An artificial ecosystem is any environment where humans deliberately assemble living organisms and physical conditions to mimic, replace, or extend the functions of a natural ecosystem. The concept covers an enormous range, from a backyard fish pond to a sealed habitat designed to keep astronauts alive on Mars. What ties them together is intent: someone chose which species to include, engineered the physical space, and set the rules for how energy and nutrients cycle. That act of deliberate design is also what makes artificial ecosystems fragile in ways that natural ones are not, and understanding why has become one of the more practically important questions in ecology.
What Counts as an Artificial Ecosystem
The term gets applied to a wide spectrum of systems, and the boundaries are blurry. At one end sit simple, familiar examples: an aquarium, a terrarium, or a planted garden bed. These are artificial in the sense that a person selected the species and maintains the conditions, but they are open systems that exchange air, water, and organisms with the outside world freely. At the other end are fully sealed enclosures like Biosphere 2 or the Chinese Lunar Palace 1, where the goal is to close every loop so that nothing enters or leaves except energy from light. Between those extremes sit constructed wetlands that treat wastewater, artificial coral reefs bolted to the ocean floor, vertical farms growing lettuce under LED lights, aquaponics systems that link fish tanks to plant beds, and synthetic microbial communities engineered in a lab to break down pollutants.
What makes the category useful is not a single definition but a shared set of problems. Every artificial ecosystem has to solve the same basic challenges natural ecosystems solve: recycling nutrients, managing waste, balancing the populations of different species, and staying stable when conditions change. The difference is that nature has had billions of years of trial and error to arrive at working solutions, while artificial ecosystems have to get there by design, often with incomplete knowledge of how the pieces interact.
Biosphere 2 and the Oxygen Problem
No discussion of artificial ecosystems is complete without Biosphere 2, the massive glass-enclosed facility in Arizona that attempted to create a self-sustaining habitat for eight people starting in 1991. It remains the most ambitious closed-ecosystem experiment ever attempted, and its most famous failure is instructive. Oxygen levels inside the structure dropped steadily over the first months, eventually falling low enough that supplemental oxygen had to be pumped in from outside. Researchers later determined that microbes in the facility’s soils were consuming far more oxygen than anticipated as they broke down organic matter, and the carbon dioxide produced by that respiration was reacting with the building’s exposed concrete to form calcium carbonate, effectively locking away carbon that should have been recycled back into the atmosphere through photosynthesis.1Eos, Transactions American Geophysical Union. Oxygen loss in biosphere 2
The lesson was stark: even with a huge physical structure, multiple biome types, and years of planning, the designers underestimated the activity of organisms they could not see. Soil microbes are among the hardest components of any ecosystem to predict, and in Biosphere 2 they effectively hijacked the atmosphere. The concrete interaction was an engineering oversight that no ecological model had flagged. Together, those two problems illustrated a recurring theme in artificial-ecosystem work: the interactions between biological and non-biological components are where surprises live.
Life Support for Space
The most demanding application of artificial ecosystems is keeping humans alive during long-duration space missions. Resupplying food, water, and oxygen from Earth becomes impractical beyond a certain mission length, so space agencies have spent decades developing bioregenerative life support systems that use living organisms to close those loops.
The European Space Agency’s MELiSSA project (Micro-Ecological Life Support System Alternative) takes this idea to its logical extreme. It is designed as a loop of interconnected bioreactors plus a higher-plant compartment, where each stage handles a different job: breaking down waste, recovering nutrients, purifying water, regenerating oxygen, and producing food.2PubMed. MELISSA: a loop of interconnected bioreactors to develop life support in space Because the system is closed, it requires precise, deterministic control of every biological process. Researchers study each unit operation with the same degree of mathematical rigor used for the overall system, covering waste degradation, water recycling, atmosphere revitalization, and food production before integrating them into a single control model.3PubMed. Dynamic aspects and controllability of the MELiSSA project: a bioregenerative system to provide life support in space
China’s Lunar Palace 1 has gone further toward real-world testing. In a 105-day experiment with three crew members, the facility maintained stable atmospheric oxygen and carbon dioxide levels, recycled water, and regenerated about 55% of the crew’s food from 21 co-cultivated plant species. The system also recovered roughly a fifth of the nitrogen from urine and degraded about 40% of solid waste, with a small amount of insect protein produced on-site.4PubMed. How to Establish a Bioregenerative Life Support System for Long-Term Crewed Missions to the Moon or Mars Those numbers fall well short of full closure, but they represent the most successful demonstration so far of a human-occupied artificial ecosystem maintaining itself over months.
