A Planet's Bookkeeping

A body goes into the roots.

The ceremony around it is elaborate and the ceremony is not what I want to look at. What I want to look at is the weeks afterward, when the singing has stopped and everyone has gone, and something entirely unremarkable happens: the body stops being a body. Its carbon leaves. Its nitrogen leaves. Its phosphorus, which was locked into every strand of its DNA and every membrane of every cell, leaves. In a warm wet forest this takes a season at most, and at the end of it there is nothing there to bury.

The Na'vi have a sentence for this. Jake Sully repeats it early, in the flat voice of a man quoting something he has not yet understood:

“

They see a network of energy that flows through all living things. They know that all energy is only borrowed — and one day you have to give it back.

Read it as theology and it is a lovely thing to say at a funeral. Read it as a claim about the physical world and it turns into something sharper: an assertion that nothing here is consumed, that the totals are conserved, that every atom in that body was on loan from a pool it must eventually return to. Stated that way it stops being a sentiment. It becomes an accounting statement — and accounting statements can be checked.

So let us check it.

This chapter is an audit. We are going to open Pandora's books, take canon's own published numbers as the entries, and see whether they balance. It requires exactly one piece of arithmetic, which I will build in a moment and which will then do all the work for the rest of the chapter. Applied to four elements it produces four answers. Three of those answers are startling. One of them, I think, cannot be made to work at all — and finding that out is far more interesting than being told the forest recycles beautifully.

The only arithmetic you need

Here is the instrument. It is embarrassingly simple, and it is the reason planetary chemistry is a quantitative science instead of a set of stories.

Pick any substance and any place it collects — carbon in the air, salt in the sea, water in a lake. Call how much sits there the reservoir, measured in tonnes or gigatonnes or moles, whatever is convenient. Call the rate at which it leaves the flux, measured in the same units per year. Divide the first by the second:

τ = M / F

What comes out is a time — the average span an individual atom spends in that reservoir before something removes it. This is the residence timeresidence time, and Pandora's Water and Weather already put it to work once: Earth's entire atmosphere holds a layer of water only two and a half centimetres deep, rains about a metre a year, and therefore turns its whole water content over every nine days. The sky is not a tank. It is a conveyor.

Now the move that makes this an instrument rather than a curiosity. Run the division backwards.

If you know how much of something a reservoir contains, and you know how fast the world destroys or removes it, then you also know how fast something must be replacing it — because if replacement did not match removal, the reservoir would be visibly draining. Rearranged:

F_source = M / τ

That is the whole trick. A concentration you can measure, plus a lifetime you can calculate from chemistry, gives you a source flux you never observed. It converts a static fact about how much of something is present into a hard requirement about the machinery that keeps it there. And if no plausible machinery can meet the requirement, then one of your inputs is wrong, and you have learned something.

Two things about that spread are worth carrying forward.

The first is that residence times are wildly unequal, even for the same element. A carbon atom in the air stays a few years; a carbon atom in seafloor limestone stays hundreds of millions. Same element, same planet, two reservoirs whose behaviour has nothing in common. Fast reservoirs respond instantly to change and store nothing. Slow reservoirs store enormous quantities and take geological time to notice anything has happened. When someone tells you a planet's chemistry is stable, the useful question is always: stable in which reservoir, and paid for by which flux?

The second is subtler and it is where popular accounts most often go wrong. The time an individual molecule spends in a reservoir is not the same as the time it takes a disturbance to that reservoir to fade. An individual CO₂ molecule leaves Earth's atmosphere in about four years, swapped out by photosynthesis and air-sea exchange. But those exchanges hand it straight back — they shuffle carbon between boxes rather than removing it. An excess of carbon relaxes only as fast as the slow permanent sinks can bury it, which takes a hundred thousand years and more. Four years and a hundred thousand years, both correct, describing different questions. Conflating them is the single most common error in writing about carbon, and keeping them apart is most of what it takes to reason well about any cycle.

