Synthetic Ecosystems

Established A closed ecological system is one where matter cycles and only energy crosses the boundary, and exactly four have ever been built at human scale. BIOS-3 in Krasnoyarsk from 1972; Biosphere 2 in Arizona from 1991; the MELiSSA pilot plant in Barcelona from 2009; and Lunar Palace 1 in Beijing from 2014. Established Everything else in this field is a microcosm, a bioreactor, a model, or a proposal. Established The four programmes together account for roughly four crewed years of sealed operation across half a century, and not one of their results has been reproduced at a second, unaffiliated facility.

Established Two numbers that almost everybody gets wrong carry most of the argument. BIOS-3 is routinely credited with “91–95% closure”; its own directors wrote that the atmosphere and water were regenerated completely while food regeneration amounted to 93%, and that up to half or more of the plant biomass was inedible — a deadlock product the system never processed. Established Biosphere 2's oxygen is quoted to a decimal point it does not have: the primary source says it fell from the ambient 21% to 14% over sixteen months, and the circulating 14.2% and 14.5% are downstream precision. Established The corrected versions are more interesting than the folklore. Gas and water closure is a solved problem; food is nearly solved; solid organic residue has never been closed by anyone.

Speculative The exotic question is whether a small closed system can be stable at all, or whether ecology forbids it. There is a serious case that it does: a sealed ecosystem is a set of coupled nonlinear dynamics with no immigration term, so a guild that goes locally extinct stays extinct, and small systems have too few individuals per function for stochastic loss to average out. Speculative There is an equally serious case that nobody has tested it, because the field's money is attached to spectacular multi-year demonstrations rather than to the cheap replicated microcosm array that would produce a size-versus-persistence curve. Handwave And there is a demonstrated method nobody has applied: whole ecosystems were shown to have heritable system-level phenotypes in 2000, which makes directed evolution of an ecosystem a twenty-six-year-old technique waiting for a user.

Established BIOS-3 is the best-documented closed system ever built, and the primary account is by the men who ran it. Gitelson and Lisovsky, writing in 2008 in the Journal of Siberian Federal University: Biology, describe a structure of 14 × 9 × 2.5 m — a volume of 315 m³ — divided equally into four compartments: a crew compartment with three cabins, galley, lavatory and control room; an algal cultivator; and two phytotrons for wheat and vegetables. Established The facility ran ten crewed closure experiments with crews of one to three people, continuing at least until 1984. Established Air regeneration by Chlorella required roughly 8 m² of exposed algal surface per person.

Established The closure figure, verbatim, from the same source: “The atmosphere and water were regenerated completely while food regeneration amounted to 93%.” That is the sentence the field has spent decades garbling. Established The widely circulated “91–95% closure” is a mangled restatement of a food figure and was never a system-wide closure percentage. Established The honest one-line summary of BIOS-3 is: gas closed, water closed, food 93%, solid organic waste not closed at all. Established Two of those four are genuine achievements that nothing since has improved on, and the fourth is the reason the programme did not scale.

Established The unclosed remainder is stated in the primary with unusual bluntness: “up to half or more of the biomass synthesized by plants is inedible, being a deadlock product.” Straw, stems, roots, chaff — lignocellulose the crew cannot eat and the loop cannot mineralise fast enough to return as nutrient. Established This is the real BIOS-3 result and it is almost never reported, because it is a negative and because it does not fit a percentage. Frontier It is also the single most transferable finding in the field: every closed-loop proposal since, including MELiSSA's, inherits the same residue and solves it the same way, which is to say by storing it. Speculative A loop with a growing pile of unprocessed lignocellulose in the corner is not closed; it is open on a delay.

Established The durations are smaller than the folklore, and the primary is precise about them. The longest continuous experiment with humans in the closed system lasted six months; the longest time one human spent inside the system was thirteen months; and the facility accumulated about a year of crewed closure in total across the whole programme. Established A tertiary account gives the flagship run as 180 days with a three-person crew in 1972–73, which is consistent. Established What is not consistent with the folklore is the implication of routine capability: one six-month run in a programme spanning a decade is a demonstration, not a technology.

Established BIOS-3 was matter-closed and energy-open by a factor of more than a thousand, and this is the field's least-discussed fact. The facility drew 400 kW from a nearby hydroelectric station to support up to three people — roughly 133 kW per person against a human metabolic rate of about 100 W. Established The lighting specification is also misreported: the primary says each phytotron originally had twenty cylindrical vertical 6 kW xenon lamps, roughly 120 kW per phytotron, doubled in one phytotron by 1991 — not “20 kW xenon lamps,” which collapses lamp count and lamp rating into one wrong number. Established Photosynthetic photon flux at plant level was about 900–1000 µmol m⁻² s⁻¹ under single lamps and 1600–1850 under doubled lamps, against roughly 2000 for full sunlight. Speculative Every closure claim in this field silently assumes unlimited cheap power, and no programme has published the design curve trading photosynthetic area against electrical input per crew member.

