In 1991, 8 people shut themselves for 2 years inside an airtight Arizona dome, and the walls quietly swallowed their oxygen for 475 days before anyone figured out why

Eight people walked into a huge glass-and-steel building in Oracle, Arizona, on September 26, 1991, and locked the door behind them. They were not coming out for the next two years.

The facility was a 3.14-acre (1.27-hectare) airtight structure built to be a self-contained world, with a rainforest, an ocean with a coral reef, mangrove wetlands, a savannah grassland, a fog desert, farmland, and human living quarters. It's the largest closed ecological system ever created. The eight-man crew was to live off the system's production, breathing its air, drinking its recycled water, and growing their own food: a test run, in effect, for long-term human life in space.

According to a 2008 paper, ‘Tightly closed ecological systems reveal atmospheric subtleties – experience from Biosphere 2’ by engineer William F. Dempster, published in Advances in Space Research, the greatest engineering achievement of Biosphere 2 was not its artificial rainforest or tiny ocean, but how well it contained air. The facility had a volume of 200,000 cubic meters and a leak rate of less than 10 percent per year, making it one of the most airtight non-pressurized enclosed structures ever built. That tightness, it seems, changed what science could detect.


The oxygen that started disappearing
A few months into the mission, something strange started to happen. Oxygen levels within the dome began to fall, slowly, steadily, silently.

Oxygen started at the normal ambient level of 20.9 percent and fell to around 14.4 percent some 475 days later, in mid-January 1993, according to the Dempster 2008 study. The average rate of decline was about 140 parts per million per day for the first 16 months of closure.

In a landmark paper in Eos, Transactions of the American Geophysical Union, titled 'Oxygen loss in Biosphere 2,' Severinghaus, Broecker, Dempster, MacCallum, and Wahlen attributed the oxygen loss to two related processes: the microbes in the soil were respiring at a rate that exceeded the ability of the plants to replace the oxygen through photosynthesis, and the carbon dioxide produced by that microbial activity was reacting with exposed concrete inside the structure, forming calcium carbonate, effectively sequestering both the carbon dioxide and the associated oxygen within the very walls of the building. The crew actually suffered real health effects at those lower oxygen levels and eventually had to supplement oxygen to keep the mission going.

Why was being sealed so tightly made the discovery possible
Dempster’s research demonstrated that computer simulations showed that if Biosphere 2 had leaked at a rate as low as one percent per day, still far tighter than most comparable research chambers, the oxygen decline would have been completely hidden. If fresh outside air had been continuously mixed in, the signal would have been diluted, and the fundamental imbalance between respiration and photosynthesis would never have been detected.

Imagine it as a slow dribble of dye into a sealed tank versus a free-flowing river. The dye disappears in the river but collects in the sealed tank. Biosphere 2 was a sealed tank, tight enough that the science actually showed up.

According to Dempster, this also speaks to something larger: the challenge of tracking where exactly carbon dioxide from fossil fuels ends up once it’s in Earth’s atmosphere. The global carbon cycle has persistent gaps: sources and sinks that don’t quite balance. The closed system, tight enough to capture slow atmospheric drift, is a precise research model that no open-air science can match.

The engineering that made it work
Keeping the dome sealed required solving a fundamental problem. Changes in temperature, humidity, and outside barometric pressure constantly push on any enclosed structure, forcing air in or out through even the tiniest gaps, Dempster says. If you don’t account for those forces, you either burst the building or constantly suck in outside air.

Biosphere 2 did this with two large expandable chambers called "lungs". Dempster’s study found that the lungs inflated to fill the internal atmospheric volume as they warmed and expanded during the day, and deflated as temperatures dropped at night. So the pressure differential between the interior and exterior was maintained within ±8 Pa (0.08 mBar), and this pressure neutralization enabled the low leak rate.

The same study measured the leak rate by two independent methods: by observing lung deflation at a controlled increased internal pressure, and by spiking the atmosphere with inert trace gases, sulfur hexafluoride, helium, and krypton, and then measuring how slowly their concentrations diluted over time. Both techniques confirmed that leakage was below 10 percent per year.

Why it still matters today
In 2007, the University of Arizona assumed management of Biosphere 2, which remains in use today for Earth systems and climate research.

But the larger legacy is what Biosphere 2 taught us about closed-system science. Dempster notes that the problems of slow atmospheric changes, trace gas buildup, and long-term ecosystem stability can’t be experimentally studied with systems that leak several percent per day. These are the sorts of things that would be critical to any real deep-space mission lasting years or decades. Before a crew trusts a closed system with their lives on a Mars mission, they need proof from something that has actually been sealed tightly enough to trap what goes wrong slowly.

The dome in the Arizona desert, no matter all the difficulties, proved that building something that tight is possible. That alone was worth the experiment.