juliensimon/substorm-onsets

Substorm Onset Events (SuperMAG)

Credit: NASA

Part of a dataset collection on Hugging Face.

Dataset description

Magnetospheric substorm onset events from five independent detection algorithms via the SuperMAG collaboration. This is the most comprehensive substorm event list available, enabling multi-algorithm comparison and consensus studies.

Magnetospheric substorms are fundamental space weather events driven by the solar wind's interaction with Earth's magnetic field. During a substorm, magnetic energy stored in the magnetotail is explosively released, accelerating charged particles that stream along field lines into the polar regions. This produces sudden auroral brightenings along with rapid changes in ground-level magnetic fields detected by magnetometer networks worldwide.

Ground-based methods detect substorms through characteristic negative bays in the SML (SuperMAG Lower) index -- a measure of the westward auroral electrojet current. Space-based methods directly observe the initial auroral brightening using ultraviolet imagers aboard the IMAGE and Polar satellites. Each algorithm has different sensitivity and false-positive rates, so researchers often require onset confirmation across multiple lists.

This dataset is suitable for tabular classification, time-series forecasting tasks.

Schema

Column	Type	Description	Sample	Null %
datetime_utc	datetime64[us]	UTC timestamp of the substorm onset -- the start of the expansion phase when energy stored in the magnetotail is explosively released; accurate to +/-1-2 minutes for ground-based detections, +/-1 minute for auroral imager detections	1975-01-04 07:24:00	0.0%
mlt_hours	float64	Magnetic Local Time (MLT) of the onset location (0-24 h, where 0/24 = magnetic midnight, 12 = magnetic noon); midnight-sector onsets (22-02 MLT) are most common	23.09	0.0%
magnetic_latitude_deg	float64	Magnetic latitude (MLAT) of the onset location in degrees; substorm onsets typically occur at 60-75 deg MLAT within the auroral oval	62.46	0.0%
geographic_longitude_deg	float64	Geographic (geodetic) longitude of the onset location in degrees (-180 to 180); suitable for plotting on standard world maps	246.69	0.0%
geographic_latitude_deg	float64	Geographic (geodetic) latitude of the onset location in degrees; use with geographic_longitude_deg for ground-track mapping	54.71	0.0%
source	str	Detection algorithm that identified this onset: newell (SML index threshold), forsyth (SME index derivative), ohtani (negative bay in SML), frey (IMAGE satellite UV imager), liou (Polar satellite UV imager)	ohtani	0.0%
method	str	Human-readable detection method category describing the instrument and technique used by each algorithm	SML bay detection (ground magnetometers)	0.0%

Quick stats

• 253,319 total substorm onset events
• 1975--2024 temporal coverage
• 5 independent detection algorithms
• 246,589 ground magnetometer detections, 6,730 auroral imaging detections
• By algorithm:
• forsyth: 120,069 events (SML/SMU expansion-recovery (ground magnetometers))
• newell: 81,914 events (SML index (ground magnetometers))
• ohtani: 44,606 events (SML bay detection (ground magnetometers))
• frey: 4,191 events (IMAGE/FUV auroral imaging (space-based))
• liou: 2,539 events (Polar UVI auroral imaging (space-based))

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/substorm-onsets", split="train")
df = ds.to_pandas()

from datasets import load_dataset

ds = load_dataset("juliensimon/substorm-onsets", split="train")
df = ds.to_pandas()

# Events per algorithm
print(df["source"].value_counts())

# Annual substorm rate by algorithm
import matplotlib.pyplot as plt
df["year"] = df["datetime_utc"].dt.year
df.groupby(["year", "source"]).size().unstack().plot(figsize=(12, 5))
plt.ylabel("Substorm onsets per year")
plt.title("Annual Substorm Rate by Detection Algorithm")
plt.show()

# MLT distribution — substorms peak near midnight
df["mlt_hours"].hist(bins=48, alpha=0.7)
plt.xlabel("Magnetic Local Time (hours)")
plt.ylabel("Count")
plt.title("Substorm Onset MLT Distribution")
plt.show()

# Find consensus events (multiple algorithms within 10 minutes)
from datetime import timedelta
newell = df[df["source"] == "newell"]["datetime_utc"]
ohtani = df[df["source"] == "ohtani"]["datetime_utc"]

Data source

https://supermag.jhuapl.edu/substorms/

Related datasets

• juliensimon/dst-index

• juliensimon/geomagnetic-kp-index

• juliensimon/auroral-electrojet-index

• juliensimon/donki-space-weather-events

• juliensimon/ae-index

• juliensimon/celestrak-sw

• juliensimon/donki

If you find this dataset useful, please consider giving it a like on Hugging Face. It helps others discover it.

About the author

Created by Julien Simon — AI Operating Partner at Fortino Capital. Part of the Space Datasets collection.

Citation

@dataset{substorm_onsets,
  title = {Substorm Onset Events (SuperMAG)},
  author = {juliensimon},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/substorm-onsets},
  publisher = {Hugging Face}
}

License

CC-BY-NC-4.0