In the second week of January 1993, inside a sealed glass building in the Arizona desert, the crew’s physician could not add up a column of numbers. Roy Walford was sixty-eight and had lived under the glass for sixteen months with seven others. The air they were breathing had been 20.9 percent oxygen when the doors closed. On day 475 it was 14.4 percent, the air of a mountain pass at 4,000 metres, and it was still falling.

Half of the eight had altitude sickness by now, sleep apnoea among its symptoms, and the crew had asked the doctors outside to be ready to overrule their own, because nobody inside trusted a judgement made inside any more. The failed arithmetic was the point at which the experiment’s central rule, that nothing goes in and nothing comes out, gave way to the people it was keeping. Refrigerated trucks of liquid oxygen came up from Tucson, and between day 475 and day 494 it was pumped through the seal, one chamber first. Mark Nelson, one of the eight, wrote afterwards that the recovery was immediate.

Nobody yet knew where the missing oxygen had gone. Finding out would take a geochemist from New York, a graduate student, and the one thing the building had been built to provide: a system so tightly closed that the atoms could not have left.

That is the spine of the story, and this entry follows it from the theatre company that dreamed the building to the data nobody kept. Each turn has more in it than one entry can hold: how eighty kilometres of glazing was made to leak less than a tenth of its air a year; how two men found twenty-six tonnes of missing oxygen by counting carbon atoms; what two years of hunger did to eight people’s blood, and what happened when they ate again; who the eight were and why five of them never left the field; which species died, which took over, and the discovery that came out of the reef; what the critics were really arguing about; the morning the marshals came; and why the founders believed the Earth was a thing that could have a second instance. Eight shorter pieces take each of them further. They are at the foot of this page.

Who ran it

Biosphere 2 came out of a theatre company. In 1969 John Allen, a Harvard MBA and metallurgist who wrote poetry as Johnny Dolphin, founded Synergia Ranch outside Santa Fe, New Mexico, with a group that called itself the Theater of All Possibilities. Over the next fifteen years the group ran projects on several continents, among them an ocean-going research ship, and developed a philosophy Allen called ecotechnics: that humanity would have to learn to build and run whole living systems, on Earth and eventually off it. One of the people drawn to the ranch was Edward P. Bass, a Texas oil heir. In 1984 Bass and Allen announced Space Biospheres Ventures, and from 1987 to 1991 it built, on a ranch north of Tucson at a place called Oracle, the largest sealed ecological system ever attempted. Bass had put in about $150 million by the time the doors closed; later accounts say $200 million. The name was a claim. The Earth was Biosphere 1, and this was to be the second one ever made.

The eight who went in were, apart from Walford, drawn from Allen’s group and its projects rather than from university science, and had trained for years at the site while the building went up around them. Walford, a professor of pathology at UCLA and the best-known scientist among them, was the doctor, and he had his own reason to be there: he had spent decades showing that mice fed less lived longer, and he wanted to try it on people.

The press noticed the theatre company before it noticed the science. In the months before closure the Village Voice ran a series by Marc Cooper describing the group’s commune origins, Allen’s authority over it, and the thinness of the crew’s scientific training, and asking whether the project was an experiment at all. Allen denied “singular and authoritarian control.” Taber MacCallum’s answer to the word cult was that “many companies and organizations fit it.” The night before the mission began the company threw a party for two thousand people, and on September 26, 1991, with the Secretary of Commerce calling it “a noble experiment,” the airlock closed.

What they were trying to find out

The stated question was whether a sealed, living system could keep eight people alive, fed and sane for two years on sunlight alone, with the air, water and food cycling through plants, soil and sea the way they cycle through the planet. Allen’s longer purpose was a prototype for a colony elsewhere. Walford’s was the diet. The company’s was a working facility that would pay for itself through tourism and research.

The crew’s own expectation was lower and more honest. MacCallum said later that “most people on the Earth thought Biosphere 2 was going to turn to green slime in a matter of weeks.” Walford called the place “the Garden of Eden on top of an aircraft carrier.” They expected things to go wrong. What they did not expect was how the going wrong would be judged.

