# Study 02 — Rocket-Launch Ionospheric Holes: the sky keeps a launch log

A rocket's second stage burns through the one layer of the sky that carries long-range radio. On the night of 2023-07-19 (04:09 UTC July 20), a Falcon 9 out of Vandenberg lit its second stage at 286 km altitude — right at the F-region peak — and its water and carbon-dioxide exhaust quenched the local ionization by up to 70%. People from Southern California to Arizona photographed the wound: a red 630.0 nm glow, visible for ~20 minutes, oxygen ions recombining where the plasma used to be (Jeff Baumgardner, Boston Univ., via spaceweatherarchive.com 2023-07-23 — VERIFIED by fetch).

That is the shape this study grades. A hole ~900 km across, punched along a published track, opening 5–15 minutes after a liftoff second the operator announces in advance, and healing by plasma diffusion in 1–3 hours (Falcon-9-class F-region holes; the Saturn-V-class founding event ran ~4 h). Around it, on the biggest events, a ring: circular shock-acoustic waves expanding at 629–726 m/s (FORMOSAT-5, 2017 — the largest on record).

The sky writes geometry, and geometry cannot lie. The eclipse study (Study 1 — see [Eclipse 2026 — Overview](Eclipse-2026-Overview.md)) proved the method on a Moon-driven clock: a frozen exact-integer law, an adversary that mimics magnitude but not shape, predictions sealed 27 days ahead. This study runs the same shear recipe on a forcing with a *weekly* clock instead of a yearly one — the fastest forcing cadence in the program.

## This morning — Starlink Group 10-49 (DATA)

Fetched **2026-08-25T20:33:34Z**. The morning launch **is** Starlink Group 10-49. Public alarm — why the heatmap is the wrong court, and where the package goes: [Study 02 — Regulatory alarm](Study-02-Regulatory-Alarm).

| Field | Measured this hour | Source |
|---|---|---|
| Mission | **Falcon 9 Block 5 \| Starlink Group 10-49** | Launch Library 2 `ad501ebd-a357-4305-90c1-c46c6feb2ecf` HTTP **200** |
| Status | Launch Successful | LL2 `status.id=3` |
| Liftoff UTC | **2026-08-25T09:33:38Z** (second precision) | LL2 `net` (yesterday's NET was `09:33:17Z`; actual is 21 s later — inside the sealed 10-minute VOID-by-slip tolerance) |
| Local | 05:33 EDT / 09:33 UTC | SpaceX `https://www.spacex.com/launches/sl-10-49` HTTP **200**; Spaceflight Now; Space.com |
| Pad | **SLC-40**, Cape Canaveral SFS | LL2 `28.56194122, −80.57735736` → **28561941 / −80577357 µ°** |
| Payload | 29 Starlink · booster B1067 flight 37 · ASOG droneship | SpaceX + Next Spaceflight |
| Jellyfish / twilight plume | Photographed FL → NY / DC / NC / GA–SC | Florida Today 2026-08-25; Capital Weather (DC); WXII; WJCL. Visual = sunlit H₂O/CO₂ exhaust while the ground is still dark. That is the F-region quench chemistry (H₂O/CO₂ + O⁺), not a TEC court. |
| NOAA SWPC nowcast (`-1` / `0`) | Feed only. Current `0` DateStamp **2026-08-25** 20:32:00 R=0 S=0 G=0. Yesterday `-1` R=2. | `https://services.swpc.noaa.gov/products/noaa-scales.json` HTTP **200**. Quiet nowcast does not void a launch hole. Storm gate remains 3×Kp at t0. |

Yesterday's freeze (`2026-08-24T12:19:23Z`) named this same NET. The weekly clock this study grades **fired**.

### The heatmap is the old court

Standard viz paints a 10–70% TEC drop as a continuous color gradient. Smoothing the depletion boundary invents spatial uncertainty the raw phases never contained. Subtract that gradient from a cosmic shear baseline and the leftover is later named as noise or as lensing.

**Affine seals integer L1/L2, not a float heatmap.** The 2023 charter 70% / 900 km numbers stay **cited founding-event integers**. They are not this morning's measured hole. Health remains PK dose 70/100. Fluids remain Φ. Physics remains clock / track / table_counts.

Mass: ***not known***. The lockout **09:35–11:30 UTC** is a precautionary charter so a plasma hole is not read as cosmological shear. Study 07 (Sgr A\* / EHT, MJD 57849–57850 = 2017 April 6–7) is the cosmic baseline the window quarantines **against**. **No intersection measured** with East Coast 09:35–10:15 UTC this morning — no in-tree EHT/Euclid visibility table intersects that window.

Phase-delay strings route through the Swift 6.4 membrane already on the nine cells.

| Quantity | Grade this hour |
|---|---|
| Spatial diameter ~900 km | **Cited founding scale / not measured.** Madrigal world TEC (kindat 3500, instrument 8000) for 2026-08-25 returned HTTP **200** and **0 bytes**. Control: 2026-08-12 experiment list HTTP **200**, 702 bytes (eclipse-day files exist). |
| Cylinder 200–300 km · Cape → coastal Carolinas | **Cited.** Northeast track from Spaceflight Now. CORS this hour is Miami (FLF1) → COLB / NCLL → HNPT → NYST / NYBP → WES2. Cape pad hourly `ccv6` / `cccs` / `ksc1` HTTP **404**. |
| TEC drop threshold 1/10 … 7/10 | **Not applied.** No Madrigal TEC numerators for this window. |
| Window lockout | Presented integers: `window_start=935` `window_end=1130` `constrained=1` (HHMM). Precautionary charter — not a measured TEC void. |
| IGR Rapid orbit | **WAIT** until **~2026-08-26T17:00Z** (17–24 h latency from 09:33 UTC). Not ingested today. Second increment calendared then. |

### CORS RINEX — LIVE integer ingest (09–11 UTC)

Public path measured: `https://geodesy.noaa.gov/corsdata/rinex/2026/237/{ssss}/{ssss}237{h}.26o.gz` (h = j/k/l = 09/10/11 UTC). NGS published **RINEX 2.11**, not RINEX 3, on this index. Hourly files exist (index HTTP **200**). Named Eastern Seaboard list only — no brute scan.

