Listening for events…
← All Docs

Feasibility — Starlink decay as a thermosphere/solar sensor

Checked 2026-06-09 against live PG16. Question (Mike): can the Starlink reentry watch (#105) be repurposed to measure the upper atmosphere expanding and contracting under solar forcing?

Verdict

GREEN for a short-window "storm pilot" now. The solar-cycle and CO₂-climate versions are NOT feasible yet — they need a multi-year baseline we don't have.

The physics is established (satellite drag has been used as a thermospheric density proxy for decades). Our novelty would be fleet-scale N and the standing-monitor framing, not new physics. Honest framing required.

What we actually have (verified)

  • celestrak source is live, 2.66M rows, but history only spans 2026-03-17 → present (~3 months) — the fetcher started in March. It stores per-object orbital elements in extra_json: norad_id, inclination_deg, eccentricity, mean_motion. (extra_json is TEXT, cast ::jsonb to query.)
  • mean_motion is the signal we need. It's revolutions/day; it rises as a satellite's orbit decays. The rate of rise, d(mean_motion)/dt, is the drag = thermospheric-density proxy. No extra derivation needed.
  • Starlink fleet is densely tracked: 10,591 distinct Starlinks; 10,391 sampled on ≥20 distinct days; avg ~53 days/object over the 83-day window. Per-object decay-rate time series are clean and abundant.
  • ~3,793 Starlinks are clearly decaying (mean_motion rose >0.01 rev/day over the window); 105k snapshots sit at low altitude (mean_motion > 15.5). These non-maneuvering / deorbiting objects are the clean-drag population.
  • A real geomagnetic storm sits in the window: Dst hit −101 nT on 2026-03-22 (moderate–strong), with 2,215 storm-level hours (Dst < −50). That is a built-in natural experiment, 5 days after the celestrak history begins.
  • Solar/geomagnetic drivers cover the window:
    • dst_index 2026-03-03 → 2026-05-27 (storm intensity; ~2-week freshness lag)
    • dscovr_solar_wind 2025-10 → present (driver)
    • silso_sunspots 1817 → 2026-05-30 (F10.7/EUV proxy — the "puffing" driver)
    • goes_xray 2026-03-17 → present (flares)
  • We do NOT ingest F10.7 or Kp directly. Not blocking: SILSO sunspots proxy F10.7, and Dst proxies Kp. Adding F10.7/Kp would tighten the driver side but is optional for the pilot.

The feasible pilot

"Does the Starlink fleet's orbital decay rate respond to the March 2026 geomagnetic storm?" — a superposed-epoch / before-after test of fleet-wide d(mean_motion)/dt around 2026-03-22, plus a daily regression of decay rate vs Dst / solar wind / sunspots across the 3-month window.

  • N: thousands of satellites × ~daily cadence. Power is not the constraint.
  • Expected effect is large: the Feb 2022 event (a storm puffing the air and dragging ~38 fresh Starlinks to reentry) shows storm-scale drag changes are tens of percent. A Dst −101 storm should be clearly detectable.

The load-bearing confound

Operational Starlinks actively station-keep (ion thrusters raise orbits), which would mask or invert the drag signal. Mitigations:

  1. Restrict to the ~3,793 monotonically-decaying objects (non-maneuvering / end-of-life). This is the primary defense.
  2. Lean on the storm transient: drag from a storm is a sharp multi-day impulse; station-keeping is slow and roughly continuous, so a superposed-epoch on the storm separates them.
  3. Sensitivity: repeat on quiet-period controls (no storm) — decay rate should flatten.

What's out of reach now (and why)

  • Solar-cycle (11-year) version: needs years of TLE history; we have 3 months. Requires a Space-Track / CelesTrak historical GP-archive backfill (external, large) — or just waiting, since this is a standing monitor.
  • CO₂ secular thermospheric contraction (the real climate signal): decadal; far beyond Starlink's 2019-onward existence. Dream-tier only.

Connection to the taxonomy

This is the open / standing-monitor quadrant the paper taxonomy flagged as empty. The celestrak archive accrues every day, so the pilot ships now on the March storm, and the same monitor grows into the solar-cycle version over time. It both breaks the snapshot-only streak and reframes #105 from "watch for reentries" to "watch the upper atmosphere breathe."

Recommended next step

Scope the storm pilot (power check is effectively passed; the real design work is the de-maneuvering filter and the superposed-epoch protocol). Hold the solar-cycle/climate version as a dream-paper seed gated on a TLE history backfill or accrued monitor time.

Live Feed