A follow-up effort, the Lunar Palace 365 experiment, extended occupation to 370 days and pushed the research further into questions about how living in such a closed loop affects the people inside it. That work is discussed in a later section.
Artificial Reefs and Constructed Wetlands
Not all artificial ecosystems are sealed boxes. Some are built directly into natural environments to restore lost ecological functions or create new ones. Artificial reefs are a prominent example. The long-running debate about whether they merely attract fish from surrounding areas or actually produce new biomass has been a real concern for fisheries managers. Simulation work has shown that attraction is not always harmful: under some conditions it disperses fish biomass across a wider area, actually making them harder to catch and reducing fishing pressure.5PubMed Central. Fish attraction to artificial reefs not always harmful: a simulation study And stable-isotope analysis of a large Mediterranean reef system demonstrated that the structures genuinely supported new biomass production, with invertebrate species depending on organic matter produced locally on the reef and fish occupying trophic positions consistent with a real food web rather than a temporary aggregation.6Marine Ecology Progress Series. Artificial reefs do increase secondary biomass production: mechanisms evidenced by stable isotopes
Constructed wetlands operate on a similar principle: assemble the right combination of plants, soil substrate, and microorganisms, and the result is a system that treats wastewater biologically while providing landscape and habitat value. Recent work has focused on enhancing these systems with advanced biotechnologies to improve nitrogen removal and reduce greenhouse gas emissions, pushing constructed wetlands toward carbon-neutral performance.7PubMed. Comprehensive review of constructed wetlands implemented with advanced biotechnologies for carbon-neutral treatment of urban wastewater In cities, constructed soils offer another avenue. A pilot study in New York mixed inorganic sediments from a municipal soil bank with compost, planted diverse crop combinations, and within two years observed clear increases in soil carbon and microbial biomass, with declines in respiration rates that suggested the soil was maturing into a stable, living system suitable for urban agriculture.8Geoderma. Carbon and nitrogen cycling in an urban constructed technosol: The artist-led carbon sponge pilot study
Aquaponics and Indoor Farming
Aquaponics is one of the more elegant small-scale artificial ecosystems because it links two production systems through a shared microbial community. Fish excrete ammonia; bacteria in the grow beds convert that ammonia first to nitrite and then to nitrate, which plants absorb as fertilizer; the cleaned water recirculates back to the fish. The microbial community doing the conversion, dominated by nitrifying bacteria from the genus Nitrospira in media-bed systems, is essential to the whole cycle.9Bioresource Technology Reports. Coupling, decoupling, and fish tank illumination shape the functional microbiome and nutrient cycling in aquaponics The broader microbial ecosystem in an aquaponics system does more than just nitrification: it mineralizes nutrients, suppresses pathogens, and mediates overall water quality in ways that are still being mapped.10Annals of Microbiology. The complex microbiome in aquaponics: significance of the bacterial ecosystem
Vertical farms are a related but distinct type of artificial ecosystem, though they are simpler in their biology. Most commercial vertical farms grow plants hydroponically under artificial light, with precisely controlled temperature, humidity, and nutrient delivery. The ecological dimension is minimal compared to aquaponics: there is no animal component and often very little microbial diversity by design. The appeal is consistency and land efficiency, but the tradeoff is energy. A life cycle assessment of a large Swedish vertical farm found that its lettuce produced about 0.98 kg of CO₂-equivalent per kilogram of edible product, which was competitive with conventionally imported lettuce and comparable to Swedish greenhouse lettuce, though open-field and polytunnel lettuce still had lower emissions because of their minimal energy needs.11Sustainable Production and Consumption. Environmental life cycle assessment of a large-scale commercial vertical farm A Finnish study echoed this pattern, finding that vertical farming with renewable energy and waste heat recovery had the lowest climate impact across all categories compared to greenhouse production, but that mineral, metal, and water-scarcity impacts could be higher for vertical farms running on average grid power.12The International Journal of Life Cycle Assessment. Assessment of climate change impact and resource-use efficiency of lettuce production in vertical farming and greenhouse production in Finland: a case study
The picture gets more uncomfortable when you zoom out beyond just carbon. One analysis found that while open-field farming has higher impacts from fertilizer, pesticide, and land use, the climate impact of energy-intensive indoor farming may cause greater overall damage to human health and ecosystems than all of those other impacts combined for field-grown produce.13The International Journal of Life Cycle Assessment. The challenges of controlled environment hydroponic farming: a life cycle assessment of lettuce In other words, the artificial ecosystem trades one set of environmental costs for another, and which comes out ahead depends heavily on the local energy grid.