**7 stations × 3 hours = 21 files. First-epoch L1/L2 as integer milli-cycles. `REFUSED_FLOAT` = 0.**

Puller: `scripts/study-02-cors-integer-phase.sh` — decimal-string shift, presented first epoch only.

| Site | Where (verified) | Hour j L1 / L2 milli-cycles | Hour k | Hour l |
|---|---|---|---|---|
| FLF1 | Miami / Richmond FL (site log) | 108457668731 / 84512448562 | 110639339950 / 86212452296 | 117983716024 / 91935337954 |
| COLB | Columbus, Ohio (NGS coord_20 HTTP 200; inland — not the Cape→NY track) | 132098480845 / 102933980505 | 131430874164 / 98146481686 | 110762098322 / 86308080798 |
| NCLL | North Carolina (ECEF) | 131465531640 / 102440817370 | 132158508246 / 102980798433 | 110587326128 / 86172031537 |
| HNPT | Maryland Eastern Shore (IGS 38.59N, 76.13W) | 105945820596 / 82555189889 | 110854825817 / 86380387239 | 121187288685 / 94431648559 |
| NYST | Saratoga NY | 24308663980 / 127742905757 | 24250055940 / 127434911542 | 21580256820 / 113405040707 |
| NYBP | New York | 119544768442 / 93151769331 | 123182283919 / 95808427016 | 121862410698 / 94781855367 |
| WES2 | Westford MA / Haystack | 120769576147 / 94106077730 | 110061794331 / 85762355177 | 105536335394 / 82236022930 |

P011 hourly files exist but the header is **Arizona** (36.15N, 109.52W). Excluded from the East Coast map. Cape stations in the presented list were **404**.

Constraint goal: raw transient baseline variance is the electronic shadow of the hole. These integers are the presented first-epoch limbs, not a TEC map. The ~900 km diameter stays cited founding scale until Madrigal posts launch-day TEC.

SHA-256 of every `.26o.gz` is in `evidence/study-02-falcon-20260825/cors/downloads.tsv`. The ledger claim is those hashes plus the 21 integer pairs.

### Madrigal Haystack — API live, this window GAP

`https://cedar.openmadrigal.org/getExperimentsService.py?code=8000&startyear=&startmonth=&startday=&starthour=0&startmin=0&startsec=0&endyear=&endmonth=&endday=&endhour=0&endmin=0&endsec=0&local=0`

| Day | HTTP | Bytes | Grade |
|---|---|---|---|
| 2026-08-25 (launch) | 200 | **0** | GAP — world TEC not posted at T+11 h |
| 2026-08-24 | 200 | **0** | GAP |
| 2026-08-23 | 200 | **0** | GAP |
| 2026-08-12 (control) | 200 | **702** | API works; eclipse-day `gps/12aug26` listed (Anthea Coster / Haystack) |

CORS L1/L2 through Madrigal is **not measured as a 2026-08-25 kindat-3500 file this hour**. NGS CORS is live. Haystack TEC for today is empty.

### Court seals this hour (2026-08-25T20:38:38Z)

| Arm | Body | HTTP | Receipt |
|---|---|---|---|
| physics lockout | `clock=1 track=1 table_counts=70 rinex_n=21 stations_n=7 window_start=935 window_end=1130 constrained=1 liftoff_hhmmss=93338 pad_lat_micro_deg=28561941 pad_lon_micro_deg=-80577357` | 200 | `CALORIE_GAME_INGEST` · `CONTRACT` · `PHYSICS_TABLE_IS_THE_COURT`. Extra integers accepted (no float). Court still grades the table appointment — membrane was **not** rolled. |
| float adversary | `clock=1.5` | 400 | `REFUSED_FLOAT` |
| health | `mass_milli=70 vol_milli=100` | 200 | `WIN` `HEALTH_DOSE_IS_RATIONAL` `70/100` |
| fluids | `phi_in=4 phi_out=4` | 200 | `WIN` `FLUID_INTEGER_FLUX` residual 0 |
| Eisenstein | `GET /language-invariant/eisenstein/seal` | 200 | `ok=true` · `AFFINE_EISENSTEIN_WIRE_SEAL_V1` · 12 checks pass. Metric-tensor quench law for this hole is **LAW FROZEN** on the integer path; no new F-region Eisenstein CALORIE invented. |
| NOAA space | MCP `noaa_goes_r_weather` `lane=space` | 200 | `isError=false` |

Swift 6.4 + CORS integer ingest: **PASS** — 7 stations × 3 hours × integer L1/L2.

Evidence: `evidence/study-02-falcon-20260825/VERDICT.json`. Claim: [study-02-falcon.html](https://affine.earth/language-game/study-02-falcon.html). IDE: [ide.html#falcon-damage](https://affine.earth/language-game/ide.html#falcon-damage). **Regulatory alarm (package ready; founder files):** [Study-02-Regulatory-Alarm](Study-02-Regulatory-Alarm) · `CONSTRAINT_DIFFERENTIAL.json` sha256 `6569066c62c1eded957b4717e1d682d05361a2ec47b07ba185219bf9400b440f` · IONEX `1ebc01833ea4e57e48061a08342a8337062661ee946cdf73a928d43195d1038a` (one-way bridge, not the seal) · FITS `e36e3b8e3aa3353c24fd10eb4c5c390235ba2a56051a2edaf9d599652313e791` (21-row table).

## Live surfaces, tools, and courts (2026-08-25)

Look does not claim. Claim does not look. The charter stays an F-region hole. The health court is PK dose. Ozone is **cited**, not a court key.

| Surface | URL | What it is |
|---|---|---|
| Look | https://affine.earth/language-game/#researcher | GET catalogs. No visitor data. Paints 15 MCP tools. |
| Claim | https://affine.earth/language-game/study-02-falcon.html | POSTs `source=researcher` integers. Not the look path. |
| IDE | https://affine.earth/language-game/ide.html#falcon-damage | Researcher file surface. Public `/codegen` is the HTML shell. |
| Walkthrough | https://affine.earth/language-game/#story/Build-a-Study-Falcon-Walkthrough | Falcon first through the loop. |
| MCP | `POST https://affine.earth/language-invariant/mcp` | Registry `earth.affine/math-court` **2026.8.24**. |
| Public clients | https://github.com/gaiaftcl-sudo/uum8dSolarResearch/tree/main/clients/math-court-mcp | Bind trees. They POST. They do not evaluate a lattice. |
| Glama | https://glama.ai/mcp/connectors/earth.affine/affine-earth-math-court-remote | Connector **approved**. **Not tested.** No invented email. Do not claim a Glama test ran. |
| Prove | `evidence/study-02-falcon-20260825/VERDICT.json` | Marker `STUDY02_FALCON_1049_TERRESTRIAL_DEBIAS`. CORS integer ingest + lockout appointment. Prior loop: `evidence/researcher-ide-falcon-loop-20260824T121306Z`. |
| CORS integer puller | `scripts/study-02-cors-integer-phase.sh` | Presented first-epoch L1/L2 milli-cycles. `REFUSED_FLOAT` on sci-notation. |

### Fifteen live MCP tools (measured `tools/list` = 15)

Do not invent a sixteenth.