Established Biosphere 2 was two orders of magnitude larger and failed in a way that was chemical rather than ecological. Mission 1 ran from 26 September 1991 to 26 September 1993 with eight crew; Mission 2 ran 6 March to 6 September 1994 with seven and ended early. Total enclosed area was 1.27 hectares. Established The biome allocation is the design decision that governs everything downstream: rainforest 1,900 m², ocean 850 m², mangrove 450 m², savannah 1,300 m², fog desert 1,400 m², and intensive agriculture 2,500 m² — under 20% of the footprint doing all of the calorie work. Established Atmospheric leak rate was under 10% per year, which for a structure of that size is an outstanding sealing result and remains the field's reference engineering achievement. Established The structure worked. The biogeochemistry did not.

Established Oxygen fell from the ambient 21% to about 14% over the first sixteen months, and the cause was concrete. Severinghaus, Broecker, Dempster, MacCallum and Wahlen reported in Eos in 1994 that the loss was due to microbial respiration of organic soil matter together with a reaction between respired CO₂ and concrete forming calcium carbonate. Established The two sinks are why the budget looked impossible from inside: soil microbes consumed O₂ and produced CO₂ as expected, but the CO₂ did not accumulate in the atmosphere because uncured concrete absorbed it, so the oxygen appeared to vanish without a trace. Isotopic analysis found the trace in the walls. Established Oxygen was injected from outside in January 1993; tertiary accounts give the endpoint as 14.5%, an equivalent altitude of 4,080 m, and a decline under 0.25 percentage points per month, and give CO₂ diurnal swings of about 600 ppm with winter peaks of 4,000–4,500 ppm. Frontier The decimal in “14.5%” is not in the primary abstract, which says 14%, and the Eos full text remains behind a paywall this pack could not open — so the defensible statement is “about 14%” and any decimal belongs to whichever secondary supplies it.

Established The farm was excellent and still could not feed eight people, which is an area result rather than a competence result. The agricultural system supplied 83% of the crew's total diet across Mission 1, at a reported productivity more than five times that of the most efficient traditional agrarian systems in Indonesia, southern China and Bangladesh. Established Crew lost an average of 16% of body weight before stabilising — and the caloric restriction was a designed experimental protocol on Roy Walford's lifespan-extension regime, with Walford himself as crew physician, not an accident of shortfall. Established The ecological outcome was worse than the nutritional one: most introduced vertebrates and virtually all pollinating insects died, cockroaches and the invasive longhorn crazy ant Paratrechina longicornis dominated, and the one clear ecological success was that corals reproduced in the ocean biome. Established Ed Bass put about $150 million into the facility between 1984 and 1991, against a reported $25 million loss in fiscal 1992.

Established MELiSSA is the most serious sustained attempt at an engineered loop, and after thirty-seven years its crew is a rat. The MELiSSA Foundation states that the concept was completed and published in 1988, formal activity began in 1989, and the programme now involves approximately fifty organisations of which fifteen partners have signed formal agreements, across Europe and Canada. Established The architecture is a compartment chain: a liquefying compartment running thermophilic anaerobic digestion at 55 °C on urine and kitchen waste; a photoheterotrophic compartment removing volatile fatty acids; a nitrifying compartment where Nitrosomonas and Nitrobacter convert ammonia to nitrate; and a photoautotrophic compartment split between Arthrospira platensis and a higher-plant chamber supplying oxygen and food. Established The pilot plant at the Universitat Autonoma de Barcelona, inaugurated in 2009, has five compartments designed and operating — waste degradation, nitrification, air revitalisation, food production, and a mock crew that is a rat isolator.

Established The programme's own words are the most exact status report available: the plant is in an “animal based phase,” which is “a preparation phase for a future man-rated facility.” The pilot-plant page publishes no loop-closure percentage, no continuous-operation duration and no crew-equivalent support figure, and describes the compartments as developed at individual level ahead of integration of the complete loop. Established MELiSSA's one deployed real-world product is terrestrial and is not a closed loop: a grey-water treatment unit at Concordia station in Antarctica, in cooperation with ESA since technology transfer began in 2005, with a daily treatment capacity of 2,400 to 2,800 litres, the equivalent grey-water production of about twenty-five people. Frontier No recycling percentage is published for it either. Speculative A water-treatment skid is a respectable engineering output for a space programme; it is not what the programme is sold as, and the gap between the two is the honest measure of how hard loop closure is.

Established The longest crewed bioregenerative closure run in history happened in Beijing and Western accounts routinely omit it. Lunar Palace 1 at Beihang University is 160 m² and 500 m³, with 58 m² of vegetation area and 42 m² of living area, described as the third bioregenerative facility built in the world and the first in China. Established Its first mission ran 105 days from 3 February to 20 May 2014 with a crew of three, producing 55% of food internally. Frontier “Lunar Palace 365” then ran 370 days with two rotating teams of four, claiming total recycling of oxygen and water and complete production of plants and mealworms — longer than Biosphere 2's Mission 2 and twice BIOS-3's longest run. Frontier The results are largely unreplicated outside Beihang, which is true of every other programme in this field as well, and is not a reason to leave the record out of the comparison.

Established The field's own summary verdict, stated flatly in a standard reference: no artificial closed ecological systems involving humans have been able to successfully support life for long periods of time. Established That is the correct headline after fifty-four years and four facilities. Frontier What it does not say, and what this brief argues, is that the reason is unknown — because the experiment that would distinguish “too small to be stable” from “never properly attempted” has not been run.

