Ecosystems keep time in centuries. A glacier is ice laid down flake by flake over tens of thousands of winters, and can lose a decade’s accumulation in one hot June. An aquifer fills over millennia and can be pumped dry inside a single working life. A mangrove takes centuries to grow the root mass that breaks a storm surge, and loses it to a bulldozer in an afternoon.
Our institutions do not keep this time. A budget runs a year, a government a term, a long-term fund at its most patient a generation. We built financial systems for the quarter and called it foresight.
This mismatch cannot be closed by asking charity or the market to guard something permanent. Permanence of finance is not permanence of substance. You can endow forest protection in perpetuity and still permit a slow, fully authorised liquidation of the forest itself — a percentage a year, each one individually defensible — until the forest is gone and the fund is intact.
This essay’s claim, stated plainly: some ecological systems should be treated not as capital but as principal — assets whose destruction cannot be compensated for by money, restoration elsewhere, or future growth. That is a claim about accounting, but not only about accounting. It asks environmental law and development assessment to make room for a category of asset they do not currently have. Underneath, it rests on an argument about what one generation owes the next. What follows tests that claim against real cases, then asks what a regime built to honour it would require.
The distinction underneath it is older than economics: capital we own, and capital we merely hold. For most of economic history, nature entered the accounts only when extracted — a forest as timber, a river as water drawn from it, a fishery as catch on a dock.
Natural-capital accounting put nature back on the page, and rightly so. But a balance sheet only tells you what a thing is worth. It does not tell you what you are allowed to do with it.
An endowment settles that question by separating two things a balance sheet does not: a return that can be spent, and a principal that is not supposed to move. A university does not fund this year’s scholarships by selling the buildings that generate them.
Some of nature works the same way a forest yields timber. Standing, it is also a watershed and a windbreak. A wetland yields nothing you can put on a truck, and still purifies water, breaks storm surges, and shelters much of a coastal economy’s fish stock. A glacier yields meltwater is a river’s dry-season pulse, the basis of an irrigation calendar whole valleys are built around. But the ice itself is the principal. Draw it down faster than snowfall replaces it, and the yield does not shrink, it disappears, and the river disappears with it. A grassland’s visible yield is forage. Its principal is the root mat and soil carbon beneath, doing more ecological work than the grass anyone sees. A mangrove yields little that shows up as revenue. Its principal — the reef of roots along a coastline — is what stands between a neighbourhood and the next storm surge.
In each case, the yield can be taken. The principal cannot be sold without ending the endowment itself. Call the part that cannot be sold a Permanent Ecological Asset. Not everything called an ecosystem qualifies, and not even whole ecosystems qualify uniformly — a forest’s timber yield may be genuinely regenerative even where its old-growth structure is not. The more honest version of this category names functions and components, not whole landscapes. “Permanent” does not mean unchanging. Glaciers move, forests migrate, climate is rewriting what any of these places will look like in a century. It means: cannot be recreated, restored, or substituted within any timeframe that matters to the people or systems that depend on it now.
Take the Mumbai mangrove belt as one attempt at that boundary. Not every patch of mangrove along the coast would qualify — isolated, already-fragmented stands with no continuous root structure do not perform the surge-breaking function a Permanent Ecological Asset requires, and treating them as inviolable would freeze development against no real ecological benefit. The stretches that do qualify are the ones that still function as a continuous barrier — root structure unbroken enough to blunt a storm surge before it reaches the settlements behind it. Drawing that line costs someone something today: a road takes a longer route, a port pays more to reclaim land elsewhere, a fishing community loses a shortcut it has used for a generation. None of that cost disappears by naming the asset permanent. It only moves — from the ecosystem, which cannot bargain for itself, to the people and institutions doing the naming, who can.
Fixing that boundary precisely is a scientific and political judgment, not a formula. But the difficulty of drawing a line is not a reason to refuse to draw one — we already separate income from capital, private land from commons, what may be sold from what is only ever held in trust.