`atc.assert_4d_deconfliction` · `corpus_bonds` · `critique_frame` · `execute_transition` · `feeds_catalog` · `game_frame_meta` · `ide_rebuild_mesh` · `membrane_health` · `noaa_goes_r_weather` · `twin.robotics.evaluate_exact_ik` · `umc_direct` · `umc_resume` · `umc_status` · `verify_jordan_bond` · `weather.convective_containment`

### Courts this study actually binds

| Court | Appointment | What it grades |
|---|---|---|
| physics | `clock=1` `track=1` `table_counts=70` | clock / track / table_counts |
| health | `mass_milli=70` `vol_milli=100` | PK dose — **not ozone**, not DU |
| fluids | `phi_in=4` `phi_out=4` | exact flux |
| NOAA space | MCP `noaa_goes_r_weather` `lane=space` | live SWPC sources |
| `weather.convective_containment` | MCP | declared track vs convective cells |
| `execute_transition` | MCP | certified membrane turn |
| `corpus_bonds` | MCP | sealed corpus bonds |
| `verify_jordan_bond` | MCP | integer Jordan admit |

Ozone 4%, soot heat, jet-slow 35/1000, and phytoplankton are **cited** literature on the claim page. The health court seals 70/100 PK dose. It did not grade ozone.

### NEW 2026-08-24 court seals (re-measured this hour; not copied from chat)

ISO stamp **2026-08-24T12:19:19Z**. Every integer POST carried `source=researcher` plus a role. A failed POST would be written as a failure. None failed.

| Arm | Body | HTTP | Receipt (verbatim markers) |
|---|---|---|---|
| physics | `clock=1 track=1 table_counts=70` role=`experimentalist` | 200 | `status=CALORIE_GAME_INGEST` · `verdict=CONTRACT` · `proven_marker=PHYSICS_TABLE_IS_THE_COURT` · `source=researcher` |
| health | `mass_milli=70 vol_milli=100` role=`pharmacist` | 200 | `verdict=WIN` · `proven_marker=HEALTH_DOSE_IS_RATIONAL` · `concentration=70/100` |
| fluids | `phi_in=4 phi_out=4` role=`designer` | 200 | `verdict=WIN` · `proven_marker=FLUID_INTEGER_FLUX` · `phi_in=4` `phi_out=4` `residual=0` |
| NOAA `lane=space` | MCP `tools/call` `noaa_goes_r_weather` | 200 | `isError=false` · `status=CALORIE_SPACE_WX` · `proven=["NOAA_SWPC_LIVE"]` · `goes_r=NOAA_SWPC_GOES_PRIMARY_XRAYS_6H` · `sources_ok=["NOAA_SWPC_SCALES","NOAA_SWPC_GOES_XRAYS_6H","NOAA_SWPC_PLANETARY_K","NOAA_SWPC_OVATION_AURORA","NOAA_SWPC_RTSW_WIND_1M","NOAA_SWPC_ENLIL","COMPUTED_SOLAR_TERMINATOR"]` · 7/7 sources |
| float adversary | `clock=1.5` | 400 | `status=REFUSED_FLOAT` · `adversary=clock=1.5` |
| empty source | `source=""` | 400 | `status=REFUSED_UNATTRIBUTED` |
| tools/list | MCP | 200 | count **15** — the names above, no extras |

Charter integers already on the claim page (founding event **2023-07-19** — keep as charter, not this seal): Vandenberg **34632000 / −120611000 µ°** · Houston **30000000 µ°** · hole **900 km** · burn **286 km** · quench **70%**.

### Fresh public datum (2026-08-24 freeze — superseded by liftoff above)

Launch Library 2 prod fetched **2026-08-24T12:19:23Z**, HTTP 200: **Falcon 9 Block 5 | Starlink Group 10-49**, then `net=2026-08-25T09:33:17Z`, status **Go for Launch**, SLC-40. That freeze was correct. Liftoff measured this morning is **09:33:38Z**. Next Vandenberg on that fetch: **Starlink Group 15-22**, `net=2026-08-26T09:27:00Z`, Minute, SLC-4E.

NOAA SWPC re-fetched **2026-08-25T20:33:34Z** — see the this-morning table. Quiet nowcast scales do not void a launch hole; the storm gate is 3×Kp at t0.

## Why this study matters to humanity

Two layers. Do not mix them.

**1. Sealed (what the courts grade).** A ~900 km F-region hole on a published launch clock destroys and quenches the ionosphere that carries long-range radio, GNSS, aviation HF, and the TEC ledger. A weekly cadence (Starship / Starlink-class manifests) is the fastest forcing in the shear program. The public can re-run the integer law on Madrigal TEC. That is a checkable atmospheric footprint for the launch industry. The win is the verified bond — clock, corridor, recovery — not custody of anything. Honest edge word: measured receipt, or *not known*.

**2. Cited (literature the TEC ledger must inform — NOT court keys).** Hyperscale launch and re-entry is unregulated geoengineering in the published argument: stratospheric soot, alumina, **cited** ozone / UV-B (melanoma, cataracts, immune), **cited** jet-stream / crop lock-in, **cited** phytoplankton baseline. The health court is PK dose **70/100**, not DU. Ozone, jet-slow **35/1000**, and phytoplankton are **cited**. The health court did not grade ozone.

This morning’s hole is terrestrial de-bias. Mass: *not known*. The win is the verified bond — clock, corridor, recovery.

---

## The forcing and its clock

The forcing is a cone of exhaust dragged along a line in the sky. Three geometric pieces, each computable before the event:

**The clock.** Pre-registration reads the TheSpaceDevs Launch Library 2 `net` field — VERIFIED by live fetch of `https://lldev.thespacedevs.com/2.3.0/launches/upcoming/` on 2026-07-16, which returned an upcoming launch with `net 2026-07-16T20:22:00Z`, `net_precision "Minute"`, pad coordinates 34.632/−120.611, and target orbit, no auth. Post-hoc ground truth is Jonathan McDowell's GCAT/LVDB launch log (planet4589.org/space/gcat/web/launch/index.html — REPORTED via search), the only online source giving liftoff UTC to the second for most launches. Cross-checks: the operator's webcast T-0 and the Spaceflight Now launch log.

Minute-level precision at freeze time is not a weakness: the response rises at T+5 to T+15 min, so the frozen clock is 10–100× tighter than the phenomenon it predicts. GCAT then sharpens it to the second after the fact.

**The track.** The hole's source is the second-stage burn arc, ~200–700 km altitude. Reconstruction is one sealed algorithm, not a menu. Start at the LL2 pad coordinates and take launch azimuth from the target-orbit inclination by spherical trig; the two azimuth solutions (ascending vs descending node) resolve by a sealed per-pad branch rule — Vandenberg takes the southbound branch, Cape and Boca Chica take the eastbound/northeast branch. Space-Track refinement — back-propagating the first post-launch element set to insertion — is mandatory iff element sets exist by T+24h and excluded otherwise, never optional; when it runs, the refined corridor cell list is recorded at T+24h beside the frozen pad+azimuth list and is the graded corridor. The corridor cell list finalizes at T+24h by this rule and never changes afterward — grading at T+7 days reads frozen cell lists only. flightclub.io publishes simulated full-telemetry trajectories per mission for cross-check — descriptive only, never altering the sealed corridor (site VERIFIED by fetch; per-mission trajectory coverage REPORTED, not individually confirmed).