Frontier There is exactly one new sealed, human-rated closure facility anywhere in this subject, and it publishes no performance figures. The Space Analog for the Moon and Mars, SAM, at Biosphere 2, is built around the 1987 Biosphere 2 Test Module — the prototype in which the original sealing technology was proven. Frontier It has a fully functional airlock, a computer-controlled and electrically actuated pressure-regulation system, living quarters, workshop, kitchen, crew quarters and toilet, with continuous monitoring of carbon dioxide, oxygen, relative humidity, temperature and barometric pressure. Its own framing is that it is not another open-air analog. Established No sealed volume, no leak rate and no completed-mission durations are published on its page, and the research-team listing was empty when this pack fetched it. Speculative A real, operating, sealed human habitat with no published numbers is the field's structural problem in miniature: the demonstration is the product, and the measurement is somebody else's job.

Frontier The most relevant experiment in the world for closed-system ecology is being run next door and the closed-system literature does not cite it. The Landscape Evolution Observatory at Biosphere 2 is described by the facility as the world's largest laboratory experiment in the interdisciplinary Earth sciences: three artificial landscapes, each 30 m long and 11 m wide at an average 10 degree slope, filled to 1 m depth with about 500 tonnes of crushed basalt from a northern Arizona volcanic crater, instrumented with over 1,800 sensors and sampling devices tracking water, carbon and energy. Frontier Its question is how a purely mineral, abiotic substrate becomes a living landscape. Speculative That is the minimum-viable-biosphere question approached from the abiotic end, at replicate scale, with instrumentation no life-support programme has ever had — and the connection has not been made in print.

Frontier Biosphere 2 as a working laboratory now produces results that only a sealed biome-scale enclosure can produce. The facility reports that tropical plants in its rainforest biome stop absorbing additional carbon dioxide once atmospheric levels reach about 600 ppm, a threshold the facility notes we are on track to hit by 2050, and that elevated atmospheric CO₂ significantly slows coral growth rates in its ocean. Established Biosphere 2 has been a University of Arizona research facility since July 2011, funded partly by a Philecology Foundation gift and NSF grants, with a portfolio spanning LEO, the rainforest, the ocean reef lab, agrivoltaics, SAM and a Center for Human Space Exploration. Speculative The rehabilitation of the site is the strongest single argument against the “famous failure” framing: the enclosure turned out to be a scientific instrument, which is not what it was built to be and is what it is good for.

Frontier MELiSSA's own current research names the loop's open end explicitly. Frossard and colleagues, in Life Sciences in Space Research in 2023, report that nitrogen and phosphorus recovery from urine is established but that much work remains on recovering nutrients from other liquid and solid organic waste. Frontier The named unsolved problems are specific: sodium-chloride removal, nitrogen and mineral balancing, nutrient-flux adequacy across diverse crops, recovery efficiency, and the disposal of nutrient solutions once they become unusable. Established That last item is the tell. A loop with a defined disposal stream is a loop with an acknowledged open end, stated by the programme itself, in the peer-reviewed literature, thirty-four years in.

Established Whole-ecosystem artificial selection is not a proposal; it was demonstrated in 2000 and nobody in life support has used it. Swenson, Wilson and Elias, in PNAS, subjected entire soil microbial ecosystems to artificial selection and reported that laboratory-initiated ecosystems exhibit phenotypic variation with a heritable proportion despite comprising thousands of species and millions of individual organisms, illustrating an important role for complex interactions in evolution. Established The standard objection to directed evolution of ecosystems — that community-level heritability is too weak for selection to act, because an offspring community is a random sample of its parent — was answered in the affirmative twenty-six years ago. Speculative So the live question is not whether whole-ecosystem selection works. It is why a demonstrated method from 2000 has never been pointed at a closed life-support compartment, and that is an institutional question rather than a scientific one.

Frontier Designed microbial consortia are where division-of-labour engineering is actually happening, and the methodology is contested. The current debate is between bottom-up assembly, where a community is built from characterised monocultures whose interactions are modelled, and top-down enrichment, where a functional community is selected out of a complex inoculum and then reduced; San Leon and Nogales argued in 2022 for merging the two in model-based design. Frontier On the applied side, co-cultivation rather than assembly from monocultures has been shown to produce more robust functionally designed consortia, which is a direct methodological bridge from gut-microbiome engineering into closed life support. Speculative The unifying idea is that a compartment should be specified by function with several organisms per function, rather than by organism — and no closed-system programme has yet published a compartment designed that way.

Established The workback plan to a multi-year closed loop has eight links, and the binding one is second. Established Link 1, a materials-inert enclosure: rebuild a Biosphere-2-class enclosure with fully cured or carbonation-inert structural materials and measure the closed carbon budget to under 1% closure error over twelve months. This directly tests whether the 1991–93 oxygen crash was a materials artefact. It is cheap and nobody has funded it. Frontier Link 2, minimum viable biosphere size: replicate microcosm arrays with n greater than 100 at logarithmically spaced volumes, measuring per-guild extinction hazard as a function of system size and functional redundancy, and output a size-versus-persistence curve. Speculative This is the genuine scientific unknown, and no rigorous minimum-viable-biosphere size theorem has been published by anyone.

