pysatellitespotter

Predict when satellites are visible to the naked eye, the way See A Satellite Tonight does. This is a Python port of James Darpinian's site, made with his permission. It powers the Home Assistant Satellite Spotter integration.

This is a fork of Willem Vooijs's see-a-satellite, with fixes for crashes on valid catalog data and on passes past the prediction window (wildekek/see-a-satellite#2). It installs as its own module, pysatellitespotter, so it can sit beside see-a-satellite in one environment, as it does when Home Assistant runs both integrations.

pip install pysatellitespotter

import aiohttp
from pysatellitespotter import fetch_site_catalog, group_transits, predict_transits

async with aiohttp.ClientSession() as session:
    satellites = await fetch_site_catalog(session)

# CPU bound (a few seconds): run it in an executor from async code.
for group in group_transits(predict_transits(satellites, 52.03, 5.56, 10)):
    t = group.primary
    print(group.start, t.satellite.name, f"mag {group.min_magnitude:.1f}",
          f"rises {t.rise.azimuth:.0f}°/{t.rise.elevation:.0f}°", "bright" if group.bright else "")

What it computes

The pipeline follows the website step by step:

• Propagation. SGP4 runs from OMM elements at one-minute steps for 5 days, and positions are rotated from the TEME frame to Earth-fixed coordinates.
• Brightness. The intrinsic magnitude at 1000 km is adjusted for range, atmospheric extinction (Green 1992) and phase angle (diffuse sphere). Starlink and BlueBird satellites use their own phase-function models (Mallama et al. 2025; Cole et al. 2025).
• Visibility. A satellite counts as visible when it is at least 10° up, brighter than the limiting magnitude for the sun's depression (Weaver 1947), and lit by the sun above the troposphere (grazing height ≥ 15 km).
• Refinement and grouping. Each pass is refined to one second. Overlapping passes are grouped, incidental Starlinks are pruned, and a "bright" label is applied.

Checked against the website, for every pass the site shows over 5 days:

• start times agree within 9 s
• end times agree within 19 s
• magnitudes agree within 0.03

A 5-day prediction for the website's catalog of about 860 satellites takes 0.8 s on one core of an Apple M5. Minutes when the sky is too bright for a satellite to be seen are skipped, so most of the time goes to SGP4 for the twilight and night hours.

Catalog

fetch_site_catalog(session, "brightest" | "starlink") downloads the website's catalog, which the site's author collects and curates. Orbital elements come mostly from Space-Track (US Space Force 18th Space Defense Squadron), and Starlink elements come from CelesTrak. Magnitudes come from Mike McCants and the MMT-9 survey (Karpov et al. 2016), and constellation phase functions come from published photometry. See James's write-up. This library only consumes the catalog. The site refreshes it every 8 hours. Cache it (dump_catalog / parse_site_catalog) rather than downloading it on every prediction.

Development

uv sync
uv run pytest && uv run mypy && uv run ruff check src tests