**The corridor.** A ±3° band around the reconstructed ground track over the T+0 to T+15 min arc. The grading grid is 1°×1°, so km-scale trajectory error is irrelevant — the corridor eats it. The corridor is materialized as an explicit list of 1°×1° cells sealed into the prediction record — the in-corridor flag {0,1} is frozen data, never geometry recomputed at grading time.

> [!NOTE]
> Nothing in the clock or the track comes from the response data. The forcing is fully specified by public manifests and orbital mechanics before the first response bin exists. That independence is what the shear cuts along.

![Diagram of the launch-hole forcing geometry: pad, second-stage burn arc, ±3° corridor, and the T+5 to T+15 minute onset clock](images/study-02-forcing.svg)
*The forcing in one frame: a published pad, a reconstructed burn arc, a ±3° corridor, and a clock that starts at a liftoff second announced in advance.*

---

## The response archive

All response data is public, raw, and named here exactly. Grading surface first, disputes and inputs after.

| Archive | URL | Format | Cadence | Auth | Sourcing |
|---|---|---|---|---|---|
| Madrigal world-gridded vertical TEC (CEDAR), instrument 8000, kindat 3500 | https://cedar.openmadrigal.org | ASCII via web services (columns `YEAR MONTH DAY HOUR MIN SEC GDLAT GLON TEC DTEC`) or HDF5 (`gpsYYMMDDg.00x`); 1°×1° global grid | 5-min bins; files post ~1–3 days after real time | None — identification triple (name/email/affiliation) only; "prior permission to access the data is not required" | VERIFIED twice: site fetched 2026-07-16, and the program's existing Madrigal ingestion client (the same path that graded Study 1's response side) already reads exactly this instrument/kindat, live-verified 2026-07-16 |
| Madrigal line-of-sight TEC, kindat 3505 | https://cedar.openmadrigal.org | HDF5 (`losYYMMDD`, tens of GB/day), per receiver–satellite pair with pierce points | 30-s native GNSS cadence, daily files | None (same triple) | REPORTED (kindat number from program seed + Madrigal convention; site verified, this kindat not individually fetched) |
| Launch Library 2 API (TheSpaceDevs) | https://ll.thespacedevs.com/2.3.0/launches/upcoming/ | JSON REST: `net`, `net_precision`, `status`, `pad.latitude/longitude`, `mission.orbit`, `launch_service_provider` | Continuously updated; poll daily, freeze at T−1h (prod rate-limited; `lldev` mirror for development) | None for read | VERIFIED by live fetch of the lldev development mirror 2026-07-16; prod endpoint REPORTED (same API, rate-limited) |
| GCAT / JSR Launch Vehicle Database | https://planet4589.org/space/gcat/web/launch/index.html | Tab-separated text catalogs + web tables; liftoff UTC to the second where available | Updated ~monthly | None | REPORTED (via search; URL from search result, not individually fetched) |
| Space-Track.org TLE archive | https://www.space-track.org | REST API, JSON/TLE; first post-launch element sets | Hours after orbit insertion | Free registered account | REPORTED (standard, not re-fetched here) |
| GFZ Potsdam Kp index, definitive series | https://kp.gfz-potsdam.de | ASCII/JSON 3-hour Kp values | 3-h values; definitive series posted after month close (nowcast sooner) | None | REPORTED (standard reference archive; not individually fetched) |
| DRAO Penticton F10.7 solar flux (via NOAA SWPC) | https://www.swpc.noaa.gov | Daily observed F10.7 flux values | Daily | None | REPORTED (standard reference archive; not individually fetched) |

Kindat 3500 is the primary grading surface. Kindat 3505 is used only for high-resolution corridor cross-sections on WIN/MISS disputes — and those cross-sections are descriptive only: they never change a sealed ledger row, exactly as reprocessed Madrigal files never regrade a sealed verdict. Space-Track is secondary — corridor from pad + azimuth alone is sufficient at 1° grid scale, and the sealed refinement rule in the track section above governs exactly when element sets enter. Kp gating reads the GFZ Potsdam definitive series: "Kp at t0" means the definitive 3-hour Kp value covering t0; if the definitive value is not yet posted at grading time, grading waits for it — nowcast Kp never seals a verdict. Definitive Kp arrives in thirds (a 4− prints as 3 2/3), so every gate on this page grades the exact integer 3×Kp: the storm gate is 3×Kp ≥ 12 → VOID-by-storm, and a baseline day qualifies only if all eight of its definitive 3-h values satisfy 3×Kp < 12 — no fractional comparison anywhere. F10.7 stratification reads the DRAO daily value for the launch date.

---

## The adversary

The adversary mimics magnitude. It cannot mimic shape. That gap is the shear — the same cut the eclipse study made against the Gannon superstorm ([Eclipse 2026 — Model shear](Eclipse-2026-Model-Shear.md)).

**Why it mimics.** Geomagnetic-storm negative phases deplete midlatitude total electron content (TEC — the column count of free electrons a radio signal must cross, measured in TEC units, TECu) by tens of TECu for hours — the Gannon G5 superstorm (2024-05-10, SSC 17:05 UT, Dst −412 nT, Kp 9) drove −20 to −25 TECu over Europe and daytime depletions exceeding −35 TECu persisting 18 h at 30–50°N. That is deeper and 6–10× longer-lived than any launch hole. Storm LSTIDs — large-scale traveling ionospheric disturbances (TIDs) — and impulsive-event TIDs travel at 300–1000 m/s (REPORTED via search) — the same band as launch shock-acoustic waves (FORMOSAT-5: 629–726 m/s). A TEC dip of 2–70% with wave activity is, pointwise, indistinguishable from a launch signature. Magnitude alone decides nothing here.

**Why shape defeats it.** Three independent locks the adversary cannot fake simultaneously:

1. **Clock.** The hole opens 5–15 min after a to-the-second public epoch chosen by a launch manifest — statistically independent of solar-wind arrival. Epoch-superposing hundreds of launches averages storm noise to zero.
2. **Corridor.** The hole (~900 km) sits ON the published second-stage ground track. Storm depletion is continental and organized by magnetic latitude; natural TIDs move THROUGH the corridor rather than being confined to it. In-corridor vs control-corridor shear ratio ~1 for storms, large for launches.
3. **Recovery.** Falcon-9-class launch holes refill in 1–3 h by plasma diffusion (Saturn-V-class events ran ~4 h); storm negative phases persist 6–24+ h (REPORTED via search). Kp/Dst gating — windows with 3×Kp ≥ 12 (Kp ≥ 4 in exact thirds) are pre-declared VOID-by-storm — removes the residual tail; per S2 this gate is disabled for adversary grading, so the adversary is graded on shape, not gated away.

