Study 31 Biosphere Cascade

Study 31 — The Biosphere Cascade Court

Page class: CHARTER. Law frozen 2026-09-01, before the chain was multiplied. Every link carries an evidence grade defined in the ontology. Producing program: reproduce/biosphere-cascade-chain.swift.

Status: LAW FROZEN · CORPUS INGESTED · LIVE ON THE COURT — STUDY31_BIOSPHERE_JOINT_LEDGER_PROVEN, verified on all nine cells 2026-09-01.

---

Why this court exists

Every other study on this wiki grades a single instrument against a single question. This one grades a chain.

The reason is that the atmospheric literature is organised link by link. One group models injection into the stratosphere. A different group models ozone response. A third field studies UV and marine biology. Each link is competently studied and nobody multiplies them, so no published number exists for the thing a person actually wants to know: what does this do to the biosphere.

That gap is not a conspiracy. It is what happens when a question spans four disciplines and no funding line spans them. But the consequence is that the cascade is unexamined, and this court examines it.

The biosphere is the system, and we are one node in it

This study does not ask what satellite reentry does to humans. It asks what it does to an integrated system in which humans are one small node among many, most of them older and more load-bearing than we are.

That framing is not sentiment. It is what makes the arithmetic different. A human-impact study can stop at UV dose and skin cancer. A system study cannot, because the biosphere is not a passive recipient of these axes — it is a term inside them:

• Phytoplankton are the primary oceanic carbon sink. Damage there raises the greenhouse column, which is one of the axes being measured. The response feeds the forcing.
• Marine biology seeds cloud condensation nuclei. Damage there changes albedo — the same physical quantity as the reflective term.
• Ice and vegetation albedo respond to both, and to each other.

So the honest object of study is a coupled system with feedback, not a chain of one-way effects. And that is precisely what no classical single-model study can measure, for a structural reason rather than a negligent one: each published group runs one model answering one question, and no funding line spans four disciplines.

What every published group actually covers

paper	mass in	ozone	radiative	biosphere
Ferreira 2024	yes	mechanism only	—	—
Maloney 2025	yes	weaker hole	1.5 K + vortex	—
Revell 2025	launch only	+0.27%	—	—
Barker 2026	yes	9% megaconstellation	+6.47 / −6.40	—
Vliex 2026	yes	85.6 mDU	4.1 mW/m²	—

The biosphere column is empty in every row. Five peer-reviewed papers on what satellite reentry does to the atmosphere, and not one carries a biological term. That is not a criticism of any of them — each answered the question it asked. It is the gap that exists between them.

And no two rows share a full axis set, so a reader cannot sum them either. The joint number does not exist in the literature and cannot be assembled from it.

And this table is why "the studies disagree" cannot function as reassurance. Each row is an instrument pointed at one axis; the biosphere column is empty in all five; and their disagreements trace to their own arithmetic — a units artifact here, a forcing-convention sign flip there, a global mean that is twelve times kinder than the regional record. A study that never modelled the coupled system cannot find it safe. It can only fail to mention it, and silence quoted as safety is the failure mode this joint ledger exists to close. The full argument: the narrow-study defence, answered.

The joint ledger — one account, every axis

This is what a closed system with exact arithmetic buys: the same integer ledger carries mass, optical depth, radiative forcing and the coupling, so the axes can be held simultaneously rather than one at a time. Producing program: reproduce/closed-system-joint-ledger.swift.

scenario	mass/yr	vs meteoric flux	regional O₃	vs ALL greenhouse forcing
2025, measured	461 t	3.1–6.3%	0.35%	0.1%
authorised 19,408	1,909 t	13.0–26.1%	3.1%	1.0%
Gen3 filed 100,000	9,840 t	67–135%	15.6%	5.3%
orbital data centres	80,000 t	548–1096%	exceeds column	43.7%

Grades travel with the numbers: mass is MEASURED (GCAT rows); the meteoric ratio is MEASURED (residence time cancels); regional O₃ is a PROJECTION × a MEASURED 12× factor; the greenhouse share is a PROJECTION over the MEASURED AGGI baseline.