What they built

The building was a glass-and-steel space frame on a footprint of 1.27 hectares, enclosing about 180,000 cubic metres of air, 30,000 tonnes of soil and four million litres of seawater. Under the glass were a rainforest, a savannah, a desert, a mangrove marsh and an ocean with a coral reef and a wave machine, together holding some 3,800 species, plus a quarter-hectare farm, a wastewater marsh, and a habitat with apartments, a kitchen, a laboratory and a library in a tower. Below ground the whole thing sat on a welded liner of stainless steel; above ground were 16,000 square metres of laminated glass, 80 kilometres of caulked joints and 16 kilometres of welds. William Dempster, the systems engineer, later calculated that every pinhole and construction flaw in all of that added up to a single hole about 19 millimetres across.

The piece of engineering that made the rest possible was a pair of domes the crew called the lungs, whose weighted membranes rose and fell with the air inside so that its pressure never differed from the outside’s by more than a few pascals, and nothing pushed air through the flaws. Dempster measured the leak two independent ways, and both put it under ten percent of the atmosphere a year, about a hundred times tighter than the best sealed chamber NASA had. That figure is the one that made everything that followed into data rather than anecdote.

Because the glass and the frame took half the light, only 45 to 50 percent of the Arizona sun reached the plants. Because the system was closed, condensation off the cooling coils was the water supply: tens of thousands of litres a day, clean enough to drink. And because the designers expected the farm and the wilderness to go out of step with the seasons, with respiration outrunning photosynthesis in winter, they installed a scrubber that blew air through a falling rain of sodium hydroxide to pull carbon dioxide out of it and store it as solid calcium carbonate, with a furnace that was meant to release it again in summer. Hundreds of sensors logged temperature, humidity, light and gas concentrations every fifteen minutes. The furnace’s heating elements failed and the summer half of the cycle never ran. The scrubber was listed in the engineering papers from the start. It would still become a scandal. How the seal itself was made, and why it is the one part of the project nobody disputes, is the first of the pieces below, The seal.

What happened

The first month. Twelve days after closure, Jane Poynter put her hand into the rice thresher and cut off the tip of a finger. It could not be dealt with inside, so she went out through the airlock, had surgery in Tucson, and was back inside in under seven hours. When she returned, she brought a duffel bag. The company did not say so until asked. Poynter has always said the bag held “some drawings and a couple of things”; the reporting said spare parts in the airlock, then food, then that the experiment was compromised. The company’s handling of a small event set the pattern for the large ones: problems were admitted, in Donella Meadows’s words, “too late, reluctantly,” through a management that “shows a consistent preference for image over truth.”

Hunger. The winter of 1991 to 1992 was an El Niño year, and southern Arizona was unusually cloudy. With half the sun already lost to the glass, the farm produced less than planned, and pests the crew could not spray for, broad mites and nematodes among them, took a share of the rest. The eight grew about 83 percent of what they ate and made up the difference from seed stock and from a reserve of grain and beans that had been placed inside before closure. Walford put them on his diet by necessity as much as by design: an average of 1,780 calories a day in the first months, rising toward 2,200 as they became better farmers, ten percent of it fat, mostly vegetables, grains and beans, with a little goat milk, eggs and meat.

In six months the men went from an average of 74 kilograms to 62 and the women from 61 to 54; MacCallum lost 54 pounds. Their blood pressure fell from 109 over 74 to 89 over 58 and their cholesterol from 191 to 123. Walford published the numbers in the Proceedings of the National Academy of Sciences in December 1992, from inside. They were, by every measure he could take, healthier. They were also hungry all the time, and in November 1992 they began eating the seed stock. “Healthy starvation” was Walford’s own name for it. What the diet did to the eight, and what the trials since have made of it, is in Healthy starvation.

Carbon dioxide. The air outside held about 350 parts per million of carbon dioxide. Inside, it swung by 600 parts per million between morning and evening, sat near 1,000 in summer, rose past 3,000 in the first winter and close to 5,000 in the second. The scrubber ran through both winters and took out 4.3 tonnes. It was not enough to flatten the curve. In 1991 Marc Cooper reported in the Village Voice that the company had quietly installed a carbon dioxide scrubber, the kind of machine a submarine carries, and he has said since that it “made the entire experiment worthless, whatever value it had.” The crew and the engineers said it had been designed in from the beginning, which the papers bear out. Both things can be true: it was in the design, and the public had been told a story in which it had no place.

The ecosystems. The hummingbirds died, and so did the honeybees, and with them went nearly every pollinator the designers had brought in. Of the small vertebrates stocked at closure, the usual count is that nineteen of twenty-five species were gone by the end. What thrived was what thrives in any disturbed place: a tramp ant, Paratrechina longicornis, that overran every other ant and the crickets and grasshoppers too; cockroaches, in numbers; and a morning glory that climbed over the rainforest canopy faster than the crew could cut it back. The desert got too much rain from condensation and grew greener than a desert should. The coral reef, which Dempster called the most fragile system in the building, survived.