Speculative Links 3 to 6 are engineering that waits on link 2 to know what to build. Link 3, guild-level redundancy: for each of roughly eight critical functions — photosynthesis, nitrification, denitrification, cellulolysis, methanogenesis control, pollination, decomposition, pathogen suppression — demonstrate at least three non-redundant organisms sustaining the function through a deliberate knockout of the dominant. Never attempted at system scale. Frontier Link 4, nutrient closure without accumulation: at least five consecutive waste-to-plant-nutrient cycles with no build-up of growth-inhibiting compounds, which is exactly the MELiSSA problem the 2023 nutrient-recycling paper names. Established Link 5, calorie closure at realistic area: 100% rather than 83% of a 2,500 kcal per person per day diet from 500 m² or less per person, with full macro- and micronutrient adequacy, sustained a year. Biosphere 2 reached 83% at roughly 310 m² per person. Speculative Link 6, energy honesty: publish the design curve trading photosynthetic area against electrical input per crew member. Until that curve exists, closure claims are unfalsifiable, because any shortfall can be absorbed by unstated wattage.

Speculative Links 7 and 8 are money and sociology, and link 8 is the one that makes this a demonstration field rather than a science. Link 7, a multi-year crewed demonstration: three years or more, four crew or more, zero material resupply, independently audited mass balance. Lunar Palace 365's 370 days is the record to beat and no customer for the successor exists. Established Link 8, independent replication: the same result at a second unaffiliated facility. Not one closure result in this field has ever been independently replicated. Speculative The field's structural pathology is visible in the ordering: link 2 is cheap and nobody runs it, because funding attaches to link 7, which is spectacular and unaffordable. Handwave A programme that spent one Biosphere 2 budget on link 2 instead would probably know by now whether closure is possible.

Established This brief waits on one other brief's output and it is not the one people expect. Genetic Engineering supplies the only route by which a compartment stops being a found community and becomes a designed one: engineered organisms with specified metabolic roles, built-in redundancy, and characterised failure modes. Frontier Every closed system built so far has used wild-type organisms chosen for convenience — Chlorella, Arthrospira, wheat, Nitrosomonas — and their interaction networks were inherited rather than specified. Speculative The step from a loop that works to a loop that can be certified runs through organisms whose genomes are known well enough to predict what they do when a neighbour dies.

Frontier Three results outside this brief's control would change its assessment most, and none is being funded as a priority. First, a per-guild extinction hazard function estimated from existing microcosm data, which is a reanalysis rather than a new experiment and would give the field its first quantitative statement about system size. Frontier Second, the physicochemical-versus-bioregenerative mass crossover point, which lives in the Equivalent System Mass modelling literature and which this pack could not retrieve — without it, the argument that bioregeneration is worth building is qualitative. Speculative Third, an independent replication of anything: a single closure result reproduced at a facility with no stake in the original would do more for this subject's credibility than another decade of first-of-a-kind demonstrations.

Speculative The dependency that is usually asserted and does not hold is a dependency on space programmes. Closed loops are needed where resupply is impossible, and no funded mission profile this century requires one; the demand signal that everyone assumes is upstream of this field is in fact absent. Established What is genuinely upstream is unglamorous: soil biogeochemistry, controlled-environment horticulture, anaerobic digestion, and the microbial ecology of defined communities. Speculative A field that believes its customer is a Mars mission and whose actual inputs are soil science and wastewater engineering will keep mis-specifying its own experiments.

Frontier The one experiment that would change everything is a microcosm array, and it costs less than a single crewed demonstration month. Build 100 or more replicate closed microcosms at logarithmically spaced volumes spanning three or four orders of magnitude, seed them identically, run them for years, and measure which functional guilds are lost, when, and as a function of what. Speculative The output is a size-versus-persistence curve and a per-guild extinction hazard rate, and it converts the central question of this subject from a debate into a number. Established Nothing about it is technically hard. It has not been done because no funder buys a graph.

Frontier The second experiment is the materials control Biosphere 2 never had. Run two sealed enclosures side by side, identical in soil, biota and volume, differing only in whether the structural concrete is fully cured or freshly poured, and measure the closed carbon and oxygen budgets independently. Speculative If the cured enclosure holds its oxygen, the most famous failure in this field was a construction-schedule decision rather than an ecological verdict, and thirty years of interpretation need revising. Frontier The counter-hypothesis is testable in the same run: soil organic-matter oxidation was also anomalously high in Biosphere 2 because the soil was imported, rich and freshly disturbed, which is an ecological choice rather than a materials one, and a soil-maturity arm would separate the two.

Speculative The third is the one this brief would most like somebody to run, because the method already exists. Take a MELiSSA-style compartment — nitrification, or waste liquefaction — and subject a few hundred replicate communities to whole-ecosystem artificial selection on a system-level phenotype: nitrate yield, or resistance to a defined perturbation. Established Swenson, Wilson and Elias showed in 2000 that ecosystem-level phenotypes have a heritable component at exactly this kind of scale. Frontier The tunable parameter is bottleneck size at propagation, and the assay has to be cheap enough to run across hundreds of replicates per generation, which for a gas or ion measurement it is. Speculative Nobody has published this and the reason appears to be that the two communities — experimental evolution and life support — do not read each other.