![Contrast of the true forcing shape versus the adversary shape: a corridor-confined launch hole that heals in hours beside a continental storm depletion that lingers](images/study-02-shear.svg)
*Same depth of depletion, different shape: the launch hole sits on its published track and refills in hours; the storm spans continents, ignores every corridor, and lingers.*

Three named adversary events stand ready for raw grading, exactly as Study 1 graded the Gannon shear:

<details>
<summary><strong>Adversary event roster (grade raw — any detector firing here is a recorded falsifier)</strong></summary>

| Window (UTC) | Event | Why it is the adversary |
|---|---|---|
| 2024-05-10 17:05 (SSC) – 2024-05-12 00:00 | Gannon G5 superstorm (Dst −412 nT, Kp 9) | Strongest storm in two decades; continental-scale depletion mimic (−20 to −25 TECu Europe; >−35 TECu daytime, 18 h at 30–50°N). No corridor confinement, hours-long recovery, clock-locked to the SSC and to no launch epoch. The detector must NOT fire. Citations: ScienceDirect S136468262500272X; arXiv:2511.16384. REPORTED via search. |
| 2022-01-15 04:14 (eruption) – 2022-01-19 08:14 (the ~100 h global propagation interval) | Hunga Tonga–Hunga Haʻapai global TIDs | Largest natural TID event on record: wavelike TEC structures at ~300–350 m/s with 500–1,000 km horizontal wavelengths circling the globe three times over ~100 h (Zhang et al. 2022), plus a ~1000 m/s near-field/LSTID component (Themens et al. 2022, GRL), spectrally overlapping launch SAWs. Tests the traveling-lag discriminant: Tonga waves hold no fixed lag to any launch epoch and cross corridors isotropically. Zhang et al. 2022, Front. Astron. Space Sci., doi:10.3389/fspas.2022.871275; Themens et al. 2022, GRL. REPORTED via search. |
| 2024-04-01 – 2024-05-31 (multiple; burn epoch list — pad, t0, reentry arc — to be extracted from Inchin et al. 2025 into the charter before sealing) | Falcon 9 second-stage deorbit burns over CONUS | The hardest mimic: chemically identical exhaust-driven depletions (Inchin et al. 2025, GRL, doi:10.1029/2024GL112364) occurring ~hours after launch epoch on descending arcs. Defeated only by the clock lock (wrong lag from t0) and corridor lock (reentry arc ≠ ascent track). Catalogued as a known-confounder class, not noise. REPORTED via search. |

</details>

> [!WARNING]
> A recorded failure is a finding. If the sealed detector fires inside an adversary window, that firing is published raw in the ledger. It is never renamed, filtered, or re-baselined away.

---

## The blind spots this study targets

Four gaps in the published record, each verifiable by reading the literature this charter cites:

**Nobody has ever pre-registered a launch hole.** From Mendillo's 1975 Skylab paper through Yasyukevich 2024 and Rajesh 2025, every published result is a post-hoc single-event or few-event case study. Falcon 9 flew 130+ missions per year in 2024–2026 (REPORTED via search); the several-hundred-launch population 2017–2026 has never been epoch-superposed against gridded TEC as a statistical ensemble with frozen-in-advance predictions. This study is that survey.

**The community's own tooling deletes the signal.** Conventional TID climatologies built on band-pass-filtered dTEC are blind to the DC hole by construction; the 2025 MRMIT study of Jiuquan launches (MDPI Remote Sensing 17:3327) detrends TEC to extract TID amplitude and period separately from the depletion (REPORTED via search; full text not fetched). This study grades exactly the thing the filters remove.

**Operational products smear the hole below representation.** IGS global ionosphere maps run at 1–2 h cadence with few-degree spherical-harmonic smoothing (REPORTED via search); a 900-km, 2–3 h hole vanishes inside that kernel. No GNSS-augmentation threat model flags scheduled launches — despite depletions up to 70% directly over CONUS, on a clock published in advance.

**Existing lag statistics are contaminated in exactly the graded dimension.** Deorbit-burn depletions were only first characterized in 2025 (Inchin et al., GRL, doi:10.1029/2024GL112364). Every earlier hole catalog conflates ascent-burn and reentry-burn signatures — so lag-from-liftoff statistics in the literature are polluted in the one coordinate (clock lag) this study's metric isolates.

---

## The shear metric

The metric is the eclipse study's discriminant pair, transplanted: corridor confinement replaces umbral confinement, the launch clock replaces the umbra's velocity lock.

**Confinement analog — corridor-confinement shear.** The quantized TEC anomaly *inside* the ±3° corridor around the reconstructed second-stage track is graded against a *control corridor* at the same geomagnetic latitude band, displaced exactly 20° in longitude at the same UT. The quiet-day baseline is sealed arithmetic, not a judgment: for each cell and UT bin, take the first 3 qualifying days in day +1, −1, +2, −2, +3, −3, +4, −4, +5, −5 order — a day qualifies only if it has no launch in the corridor and all eight of its definitive 3-h Kp values satisfy 3×Kp < 12 — and the baseline is the integer median of those 3 values (K = 3 sealed; any even-count median anywhere in this charter is the lower median, so every median is an exact integer). Fewer than 3 qualifying days, or a corridor baseline ≤ 0 deci-TECu (possible in sparse night cells), is VOID-by-baseline. The control corridor is chosen by a sealed integer rule, not by eye: two candidates, displaced +20° and −20° in longitude; count the non-absent kindat-3500 cells in each candidate over the same UT bins on the sealed lookback window (days −12 through −3 relative to the frozen net date); the larger count wins, ties go eastward. Over CONUS for Vandenberg this lands eastward, where the receiver mesh is dense and a westward displacement would sit over the receiver-free Pacific; if both candidates are oceanic/data-absent, the verdict is VOID-by-control. Both cell lists — in-corridor and control — are sealed into the T−1h freeze as explicit 1°×1° cell enumerations, so no spherical trigonometry runs at grading time and the in-corridor flag {0,1} is frozen data, not recomputed geometry. A launch hole: in-corridor depletion 10–70% with control ~0. A storm negative phase: both corridors depleted equally.

**Traveling-lag analog — clock-locked lag structure.** Two traveling signatures, both keyed to t0: (a) the hole onset propagates along track at vehicle speed (~7 km/s); at that speed the vehicle traverses the entire graded arc inside a single 5-min kindat 3500 bin, so this signature is resolvable only on kindat 3505 30-s data and is descriptive — it is not part of the sealed WIN rule; (b) the SAW ring expands at 600–1000 m/s acoustic speed from the burn arc (FORMOSAT-5: 629–726 m/s). Natural mimics fail the lag test: MSTIDs cross at 100–300 m/s with no fixed lag to t0 (REPORTED via search); Tonga-class Lamb waves at ~330 m/s are locked to the eruption clock, not the launch clock; deorbit burns show the right chemistry at the wrong lag (hours).