Synthetic Microbial Ecosystems
At the smallest scale, researchers are building artificial ecosystems out of carefully selected microbial species. These synthetic communities are designed so that each member contributes a specific metabolic capability that complements the others. A striking example is petroleum cleanup. A recently assembled three-species bacterial consortium achieved about 86% degradation of total petroleum hydrocarbons, far outperforming any of its individual members or pairings. It was especially effective against long-chain hydrocarbons (the hardest fraction to break down), where degradation rates jumped by roughly 29 to 40 percentage points compared to single strains.14Chemical Engineering Journal. Enhancing petroleum hydrocarbon degradation by a synthetic bacterial consortium: Insights into functional complementarity and regulatory coordination
These microbial ecosystems are “artificial” in the same sense as any other example here: someone chose the members and controlled the conditions. But they raise distinct questions about predictability and containment that more traditional artificial ecosystems do not, which is why they increasingly appear in discussions of biosafety governance.
Why Stability Is So Hard to Engineer
A common assumption is that adding more species to an artificial ecosystem makes it more stable, the way a diverse natural forest seems more resilient than a monoculture plantation. The evidence is more complicated. Experimental work has found that while increasing the number of functional groups in a plant community can improve some ecosystem properties like productivity, it does not reliably improve resistance to disturbance. What matters more is which specific functional groups are present, their identity, rather than how many there are.15Oikos. Stability of ecosystem properties in response to above‐ground functional group richness and composition For designers of artificial ecosystems, this is a crucial finding: you cannot just throw more species in and expect stability. You need the right species for the specific disturbances you expect.
Mesocosm experiments, essentially medium-scale artificial ecosystems used as research tools, have illustrated how quickly things can cascade. In a subtropical lake simulation, warming the water by just a few degrees in the presence of fish triggered a chain reaction: zooplankton populations crashed, phytoplankton bloomed, phosphorus concentrations tripled, and the system lurched toward the kind of eutrophic state that lake managers dread.16PubMed. Fish-mediated plankton responses to increased temperature in subtropical aquatic mesocosm ecosystems: Implications for lake management A small environmental shift cascaded through multiple trophic levels simultaneously via both top-down and bottom-up pathways. In a natural lake, surrounding watersheds and connected waterways buffer against that kind of collapse. In a mesocosm, there is no buffer.
Rapid evolution adds another layer of unpredictability. Populations placed in novel environments can adapt within just a few generations, growing substantially larger and spreading faster than populations without the opportunity to evolve. One experimental study found that evolving populations grew three times larger and spread about 46% faster within only six generations.17PubMed Central. Rapid adaptive evolution in novel environments acts as an architect of population range expansion For artificial-ecosystem designers, this means the organisms you put in are not the organisms you will have a year later. Adaptation is constant, and it can either rescue a failing system or undermine a carefully calibrated one.
What Happens to People Who Live Inside One
The Lunar Palace experiments in China have produced some of the only data on how living inside an artificial ecosystem for months affects human biology. The 105-day trial found that the high-plant, high-fiber diet and structured daily labor in the plant cabin shifted the crew members’ gut microbiota in consistent ways: microbial diversity increased, and bacteria associated with fiber fermentation and gut health became more abundant.18PubMed. The influence of bioregenerative life-support system dietary structure and lifestyle on the gut microbiota: a 105-day ground-based space simulation in Lunar Palace 1 The gut microbiota of the three crew members converged over time, likely a result of shared food, shared air, and shared space.