Frontier Three measurements that would cost little and are simply missing. One, a mass balance on inedible biomass in any operating closed system, stating how much accumulates per crew-day and what happens to it — the deadlock product has been named since at least 2008 and never quantified in a published loop. Established Two, the energy-versus-area design curve, per crew member, for every facility that has ever claimed closure. Speculative Three, a re-seeding mass budget: if a system is permitted periodic resupply from a frozen library, what is the minimum re-seeding mass per crew-year? That reframing makes the problem tractable and the field has not adopted it.

Established The enclosure is the solved part and it is worth saying so clearly. Biosphere 2 achieved an atmospheric leak rate under 10% per year across 1.27 hectares under sealed glass, and SAM inherits the same technology from the 1987 Test Module in which it was proven. Established Sealing is not the bottleneck. Frontier What the Biosphere 2 experience adds is that the enclosure's materials are part of the biogeochemistry, not a neutral boundary: uncured concrete was an active participant in the carbon cycle, and no closed-system design specification this brief could find treats structural material as a chemical component of the ecosystem.

Established The power and area requirements are the honest constraints and both are large. BIOS-3's 400 kW for up to three people is roughly 133 kW per person, against a metabolic 100 W; its lamps ran at 20 × 6 kW per phytotron and still delivered only about half of full-sun photon flux at plant level. Established Biosphere 2 got 83% of a diet from 2,500 m² serving eight people, roughly 310 m² each, with sunlight free. Speculative Any off-planet version of this trades one against the other — area against watts — and the exchange rate has never been published, which means no proposal in this field can currently be costed by an outsider.

Frontier Waste processing is the engineering gap that nobody has closed, and it has two halves. The liquid half is nearly solved: MELiSSA's thermophilic anaerobic digestion at 55 °C plus nitrification is a working chain, and urine-derived nitrogen and phosphorus recovery is established. Established The solid half is not: lignocellulosic residue, which is up to half or more of synthesised plant biomass, has no demonstrated route back into the nutrient loop at closed-system rates. Speculative Candidate routes exist — fungal degradation, insect conversion of the kind Lunar Palace's mealworms and the earlier closed-loop silkworm work represent, thermophilic composting, or pyrolysis with the carbon returned as CO₂ — and none has been demonstrated at the throughput a crew generates.

Established Industrial civilisation failed to close the same lignocellulose loop, expensively, and the record is instructive. Cellulosic ethanol was the promised escape from the food-versus-fuel trap; most plants were cancelled or closed in the early 2010s as technical obstacles proved too difficult, and as of 2018 exactly one remained operational, against a 2007 estimated production cost of about $2.65 per gallon, roughly two to three times corn ethanol. Speculative That is the same molecule, the same recalcitrance, with billions of dollars and no crew depending on it. Speculative A closed life-support programme claiming it will process straw should be asked what it knows that the cellulosic ethanol industry did not.

Frontier Controlled-environment agriculture is the terrestrial industry that shares this field's core trade and has real numbers. A Finnish life-cycle assessment found direct energy of 10–95 MJ/kg for vertical farming against 120–133 MJ/kg for greenhouse lettuce, with the best vertical-farm case beating every greenhouse case on climate impact — but only with renewable electricity and waste-heat recovery, and impacts rose 150% without heat recovery. Established This is the same energy-versus-area exchange that closed systems refuse to publish, worked out in commercial conditions with published boundaries. Speculative Future Agriculture owns that trade; this brief inherits it and should be reading its numbers.

Established Four neighbours own pieces of the problem this brief cannot solve alone. Microbiome Engineering owns defined-consortium design and the colonisation-resistance question, which is a compartment-stability question in different clothing. Established Biological Energy Systems owns anaerobic digestion and the energetics of turning residue into something useful. Speculative Space Habitats and Terraforming are the customers this field believes it has: the first would need a closed loop only for very long missions, and the second is this problem at planetary scale with the failure modes correspondingly slower and larger. Speculative Designer Organisms is where the redundancy engineering of link 3 would have to come from.

Established The defining institutional fact is that every result in this field belongs to exactly one facility. BIOS-3 was the Institute of Biophysics in Krasnoyarsk; Biosphere 2 was a private venture and is now the University of Arizona; MELiSSA is ESA with a pilot plant in Barcelona; Lunar Palace 1 is Beihang. Established Not one closure result has been reproduced anywhere else. Speculative A subject with zero independent replications is not yet a science; it is a sequence of demonstrations, and it should be funded and read as such until that changes.

Frontier The funding structure selects for the wrong experiment, and the mechanism is legible. A multi-year crewed closure is spectacular, photogenic, expensive and produces a single data point with n = 1 and no control. Established A hundred-microcosm array produces a curve, costs a fraction as much, and photographs badly. Speculative The field's entire recorded output is the first kind, which suggests the selection is happening at the funder rather than at the investigator. Frontier MELiSSA's thirty-seven-year trajectory — concept in 1988, pilot plant in 2009, five individually-operating compartments and a rat in 2026 — is what patient, serious, well-managed funding produces when the programme's own milestone is an integrated demonstration rather than a measurement.