What the join shows that no single axis does

Read the authorised row — the fleet that is already approved. Every axis alone looks survivable: 1.0% of greenhouse forcing is small; 3.1% regional ozone sits inside historical variation; 13–26% of the meteoric flux is a fraction of nature's own. A reader shown any one of those would reasonably move on.

Together they say something else: an already-authorised fleet moves three independent planetary accounts by percent-scale amounts at once, and no instrument reports any of the three.

And at the orbital-data-centre scale two of the four axes leave their valid ranges entirely — the ozone chain exceeds the column, the meteoric ratio passes 1000%. Two independent axes breaking at the same scale is not a coincidence. It is the signature of instruments being used far outside the regime they were built for, on every axis simultaneously.

The coupling terms are named, graded, and left out of the sum

Every feedback above is an established mechanism, and none has a measured transfer function at this forcing. So none is summed into the ledger and no number is invented for any of them. They are named, graded NOT_KNOWN, and excluded from the arithmetic.

That omission is one-directional, and it is stated plainly: every named feedback, if quantified, would ADD to the totals rather than subtract. The joint ledger is therefore a floor, not an estimate.

The radiative axis, against a measured baseline

Producing program: reproduce/radiative-baseline.swift. Corpus: NOAA Annual Greenhouse Gas Index, aggi.txt, pinned — 46 years, 1979–2024, published at three decimals of W/m² and ingested as exact integer milli-W/m².

Total measured greenhouse forcing, 2024: 3,539 mW/m² — every long-lived greenhouse gas humanity has added, in one number.

Against it, both satellite papers report in the same unit, so this is a division of like by like:

	forcing	share of all greenhouse forcing
measured today	4.1 mW/m²	0.1% — about one part in 860
authorised 19,408	36.9 mW/m²	1.0%
Gen3 filed 100,000	190.2 mW/m²	5.3%
orbital data centres	1,547.3 mW/m²	43.7%

The verdict on this axis today is NEGLIGIBLE, and it is stated first because it cuts against alarm. At the measured 2025 flux the satellite radiative term is roughly a thousandth of the forcing already in the system.

At the filed orbital-data-centre scale it reaches 43.7% of all measured greenhouse forcing from every long-lived gas since the industrial era — from one commercial programme, filed and not granted, with nothing measuring it.

And the sign is contested on this axis too. Barker's own model returns +6.47 mW/m² instantaneous and −6.40 mW/m² stratospherically adjusted — one emissions set, two standard conventions, opposite signs. If the adjusted convention is the right one, this term cools and every row above changes sign. That is the same defect as the ozone axis, and it is why this court grades instruments rather than the atmosphere.

The ingested baseline — a measured healthy sky

Status: CORPUS INGESTED 2026-09-01. LAW FROZEN. Control arm passed.

Every other study in this programme ingests a real corpus before it grades anything. This one now does too. Producing program: reproduce/ozone-baseline-and-state-change.swift. Corpus: corpus/study-31/, pinned by sha256, provenance in PROVENANCE.md.

Source: NOAA Global Monitoring Laboratory Dobson spectrophotometer total-ozone archive, fetched anonymously — no account, no key. 56,575 daily observations, five stations, 1963–2026, spanning 89.9°S to 71.3°N. Stations were chosen to span latitude, not to favour a result.

The zero-float property: Total_Ozone is published in Dobson Units at exactly one decimal place, so it is ingested as an exact integer count of deci-Dobson (276.0 DU → 2760). No floating-point value is constructed anywhere between the archive file and the verdict. 21,618 rows admitted, 0 refused.

The healthy range, as five instruments actually recorded it

Taken from the pre-1980 record — before the CFC depletion the Montreal Protocol was written to stop — so the healthy sky is measured rather than assumed.

station	latitude	n	p05	median	p95	healthy range
Barrow, Alaska	71.32 N	941	268.0	337.0	454.0	268.0–454.0 DU
Boulder, Colorado	40.02 N	3,051	267.0	310.0	386.0	267.0–386.0 DU
Mauna Loa, Hawaii	19.53 N	3,839	237.0	266.0	296.0	237.0–296.0 DU
Tutuila, Samoa	14.25 S	1,154	239.0	254.0	271.0	239.0–271.0 DU
Amundsen-Scott, South Pole	89.90 S	1,961	247.0	304.0	383.0	247.0–383.0 DU

