juliensimon/cdaw-lasco-cme-catalog

SOHO/LASCO Coronal Mass Ejection Catalog (CDAW)

Credit: NASA/SDO

Part of a dataset collection on Hugging Face.

Dataset description

The SOHO/LASCO CME Catalog from the NASA CDAW Data Center — every coronal mass ejection manually identified in the SOHO Large Angle and Spectrometric Coronagraph (LASCO) C2 and C3 white-light coronagraph images from January 1996 through the current data release.

Coronal mass ejections are the largest energy-release events in the heliosphere: ~10^16 g of coronal plasma launched at 50-3000 km/s, carrying enough magnetic energy to drive the strongest geomagnetic storms when directed at Earth. The CDAW catalog is the canonical human-curated CME record used for solar-cycle statistics, space-weather forecasting model training, and event-correlation studies with in-situ measurements at L1 (e.g. OMNI, DSCOVR) and at planetary missions. Each entry records the CME first-appearance time in the LASCO C2 field of view, position angle and angular width on the plane of sky, multiple speed estimates (linear and second-order fits), acceleration, estimated mass and kinetic energy, and curator-added remarks flagging data quality and instrument coverage.

Halo CMEs (angular_width = 360 degrees) are the operationally critical subset: their full-360 appearance in coronagraph images indicates the CME is launched along the Sun-spacecraft line, either Earth-directed (driving terrestrial storms) or anti-Earth-directed (back-side events detectable only at L1 or planetary spacecraft). The is_halo column flags these directly. The is_poor_event flag distills the most common quality concern from the remarks field for filtering.

This dataset complements juliensimon/donki-space-weather-events (NASA DONKI's curated CME catalog, used by NOAA SWPC), juliensimon/space-weather-indices (Kp/Ap/F10.7 indices for geo-effective response), juliensimon/omni-solar-wind-parameters (in-situ solar wind at L1 for CME arrival validation), juliensimon/dst-index (Dst geomagnetic storm intensity), juliensimon/solar-flares (the X-ray events typically associated with CME launches), and juliensimon/parker-solar-probe-encounters (PSP in-situ CME observations).

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

Schema

Column	Type	Description	Sample	Null %
date_utc	object	First-appearance date of the CME in the LASCO C2 field of view (UTC)	1996-01-11	0.0%
time_utc	object	First-appearance time (UTC, HH:MM:SS); together with date_utc this is the closest available proxy for CME launch time	00:14:36	0.0%
datetime_utc	datetime64[us, UTC]	Combined first-appearance datetime in UTC (date_utc + time_utc)	1996-01-11 00:14:36+00:00	0.0%
year	int64	Calendar year of first appearance (integer, derived from datetime_utc)	1996	0.0%
central_position_angle_deg	float64	Central position angle (degrees, 0-360); measured counterclockwise from solar north, marking the angular center of the CME front; null for halo CMEs (full-360 events)	267.0	2.4%
angular_width_deg	float64	Angular width of the CME in the plane of sky (degrees); width = 360 marks a 'halo CME', which are the most geo-effective and often Earth-directed	18.0	0.0%
linear_speed_kms	float64	Linear-fit projected speed in the plane of sky (km/s); CMEs span ~50 to >3000 km/s, with the fastest events driving strongest geomagnetic storms	499.0	0.3%
second_order_speed_initial_kms	float64	Speed (km/s) from a quadratic time-distance fit, evaluated at first appearance	571.0	1.7%
second_order_speed_final_kms	float64	Speed (km/s) from a quadratic fit at the last observed measurement	426.0	1.7%
second_order_speed_20rs_kms	float64	Speed (km/s) from a quadratic fit evaluated at 20 solar radii — closer to the constant-speed asymptote in the outer corona	0.0	1.7%
acceleration_ms2	float64	Plane-of-sky acceleration (m/s^2) from the quadratic fit; positive = accelerating outward, negative = decelerating	-64.3	1.7%
mass_g	float64	Estimated CME mass in grams; an asterisk in the source file (now stripped) marked a poor mass measurement; null where mass is not measured	71000000000000.0	44.2%
mass_quality_flag	bool	True if the source file marked the mass measurement as poor quality (asterisk in the raw catalog)	False	0.0%
kinetic_energy_erg	float64	Estimated CME kinetic energy in erg = mass * (linear_speed)^2 / 2; null where mass is not measured	2.5e+28	44.3%
kinetic_energy_quality_flag	bool	True if the source file marked the kinetic energy as poor quality	False	0.0%
remarks	str	Free-text annotation: data-quality notes ('Poor Event', 'Very Poor Event'), instrument coverage ('Only C2', 'Only C3'), and event-curation notes ('Newly inserted on YYYY/MM/DD')	272 Only C3	0.0%
is_halo	bool	True if the CME is a halo event (width = 360 deg) — most likely to be Earth-directed and geo-effective	False	0.0%
is_poor_event	bool	True if the remarks contain 'Poor Event' or 'Very Poor Event' — measurements should be treated with caution	False	0.0%

Quick stats

• 42,744 coronal mass ejections manually identified in SOHO/LASCO C2 and C3 images
• 1,015 halo CMEs (angular width = 360 deg) — the geo-effective Earth-directed subset
• Median linear speed: 319 km/s | fastest recorded: 3387 km/s
• Catalog spans 31 years across two solar cycles and counting (1996+)
• 27,068 events flagged by curators as 'Poor Event' or 'Very Poor Event' — exclude from statistical analyses

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/cdaw-lasco-cme-catalog", split="train")
df = ds.to_pandas()

from datasets import load_dataset
import matplotlib.pyplot as plt

df = load_dataset("juliensimon/cdaw-lasco-cme-catalog", split="train").to_pandas()
df["datetime_utc"] = df["datetime_utc"].astype("datetime64[ns, UTC]")

# Annual CME counts traces the solar cycle (peak ~2000 and ~2014)
yearly = df.groupby("year").size()
halo_yearly = df[df["is_halo"]].groupby("year").size()
fig, ax = plt.subplots(figsize=(11, 5))
yearly.plot(ax=ax, label="All CMEs")
halo_yearly.plot(ax=ax, label="Halo CMEs", color="red")
ax.set_ylabel("CMEs per year")
ax.set_title("SOHO/LASCO CME rate vs solar cycle (1996+)")
ax.legend()
plt.tight_layout()
plt.show()

# Fastest 10 halo CMEs — the operationally most significant events
fast_halo = (df[df["is_halo"] & ~df["is_poor_event"]]
             .nlargest(10, "linear_speed_kms")
             [["datetime_utc", "linear_speed_kms", "angular_width_deg", "remarks"]])
print(fast_halo)

Data source

https://cdaw.gsfc.nasa.gov/CME_list/UNIVERSAL/text_ver/univ_all.txt

Update schedule

Weekly

Related datasets

• juliensimon/donki-space-weather-events

• juliensimon/space-weather-indices

• juliensimon/omni-solar-wind-parameters

• juliensimon/dst-index

• juliensimon/solar-flare-events

• juliensimon/parker-solar-probe-encounters

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{cdaw_lasco_cme_catalog,
  title = {SOHO/LASCO Coronal Mass Ejection Catalog (CDAW)},
  author = {juliensimon},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/cdaw-lasco-cme-catalog},
  publisher = {Hugging Face}
}

License

CC-BY-4.0