Established Publication norms in this field are unusually weak and the fix is administrative. The MELiSSA pilot plant publishes no closure percentage, no continuous-operation duration and no crew-equivalent figure; SAM publishes no sealed volume, leak rate or mission duration; the Concordia unit publishes a throughput but no recycling percentage; and no programme anywhere publishes its energy-per-crew-member curve. Speculative None of this is secrecy. It is the ordinary result of nobody being required to report a standard set of numbers, in a field where the reporting standard would take an afternoon to write. Frontier A funder able to attach six mandatory reported quantities to a grant — sealed volume, leak rate, crew-days, energy per crew-day, mass in, mass out — would improve this subject more than any single experiment on this page.

Established Crewed closure experiments are human-subjects research and have not always been run that way. The Biosphere 2 crew's 16% average weight loss was not an accident: the diet followed Roy Walford's caloric-restriction protocol with Walford as crew physician, and the physiological results were published as a gerontology study. Frontier That is a legitimate design and it means Mission 1 was simultaneously an engineering demonstration and a nutrition trial on eight people who could not leave. Speculative The oxygen decline compounds it: a crew was living at a physiological equivalent of about 4,080 m for months while the cause was unknown, which no modern protocol would permit without a defined abort threshold.

Frontier Animals are the field's quiet ethical load and the framing hides it. MELiSSA's crew is a rat isolator, described by the programme as an animal-based phase preparing for a future man-rated facility; Lunar Palace ran mealworms as a protein loop; the closed-system literature's protein routes run through insects and small vertebrates almost exclusively. Speculative Calling an animal a “mock crew” is accurate engineering language and it also does the work of not describing an animal living permanently inside a sealed metabolic experiment. Established Biosphere 2's ecological outcome — most vertebrates and virtually all pollinators dead — is the same question at ecosystem scale, and it was foreseeable.

Speculative The third problem is what a designed ecosystem means outside the enclosure. A synthetic ecosystem that works is by construction a set of organisms selected or engineered to be stable together, and the containment question for such a community is not the same as for a single engineered organism, because the unit that might escape is a functioning consortium with its own metabolic redundancy. Handwave No regulatory framework anywhere addresses a designed multi-species community as a release object. Frontier The nearer-term version is less exotic and more likely: whole-ecosystem artificial selection produces communities nobody has characterised, by design, since the whole point is to select on function rather than composition.

Speculative If closure is achievable, the change is not primarily about space. A system that can hold a stable ecology at small scale is a testbed for the question every planetary-scale intervention needs and cannot run: what does a coupled biogeochemical system do when you push one variable? Frontier Biosphere 2 already demonstrates the value in miniature — a rainforest saturating at about 600 ppm and corals slowing under elevated CO₂ are results a sealed enclosure produced and no field site could. Speculative The civilizational product of this field may turn out to be experimental Earth-system science rather than life support, and that is not a consolation prize.

Speculative If closure is genuinely forbidden at small scale, that is a more consequential result than success would be. It would mean that every long-duration settlement scenario requires either a system large enough that guild extinction is negligible — converting an ecological question into a mass-budget one — or permanent active management with a re-seeding library, which is not self-regulation but husbandry. Handwave The second is achievable now and the field has not adopted it, largely because “closed” is a binary the literature will not give up. Speculative A civilization that can only live off Earth with a freezer full of spores and a technician is a different civilization from one that can build a biosphere, and the difference is worth knowing before anyone leaves.

Frontier The near-certain civilizational output is the boring one. Three decades of closed-system work have produced soil biogeochemistry, controlled-environment horticulture, a grey-water plant in Antarctica, and an instrumented hillside in Arizona. Established That is a real return on the money. Speculative It is also an argument for funding this field honestly, as terrestrial ecological engineering with a speculative space option attached, rather than as a space programme that keeps producing terrestrial results by accident.

Horizons here track two clocks: the measurement clock, which could move quickly and cheaply if anyone chose, and the demonstration clock, which is set by whether a mission profile ever needs this.