Instrument built. Now the specimen.

Weighing the sky

You can weigh an atmosphere from the ground without ever leaving it.

The pressure you feel at sea level is the weight of the entire air column standing on every square metre above you. So the total mass of an atmosphere is just that pressure, multiplied by the surface area of the world, divided by its surface gravity. Three numbers, all of which canon supplies for Pandora: a surface pressure around nine-tenths of Earth's, a radius of about 5,720 kilometres, a surface gravity of about 0.8 g.

Run it and Pandora's atmosphere comes out at roughly 4.8 × 10¹⁸ kilograms — about ninety-three percent of Earth's, which is unremarkable and rather satisfying. A slightly smaller world holding slightly less pressure ends up with about the same amount of air.

Then you apply the composition, and it stops being unremarkable.

What’s Really in the Air? went through what Pandoran air does to a human body and why twenty seconds is the number that matters. I am after something different: not what the air does, but how much of it there is. Canon's published breakdown puts carbon dioxide at roughly eighteen percent of the atmosphere. On Earth it is four hundredths of one percent.

Two hundred and eighty-five thousand gigatonnes of carbon, hanging in the air. That is not a slightly enriched atmosphere. That is more carbon in Pandora's sky than exists in the whole of Earth's atmosphere, vegetation, soils and surface ocean combined, several times over. And canon is emphatic that the moon has been like this for a very long time — the Na'vi are described as morphologically and genetically stable for something on the order of twelve million years, individual Hometrees stand for twenty thousand, and nothing in the record hints at a biosphere in the middle of a chemical transition.

So the audit has its first entry, and it is a large one. Roughly 285,000 gigatonnes of carbon, in a reservoir alleged to have held steady for twelve million years.

Which means there had better be a source and a sink, and they had better be equal to about twelve significant figures, because a mismatch of even a tenth of a percent per year empties or doubles that reservoir in a geological instant.

The fast loop and the slow one

Carbon on any living world runs two circuits at once, and almost everything interesting comes from the difference in their speeds.

The fast circuit is biology. Photosynthesis pulls carbon out of the air; respiration and decay push it back. On Earth these two flows are enormous — around 120 gigatonnes a year fixed by land plants, most of it returned within months by the plants' own respiration and by the microbes eating what falls. Add the ocean's two-way gas exchange, another ninety-odd gigatonnes each way, and something like two hundred gigatonnes of carbon crosses in and out of the atmosphere every year.

Against a reservoir of 875, that is a residence time of about four years. The air's carbon is completely replaced within a presidential term.

Now look at what those vast flows actually accomplish. Gross uptake: 120. Plant respiration back out: 60. Microbial decay of everything that fell: 57. Net retained by the land, after all that: two or three gigatonnes. The gross flows are forty times larger than the net result, and the net result is what determines the atmosphere's future.

This is the structural fact that makes carbon accounting hard and worth doing. You are trying to measure a small difference between two huge, noisy, seasonally swinging quantities. It is like weighing a letter by weighing a truck with and without it. But that small difference is the entire story, because it is the only part that accumulates.

The slow circuit is geology, and it is tiny. Some fraction of the carbon that life fixes never gets eaten: it sinks, gets buried in sediment, and is compressed into rock. Earth buries something like two-tenths of a gigatonne of carbon a year this way — a rounding error against the fast loop's two hundred. Volcanoes and metamorphism return roughly the same amount, exhaling carbon that was buried tens of millions of years earlier.

Two-tenths of a gigatonne a year, against a sedimentary carbonate reservoir north of sixty million gigatonnes. Residence time: hundreds of millions of years.

And here is the point of separating them. The fast loop is huge and nearly balanced, so it can shuffle carbon around quickly but cannot change the totals. The slow loop is minuscule and not required to balance in the short run, so it sets where the totals end up over millions of years. Climate on geological timescales is decided by the small circuit, not the large one.

