Life-support systems

Water, oxygen, food and energy: the backbone of a Mars settlement

A settlement survives only when its resource loops are measurable, repairable and backed by reserves.

Evidence-led
Original public guide
Updated 4 August 2026

Discover Arcadia Explore the complete series

Closed-loop diagram linking water extraction, oxygen production, food growth, waste recycling and power on Mars

A Martian settlement does not have four separate problems called water, oxygen, food, and energy. It has a dependency network in which every resource is a stock, a flow, a quality requirement, and a capability that must be restorable after failure. This chapter rebuilds that architecture around measurable balances and a simple rule: autonomy is not declared; it is accounted for.

Martian greenhouse combining crops with water-management equipment. — Conceptual visualization of vital-resource coupling: food, water, air and energy are not four independent stocks but interacting loops that impose mass, power and reliability trade-offs.

Martian vital-resource network linking fission and solar power, water, oxygen, food, storage, and ECLSS. — Vital resources form a coupled network: production is not enough; distribution, storage, measurement, and restart capability matter as well.

Water: close the loop without forgetting makeup

A high recovery rate never removes the need for makeup water. A town must know how much water it truly loses, where that water goes, and the quality of what returns to the loop.

Recovery boundary. A recovery percentage must state which streams are counted: urine, condensate, hygiene water, process water, brines, and humidity do not always share the same denominator.

Potable quality. Recovered water is not automatically potable water; sensing, analysis, catalytic treatment, and the ability to reprocess off-spec water are separate functions.

Local ice. Access to subsurface ice turns water from imported mass into a mining resource, while adding excavation, heating, purification, and geological variability.

Buffer storage. Emergency storage should be sized in days of survival rather than only as a percentage of annual consumption.

Water as shielding. Water reserves can also contribute to radiation shielding, provided that this role does not make maintenance or access unsafe.

Reproducible calculation — Firm power

P_ferme = P_installée × disponibilité

With 40 kW electric installed and 0.95 availability, expected firm power is 38 kW. This does not replace a peak-load balance, but it forces nominal power to be separated from power that is actually available.

Oxygen, CO2, and buffer gases

A habitat atmosphere is a composition to maintain, not merely a tank of oxygen. Partial pressures, humidity, contaminants, and buffer gases jointly determine what the crew actually breathes.

O2 production. Water electrolysis can provide breathing oxygen; on Mars, local pathways can also produce oxygen from Martian resources, but at different maturity levels.

CO2 scrubbing. Metabolic CO2 production requires scrubbers that can handle activity peaks and remain controllable during partial failures.

Nitrogen and argon. Buffer gases affect total pressure, fire risk, and leak reserves; their logistics must not disappear behind the focus on oxygen.

Humidity and condensate. Water released by breathing, perspiration, and cooking becomes a resource when condensate is captured, but a mold and corrosion risk when it is not.

Leak detection. Pressure decline is only a signal; localization, estimated leak rate, and the ability to isolate a volume determine the time actually available to act.

Feed a population without exhausting power

Food has to be treated as a system combining stocks, crops, water, light, nutrients, labor, and microbiological safety. A greenhouse that produces heavily while consuming the entire power reserve can weaken the settlement.

Transitional stores. Shelf-stable food remains essential during agricultural ramp-up, crop disease, or production failures.

Fresh crops. Crops provide nutrition, variety, and behavioral benefits, but require volume, climate control, water, lighting, and repeated operations.

Protein and fat. A food architecture cannot be evaluated only in kilograms of plants: amino acids, fats, micronutrients, and menu acceptability matter.

Seeds and diversity. Seed stocks must preserve genetic diversity, germination capability, and the ability to recover after contamination or loss of a variety.

Cooking and processing. Milling, cooking, fermenting, and preserving change power, water, waste, and crew-time balances; the kitchen is part of the food factory.

Reproducible calculation — Annual water makeup

m_appoint = N × q × 365 × (1 − R)

For N residents with a q kg/day flow inside the stated boundary and recovery ratio R, this relation computes minimum makeup before other losses. Each symbol must use the actual flow and recovery boundary.

Store, distribute, and measure

A resource that is produced but inaccessible at the right place and time is not an available resource. Storage and distribution therefore deserve the same engineering rigor as production machines.

Segmented tanks. Multiple isolatable volumes limit the effect of a leak or contamination event, at the cost of additional valves, sensors, and procedures.

Local loops. An extended settlement may prefer sub-networks able to operate temporarily on their own instead of one central network crossing the whole base.

Inventory metrology. Level, mass, pressure, conductivity, or chemical analysis reveal what actually exists; sensor errors become decision errors.