**Quantization — exact integers only, per program law.**

- TEC and dTEC stored as integer deci-TECu: quantization is decimal-string manipulation of the printed Madrigal ASCII field — shift the printed decimal point one place right, then round half-away-from-zero on the printed digits — never a multiply of a binary-float parse, which can misround a printed x.x5; applied once at ingestion and sealed with the Form, so identical files yield identical integers on any machine
- time as integer 5-min bin index from the liftoff epoch (matches kindat 3500 cadence exactly)
- space as integer 1°×1° cell with binary in-corridor flag {0,1}, read from the sealed cell list frozen with the prediction — frozen data, never recomputed geometry
- corridor-to-integer reduction, sealed: per bin, the corridor statistic is the lower median of the in-corridor cell anomalies (anomaly = cell deci-TECu minus that cell's baseline deci-TECu; cells with absent data or absent baseline are excluded from both medians) and the corridor baseline is the lower median of the same surviving cells' baselines; identically for the control corridor
- one sealed grading unit everywhere: percent-of-baseline, always evaluated by integer cross-multiplication, never by division — depth passes an integer threshold T iff 100 × statistic_deci ≤ −T × baseline_deci; a corridor baseline ≤ 0 deci-TECu is VOID-by-baseline
- the WIN condition, stated once and in full, in the sealed unit: (a) depletion — 100 × in-corridor statistic ≤ −T_stratum × in-corridor baseline in at least one bin of +1..+18; (b) recovery — 100 × in-corridor statistic ≥ −R × in-corridor baseline for 2 consecutive bins at or before bin +36 (T+180 min); (c) both controls quiet — for the shifted-corridor control AND the shifted-epoch control (same pad, same UT bins, on the first day with no launch from that pad taken in day +1, −1, +2, −2, +3, −3, +4, −4, +5, −5 order; order exhausted → VOID-by-control), 100 × control statistic > −U × control baseline in every bin of +1..+18 — for sealed integers T_stratum, R, U; no division anywhere, so a ~0 or data-absent control can never produce an undefined ratio (a data-absent control is VOID-by-control, never a free pass)
- onset defined as an integer: the smallest bin in +1..+18 whose corridor statistic passes T_stratum; expected onset bins +1..+3 is graded in S4 across the ledger, not in the per-launch WIN rule

No floats anywhere in a sealed verdict. Same inputs, same integers, on any machine.

**Threshold derivation plan — frozen from history, then sealed.**

1. Pull kindat 3500 for N ≥ 100 Falcon 9 launches 2018–2025 (GCAT times × LL2/GCAT pad+azimuth corridors), storm-gate-clean windows only (3×Kp < 12).
2. Build the in-corridor depletion-depth distribution — the per-bin corridor statistic graded against corridor baseline by cross-multiplication — at bins +1..+18 (T+5–90 min — the same sealed window the LAW grades) and the matched null distribution from shifted-epoch controls and shifted-corridor controls, both per the sealed rules in the Quantization block.
3. Set the WIN threshold table — one integer T_stratum per (local-time class, F10.7 band) stratum, in the sealed unit (integer percent-of-baseline), F10.7 from the DRAO/SWPC archive named above — at the depth where in-corridor exceedance separates from null at a fixed integer false-positive budget (e.g. ≤ 5/100 controls).
4. Validate bracketing as decidable integer inequalities in the sealed unit (percent-of-baseline), each event graded in its own (local-time, F10.7) stratum. Pin each event's depth as an integer percent fixed by its citation now, not chosen at seal time: Falcon Heavy 17 (its printed ~17%); FORMOSAT-5 10 (the minimum of its published 10–70% range); Starship by integer floor division of its published range-minimum 10 TECu against the background value printed in Yasyukevich et al. 2024, computed once and sealed. Then require FH_depth − d ≤ T_stratum ≤ FH_depth + d for a sealed integer margin d, and T_stratum < FORMOSAT5_depth and T_stratum < Starship_depth. Skylab (1973) predates GNSS TEC and is outside kindat 3500 coverage — it is explicitly exempt from the bracketing check.
5. Seal the threshold table into the standing Form **before** the first pre-registered launch. After sealing, the threshold never moves.

---

## Historical corpus

Five published events anchor the corpus — a half-century arc from the founding observation to the newest depletion class.

| Date (UTC) | Event | Clock | Response | Citation |
|---|---|---|---|---|
| 1973-05-14 | Skylab 1 / Saturn V — founding event of the field | Liftoff 17:30 UTC; rapid electron-content decrease within ~1000 km of the burning S-II stage over the following tens of minutes | Hole spanning ~2000 km (1000 km radius), lasting ~4 h; enhanced recombination from H₂ and H₂O exhaust | Mendillo, Hawkins & Klobuchar (1975), *Science* 187(4174):343, doi:10.1126/science.187.4174.343 — REPORTED via search (science.org + PubMed 17814266) |
| 2017-08-24 | FORMOSAT-5, Falcon 9 (Vandenberg SLC-4E) — largest circular SAW on record + giant hole | Liftoff 18:51 UTC (VERIFIED via Wikipedia); circular shock-acoustic waves from T+5 min, ~20 min duration; hole develops after | SAWs 629–726 m/s, wavelengths 390–450 km, period 10.28 ± 1 min, ~1500 km extent (114–128°W, 26–39°N); hole ~900 km diameter, 10–70% depletion vs reference days; recovery 2.5–3 h | Chou et al. (2018), *Space Weather*, doi:10.1002/2017SW001738 — numbers REPORTED via search (Wiley full text 403 on direct fetch) |
| 2018-02-06 | Falcon Heavy demo (KSC LC-39A) — trajectory-carved hole below F-peak | Liftoff 20:45 UTC (VERIFIED via Wikipedia); hole carved along trajectory ~T+10 min | ~2 TECU depletion, ~17% of ~12 TECU background, persisting >1 h without recovery; small because flight altitude stayed below 200 km | Chou et al. (2018), *GRL*, doi:10.1029/2018GL078088 — REPORTED via search |
| 2023-07-20 | Falcon 9 Starlink Group 6-15 (Vandenberg) — photographed red-glow hole over the US Southwest | Liftoff 04:09 UTC (21:09 PDT July 19); red 630.0 nm glow visible ~20 min from S. California to Arizona | Second-stage burn at 286 km, near the F-region peak; local ionization quenched up to 70%; red airglow from O⁺ recombination triggered by H₂O/CO₂ exhaust | spaceweatherarchive.com 2023-07-23 quoting Jeff Baumgardner (Boston Univ.) — VERIFIED by fetch; launch time REPORTED via space.com/Spaceflight Now search results |
| 2023-11-18 | Starship IFT-2 (Boca Chica) — first non-chemical (explosion-driven) depletion | Liftoff 13:02:50 UTC (VERIFIED via Wikipedia); booster breakup ~T+3.5 min at ~90 km; ship explosion ~T+8 min at ~150 km | −10 to −15 TECU after a +1–4 TECU pulse; depletion persisted 30–40 min in some series; 2000-km V-shaped disturbances propagating northward | Yasyukevich et al. (2024), *GRL*, doi:10.1029/2024GL109284 — REPORTED via search; times VERIFIED via Wikipedia |

The corpus proper is the several-hundred-launch Falcon 9 population 2018–2025 (threshold derivation, step 1 above); these five are the published fixed points the frozen threshold must bracket.