That judgment still needs a process, not only a principle. Who decides where the line falls — a ministry, a court, an independent scientific panel — shapes the boundary as much as any ecological criterion does, since each carries its own incentives and its own blind spots. The evidentiary standard matters just as much: whether a stretch of mangrove or a glacier’s mass balance must be shown irreplaceable beyond reasonable doubt, or only more likely than not, decides how much gets protected before the fact rather than mourned after it.
And whoever loses that argument — the planner told to reroute a road, the community denied a shortcut it has used for a generation — needs a real appeal, heard by a body other than the one that drew the line in the first place. Without that, the designation lasts only as long as the officials who made it, and is only as durable as the next change of government.
This is where our method of judging development runs into trouble first — and it is a narrower trouble than “assessments only look at one site.” Cumulative and regional impact assessments exist precisely to weigh a region’s carrying capacity rather than a single project’s footprint, and on paper they catch exactly this problem. In practice they are the exception: triggered at a regulator’s discretion, bundled into a plan or corridor study rather than required for every subsequent clearance, and built on a baseline rarely revisited as approvals accumulate against it.
A mine has a lease boundary, a factory a fence, a road a right of way — and most clearances are still granted at that scale. Groundwater does not know where the line is; a migratory bird crosses it twice a year without noticing; a river carries whatever enters it upstream to whoever lives downstream, unbilled. Where a cumulative or regional assessment was done, it was usually done once — for the corridor or the basin — and not kept current as each subsequent, lawful decision draws on the capacity it once measured. The instrument for catching a systemic loss exists; what is missing is the discipline of applying it to every decision, and updating it as the ground shifts beneath it.
Mumbai makes that gap visible. The city has run the experiment for three decades, one project at a time. Its mangrove belt has been cleared in increments — for reclamation, for informal settlement, for the airport’s expansion, now for infrastructure — each clearance judged against its own footprint, with no single, current, standing assessment of the belt as a whole. In 2026, India’s Supreme Court cleared the way for a coastal road project — the Versova–Bhayandar coastal road, for example — to remove more than forty-five thousand mangrove trees along the western shoreline — the same belt that Koli fishing families and coastal engineers describe as the city’s first line of defence against tidal flooding.
Considered alone, the approval was unremarkable: an appraisal weighed the trees against the road, found the road’s public benefit larger, and approved it. It is the same procedure coastal cities everywhere use to trade a standing buffer for a road, a port, a housing block, one project at a time. What no single appraisal is built to see is the belt itself — what this clearance means added to every one before it, on a coastline where several remaining mangrove patches are already too fragmented to function as a continuous barrier.
Mumbai’s 2005 flooding, which killed many, is usually explained by heavy rainfall, undersized drains, and a built-up floodplain — mangroves are one line in a longer account, not the whole of it. But the loss of a wave-breaking, surge-absorbing coastline belongs on that account. The protective function of a healthy mangrove belt is well established in coastal science; a city that keeps it off the ledger is not proving the mangroves didn’t matter. It is choosing not to look until a storm makes the omission visible. The mangrove’s price does not change across any of these approvals. Only the coastline’s protection does — a little more with each permit — until a city discovers what it spent only when the water arrives to collect.
Hydrologists have started reaching for this same language themselves. An assessment by the UN University’s Institute for Water, Environment and Health described several river basins as being in a state of water bankruptcy — persistent overdraft beyond what rain and rivers can replace, with losses in some basins judged beyond recovery. That is not a metaphor borrowed for effect. It is the plainest way a hydrologist has to say some of this water has already been spent past the point where spending it further means anything.
Assessments also run on the calendar of the regulator, not the calendar of the system. A little groundwater drawn down this year looks like nothing. Drawn down for twenty years past the rate of recharge, it looks like a dead well. A forest can absorb fragmentation for decades, then — past a threshold nobody marked on a map — stop recovering between disturbances.