The thermostat that should have eaten it

There is a machine on Earth that removes carbon dioxide from the air, and its most remarkable property is that it works harder when the planet is warmer.

Rain dissolves a little CO₂ and becomes a weak acid. That acid falls on rock. Where the rock is silicate — most of a continent's crust — the acid slowly takes it apart, and the products wash down rivers to the sea as dissolved calcium and bicarbonate. Marine organisms take those ions and build carbonate shells. The shells sink. Over millions of years, the carbon that started in the air ends up as limestone on the seafloor, and eventually gets subducted, cooked, and breathed back out by volcanoes.

That closes the loop, and by itself it is only a loop. What makes it a thermostat is the temperature dependence. Chemical reactions run faster when warm, and a warmer world also runs a more vigorous water cycle with more rain flowing over more rock. So silicate weatheringsilicate weathering accelerates as the planet heats — pulling down more CO₂, thinning the greenhouse, and cooling it back. Let the planet cool and weathering slows, volcanic CO₂ accumulates unopposed, and the greenhouse thickens again.

A negative feedbacknegative feedback with no thermostat in it. Nothing measures the temperature; nothing decides anything. The regulation is a property of the chemistry.

This is not a speculative mechanism. It is the accepted answer to the faint young sun paradoxfaint young sun paradox — the awkward fact that the early Sun was some twenty-five to thirty percent dimmer than today, cold enough to freeze Earth's oceans solid, and yet the geological record shows liquid water and life throughout. The carbonate–silicate cyclecarbonate–silicate cycle held the planet in the habitable band by letting CO₂ pile up under a weak sun and drawing it down as the sun brightened.

Now bring that machine to Pandora and notice what it does to our audit.

Weathering rate depends on temperature, on rainfall, and on the CO₂ pressure driving the acid. Pandora, from everything canon shows, is warm and wet essentially everywhere: rainforest across enormous areas, no depicted dry season, persistent cloud and fog. And its CO₂ pressure is not slightly elevated but roughly four hundred times Earth's. Every single dial that accelerates silicate weathering is turned up hard.

Which means Pandora's atmospheric carbon should be draining fast. Even on Earth, with its far gentler forcing, the weathering feedback relaxes a carbon perturbation on a timescale of a hundred thousand to a million years. Load the dials the way Pandora loads them and the drawdown time gets shorter, not longer.

Canon's biosphere has been chemically stable for twelve million years. That is at least ten thermostat time-constants, and plausibly a hundred.

The most abundant thing that nobody can have

Nitrogen makes up over half of Pandora's air and nearly four-fifths of Earth's. It is, by a wide margin, the most abundant element available to any organism standing anywhere on either world.

It is also, on both worlds, one of the two things most likely to be limiting how fast anything grows.

That combination sounds like a contradiction and it is the whole lesson. Nitrogen gas is two nitrogen atoms held together by a triple bond, and that bond costs 941 kilojoules per mole to break — among the strongest in ordinary chemistry. So the atmosphere is not a nitrogen supply. It is a nitrogen vault: full to the ceiling, sitting in the open, and effectively sealed. Every organism on Earth is standing in a room that is four-fifths food it cannot eat.

Life found one key. A single enzyme, nitrogenasenitrogenase, can pry that bond apart, and it is a spectacularly expensive tool to own. The reaction consumes about sixteen ATP per molecule of nitrogen reduced — an enormous metabolic bill. Worse, the iron-and-molybdenum cluster at the enzyme's heart is destroyed permanently by oxygen, so an organism that wants to run it must also build and maintain an oxygen-free compartment to run it in, on an oxygen-bearing planet.

This is why nitrogen fixationnitrogen fixation is rare, patchy, and always somebody's expensive specialisation rather than a general capability. And it is why nitrogen limits growth in most ecosystems: not because it is scarce, but because it is costly. Scarcity and cost feel similar from the outside and behave completely differently. A scarce resource runs out. A costly one is abundant forever and available only to whoever can pay.