The crew. Around the sixth month the eight split into two fours. Poynter’s account, which the others dispute in degree but not in kind, is that the groups barely spoke for the rest of the mission, and that people who had been close friends became “implacable enemies.” Nelson, who disputes the degree, allows that one of the eight remembered being spat on twice, and that they held group therapy sessions to keep the fight in the open. The split was about the project’s management and its honesty with the outside, not about the work, and the work went on: the farm was planted and harvested, the animals were fed, the systems were kept running, and nobody asked to leave. Who they were, and where they went, is in The eight.

The advisory committee. Early in the mission, with its credibility already damaged, the company appointed a Scientific Advisory Committee of eleven outside scientists, chaired by Thomas Lovejoy of the Smithsonian and including Eugene Odum, one of the founders of ecosystem ecology. On February 5, 1993, three weeks after the oxygen went in, six of the eleven voted to dissolve it. Lovejoy said he was “frustrated by the lack of progress.” Bass said the cause was personality conflicts with the management. Odum, who thought most of the panel considered the programme a success, said: “What we’ve seen is not a scientific failure; it’s a failure of human relationships.”

Oxygen. The oxygen had been falling since the first weeks, faster in the dark months, at an average of about 140 parts per million a day. The injection in January 1993 took the air back to about 19 percent; it fell again through the spring and summer, and more was added in August. On the engineering charts the two years of oxygen look like a mountain’s two slopes with a cliff between them.

The missing oxygen

The obvious explanation for the loss was respiration: something was breathing in oxygen and breathing out carbon dioxide faster than the plants could reverse it. The obvious explanation did not fit. If respiration had consumed the oxygen, the carbon dioxide should have risen by the same amount, mole for mole, and it had not risen nearly enough. Stranger still, the building’s atmosphere was shrinking: the lungs were slowly falling. Gas was disappearing from the system, and the system was sealed.

Wallace Broecker of Columbia University’s Lamont-Doherty Earth Observatory, the geochemist who had put the phrase “global warming” into print in 1975, took the problem on with his graduate student Jeff Severinghaus, and they did what the closure made possible: an audit. They counted the carbon in the air, the plants, the scrubber’s chalk and the building’s concrete, and read its isotopes to tell which carbon was which.

The soil was the source. The farm beds had been filled rich with compost, and its microbes were respiring organic matter on a scale nobody had budgeted for. The sink was the walls. Carbon dioxide from the air was reacting with some 15,800 square metres of exposed, still-curing concrete and turning to calcium carbonate, about 26 tonnes of it, six times what the scrubber had caught, and every molecule took with it the oxygen that respiration had bound. The oxygen had not leaked. It was in the basement, as stone.

Severinghaus, Broecker, Dempster, MacCallum and Martin Wahlen published the finding in Eos in January 1994. Dempster’s explanation of why the audit worked is also the reason the building was worth building: “It can hardly be overemphasized how powerful is the tool of material closure. It tells you that the atoms you begin with are in there somewhere.” The concrete has since been sealed. There has never been a third closed mission. The full accounting, atom by atom, is in The audit.

After

The eight came out at a quarter past eight on the morning of September 26, 1993, two years to the day after they went in, thinner, to flute music and speeches. The press met them, Meadows wrote a month later, with “soundbites and cruel cartoons,” one of which was captioned “Snow Job and the Several Dwarfs.” The company had spent two years playing down whatever was wrong, and when the crew walked out alive and the worst that could be said was that they had been hungry and short of breath, nobody was disposed to count that as a result.

A second crew of seven went in on March 6, 1994, for a planned ten months. On April 1 Bass, who had decided the company’s finances were out of control, obtained a restraining order; federal marshals arrived, the locks and the computer codes were changed, Allen and the senior management were removed, and a forty-year-old investment banker named Steve Bannon was put in to run the company. He estimated the year’s deficit at sixteen to twenty million dollars.

In the small hours of April 5 (April 4 in some accounts), Abigail Alling and Mark Van Thillo, who had been inside for the whole of the first mission, came back and opened one airlock and three emergency doors from outside, breaking five panes of glass. The doors stood open for about fifteen minutes and roughly a tenth of the air was exchanged. Alling said the project was “in an emergency state,” that the new crew deserved the choice to stay or go, and that “in no way was it sabotage.” Both were charged. What followed, in Zimmer’s summary, was lawsuits and quiet settlements. The second crew chose to stay, and came out on September 6, five months early. Space Biospheres Ventures was dissolved on June 1, 1994. That spring has a piece of its own, The year it fell apart.