• 10 yr: Frontier SAM runs crewed analog missions and either publishes leak rate, sealed volume and closure figures or does not — and which it does is the most informative thing about this field's norms available in the next decade. Frontier MELiSSA integrates more compartments and continues to report progress toward loop closure without a closure percentage; the animal-based phase persists. Speculative The minimum-viable-biosphere microcosm array remains unfunded, which is the base case. Speculative Most likely surprise: a Chinese facility publishes an integrated multi-year closure figure with independently reported mass balance, and the Western literature takes years to absorb it.
• 25 yr: Speculative A materials-inert enclosure test settles whether the Biosphere 2 oxygen crash was a construction artefact, most plausibly as a side-effect of a lunar or Mars surface-habitat programme rather than as a dedicated experiment. Speculative Whole-ecosystem artificial selection is applied to a life-support compartment for the first time, twenty-six years after the method was demonstrated. Frontier Calorie closure at 100% for a small crew is achieved somewhere, at an area and power cost that would be uneconomic anywhere except off Earth. Speculative The inedible-biomass deadlock is still open.
• 50 yr: Speculative A three-year crewed closure with independently audited mass balance is plausible on this horizon and only if a mission actually needs it; without a customer, the record stays at 370 days indefinitely. Speculative A published size-versus-persistence curve, if anyone runs the array, would by then have converted “can a small system be closed?” into a design parameter. Handwave Ecosystems evolved in the laboratory for stability under closure — selected as wholes, over hundreds of generations — are coherent on this horizon and would be the first genuinely synthetic ecosystems rather than assembled ones.
• 100 / 250+ yr: Handwave A self-regulating multi-generation closed biosphere requires guild-extinction hazard low enough over decades that no re-seeding is needed, which on current understanding means a system far larger than anything built and possibly larger than anything that can be launched. Handwave The honest terminal position is a declared tie: nothing in physics or ecology forbids closure, nothing in the record demonstrates it, and the discriminating experiment is cheap and unrun. Speculative If it is ever run and the answer is that small systems cannot be stabilised, the practical form of off-Earth life becomes managed husbandry with a cryopreserved library — which works, and which the field's vocabulary currently refuses to call success.

• Depends on This brief waits on one upstream result and it is genetic rather than architectural. Genetic Engineering supplies the only path from a compartment populated by convenient wild-type organisms — Chlorella, Arthrospira platensis, Nitrosomonas, wheat — to one specified by function, with characterised redundancy and predictable behaviour when a member is lost. Every closed system ever built inherited its interaction network rather than designing it, which is why none of them can state a failure mode in advance. The redundancy engineering that link 3 of this brief's workback plan requires — three or more non-redundant organisms sustaining each of about eight critical functions through a deliberate knockout of the dominant — is a genome-scale design problem before it is an ecological one, and it is the step at which a demonstration becomes a certifiable system.
• Enables What this field would unblock, if link 2 ever closed, is not a spacecraft. A published size-versus-persistence curve and a per-guild extinction hazard function would give planetary-scale intervention its first quantitative statement about how much ecosystem is enough, which is the missing parameter in restoration ecology, in rewilding, and in every argument about whether a degraded system can recover unaided. Downstream of a solved inedible-biomass loop sits terrestrial waste processing at municipal scale, because the closed-system version of the problem is the same chemistry with a harder deadline. And downstream of a genuinely closed multi-year loop sits the only settlement architecture that does not require a supply chain — which is why the field exists, and which remains the least likely of its outputs to arrive first.
• Adjacent Controlled-environment agriculture is the nearest neighbour and the one with real published economics: the energy-versus-area trade that closed systems will not publish is worked out commercially in vertical farming, with life-cycle boundaries and measured megajoules per kilogram. Microbiome engineering owns defined-consortium design and the colonisation-resistance problem, which is compartment stability in a different vocabulary and with far more money behind it. Anaerobic digestion and biological energy systems own the residue chemistry. Experimental evolution owns whole-ecosystem artificial selection, demonstrated in 2000 and never applied here. And Earth-system science is adjacent in an unexpected direction: the Landscape Evolution Observatory next door to Biosphere 2 is running the abiotic-to-living-landscape experiment at replicate scale with 1,800 sensors, and the closed-life-support literature does not cite it.

Established “BIOS-3 achieved 91–95% closure.” This is the field's most durable garbled number and it is a food figure wearing a system's clothes. The programme's own directors wrote: the atmosphere and water were regenerated completely while food regeneration amounted to 93%. Established There is no system-wide closure percentage for BIOS-3 and there never was. Established The correct summary has four parts, not one: gas closed, water closed, food 93%, solid organic waste not closed at all. Speculative A single percentage for a system with four qualitatively different loops is not a rounding error — it is a category mistake that hides the only unsolved part.

Established “BIOS-3 ran a 180-day mission, so half-year closure is a routine capability.” It ran one. The primary states that the longest continuous crewed experiment lasted six months, that the longest single human stay was thirteen months, and that the facility accumulated about a year of crewed closure in total. Established Ten experiments, crews of one to three, roughly a year of aggregate closure over a decade. Speculative One six-month run in a fifty-year-old programme is a result; treating it as a demonstrated capability is how a field talks itself into believing a problem is solved.

Established “Biosphere 2's oxygen fell to 14.5%” — or to 14.2%, depending on the source. Neither decimal is a primary figure. The Severinghaus et al. 1994 abstract says oxygen decreased during the first sixteen months of closure from the ambient 21% to 14%, enough to cause health problems in the human occupants; an independent reconstruction of the same abstract gives “approximately 14%.” Frontier Both circulating decimals are downstream precision the primary abstract does not carry, and the Eos full text is behind a paywall this pack could not open, so a decimal in the body cannot be excluded. Established The defensible published statement is that oxygen fell from 20.9% to about 14% over sixteen months. Speculative False precision on a famous number is a small thing that reveals a large one: almost nobody quoting this figure has read the source, because the source is closed-access and cheap to cite.