The Himalaya show the same pattern at greater scale. At Langtang Lirung, on the Nepal–Tibet border, a mass of glacier ice and rock broke away high on the mountain. The collapse triggered an avalanche; the avalanche released a sudden surge of water and debris; the surge became a catastrophic flash flood downstream — a sequence a US Geological Survey assessment, reported by the Guardian, traced precisely: glacier and rock collapse, then avalanche, then a sudden water-and-debris release, then flood. None of it arrived with the kind of warning monsoon-tuned flood monitoring is built to give. The instability had been accumulating for years, in a form no seasonal monitoring system was watching for, until it resolved itself in an afternoon. The system had been built to answer a different question than the one the mountain eventually asked.
Ecological systems are not straight lines. Nothing happens, and nothing happens, and then something happens.
An impact can often be mitigated. A cut into the principal cannot always be repaired — mitigation is a conditional answer, not a universal one. If what is lost can genuinely be recreated, restore it. If it cannot, no sum spent elsewhere replaces it. The taxonomy needs to be more precise than “capital” and “principal” alone allow. Not everything in nature carries the same rule, and few whole ecosystems sit neatly in one category.
Consumable capital can be depleted outright — a mineral seam is a stock, not really “capital” in the endowment sense — provided the depletion is booked as a drawdown, never disguised as income. Regenerative capital can be harvested — a well-managed forest’s timber yield, a fishery — provided the taking stays inside what regenerates within a human generation. Permanent capital cannot be liquidated at all, because its function cannot be recreated within any timeframe that means anything to the people who would need it back: a glacier’s accumulated ice, an aquifer’s fossil water laid down over an ice age, an undisturbed grassland’s root mat, a mature mangrove’s root system. The same forest can hold all three at once — regenerative timber alongside old-growth structure or watershed function that will not regenerate on any usable timescale. The honest unit of classification is the function, not the landscape it happens to sit inside.
The circular economy is a genuine improvement, and a partial one. Take, make, use, discard becomes reuse, recover, recycle — a loop worth closing. But a closed material loop protects nothing on its own. A country can recycle every gram of its plastic and still drain its aquifers, recover every tonne of metal and still lose its topsoil. Circularity earns its name only once it starts where an endowment starts: asking what may be drawn down, what must regenerate, and what does not move at all.
None of this starts from zero. A few jurisdictions have already reached for something like a principal that cannot be spent. England’s biodiversity net gain law, in force since 2024, requires most developments to leave measurably more habitat than they found. The United States has run wetland mitigation banking under the Clean Water Act for decades, trading destroyed wetlands for restored or created ones elsewhere. The European Union’s Habitats Directive bars any plan or project from damaging the integrity of a designated site, precedent or convenience notwithstanding. Courts in New Zealand, Colombia, and Ecuador have gone further still, recognising a river’s legal personhood rather than only regulating what may be done to it.
Each of these is closer to a principal than a market price is. And each still falls short of one in the same way: net gain and mitigation banking assume, by design, that ecological function is fungible — that a wetland destroyed here can be answered by a wetland built there, on different soil, on a different timeline, at a scale no regulator is required to keep re-measuring as approvals accumulate.
Compensatory afforestation carries the same assumption in a blunter form. It typically permits distance between what is cleared and what is planted, so the accounting balances — trees for trees, hectares for hectares — while the specific, located function the original stand performed goes uncompensated. A mangrove protects the coastline behind it, not some coastline. Move the replacement planting inland, or fifty kilometres up the shore, and the ledger closes while the neighbourhood behind the original belt is left undefended.
The valuation on both sides of that ledger is also usually built from routine, annual yield — carbon sequestered, timber grown, fish landed. That is the wrong basis for an asset whose real work is done rarely and catastrophically. A mangrove’s storm-surge attenuation, a wetland’s flood buffering, do not show up as revenue in an ordinary year, so a standard cost-benefit appraisal prices them near zero and then leaves them out of the compensation calculation entirely. The asset is worth most exactly when it is not earning — and the accounting that clears its removal was never built to see that.