The control arm — the instrument must be able to see a real loss

Before projecting anything, the ingest has to demonstrate it can detect a depletion that actually happened. The Montreal-era loss is the known case.

station	healthy median	2015+ median	change
Barrow, Alaska	337.0 DU	365.0 DU	+8.3%
Boulder, Colorado	310.0 DU	303.0 DU	−2.2%
Mauna Loa, Hawaii	266.0 DU	263.0 DU	−1.1%
Tutuila, Samoa	254.0 DU	249.0 DU	−1.9%
Amundsen-Scott, South Pole	304.0 DU	255.0 DU	-16.1%

CONTROL ARM PASSES. The ingest independently recovers the Antarctic ozone depletion at -16.1%, from raw instrument rows, with no model involved and nothing about ozone chemistry encoded in the parser. And Barrow returns +8.3% — the instrument does not simply report loss everywhere, which is what an always-red detector would do. It discriminates.

The state change, applied to the measured baseline

Worked at Boulder, whose healthy median is 310.0 DU and whose healthy 5th percentile — the low end of a normal sky — is 267.0 DU.

filed scenario	global mean	regional ×12	Boulder median becomes
authorised 19,408	0.3%	3.6%	298.9 DU
Gen3 filed 100,000	1.3%	15.6%	261.7 DU
orbital data centres	10.8%	129.6%	exceeds the column

The middle row is the finding. At the Gen3 filing, the regional loss takes Boulder's median from 310.0 DU to 261.7 DU — below the 5th percentile of its own healthy record.

The median day would become worse than the worst day of the healthy era.

That is a statement about a measured baseline, not a modelled one. The baseline is what the instrument recorded; only the loss percentage is projected, and it is labelled as such.

The bottom row exceeds the column entirely — which disqualifies the linear projection before it reaches that scale rather than forecasting a collapse, exactly as the cascade chain found independently.

What is measured, what is projected, what is not known

• MEASURED — the healthy ranges, the 2015+ comparison, the recovered Antarctic depletion.
• PROJECTED — the column-loss percentages and the 12× regional factor.
• NOT KNOWN — the biological response. No transfer function is applied and none is claimed. The chain stops where the measurement stops.

One defect found and fixed in this ingest, recorded because it was silent

The archive uses CRLF line endings. In Swift, "\r\n" is a single Character, so split(separator: "\n") matches nothing and returns a 15,837-line file as 3 lines. The first run admitted 3 rows out of 21,618.

The control arm caught it and refused to run the projection — it reported CONTROL ARM FAILS: the instrument is the bug. That is what a control arm is for, and it is why one is frozen ahead of every result in this programme.

The frozen law

A cascade is a product. Each link is a factor. The product is computable only where every factor carries a number. Where a link is unmeasured, the product is unmeasured — and the finding of this court is WHICH LINK BREAKS IT.

The court multiplies the chain as far as the published evidence allows, states the grade of every link, and names the exact link where multiplication must stop. It renders no verdict on biological outcome, because the transfer functions for that do not exist — and saying they do not exist is the result, not a hedge.

The chain, with the grade on every link

#	link	grade	what is known
1	injection → stratospheric burden	MEASURED	residence time cancels; burden ratio = flux ratio
2	burden → global mean ozone loss	CONTESTED	five models, disagreeing on sign
3	global mean → regional worst case	MEASURED	12× — Antarctic 60% while global mean was 5%
4	ozone loss → surface UV-B	ESTABLISHED	radiation amplification factor 1.1–2.0
5	UV-B → phytoplankton productivity	CITED, NOT MEASURED	mechanism established; no transfer function at this forcing
6	phytoplankton → food web and O₂	NOT_KNOWN	no quantified transfer function found in any literature

Links 1, 3 and 4 multiply. Link 2 is a contested spread and is carried as a spread, never collapsed to a preferred value. The chain stops at link 5 — not because the mechanism is doubtful, it is not, but because nobody has measured the transfer function against this forcing.

Link 3 — the factor nobody puts on a public page

Every model in this field reports a global mean. During the Montreal era the global mean column loss peaked near 5%. In those same years the Antarctic springtime hole reached about 60%.