---

## Pre-registration

**Target.** Every upcoming Falcon 9, Falcon Heavy, and Starship launch appearing in Launch Library 2 with `net_precision` at Minute or better. Priority tiers:

1. **Vandenberg** — corridor over the dense CONUS GNSS receiver mesh where kindat 3500 quality is best; both flagship published events (FORMOSAT-5, Starlink 6-15) are Vandenberg.
2. **Starship from Boca Chica** — largest response class (10–15 TECu), lower cadence.
3. **Cape launches** — graded only on the first ~1000 km of track before the corridor goes oceanic and data-sparse.

**Cadence.** Several launches per week — Falcon 9 alone flew 130+ missions/yr in 2024–2026 (REPORTED via search), and the live LL2 feed (VERIFIED 2026-07-16) showed a Falcon 9 at Vandenberg within hours of the fetch. This is the fastest forcing clock of the five shear studies: WIN/MISS grading lands within ~2 days of each launch (Madrigal posting lag), tens of graded predictions per month. Poll LL2 daily; freeze each prediction at T−1h with the then-current `net` plus the explicit in-corridor and control cell lists (finalized, if Space-Track elements exist by T+24h, by the sealed refinement rule in the track section — after which nothing moves; corridor geometry never runs at grading time); reconcile post-hoc against GCAT's to-the-second liftoff, which then defines bin 0; scrubbed launches are VOID-by-scrub, never MISS.

**Standing law, not per-event forms.** One standing SEALED Form — the launch-hole invariant:

> **LAW.** For every pre-registered epoch t0 (bin 0 = the GCAT liftoff second when available, else the frozen net; a frozen-net-to-actual delta greater than 10 minutes is VOID-by-slip) with frozen corridor cell list C: if the storm gate fails (3×Kp ≥ 12 for the GFZ definitive 3-hour value covering t0), the prediction is VOID-by-storm, not MISS; otherwise the verdict is WIN iff the sealed WIN condition of the Quantization block holds — in-corridor depletion passing T_stratum by integer cross-multiplication in at least one bin of +1..+18 (T+5–90 min), recovery per the sealed R-rule for 2 consecutive bins at or before bin +36 (T+180 min), and BOTH the shifted-epoch and shifted-corridor controls quiet under U in every bin of +1..+18 — and MISS otherwise. The complete VOID enumeration is sealed here, once: VOID-by-scrub (no liftoff), VOID-by-slip (net-to-actual delta > 10 min), VOID-by-storm (3×Kp ≥ 12 at t0), VOID-by-baseline (fewer than 3 qualifying baseline days, or corridor baseline ≤ 0 deci-TECu), VOID-by-control (both control candidates data-absent, or the shifted-epoch day order exhausted), VOID-by-crowding (the multi-event exclusion rule of Honest limits), VOID-by-data (no Madrigal file at T+7 days). No other VOID class exists; S3 and Honest limits reference this list.

Each launch grades the same sealed law again. This is the eclipse-study pattern — pre-registered prediction Form plus automated watch — reused with a weekly rather than yearly clock. The program's existing Madrigal ingestion path grades the response side unchanged.

---

## Success criteria

All four are decidable — integer verdicts against sealed rules, no judgment calls after sealing.

- **S1 — Threshold frozen from history.** An integer WIN threshold table in the sealed unit (percent-of-baseline, graded by cross-multiplication), keyed by (local-time class, F10.7 band), is derived from ≥ 100 historical Falcon 9 launches (storm-gate-clean, 3×Kp < 12) against shifted-epoch and shifted-corridor nulls at a fixed integer false-positive budget (≤ 5/100 controls), satisfies the per-stratum integer bracketing inequalities of derivation step 4 with every pin fixed by its citation rather than chosen at seal time (Falcon Heavy within sealed margin d of its stratum's threshold; FORMOSAT-5 at its pinned range-minimum 10 and Starship at its pinned background-converted integer, both strictly cleared; Skylab exempt as archive-unreachable), and is sealed into the standing Form dated **before** the first frozen prediction. Decidable by the seal record's timestamp and the bracketing inequalities.
- **S2 — Adversary graded raw.** The adversary run is a sealed finite computation, sealed with the Form: synthetic probe epochs at every 5-min bin index inside each adversary window, crossed with the full historical-corpus corridor set (every frozen corridor-and-control cell-list pair from derivation step 1), each (t0, corridor) probe graded by the sealed detector with baseline and control rules unchanged and with the storm-gate VOID disabled for adversary grading — an override that is itself sealed with the Form, since otherwise the storm gate, not shape discrimination, would trivially guarantee zero WINs on Gannon. It must produce zero WINs over the enumerated adversary windows: Gannon 2024-05-10 17:05 to 2024-05-12 00:00 UTC; Tonga 2022-01-15 04:14 to 2022-01-19 08:14 UTC (the ~100 h propagation interval); and the CONUS deorbit-burn class, which grades only after the specific burn epoch list (pad, t0, reentry arc) has been extracted from Inchin et al. 2025 into the charter and sealed — until that list is sealed, that window is not gradeable. Any firing is published raw in the ledger as a falsifier. Decidable by the ledger rows — the same finite integer computation on every machine.
- **S3 — Prospective predictions sealed and graded.** Every pre-registered launch frozen at T−1h receives exactly one verdict — WIN, MISS, or VOID (exactly one class from the LAW's sealed VOID enumeration: scrub, slip, storm, baseline, control, crowding, data) — from the sealed law at T+7 days, with the integer anomaly and baseline pairs (in-corridor and control, deci-TECu — depth grades by cross-multiplication and is never stored as a divided percent) and onset/recovery bin indices recorded for each. Decidable by the presence and completeness of the ledger row.
- **S4 — Shape separates on the prospective ledger.** After the first N non-VOID graded launches (integer N sealed with the Form, e.g. N = 100): WINs ≥ W (integer W sealed with the Form) while shifted-epoch and shifted-corridor control firings ≤ floor(N/20), and the sealed onset structure holds with a sealed quantifier — onset (the smallest bin in +1..+18 whose corridor statistic passes T_stratum) falls in bins +1..+3 on ≥ K of the WINs, integer K sealed with N and W; recovery by bin +36 is already part of every WIN by the sealed WIN condition. Otherwise S4 is FAIL and the MISS ledger stands exactly as recorded. Decidable by the same integer counts the threshold was defined on.