Columbia University ran the facility from the end of 1995, cut openings in the seal, and used it as a very large greenhouse for studies of carbon dioxide, coral and drought, until 2003. In 2007 the land was sold to a developer for $50 million, with the University of Arizona taking over the research; the university took full ownership in 2011, with Bass’s money behind it again, and runs it today as an Earth-systems laboratory. It has never been sealed again.

In 2019 Carl Zimmer, writing in the New York Times, went looking for the two years of data, the fifteen-minute readings from hundreds of sensors that had been the project’s whole claim to science. Mark Nelson told him only a small fraction had ever been analysed. The university’s deputy director said it “was not transferred any historical data archives.” Nelson wrote: “It’s shameful to simply accept that these important data are gone.” Walford died in 2004, at seventy-nine, of motor neurone disease, still on his diet. Nelson and Poynter wrote books. Poynter’s title gives the length of the experiment to the minute: Two Years and Twenty Minutes.

What it taught

Start with what it was not. It was not a failure to keep eight people alive; all eight lived, worked, and came out with better blood than they went in with. It was not a leak; the seal held to a standard no one has matched at that size. And it was not, in the end, a mystery; the oxygen was found. The legend of Biosphere 2 as the sealed world that broke is built from a true story told by people who had been promised it would not break. The fairest verdict came from two ecologists asked to review it in 1996, Joel Cohen and David Tilman: “No one yet knows how to engineer systems that provide humans with the life-supporting services that natural ecosystems produce for free.” Nobody knew before, either. The difference is that now there was a measurement.

In a closed system the unmeasured sink always finds you. The designers instrumented the air to the quarter hour and lost nearly a third of its oxygen to a surface nobody had put on the ledger. The soil was chosen for fertility and respired accordingly; the concrete was chosen for strength and absorbed accordingly; neither was wrong for its purpose, and together they bled the building for sixteen months. The lesson is not that they should have known. It is that closure turns every material in the system into a participant, and the materials you have not measured are the ones that will surprise you. Any sailor who has found the bilge full from a hose nobody counted as plumbing knows this.

The closure was the instrument. The same tightness that made the loss alarming made it solvable. Because the atoms could not have left, the missing oxygen was an accounting problem, and accounting problems have answers. Severinghaus and Broecker did not find the oxygen with a better sensor. They found it by summing the carbon in every compartment until the books balanced. A leaky system would have hidden the whole thing; Dempster later showed that at a leak of one percent a day the decline would have been invisible. If you want to learn from a system, seal it, and if you cannot seal it, measure what crosses the boundary before you measure anything else.

Decide before you shut the door whether it is a demonstration or an experiment. The scrubber was designed in; the grain reserve was prudent; the supplies in the airlock were trivial; the oxygen injection was the right call. Each was defensible as engineering and each became a scandal, because the company had sold the public a closed box in which none of them could exist. An experiment publishes its rules and its exceptions and is judged on what it learned. A demonstration promises an outcome and is judged on whether it delivered. Space Biospheres Ventures tried to run both at once, and when the outcome slipped it had no standing left to claim the learning. Meadows’s version was shorter: “in science experiments that don’t work are nothing to hide. They’re the source of learning.”

Eight people are a life-support system too, and the least instrumented one. The air was measured every fifteen minutes; the crew’s cohesion was not measured at all, and it failed at month six. Odum’s judgement that this was “a failure of human relationships” rather than a scientific one is kind to the science and exactly right about the rest. What is remarkable is what held: nobody left, the farm was planted, the systems ran. The group that could not speak to each other still kept each other alive. Whoever plans the next sealed habitat should budget for that split as surely as for the carbon dioxide.

A failure that is measured is worth more than a success that is not, if the measurements survive. The leak rate, the oxygen curve, the diet study and the carbon audit are all published and all still cited; they are the reason anyone writes about this building thirty years on. The rest, two years of fifteen-minute readings from hundreds of sensors, was left in a company that dissolved, and no one can now say where it is. The experiment produced its data and then lost custody of it. Whatever else you plan for, plan for who holds the record when the organisation that made it is gone.