Now Pandora, where this gets uncomfortable.

Canon shows a biosphere of extraordinary productivity: forests with canopies well over a hundred metres, trees whose trunks are tens of metres across, standing biomass that on any reasonable reckoning must exceed Earth's terrestrial vegetation by a large multiple. Every gram of that is protein and nucleic acid, and every gram of protein needs fixed nitrogen.

And canon names no fixer. Not one organism, not one mechanism, not one mention of nodules or symbionts or a nitrogen-processing tissue anywhere in the published record. There is also no mention of soil fertility, no fertiliser, no agriculture, no nutrient limitation of any kind — the Na'vi are gatherers and hunters, and nothing in the material describes them managing soil at all.

Two honest observations about that gap. First, lightning is a real abiotic route — the plasma in a discharge channel is hot enough to break N₂ directly — and Pandora has the storms for it, though Earth's lightning contributes only five to ten teragrams a year, a small share of the total. Second, the root network canon does describe is repeatedly said to carry signals, memory, and electrochemical traffic. It is never said to carry nutrients. Those are very different claims and fans routinely merge them; The Wood-Wide Web Revisited deals with what such a network can actually do.

The element with no way home but stone

Three of our four elements have an atmospheric reservoir. Phosphorus does not, and everything about phosphorus follows from that absence.

There is no stable phosphorus gas at planetary surface conditions. None. So while carbon can be pulled from the air by any leaf and nitrogen can be pulled from the air by anything willing to pay nitrogenase's bill, phosphorus has exactly one entrance to the biosphere — the chemical weathering of phosphate minerals in rock — and exactly one exit, sinking into marine sediment.

That single structural difference converts a nutrient question into a geological one. A world's long-term fertility depends on its capacity to lift buried rock back into the reach of rain. Uplift, erosion, exposure: the machinery A Crust That Moves spent a chapter trying to identify. If that machinery stops, phosphorus keeps draining to the seafloor and never comes back, and the biosphere thins out no matter how much sunlight and water and carbon it has.

Life's response to this constraint is not to find more phosphorus but to stop losing any. Old tropical ecosystems on Earth run phosphorus at almost unbelievable efficiency — mycorrhizal networks intercepting it, phosphatase enzymes scavenging it from dead tissue before it can wash away, whole nutrient economies operating in the canopy rather than the soil so a released phosphate atom is recaptured metres above the ground. Canopy soilCanopy soil is not a curiosity; it is a phosphorus-retention strategy.

For Pandora the phosphorus gap in canon is the largest of the four and the least discussed. A forest of that productivity standing on ground that has been forest for twenty thousand years and more has a phosphorus budget that must be either extraordinarily tight or continuously resupplied. Volcanic ash is a genuinely plausible resupply — tephra weathers fast and carries apatite, and Pandora has no shortage of eruptions — but ash falls locally and episodically, and a planet-wide budget cannot be balanced by a regional subsidy without someone doing the arithmetic. Nobody has.

Two days

Now the fourth element, and the place where the audit stops being an audit and becomes a finding.

Canon is specific about hydrogen sulfide in Pandora's air, and it needs the gas for two separate jobs. The careful sources place it below one percent of the atmosphere. And the films need it lethal — part of why an unmasked human collapses in twenty seconds is that H₂S jams the final enzyme of the respiratory chain, the same target cyanide attacks. What’s Really in the Air? follows that mechanism down to the mitochondrion. Both jobs are load-bearing: the concentration and the lethality are each doing narrative work.

Take our instrument to it.

The reservoir first. Below one percent of an atmosphere of 4.8 × 10¹⁸ kilograms is, in sulfur terms, something like 4 × 10⁷ teragrams of sulfur suspended in the sky. Large, but so far only a number.