Quality traceability. A batch of water, gas, or food should be traceable to its origin, processing steps, and quality results when an anomaly appears.

Distribution losses. Pipes, seals, purges, and transfers create losses distinct from the main process and should be accounted for separately.

From four people to a town

The most important scaling factor is not the 250-fold multiplication from four to a thousand people. It is the emergence of separate services, redundant networks, specialized staff, and maintenance that can no longer rely on a few generalists.

Four-person outpost. At very small scale, the same machines cover several functions and each major failure can involve almost the entire crew.

Twenty-person base. Specialized teams become possible and some systems can be duplicated without duplicating every interface.

Hundred-person community. Logistics begins to resemble small municipal infrastructure: reserves, distribution, laboratory services, workshops, and on-call coverage must be organized.

Thousand-person town. Multiple districts, production units, and strategic stocks reduce the consequences of a single loss while requiring coordination and standards.

Growth margin. Interfaces should accept additional trains without shutting down the entire settlement, otherwise every expansion becomes a high-risk operation.

Reproducible calculation — Buffer stock

M_tampon = débit_journalier × jours_d_autonomie

A critical flow of 2,400 kg/day and seven days of reserve gives 16,800 kg. That becomes an architectural choice involving number of tanks, isolation, location, and ability to preserve part of the stock after an incident.

Coupled failures and degraded modes

Vital resources become most fragile when a failure crosses networks: power loss, reduced ventilation, rising CO2, then restricted activity. Scenarios therefore need to test chains rather than isolated components.

Long power outage. Load shedding must preserve vital functions and stocks while preventing a sudden restart from collapsing the grid again.

Water contamination. The network must isolate a suspect batch, retain a safe supply, and identify the process or sensor that caused the problem.

Common-mode air fault. Two identical scrubbers sharing the same software, consumable, or sensor are not necessarily two independent protections.

Seasonal crop failure. Disease or lighting loss does not immediately become famine if reserves and food diversity were designed to bridge several crop cycles.

Isolated safe haven. A credible safe haven has its own air, water, power, and communications to endure while the rest of the base is isolated.

Toward measurable autonomy

The word autonomy should be replaced by concrete indicators: days of stock, fraction of mass produced locally, mean repair time, share of consumables regenerated, and dependencies that still cannot be manufactured locally.

Critical imported mass. A tiny irreplaceable part can be more strategic than tons of locally available material; inventories should rank dependencies by criticality.

Diagnostic capability. Local production is useless if the settlement cannot measure quality, identify drift, or certify a repaired part.

Human skills. Operators need to understand physical mechanisms rather than merely follow an interface, because unanticipated situations will be inevitable.

Operational learning data. Every failure, repair, and drift event should feed a local history that can change maintenance, stocking, and architecture.

Reversible error budget. A mature infrastructure favors decisions whose consequences can be detected and corrected before available margins are exhausted.

Reproducible calculation — Availability of a vital train

A = MTBF / (MTBF + MTTR)

With a 2,000-hour MTBF and 10-hour MTTR, A ≈ 0.995. This value does not cover common-cause failures or spare shortages; it is only one piece of the reliability argument.

At one thousand residents, breathing becomes an industrial activity

Vital resources are often shown as four stocks: water, oxygen, food and power. A settlement experiences them as continuous flows that enter, circulate, transform, leak and require restoration. NASA’s carbon-dioxide technical bulletin gives design-reference values of about 0.82 kg O₂ consumed and 1.04 kg CO₂ produced per person per standard day with exercise. These are design loads rather than universal biological constants.

Scaling of metabolic oxygen and carbon dioxide flows from one to one thousand people — City-scale life support requires industrial machines, piping, inventory and maintenance. This view accompanies “At one thousand residents, breathing becomes an industrial activity” and locates the elements whose technical dependencies are developed in the surrounding text.

Scale arrives quickly. For 100 people, reference oxygen consumption is 100 × 0.82 = 82 kg/day, about 29,930 kg/year, nearly 30 tonnes. CO₂ production is about 104 kg/day. At 1,000 residents the same reference becomes roughly 820 kg O₂ consumed and 1.04 tonnes CO₂ produced every day. Recycling does not make the compressors, sorbents, pipework and storage disappear.

Water has the same scaling issue but more complex loops. NASA demonstrated about 98% overall water recovery on ISS with the brine processor. A Mars settlement cannot simply assume “2% makeup” of one fixed number: hygiene, crops, industry, maintenance losses and mobile habitats change the flow network. Closure has to be defined stream by stream and mode by mode.

