juliensimon/starlink-fleet-data

Starlink Fleet Data

Part of the Orbital Mechanics Datasets collection on Hugging Face.

Daily health snapshots of the SpaceX Starlink constellation, derived from CelesTrak GP (General Perturbations) data. Currently tracking 10,119 satellites (8,738 operational, 1,336 raising, 44 deorbiting, 9,141 ISL-capable).

Dataset description

This dataset tracks the Starlink constellation's health over time through daily aggregate snapshots per orbital shell. Each day, the pipeline fetches the latest TLE/GP data from CelesTrak, classifies every satellite's operational status using orbital mechanics heuristics, and appends per-shell summary statistics to a growing time series.

The daily_snapshots config contains historical daily aggregates — one row per shell per day — enabling trend analysis of constellation growth, shell fill rates, deployment cadence, and ISL (inter-satellite laser link) rollout.

Starlink is the largest satellite constellation ever built, representing a fundamental shift in how broadband internet is delivered globally. SpaceX deploys satellites into five distinct orbital shells, each defined by its inclination and target altitude, to provide overlapping coverage from the equator to high latitudes. The constellation operates in low Earth orbit (LEO) at altitudes between 328 km and 570 km, where orbital periods of roughly 90 minutes mean each satellite circles the Earth about 16 times per day. This low altitude reduces signal latency to 20-40 ms round-trip -- competitive with terrestrial fiber over long distances -- but also means satellites experience significant atmospheric drag and must periodically reboost or be replaced.

Understanding constellation health requires tracking each satellite through its lifecycle: initial deployment to a parking orbit, orbit raising via Hall-effect ion thrusters, operational service at target altitude, and eventual controlled deorbit. The mean motion derivative (the rate at which orbital period changes) serves as a reliable proxy for whether a satellite is actively thrusting upward or decaying. Satellites with anomalous eccentricity may indicate propulsion failures or collision avoidance maneuvers. The inter-satellite laser link (ISL) capability, rolled out starting in 2022, enables direct satellite-to-satellite routing without ground relay, dramatically improving service in oceanic and polar regions.

This dataset is particularly valuable for space situational awareness research, orbital debris modeling, constellation economics analysis, and radio frequency interference studies. The daily time series format enables researchers to measure deployment cadence, correlate shell fill rates with SpaceX launch schedules, track the fleet's deorbit rate as older v1.0 satellites age out, and monitor the ISL rollout's progress toward full mesh networking capability.

Config: daily_snapshots

Historical daily aggregates per orbital shell. Currently 7,255 rows spanning 2019-05-25 to 2026-04-04.

Column	Type	Description
date	datetime	Snapshot date (UTC)
shell_id	int	Shell identifier (0-4)
shell_name	string	Human-readable shell name, e.g. "Shell 3 (53deg / 550km)"
total_count	int	Total satellites in this shell
operational_count	int	Satellites at operational altitude
raising_count	int	Satellites actively raising orbit
deorbiting_count	int	Satellites deorbiting
isl_operational_count	int	ISL-capable satellites in this shell
new_launches	int	New launches detected this day (reserved)

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/starlink-fleet-data", "daily_snapshots", split="train")
df = ds.to_pandas()

# Constellation growth over time
growth = df.groupby("date")["operational_count"].sum()
print(growth.tail(10))

# Per-shell fill rates
latest = df[df["date"] == df["date"].max()]
for _, row in latest.iterrows():
    print(f"{row['shell_name']}: {row['operational_count']} / {row['total_count']}")

Status classification

Each satellite is classified into one of six statuses based on its orbital parameters and a single-snapshot heuristic:

Status	Criteria
operational	Altitude within the shell's operational band
raising	Below or above operational band, orbit changing toward target
deorbiting	Below operational band with strong positive mean motion derivative (orbit shrinking) or very low altitude (<300 km)
decayed	Altitude below 150 km, or stale epoch (>14 days) with altitude below 250 km
anomalous	Eccentricity > 0.005 (unusual for Starlink's near-circular orbits)
unknown	Does not match any classification rule

The classifier uses mean_motion_dot (first derivative of mean motion) as a proxy for orbital maneuvering direction: positive values indicate orbit decay (shrinking), negative values indicate active thrust (raising).

Shell assignment

Satellites are assigned to shells by inclination:

Shell	Inclination	Target altitude	Inclination range
Shell 1 (33deg)	33.0deg	328 km	< 38deg
Shell 2 (43deg)	43.0deg	340 km	38deg - 48deg
Shell 3 (53deg)	53.0deg	550 km	48deg - 60deg
Shell 4 (70deg)	70.0deg	570 km	60deg - 80deg
Shell 5 (97.6deg)	97.6deg	560 km	> 80deg

Update frequency

Updated daily at 08:00 UTC via GitHub Actions. Each run fetches the current CelesTrak GP snapshot, classifies all satellites, and appends the day's per-shell aggregates to daily_snapshots.parquet. Re-runs on the same day are idempotent (existing rows for that date are replaced).

Data source

All orbital data comes from CelesTrak GP data (NORAD/18th Space Defense Squadron). Only objects with names matching STARLINK-* are included (Starshield, debris, and unidentified objects are excluded).

Related datasets

• space-track-tle-history — 232M historical TLEs (1959-present)
• space-track-satcat — NORAD satellite catalog
• space-launch-log — Global launch history from GCAT
• starlink-ground-stations — Starlink gateway and PoP locations

See it in action

This dataset powers the fleet dashboard in Starlink Viz — constellation growth, shell fill rates, ISL coverage, and more.

Pipeline

Source code: juliensimon/space-datasets

Support

If you find this dataset useful, please give it a ❤️ on the dataset page and share feedback in the Community tab! Also consider giving a ⭐️ to the space-datasets repo.

Citation

@dataset{starlink_fleet_data,
  author = {Simon, Julien},
  title = {Starlink Fleet Data: Daily Constellation Health Snapshots},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/juliensimon/starlink-fleet-data},
  note = {Based on NORAD/18th Space Defense Squadron GP data via CelesTrak (Dr. T.S. Kelso)}
}