://github.com/erzanugroho/SeismicID

Important notice

This dataset and model outputs are experimental research artifacts. They are not earthquake early warning, official hazard advice, or replacement for BMKG / official authorities.

Files

file	rows	description
`data/area_labels.parquet`	3128	Grid cell labels and region metadata.
`data/current_forecasts.parquet`	50048	Current forecast probabilities per cell/horizon/magnitude threshold.
`data/evaluation_results.parquet`	5	Stored model evaluation metrics.
`data/forecast_archive/part-*.parquet`	1410728	Sharded archived forecast snapshots. Includes `forecast_date` and `run_id` columns.
`data/historical_events.parquet`	58676	Cleaned historical earthquake catalog used by SeismicID.
`data/historical_events_sample.csv`		Small CSV preview of historical_events.parquet.
`data/model_metadata.parquet`	4	Model version metadata.
`data/realtime_events.parquet`	1424	Recent ingested earthquake events stored by runtime app.

Main columns

`historical_events.parquet`

—event_id: source-prefixed event identifier
—time: UTC event time
—lat, lon: epicenter coordinates
—depth: depth in km where available
—magnitude: earthquake magnitude
—mag_type: source magnitude type
—source: catalog source, e.g. USGS/BMKG/EMSC depending on ingest
—place: source place text

`forecast_archive.parquet` / `current_forecasts.parquet`

Forecast columns use this pattern:

text

label_h{horizon_days}_m{magnitude_threshold_without_decimal}

Examples:

—label_h7_m45: probability for M≥4.5 within 7 days
—label_h30_m50: probability for M≥5.0 within 30 days
—label_h60_m60: probability for M≥6.0 within 60 days

Loading example

python

from datasets import load_dataset

ds = load_dataset("parquet", data_files={
    "historical_events": "data/historical_events.parquet",
    "current_forecasts": "data/current_forecasts.parquet",
    "forecast_archive": "data/forecast_archive/part-*.parquet",
})
print(ds)

Or with pandas:

python

import pandas as pd

events = pd.read_parquet("data/historical_events.parquet")
forecasts = pd.read_parquet("data/current_forecasts.parquet")

Citation

If you use this dataset, please cite the SeismicID repository:

bibtex

@misc{seismicid,
  title = {SeismicID: Indonesian Earthquake Risk Forecasting},
  author = {Erza Nugroho},
  year = {2026},
  url = {https://github.com/erzanugroho/SeismicID}
}

License

Recommended dataset license: CC-BY-4.0, assuming upstream source terms allow redistribution with attribution. Verify upstream catalog licensing before public release.