Now the sink, and this is where it turns. Hydrogen sulfide is a reduced gas, and Pandora's air is eighteen percent oxygen. Reduced gases do not last in oxidising atmospheres. Hydroxyl radicals — which any sunlit, oxygen-bearing, water-bearing atmosphere manufactures continuously — strip H₂S apart within days, and the chain runs on to sulfur dioxide and then to sulfuric acid aerosol. The measured tropospheric lifetime of H₂S on Earth is one to three days.

Call it two.

F_source = M / τ = (4 × 10⁷ Tg S) / (2/365 yr) ≈ 7 × 10⁹ Tg S/yr

Earth's volcanoes, all of them, emit about 20 teragrams of sulfur a year.

The requirement is roughly four hundred million times that.

I want to be careful about what this does and does not show, because a number that large invites showing off, and showing off is not the point.

It does not show that canon is stupid. It shows that two independently reasonable statements — the air contains a fraction of a percent hydrogen sulfide and the air contains eighteen percent oxygen — are chemically incompatible at that ratio, and that nobody noticed because the incompatibility only appears when you multiply the concentration by an atmospheric mass and divide by a photochemical lifetime. None of those three steps is obvious. All three are required.

Nor is it a small overshoot that a generous assumption might absorb. Grant a moon a hundred times Earth's volcanic sulfur output — genuinely generous, and not unreasonable for a body heated the way What Keeps Pandora Volcanically Alive? describes — and you have closed a factor of a hundred out of four hundred million. Sustaining anything near the required rate would also strip the moon's entire mantle sulfur inventory in a geological eyeblink, and bathe the surface in sulfuric acid rain that no canopy would survive.

So something in the stated numbers has to give. There are exactly two places it can give, and both are visible on the two dials above.

Lower the concentration. Drop the sulfide from one percent to a few parts per million and the required flux falls into a range a volcanic world could plausibly sustain. The books close. But watch the lethality readout while you do it: below a few hundred parts per million, hydrogen sulfide no longer drops a human in seconds. You have fixed the chemistry by breaking the toxicology, and canon needs both.

Lengthen the lifetime. If some unrecorded feature of Pandoran air scavenged hydroxyl radicals, H₂S could persist longer and need less replacing. But the lifetimes required are measured in hundreds of thousands of days, not days, and no known atmospheric chemistry does that in the presence of that much oxygen.

Which leaves the repair that actually works, and canon has left the door open for it: the high sulfide is local. Volcanic vents, fumarole fields, sulfurous basins, the Ashlands — regions where the gas is genuinely concentrated and genuinely lethal, sitting inside a global atmosphere carrying only traces. That is exactly how Earth's volcanic regions behave. It preserves every scene canon needs, it costs nothing narratively, and it makes the sulfur budget close. It requires giving up only one thing: the idea that the whole planet's air is at that concentration.

The world that ran out of air

In 1991, eight people sealed themselves inside a glass building in the Arizona desert and tried to live in a closed ecosystem for two years.

Biosphere 2 was a hectare and a quarter of engineered biomes under sealed glass — rainforest, savanna, mangrove, ocean, farm — and its purpose was to see whether a small world could be made to keep its own books. It failed in a specific and enormously instructive way.

The oxygen started falling. Slowly at first, then unmistakably: 20.9 percent at the start, 14.5 percent sixteen months in, which is the equivalent of living at four thousand metres of altitude. The crew were breathing badly and sleeping worse, and eventually oxygen had to be injected from outside — which ended the experiment as a closure test, whatever else it achieved.

Here is what makes it a lesson rather than an anecdote. Everyone could see the oxygen leaving. Nobody could find the carbon dioxide it should have become.

Because that is the constraint: if soil microbes were consuming oxygen by respiring organic matter, then every molecule of oxygen consumed should have appeared as a molecule of CO₂. Oxygen fell by a great deal. CO₂ rose by far less than it should have. Mass balance said flatly that a carbon sink existed which nobody had accounted for, and the discrepancy was too large to argue with.