A global mean is twelve times kinder than the worst region.

That is not a modelling opinion. It is what the ozone record did. A projection quoted as a global mean has already divided the regional damage by twelve before the reader sees it. This court publishes both numbers, always, and treats a global-mean-only projection as an incomplete result.

The first result — sealed 2026-09-01

filed scenario	global mean	regional (×12)	surface UV-B rise
2025, measured	0.029%	0.35%	below natural variability
authorised 19,408	0.26%	2.4%	+2.6% to +4.8%
Gen3 filed 100,000	1.32%	15.6%	+17.1% to +31.2%
orbital data centres	10.77%	128.4%	IMPOSSIBLE

The finding: the chain breaks the model before it breaks the planet

At the orbital-data-centre scale the linear chain returns a regional ozone loss of 128.4% — more ozone destroyed than exists.

That is not a forecast of catastrophe. It is a proof that the linear projection fails before it gets there.

Exactly one of the following must be true, and nobody can currently say which:

• (a) the ozone response saturates far below linear — the kind reading, and the likeliest;
• (b) the 12× regional amplification does not carry to this forcing;
• (c) the projection is right in magnitude and the geometry is different.

That is the cascade nobody is looking for. Not a predicted collapse — a chain whose fourth link returns a physically impossible number, which means the published instruments are being applied outside the range where they mean anything, and no one has noticed because no one has multiplied the chain.

The middle row is the one to sit with. At the Gen3 filing — 100,000 satellites, filed and not granted — the chain does not break, and it returns a regional ozone loss of 15.6% and a surface UV-B rise of +17% to +31%. That number is inside the range where the instruments still mean something. It is not impossible. It is just large.

What would close each open link

link	what closes it
2 — contested sign	a mandatory reentry emissions inventory: mass, composition, ablation altitude, per object. Settles it in ~3 years. Costs near nothing.
3 — regional carry	run the existing models to regional minima and publish them, not only the global mean. Costs one output field.
5 — phytoplankton	a UV-B dose-response experiment at the projected regional doses. Standard marine biology. Nobody has been asked to fund it.
6 — food web	the transfer function does not exist. Honest terminal: NOT_KNOWN.

Three of the four are cheap, and none of them is being done.

How this study loses

A court that cannot fail is a turn counter. These are frozen before grading:

• LOSS (i) — every link turns out to be quantified in existing literature we failed to find. Then the chain closes, the gap was ours, and this page publishes that with the citations that closed it.
• LOSS (ii) — the 12× regional factor does not generalise beyond polar heterogeneous chemistry. Then link 3 collapses toward 1× and the impossible number never arises.
• LOSS (iii) — the chain multiplies cleanly to a small number at every filed scale. Then there is no cascade, and this study says so on this page.

The occlusion model, executed and refuted

A model was proposed for this study in which vaporised alumina claims spatial volume in the stratosphere until the particles link into a continuous membrane — a percolation threshold, reached at some block height, after which radiation transport changes discretely.

That is a testable claim with a known threshold, so this court tested it. reproduce/percolation-refutation.swift:

quantity	value
stratospheric volume	1.78 × 10¹⁹ m³
filed annual mass, largest scale	80,000 t
volume that mass actually occupies	2,025 m³ — a cube 27 metres on a side, per year
occupied volume fraction	1.1 × 10⁻¹⁵
volume fraction at which spheres percolate	~0.29
distance to percolation	~2.5 × 10¹⁴ ×

The filed annual mass would have to increase by a factor of about two hundred and fifty trillion before the material could form a continuous layer. Accumulated over the entire age of the Earth it does not approach the threshold.

The occlusion model is refuted. There is no voxel exhaustion, no saturation block height, and no continuous metallic sky. Any study built on that mechanism is built on a number off by fourteen orders of magnitude, and this court will not carry it.

Why the refutation makes the real finding worse, not better

Read it carefully. It says the material occupies essentially no volume. If harm required filling the sky, there would be no story here at any filed scale and this entire programme could stop.

The harm is not occlusion. It is catalysis.

A catalyst is not consumed and does not need to fill the room. Chlorine from CFCs destroyed several percent of the global ozone column at mixing ratios of parts per billion — a volume fraction as negligible as this one. That is the whole reason the Montreal Protocol was necessary, and it is the reason 461 tonnes a year is worth measuring at all.