Established “The oxygen finding was published in Science.” It was published in Eos, Transactions American Geophysical Union 75(3):33–37 in 1994 — AGU's news-and-comment publication, not a peer-reviewed research article. Established It is recorded as closed access with no open-access location, which is part of why the miscitation persists. Speculative This does not make the finding wrong; the two-sink explanation is stated in the primary's own abstract. It makes the field's citation practice worse than its science.

Established “Biosphere 2 failed because of human error, or because the crew cheated.” Wrong on both counts, and it is as much an enthusiast's misconception as a sceptic's. The oxygen decline had a specific, isotopically resolved cause: CO₂ from soil respiration was absorbed by uncured concrete, decoupling the oxygen and carbon-dioxide budgets so the loss was invisible in atmospheric CO₂. Established The much-mocked interventions — injected oxygen, chemical scrubbers, a crew member leaving for hand surgery — were responses to that chemistry, not causes of it. Speculative The uncomfortable version for both camps is that the enclosure's building material was an unlisted member of the ecosystem, and nobody's design document had a line for it.

Established “Biosphere 2 leaked badly” and “the crew starved because the farm didn't work.” The leak rate was under 10% per year across 1.27 hectares, which is an outstanding sealing result and remains the reference achievement in this field. Established The farm delivered 83% of the diet at a reported productivity more than five times that of the most efficient traditional agrarian systems, and the caloric restriction was a pre-planned protocol under Walford, not a shortfall. Established The correct statement is that an excellent farm on 2,500 m² could not feed eight people. Speculative That is an area problem and it is the single most transferable number in the whole record: about 310 m² per person got you to 83%, in Arizona sunlight, for free.

Established “Any of these systems was closed.” No closed ecological system has ever been energy-closed. BIOS-3 drew 400 kW for three people; Biosphere 2 required continuous external power for cooling and air handling. Established “Closed” in this field always means materially closed, and the energy term is invariably unstated. Speculative Because it is unstated, closure claims are unfalsifiable: any shortfall in biological performance can be made up with wattage nobody has to report. Handwave Publishing a single design curve — photosynthetic area against electrical input per crew member — would retire more bad argument in this subject than any experiment on this page, and it requires no new data at all.

Established “MELiSSA is a closed loop.” MELiSSA is a research programme toward a closed loop, begun in 1989, with a pilot plant since 2009 whose five compartments are described as developed at individual level ahead of integration. Established The programme's own current statement is that the plant is in an animal-based phase preparing for a future man-rated facility, with a rat isolator as the mock crew and no published closure percentage, continuous-operation duration or crew-equivalent figure. Speculative That is not a criticism of MELiSSA, which is the most methodical programme this subject has. It is a criticism of everyone who cites it as a working closed loop.

Frontier “The longest closure experiment was Biosphere 2's two years.” Biosphere 2 Mission 1 ran 730 days with eight crew; Lunar Palace 365 ran 370 days with rotating crews of four and claims fuller recycling of oxygen and water. Frontier Neither is straightforwardly the record, because they differ in crew-days, closure completeness and independent verification — and neither has been independently verified at all. Speculative Claims of a single record-holder in this field are usually parochial rather than wrong, and the Chinese result is the one most often left out.

Speculative “Biosphere 2 was pseudoscience.” Steelmanned, the dismissal is strong: the mission had no controls and no pre-registered hypotheses, and its origins in the Institute of Ecotechnics, its theatrical framing and the 1994 management collapse were not those of a research programme. Established Steelmanned the other way: it produced a genuine, isotopically resolved biogeochemical discovery that no other facility could have produced, and it has been a normal university research facility since July 2011, running an ocean lab, an agrivoltaics programme, a sealed Mars analog and the largest laboratory experiment in the Earth sciences. Established Both accounts are true simultaneously and the field would be better off holding them together than choosing.

Handwave “Directed evolution of whole ecosystems is science fiction.” It was demonstrated in 2000: soil microbial ecosystems were subjected to artificial selection and showed phenotypic variation with a heritable component despite comprising thousands of species and millions of individuals. Speculative The standing objection — that an offspring community is a random sample of its parent, so community-level heritability is too weak for selection to act — was answered a quarter of a century ago, and bottleneck size is a tunable parameter. Handwave What remains genuinely handwaving is the leap from soil microcosms to a human-rated compartment, which is several orders of magnitude in complexity and has no demonstrated intermediate. Speculative The interesting question is institutional, not scientific: a demonstrated method has sat unused next to a field that needs it for twenty-six years.

Speculative “Closed loops are the enabling technology for space settlement.” The strongest case against this whole programme is that closed loops are needed only where resupply is impossible, and for every mission humanity is realistically going to fly this century — low Earth orbit, cislunar, Mars on a 26-month synodic cadence — physicochemical regeneration plus resupply is lighter, more reliable, more testable and better understood. Frontier On that reading bioregeneration is a solution waiting for a mission profile nobody is flying, and the field's real product is terrestrial. Frontier The counter is that a closed loop is not a mass-saving technology but a risk technology, valuable precisely when resupply fails, so its worth is an option value that mass-budget comparisons systematically undercount. Speculative This brief declares the tie: the crossover point lives in the Equivalent System Mass literature, no figure could be retrieved for it here, and until somebody publishes one both positions are arguing from intuition.