juliensimon/sentry-impact-risk

NASA Sentry: Earth Impact Risk Assessment

Credit: NASA/Johns Hopkins APL

Part of a dataset collection on Hugging Face.

Dataset description

Near-Earth objects with non-zero Earth impact probability from NASA JPL Sentry system.

The Sentry system, operated by NASA's Center for Near-Earth Object Studies (CNEOS) at the Jet Propulsion Laboratory, continuously monitors the most current asteroid catalog for possibilities of future Earth impact. Objects are listed when their orbits bring them close enough that an impact cannot be ruled out.

Each record includes the cumulative and maximum Palermo Scale rating (a logarithmic measure comparing the impact probability to the background risk), the Torino Scale rating (an integer 0-10 for public communication), the number of potential impact scenarios, estimated diameter, and the year range of possible impacts.

The Sentry system works by propagating each asteroid's orbit forward in time while sampling the uncertainty region -- the cloud of possible positions and velocities consistent with the available observations. When any of these virtual asteroids pass close enough to Earth that an impact geometry exists, it registers as a potential impact event. The cumulative impact probability is the fraction of all sampled orbits that result in a collision, typically an extremely small number.

The Palermo Scale provides the key metric for assessing whether an impact scenario merits attention. It is defined as the logarithm of the ratio between the calculated impact probability and the background impact frequency for objects of similar energy. A Palermo Scale value of 0 means the event is exactly as likely as the average random impact risk; negative values (the vast majority) indicate risk below the background level. Only a Palermo Scale above -2 is generally considered noteworthy by the planetary defense community.

Understanding this dataset requires appreciating that the presence of an object on the Sentry list does not mean an impact is expected -- it means an impact cannot yet be ruled out. The most common outcome, by far, is that follow-up observations eliminate the risk entirely.

This dataset is suitable for tabular classification, tabular regression tasks.

Schema

Column	Type	Description	Sample	Null %
designation	str	Asteroid designation: either a permanent number (e.g. '29075') for numbered asteroids, or a provisional MPC designation (e.g. '2024 YR4') for unnumbered ones	29075	0.0%
palermo_scale_max	float64	Maximum Palermo Scale over any single impact solution in the monitored time window; less conservative than the cumulative value; useful for identifying the single most hazardous potential impact date	-0.95	0.0%
last_observation_jd	float64	Same as last_observation expressed as a Julian Date (TDB timescale)	2461318.5	0.0%
last_observation	datetime64[us]	UTC date of the most recent astrometric observation incorporated into the orbit solution; more recent observations typically reduce impact probability	2026-10-05 00:00:00	0.0%
palermo_scale_cum	float64	Cumulative Palermo Scale: logarithm of the ratio of cumulative impact probability to the background impact frequency for objects of equivalent energy; PS=0 equals background risk; PS>-2 merits attention by planetary defense community; dataset is sorted by this column descending	-0.95	0.0%
v_infinity_kms	float64	Hyperbolic excess speed (v_inf) at Earth encounter in km/s; combined with Earth's gravity well this yields the actual impact speed (typically 11-30 km/s); determines kinetic energy: higher v_inf means more destructive impact for the same mass	14.1	0.2%
diameter_km	float64	Estimated diameter in km derived from absolute magnitude assuming a geometric albedo of 0.154; null if H is unavailable; range in this table is typically 0.01-10 km	1.3	0.0%
impact_probability	float64	Total cumulative probability of at least one Earth impact across all monitored years, summed over all virtual impactor solutions; typically 10^-7 to 10^-1; the vast majority of entries are below 1 in 10,000	0.0003633	0.0%
full_name	str	Full formatted name combining number (if assigned) and provisional designation or name (e.g. '29075 (1950 DA)', '2024 YR4')	29075 (1950 DA)	0.0%
absolute_magnitude	float64	Absolute magnitude H; proxy for size: H=18 ~ 1 km, H=22 ~ 140 m, H=25 ~ 40 m; actual diameter depends on unknown albedo; null for some objects	17.94	0.0%
torino_scale	float64	Maximum Torino Scale integer (0-10) across all impact solutions; public communication metric combining probability and kinetic energy: 0 = no hazard, 1 = routine monitoring, 4-7 = threatening, 10 = certain global catastrophe; historically only a few objects have briefly exceeded 1	0.0	0.1%
n_potential_impacts	int64	Count of distinct impact geometries (virtual impactors) identified by Sentry's orbital sampling; a larger number reflects broader orbital uncertainty, not necessarily higher probability	1	0.0%
year_range_min	Int64	Earliest calendar year in which an impact solution exists; the first year where the object's uncertain orbit intersects Earth's orbit	2880	0.0%
year_range_max	Int64	Latest calendar year in which an impact solution exists; Sentry monitors orbits forward up to ~100 years	2880	0.0%

Quick stats

• 2,223 objects with non-zero impact probability
• Highest cumulative Palermo Scale: -0.95
• Maximum Torino Scale: 0
• Closest potential impact year: 2026
• Highest impact probability: 2010 RF12 (1.03e-01)

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/sentry-impact-risk", split="train")
df = ds.to_pandas()

from datasets import load_dataset

ds = load_dataset("juliensimon/sentry-impact-risk", split="train")
df = ds.to_pandas()

# Objects ranked by Palermo Scale (most hazardous first)
print(df[["designation", "palermo_scale_cum", "torino_scale",
          "impact_probability", "diameter_km"]].head(10))

# Objects with Torino Scale > 0
elevated = df[df["torino_scale"] > 0]
print(f"{len(elevated)} objects with elevated Torino Scale")

# Large objects (> 100m) with upcoming potential impacts
large = df[(df["diameter_km"] > 0.1) & (df["year_range_min"] <= 2050)]
print(large[["designation", "diameter_km", "year_range_min", "impact_probability"]])

Data source

https://cneos.jpl.nasa.gov/sentry/

Related datasets

• juliensimon/neo-close-approaches

• juliensimon/fireball-bolide-events

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{sentry_impact_risk,
  title = {NASA Sentry: Earth Impact Risk Assessment},
  author = {Simon, Julien},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/sentry-impact-risk},
  note = {Derived from NASA JPL Center for Near-Earth Object Studies (CNEOS), https://cneos.jpl.nasa.gov/sentry/},
  publisher = {Hugging Face}
}

License

CC-BY-4.0