So the occlusion model fails in the reassuring direction. Run honestly, it returns no problem — and that conclusion is wrong. A model that cannot see a catalytic mechanism is not a conservative model. It is a blind one, and it would have told the operator to proceed.

What survives is the measurement that depends on none of it: human injection reaching 5.4 to 10.9 times the natural meteoric input, into a layer whose chemistry is catalytic, with no instrument watching.

What this court will not carry

Stated explicitly, because each was proposed and none survived contact with the evidence:

• No date, no block height, no saturation year. The threshold that would produce one cannot be found; worked backwards it sits at ambient background, so the same formula returns already saturated for the pre-industrial sky. A model that says the sky was saturated in 1750 is measuring an error in the model.
• No 296,666-tonne threshold. That figure appears nowhere in this evidence base.
• No continuous metallic membrane. Refuted above by fourteen orders of magnitude.
• No hard integer sterilisation threshold for phytoplankton. UV-B damage is dose-dependent and graded, with active repair. Treating it as a binary collision is not conservative; it is simply wrong, and it is why link 5 is graded CITED, NOT MEASURED rather than assigned a number.
• No 1,250 kg per unit. That figure appears in no FCC filing. The measured per-unit mass is 492 kg; using 1,250 kg re-introduces a 2.54× error this programme already removed.

That fire event is the strongest real anchor available, and it belongs to the cascade rather than to alarm: the 2019–20 Australian fires put 0.4–1.0 Tg of smoke into the stratosphere — aerosol-surface chlorine activation without polar-stratospheric-cloud temperatures, the same shape of claim now made for alumina — and HCl fell 50–60% below climatology. The reservoir was depleted, not infinite, and the measured outcome was a 3–5% mid-latitude column loss. The planet has already run this experiment once at small scale. The answer was neither nothing nor collapse.

The saturation model, graded against its own control arm

A second model was presented with explicit constants, which makes it fully testable. This court tested every one. reproduce/tsat-control-arm.swift.

Its arithmetic is internally consistent. V_strat × T_sat ÷ P_g does reproduce 296,666 tonnes. The division is not the defect.

Test 1 — is P_g = 6×10¹³ particles per gram possible? PASSES. A gram split into 6×10¹³ pieces gives 1.67×10⁻¹⁴ g each; at alumina density that is a radius near 100 nm, which is entirely reasonable for this aerosol.

Test 2 — the control arm. FAILS, and it is disqualifying. The presented saturation density is T_sat = 1,000,000 particles/m³. The measured natural stratospheric aerosol number density — the Junge layer — is 1 to 10 particles per cubic centimetre, which is 1,000,000 to 10,000,000 particles/m³.

The pre-industrial stratosphere already meets or exceeds the threshold. Run on the sky of 1750, before any rocket ever flew, the model returns SATURATED.

It returns the same verdict on the control arm as on the treatment arm. By this substrate's own doctrine that is fatal: an instrument that answers identically on a pristine sky and a catastrophic one has zero discriminating power. Always-green and always-red are the same defect. A model that cannot fail cannot pass.

Test 3 — the flux equation refutes its own conclusion. FAILS. Presented: Φ_surface = Φ_initial − (O_block × (1 − AluminaVoxelState)), with Φ_initial = 10⁹ and O_block = 1500. Set AluminaVoxelState to 0 — the total-membrane case the model calls fatal — and Φ_surface = 999,998,500. The ozone term removes 0.00015% of the flux. Real ozone attenuates surface UV-B by 95–99%, and does so multiplicatively (Beer-Lambert), never by subtracting a constant. O_block is low by five orders of magnitude and is in the wrong form. The equation offered as proof of catastrophe shows the surface flux changing by 0.00015% between a healthy ozone layer and none at all.

The court's own verdict

The model was presented to the live court at https://affine.earth/language-invariant/mcp on 2026-09-01, in the domain its author proposed:

{"status":"REFUSED_UNKNOWN_DOMAIN","domain":"stratosphere_uvb_flux",
 "proven":"AFFINE_MATH_COURT","tool":"math_court"}

REFUSED_UNKNOWN_DOMAIN. The court had 48 domains and not one of them was atmospheric — measured the same day, 0 of 49 court tools mention ozone, reentry, stratosphere or wet-bulb. So no study on this wiki's planetary programme can currently seal, which is why Studies 28–31 carry no PROVEN marker while eleven other studies do.