Check the story before you learn from it. The cost is $150 million in some accounts and $200 million in others. The oxygen’s low point is 14.2, 14.4 or 14.5 percent depending on the source and the day; the engineering paper says 14.4 on day 475. The “nineteen of twenty-five vertebrates” figure is repeated everywhere and traces to the mid-1990s review committee rather than to a published census. The scrubber was “secret” in the newspaper and in the design documents at the same time. The duffel bag held drawings in Poynter’s telling and spare parts in the reporting; the three-month food reserve was a separate matter that the retellings fold into it. What the record supports without argument is the oxygen curve, the leak rate, the diet, the carbon audit, and eight people who walked out on time. That is enough to learn from.

Sources

• Jeffrey P. Severinghaus, Wallace S. Broecker, William F. Dempster, Taber MacCallum and Martin Wahlen, “Oxygen loss in Biosphere 2,” Eos, Transactions of the American Geophysical Union 75 (3), 33–37, January 1994. The carbon-isotope audit that found the oxygen in the concrete. doi:10.1029/94EO00285
• William F. Dempster, “Biosphere 2 engineering design,” Ecological Engineering 13 (1999), 31–42. The sealed volume, the lungs, the two leak measurements, the scrubber’s 4.3 tonnes and the concrete’s 26, and the sentence about material closure. ecotechnics.edu
• William F. Dempster, “Tightly closed ecological systems reveal atmospheric subtleties: experience from Biosphere 2,” Advances in Space Research 42 (2008), 1951–1956. The 20.9 to 14.4 percent figure on day 475, the 140 ppm a day, the 19 mm hole, and the simulation of what leakier systems would have hidden. ecotechnics.edu
• Roy L. Walford, Steven B. Harris and Mark W. Gunion, “The calorically restricted low-fat nutrient-dense diet in Biosphere 2 significantly lowers blood glucose, total leukocyte count, cholesterol, and blood pressure in humans,” Proceedings of the National Academy of Sciences 89 (23), 11533–11537, December 1992. The 1,780 calories and every blood and weight figure. pmc.ncbi.nlm.nih.gov
• Joel E. Cohen and David Tilman, “Biosphere 2 and biodiversity: the lessons so far,” Science 274, 1150–1151, November 15, 1996.
• Mark Nelson, “Biosphere 2: What Really Happened?”, Dartmouth Alumni Magazine, May–June 2018. A crew member’s own account: the doctor’s arithmetic, the request that outside physicians stand ready to overrule him, the altitude sickness, the refrigerated trucks, the diet’s calories and the factions. Also his “Some ecological and human lessons of Biosphere 2,” European Journal of Ecology 4 (1), 50–55, 2018, and Pushing Our Limits: Insights from Biosphere 2, University of Arizona Press, 2018. dartmouthalumnimagazine.com, doi:10.2478/eje-2018-0006
• Jane Poynter, The Human Experiment: Two Years and Twenty Minutes Inside Biosphere 2, Basic Books, 2006. The finger, the duffel bag, the factions and the seed stock, from inside.
• Donella Meadows, “Biosphere as theater,” column of October 28, 1993, and her April 1994 column on the break-in. The press reaction, the 30 percent organic soil, and the lessons quoted above. donellameadows.org, donellameadows.org
• Carl Zimmer, “The Lost History of One of the World’s Strangest Science Experiments,” New York Times, March 29, 2019. Bannon, the break-in, Columbia, the University of Arizona, and the missing data archive. carlzimmer.com
• “Biosphere advisory panel disbands in frustration,” Associated Press, Deseret News, February 16, 1993. Lovejoy, Bass and Odum on the committee’s end. deseret.com
• Retro Report, “Biosphere 2: a faulty Mars survival test gets a second act,” with Marc Cooper, Jane Poynter and Taber MacCallum on the scrubber, the duffel bag and the cult question. retroreport.org
• Christopher Klein, History.com, on the Synergia Ranch origins, the second mission and Alling’s statements; Earth magazine, “Benchmarks: September 26, 1991: Crew sealed inside Biosphere 2,” on the El Niño winter and the Columbia years. history.com, earthmagazine.org
• “Biosphere II lives on,” Mongabay, June 2007, for the nineteen-of-twenty-five count and the 2007 handover; James K. Wetterer and colleagues, “Ecological dominance by Paratrechina longicornis, an invasive tramp ant, in Biosphere 2,” Florida Entomologist 82 (3), 1999, for the ants. mongabay.com