The longer Lunar Palace 365 experiment (370 days, eight crew members across phases) pushed this further. Researchers identified four specific gut bacteria that were associated with positive mood states during the long enclosure. These bacteria appeared to influence mood through multiple biochemical pathways, including producing short-chain fatty acids from dietary fiber and converting amino acids into neurotransmitter precursors. Animal experiments later confirmed that these microbes had measurable mood-regulating effects.19PubMed Central. Positive mood-related gut microbiota in a long-term closed environment: a multiomics study based on the “Lunar Palace 365” experiment The practical implication for future artificial-ecosystem habitats is tantalizing: the diet the ecosystem provides might shape not just physical health but psychological resilience.
Meanwhile, salivary microbiota and immune markers told a somewhat reassuring story. Oral inflammatory markers actually decreased after the crew entered the enclosed environment, and while individual differences in salivary microbiota shrank as the crew shared the same air and space, the changes reversed quickly after they left. The enclosed environment did not cause lasting harm to oral immunity or microbial composition.20PubMed Central. Effects of Long-Term Enclosed Environment on Human Health Based on the Analysis of Salivary Microbiota and Cytokines Similar observations came from the MARS500 experiment, a 520-day simulated Mars mission in Russia, where the gut microbiota of six crew members remained dynamic and individualized throughout the mission despite the strictly controlled shared environment.21PubMed Central. Temporal dynamics of the gut microbiota in people sharing a confined environment, a 520-day ground-based space simulation, MARS500
Regulation and the Containment Problem
When an artificial ecosystem stays inside a building or a sealed enclosure, governance is straightforward. The complications begin when engineered organisms are meant to function in the open environment, where they interact with natural ecosystems in ways that cannot be fully predicted. European regulatory frameworks, for instance, are built around a binary distinction between “contained use” and “deliberate release,” which maps poorly onto real-world applications of synthetic biology. A biosensor designed to detect arsenic in well water in South Asia, for example, is meant to function outside a lab but is not being “released” in the way a genetically modified crop is. Insects engineered for sterility are released specifically to die out. The existing containment-or-release framework struggles with these intermediate cases.22PubMed Central. Synthetic biology regulation in Europe: containment, release and beyond
Proposals for reform have emphasized moving toward a more flexible assessment model that accounts for degrees of containment rather than treating it as a binary. More broadly, ethical frameworks for synthetic biology, which encompasses many of the most advanced artificial-ecosystem projects, stress the precautionary principle and the need to balance innovation against environmental risk before projects move to field scale.23Journal of Biosafety and Biosecurity. Ethical framework on risk governance of synthetic biology For synthetic microbial consortia like the petroleum-degrading communities described earlier, the question of what happens when a lab-designed ecosystem meets an uncontrolled wild one is not theoretical. It is the central regulatory challenge.
Terraforming as the Ultimate Artificial Ecosystem
If you take the logic of artificial ecosystems to its furthest conclusion, you arrive at terraforming: deliberately engineering an entire planet’s biosphere. The idea has been modeled in some detail for Mars, where the proposed process mirrors ecological succession on Earth. One framework compares it to descending a mountain: starting with a polar desert at the top, passing through tundra and boreal forest, and arriving at temperate biomes where moisture determines whether you get grassland or forest. Each stage in the sequence corresponds to a warmer, wetter climate and a more biologically diverse community.24PubMed. The biological terraforming of Mars: planetary ecosynthesis as ecological succession on a global scale The model suggests that the search for candidate colonizing species should follow the same succession sequence, starting with extremophiles suited to near-Martian conditions and gradually introducing more complex organisms as the climate shifts.
Whether this is achievable on any timescale that matters to humans is an open question. But the framework is useful because it highlights what every artificial-ecosystem project eventually confronts: you are not building a static thing, you are launching a process. The organisms evolve, the conditions shift, and the system moves in directions the designer did not fully anticipate. The gap between what we can design and what we can control is the defining challenge of every artificial ecosystem, from a kitchen aquaponics setup to a hypothetical Martian biosphere.