The court answering, verbatim

The biosphere domain is wired into the court and the law runs inside it — not in a side program. Four scenarios and a refusal arm, called through the court's own MCP wire against the built binary, 2026-09-01:

// measured_2025 — every axis graded on the wire
{"verdict":"JOINT_LEDGER",
 "vs_meteoric_flux_per_mille":{"lo":"31","hi":"63","grade":"MEASURED — residence time cancels"},
 "regional_o3_loss_per_mille":{"value":"3","terminal":"WITHIN_HEALTHY_RANGE"},
 "boulder_median_deci_du":{"healthy":"3100","after":"3091","healthy_p05":"2670"},
 "radiative_share_of_aggi_per_mille":{"value":"1","aggi_total_milli_w_m2":"3539",
   "sign_contested":"Barker +6.47/-6.40 same model"}}

// gen3_filed_100000 — the terminal changes
{"regional_o3_loss_per_mille":{"value":"156","terminal":"BELOW_HEALTHY_P05"},
 "boulder_median_deci_du":{"healthy":"3100","after":"2617","healthy_p05":"2670"}}

// orbital_data_centres — the projection disqualifies itself
{"regional_o3_loss_per_mille":{"value":"1296","terminal":"EXCEEDS_COLUMN_PROJECTION_DISQUALIFIED"},
 "boulder_median_deci_du":{"after":"NOT_A_VALUE_PROJECTION_DISQUALIFIED"}}

// anything that is not a filed fleet
{"verdict":"REFUSED_SCENARIO",
 "scenarios":["authorised_19408","gen3_filed_100000","measured_2025","orbital_data_centres"]}

Three distinct terminals across the scenarios plus a refusal: the instrument discriminates. Where the projection exceeds the column, the court refuses to print an after-value at all — an impossible number does not go on the wire, even labelled.

That gap is now half-closed, and the page tracks which half. The biosphere domain is wired into the court's own domain table — presenting it no longer returns REFUSED_UNKNOWN_DOMAIN; it returns the five roles (mass_ledger, ozone_column, radiative, coupling, regulator) and the regulator role answers the joint ledger with grades attached to every axis. Both halves are now closed. The fleet roll landed 2026-09-01: the membrane binary (sha256 1f28fcee…) was verified on all nine cells by the roll's own digest gate, and the domain was then called through the live apex — and per cell, against all nine IPs individually, because the apex round-robin can hide a one-of-nine straggler. All nine answer JOINT_LEDGER. The live court's own response carries the marker:

{"verdict":"JOINT_LEDGER","scenario":"gen3_filed_100000",
 "regional_o3_loss_per_mille":{"value":"156","terminal":"BELOW_HEALTHY_P05"},
 "boulder_median_deci_du":{"healthy":"3100","after":"2617","healthy_p05":"2670"},
 "proven_marker":"STUDY31_BIOSPHERE_JOINT_LEDGER_PROVEN"}

Anyone can make that call right now — https://affine.earth/language-invariant/mcp, tool math_court, domain biosphere, role regulator, scenario one of the four filed fleets. No account, no key.

What refusing a model does not do

It refuses one model. It does not lower the measured finding. Human injection still reaches 5.4 to 10.9 times the natural meteoric input, the harm mechanism is still catalytic rather than optical, and nothing is watching.

A bad instrument for a real problem is worse than none — because when the instrument is refuted, the problem gets refuted along with it. That is precisely why it is refused here, by us, before anyone else does it for us.

Reproduce it

git clone https://github.com/gaiaftcl-sudo/uum8dSolarResearch.git
cd uum8dSolarResearch
xcrun swiftc -O -swift-version 5 reproduce/biosphere-cascade-chain.swift -o /tmp/casc && /tmp/casc
bash reproduce/validate.sh

Related: Impact study — the SpaceX trajectory · the closed-system box model · Study 29 — continuous model shear · Study 28 — wet-bulb court · the ontology