Legal personhood solves a different problem: who may speak for the river in court, not what may substitute for it.
What is missing, across all of these, is the distinction this essay is trying to draw — not that ecological loss must be offset, but that some of it cannot be offset at all, because there is no site, no fund, and no law anywhere that can grow a five-hundred-year-old forest back on schedule.
Naming a Permanent Ecological Asset is not only a scientific act. It is a governance act, and the harder one.
Someone has to hold fiduciary responsibility for capital no one is allowed to spend. Someone has to draw its boundary, and someone has to enforce that boundary against a private owner, a state-owned company, or an elected government that later decides otherwise — none of it obvious, least of all when protecting the principal imposes real costs today on people who did not create the shortfall.
Identifying the capital is the easier half of this problem. Establishing legitimate authority over it is the harder one. The hardest version of that authority is the independent trustee itself: a fiduciary insulated enough from the electoral cycle to hold a boundary a future government might want to move is, by the same design, insulated from the people that cycle answers to — and a rule no government can revisit is difficult to distinguish, procedurally, from one no government ever really agreed to.
Once some ecological capital is understood as permanent, the burden of proof has to move with it. A small, local, reversible act can still be judged the ordinary way. But the more irreversible the act, the wider its reach through a connected system, the less is known about it — the higher the bar should climb before it is allowed.
This is not an argument for doing nothing. It is an argument against mistaking a deadline for knowledge. Where permanent capital is at stake, the question is not only whether harm has been shown — it is whether enough is known to permit the reduction at all.
Absence of evidence is not evidence of absence. A monitoring system tuned to the wrong signal can record nothing wrong for years, right up until the morning it has nothing left to record.
Putting nature on the balance sheet was the right first step. It is not the last one. A balance sheet can record a loss, can record a restoration. What it cannot do, on its own, is protect a principal it has no rule against spending.
A Permanent Ecological Asset needs more than an entry — it needs a covenant specifying at least six things: a baseline ecological inventory, before anyone argues what may be taken; a prohibition on net loss for assets designated permanent, not a target but a rule; assessment that runs cumulatively across a system rather than project by project; an independent trustee or public fiduciary answerable for the principal, not any single approval; transparent accounting for every drawdown, in the language of a ledger rather than a press release; and a reversal burden — where harm may be irreversible, the default is refusal until uncertainty is resolved, not permission until harm is proven.
Specifying how that covenant gets paid for, and who holds the obligation to fund it, is a separate argument. But any such mechanism has to rest on the distinction made here — principal from yield — or it has nothing to enforce.
Every generation inherits a balance sheet it did not write — forests, aquifers, reefs, wetlands, and constructs such as Tamil Nadu’s eri tanks, kept alive for centuries by the communities who use them. It is entitled to the income. It is a trustee of the principal.
A trustee can keep every rule and still fail the trust: inherit a forest and hand down cash, inherit an aquifer and hand down a good harvest drawn from a water table that will not refill. The ledger shows growth. The land shows loss. The question that matters is not how much a generation produced. It is what it inherited, what it spent, what it put back, and what it left standing.
A glacier does not present its bill in advance. Neither does an aquifer, a grassland, or a mangrove. They fail quietly, for a long time, and then all at once — and by the time the failure is visible, no covenant written afterward can undo it.
There may be a task waiting behind all the others: not simply that some ecological capital must be financed forever, but that some of it cannot be treated as available at all, under any financing, at any price. If that is true, the question at the end of a generation is not how much wealth it made, but what it received, what it spent, what it put back, and what is still standing for the ones who come after.
I do not know how to answer that question for my own generation yet. I am not sure any generation has known, while it was still living inside the answer.