![Animation of the discriminating shape: a corridor-confined depletion opening at T+5 minutes and refilling by T+3 hours while the control corridor stays flat](images/study-02-shear.gif)
*The shape the sealed law grades, in motion: open on the clock, confined to the corridor, gone within hours — while the control corridor never moves.*

---

## What this protects — people, ecology, and the planet

A detector is only worth building if someone stands downstream of the thing it detects. Here the downstream is concrete: aircraft on satellite-guided approaches, ships and tractors steering by GNSS, oceanic flights on HF radio, and a public argument about what rockets do to the upper atmosphere that currently runs on estimates instead of measurements.

**The hazard is real and currently unlogged.** The FORMOSAT-5 event of 2017-08-24 — the largest circular launch shock ring on record, with its ~900 km hole — created plasma gradients sharp enough to introduce ~1 m of range error into GPS navigation and positioning (Chou et al. 2018, *Space Weather* — REPORTED via search). It happened on a magnetically quiet day: every magnitude-based space-weather dial read calm, the operational global ionosphere maps smeared the hole below representation (see the blind spots above), and no GNSS-integrity product logged it. That is the ruler-read failure in a single event — a magnitude instrument said *nothing happened* while a navigation-relevant hole sat over the western United States for hours. Aircraft flying RNP approaches and SBAS vertical guidance ride on exactly the quantity the hole perturbs, and the one class of ionospheric hole that arrives on a schedule published days in advance is the one class no current product watches for.

**Magnitude rulers also fail in the loud direction.** During the 2003 Halloween storms, WAAS localizer performance with vertical guidance was unavailable to aviation across the continental United States for ~15 h on 2003-10-29 and ~11 h on 2003-10-30 (REPORTED via search). During the Gannon storm of 2024-05-10, GNSS positions wandered by up to ~230 ft and GPS-guided planting idled across the Midwest at a cost later estimated above $500 million — $565 million in one Kansas State analysis — in a country where about 70% of planted acres steer by GPS guidance (REPORTED via search). After a decade of such events, operators are primed to blame any GNSS anomaly on storms, and a magnitude-only read cannot tell them otherwise: it files a corridor-confined hole that heals in 1–3 h under the same heading as a continental storm phase that persists 6–24+ h. The correct operational responses differ — wait out a launch window versus stand down for a day. A validated shape-read is the more trustworthy warning in both directions: it names the corridor, the clock, and the expected recovery hour, and it stays silent through the storms. A detector that fires on both teaches everyone to ignore it; a false alarm erodes exactly the trust an honest instrument exists to earn.

**HF over the oceans.** Long-haul flights over oceans and poles lean on HF radio where satellite coverage thins; poleward of 82°N satellite communication is unavailable and HF is the link (Xue et al. 2024, *Space Weather* — REPORTED via search). In January 2005, United Airlines diverted 26 flights off polar routes over several days to avoid HF blackout risk, paying for it in fuel, cargo capacity, and crew time (REPORTED via search). HF long-range propagation reflects off the same F-region plasma a launch hole removes — the usable frequency band falls with the electron density under the reflection point; that is propagation physics applied to the depletion numbers already cited, not a new measurement. A ~900 km, up-to-70% depletion parked across an oceanic corridor for 1–3 h is a brief, real, *schedulable* HF degradation — and today no dispatcher product marks it.

**The accountability ledger.** Global orbital launch attempts ran 146 in 2021, 186 in 2022, 223 in 2023, and 259 in 2024 — a record four years running, roughly a fifth more launches each year (REPORTED via search). The debate over what that cadence does to the upper atmosphere currently runs on modeled inventories: one 2022 study (Maloney et al., *JGR Atmospheres*, doi:10.1029/2021JD036373) projected that ~10 Gg/yr of rocket black carbon injected near 30°N could drive Northern Hemisphere **cited** ozone loss of up to 16 DU in its scenarios (REPORTED via search). Those ozone / UV-B / jet-stream / phytoplankton numbers are **cited** literature the TEC ledger must inform — they are not health-court keys (the health court is PK dose 70/100). Scenarios need checking against per-event measurements, and the ionospheric hole is the one per-launch upper-atmosphere effect that is directly and publicly measurable with existing instruments. This charter's ledger — dated, gridded, integer-graded, MISSes published beside WINs — is that measurement, one launch at a time. Whatever the mesosphere-and-ozone accountability debate concludes, it should conclude on honest per-launch records rather than estimates. The sky writes the record either way; this study proposes to keep it.

---

## Honest limits

What this study is NOT:

- **Not a TID climatology and not a wave study.** It grades one shape — the clock-locked, corridor-confined, fast-recovering depletion — and deliberately ignores the band-passed wave field the existing literature focuses on.
- **Not a forecast of ionospheric weather.** The study predicts the response to a manifest, not the state of the ionosphere. Storm-gated windows (3×Kp ≥ 12) are VOID by rule, so nothing here speaks to storm-time behavior.
- **Not uniform coverage.** Kindat 3500 exists only where ground receivers exist. Eastward Cape corridors go oceanic within ~1000 km; roughly half the launch population grades on a corridor stub. The Vandenberg tier and kindat 3505 cross-sections mitigate, not cure, this.
- **Not sensitive at the floor.** 1°×1°/5-min gridding plus ~1 TECu DTEC noise puts Falcon-Heavy-class holes (~2 TECu) near the detection floor; low-altitude-insertion profiles (burn < 200 km, below the F region) may be structurally MISS-prone. The threshold-derivation step must decide whether to gate on insertion-profile class or accept a lower per-launch win rate — and that decision seals with the Form.
- **Not immune to trajectory error.** Dogleg and dual-burn missions break azimuth-from-inclination; GTO coast-phase second burns can land the hole 1000+ km downrange of the modeled arc. The ±3° corridor absorbs most, not all, of this.
- **Not immune to clock churn or crowding.** Nets slip by hours routinely (handled by the T−1h freeze + GCAT reconciliation, with slips beyond the sealed 10-minute tolerance VOID-by-slip), but back-to-back same-pad launches and deorbit burns can put a second same-chemistry hole in a graded corridor within the grading window. The multi-event exclusion rule seals here, verbatim: if any other launch or catalogued deorbit burn deposits exhaust inside the graded corridor between bin −36 and bin +36 of t0, the prediction is VOID-by-crowding — the class of that name in the LAW's sealed VOID enumeration — never MISS.
- **Not built on an immutable archive.** Madrigal files post 1–3 days late and get reprocessed (`.001` → `.002`); grading runs once against the newest file version present at T+7 days, absence at T+7 days is VOID-by-data, the pinned version is recorded in the ledger row, and later reprocessed versions never regrade a sealed verdict.
- **Not a single-threshold world.** A 70% depletion of a 5-TECu nighttime background is 3.5 TECu; 17% of a 12-TECu daytime background is 2 TECu. Percent and absolute thresholds diverge, so the sealed threshold is an integer table stratified by local-time class and F10.7 band (archive named in the response table, each bracketing event assigned to its stratum by liftoff local time) — a single unstratified number would be systematically wrong for one class.

---

**Status: LIVE CLAIM — apex POSTs named physics / health / fluids receipts. Charter corpus watch is a later Form, not an open study.**