It took until 1994 to find it, and the answer was the building. Biosphere 2's structure contained a great deal of exposed, unsealed concrete, and curing concrete absorbs carbon dioxide, converting calcium hydroxide into calcium carbonate. The walls were eating the CO₂. Roughly 750 kilomoles of it. Soil respiration was pulling oxygen down and the concrete was quietly removing the evidence, so the loop stayed broken while looking, from the inside, like nothing was wrong at all.

And the deep reason closure matters at all, for Biosphere 2 and for Pandora and for us, is that planets are effectively sealed. Earth gains perhaps forty thousand tonnes a year of meteoritic dust and loses roughly a hundred thousand tonnes to atmospheric escape, mostly hydrogen and helium. Against an atmospheric mass of 5 × 10¹⁸ kilograms, that is a leak of about one part in a hundred trillion per year.

Nothing arrives. Nothing leaves. Recycling is not a virtue on a planet — it is the only available option.

Honest edges

Let me separate what this chapter has established from what it has only argued.

The science is not in question. Residence times, the two carbon circuits, the weathering thermostat, the nitrogen triple bond and nitrogenase's cost, phosphorus's missing gas phase, hydroxyl oxidation of reduced sulfur, the Biosphere 2 post-mortem — all of that is ordinary, measured earth-system science, tested on a planet we can put instruments on.

What canon states is much narrower: an atmospheric composition, a surface pressure, a radius and a gravity, vigorous volcanism, abundant water, an enormously productive biosphere, roughly twelve million years of biological stability, and a funeral rite that returns bodies to the roots. That is close to all of it.

Every reservoir mass in this chapter is my arithmetic from those inputs. Every flux is either an Earth measurement or an inference, because canon supplies no fluxes whatsoever. The sulfur result is the firmest conclusion here, and it rests on canon's own stated concentration meeting canon's own stated oxygen level — but it would dissolve entirely if the intended reading was always that high sulfide is local rather than global, which the text permits and never says.

Borrowed, and spent

Go back to the body in the roots.

The audit is finished and the verdict on the Na'vi precept is not the one I expected when I started. About matter, they are exactly right — not poetically, not approximately, but as a correct statement of physics. Every carbon atom in that body was borrowed. It had been in other bodies before, many times, and it will be in others after. The planet is a closed system to a precision of one part in a hundred trillion, so there is no other way it could work. Nothing is consumed. Nothing is used up. The books close, always, on atoms.

And about energy they are wrong, in a way that is worth being precise about, because the error is the most interesting thing in the sentence.

Energy is not borrowed. Energy is spent.

It arrives as sunlight — photons from two stars, few in number, each carrying a lot of energy, which is to say arriving in a highly ordered, low-entropy form. A leaf captures some. It does work: builds sugar, then wood, then a body, then a thought. And at every single one of those transfers most of it degrades into heat, and the heat radiates away into space as infrared: many photons, each feeble, thoroughly disordered.

That flow runs one way and only one way. The atoms make a ring; the energy makes an arrow. And the arrow is what pays for everything — all the order, all the structure, all the improbable persistence of a forest, is funded by the gap between the sunlight coming in and the heat going out. A biosphere does not keep its entropy low by avoiding entropy. It keeps it low by exporting the entropy outwardexporting the entropy outward, continuously, forever.

Which means the forest above that grave is not a system that has learned to need nothing. It is a system that needs one thing absolutely and continuously, and it is not matter. Pandora could recycle its atoms perfectly for another twelve million years and still be dead within days if the light went out.

Matter is only borrowed, and one day you have to give it back. That part is exact.

The energy was never a loan. It was a gift that arrives every morning, is spent by evening, and does not accumulate — and the debt the Na'vi describe so carefully is the smaller of the two obligations a living world is under.