Resources need buffers. A water plant can stop for hours without immediate danger if tanks exist. Some atmospheric contaminants allow far less inertia. Inventory should therefore be expressed as days of autonomy after failure, not only litres or kilograms. The same tank can be generous for four people and trivial for one hundred.

Power is the hidden multiplier. More water closure may require heat, pumps and maintenance; oxygen production needs compression and electrolysis; food production needs light or transparent area, water circulation and thermal control. The optimum architecture does not maximize every recycling percentage; it minimizes whole-system risk and dependence.

At city scale, life-support and industrial flows merge. CO₂ may become plant or chemical feedstock; process water may return to treatment; waste heat can warm a greenhouse; recovered nutrients can re-enter agriculture after health controls. Every connection improves efficiency but also creates a path for contamination or failure. Isolation valves, analysis and rapid decoupling therefore become essential.

Going from four to a thousand people is not simply multiplying everything by 250. Some systems gain economies of scale while others become enormous single points of failure. A settlement becomes durable when vital resources stop being mission consumables and become repairable, measured, redundant industrial services.

Flows must be converted into safety inventory. For 100 residents, a reference metabolic oxygen demand of 82 kg/day becomes 574 kg over seven days. That is not a storage recommendation; it shows how one week of autonomy immediately becomes industrial mass. If local production is available the reserve may be smaller, but it must cover a credible plant-repair duration rather than an arbitrary number of days.

Peak flow is not average flow. Showers, kitchens, greenhouses and workshops create demand peaks. A network sized only to daily average use may lose pressure or treatment capacity for several hours. Tanks and accumulators decouple production from demand, allowing some plants to operate efficiently at steady state while the settlement consumes irregularly.

At 1,000 residents, a small percentage error becomes large. An extra 1% loss on a 100-tonne-per-year stream is one tonne annually. Efficiency therefore has to be monitored by mass balance: measured inputs, useful outputs, releases, inventories and uncertainty. A Martian town cannot manage resources only with tank-level indicators; it needs physical accounting.

Primary and institutional sources : www.nasa.gov ; ntrs.nasa.gov ; www.nasa.gov ; ntrs.nasa.gov.

What ISS operations, Mars demonstrations, and 2026 architectures actually allow us to claim

A nearly closed loop still leaks

The ISS 98% water-recovery milestone is important because it shows that part of a brine stream once discarded can be recovered. But that percentage must not become a natural law for a settlement. It depends on accounting boundaries, system configuration, accepted water quality, and operations. On Mars, marginal losses become strategic because they accumulate over years. A system losing 2% of a 1,000 kg/day defined process stream loses 20 kg/day, or 7,300 kg over an Earth year if flow and recovery remain unchanged. This does not predict a town's real consumption; it shows why every percentage must be converted into mass and time before it can be judged.

MOXIE demonstrated a reaction, not an oxygen utility

MOXIE proved on Mars that oxygen can be produced from atmospheric CO2 with an electrochemical device. The lesson is profound: a planetary resource can be transformed locally. A town, however, must cross several additional orders of magnitude in continuous flow, lifetime, dust filtration, compression, storage, maintenance, hot-component replacement, electrical supply, and purity control. Moving from demonstrator to utility is not a simple multiplication of nominal flow. It requires independent trains, reserves that bridge a train outage, and restart procedures. If oxygen also serves propulsion, breathing and industrial networks must be separated or protected so a propellant campaign cannot endanger the habitat.

Surface power changes the status of every other resource

Water, oxygen, and food are often described as material stocks; they are also indirect energy stocks. Mining ice, heating regolith, electrolyzing water, lighting crops, and compressing gases require power at the moment the operation must occur. NASA's selection of fission as the primary surface-generation technology for initial human Mars missions does not mean solar disappears. It signals how heavily continuity weighs in the initial architecture. A town may later combine fission, solar, storage, and load shedding, but it must measure every resource's dependence on electricity. A grid failure is not only an electrical failure: hours later it can become a ventilation, pumping, food-refrigeration, or water-treatment failure.

Real autonomy is measured in days to shortage

Saying that a town produces 80% of its food reveals little if the remaining 20% includes a vitamin, medicine, or additive with no substitute. The useful metric is time to shortage for each critical dependency. Six months of an ordinary consumable may matter less than five years of an irreplaceable component whose failure would stop the water loop. The same reasoning applies to seeds, nutrients, membranes, catalysts, and power electronics. In practice, the base needs a dependency register linking each item to stock, consumption, resupply delay, repairability, and substitute pathways. That accounting turns autonomy from a slogan into an operational variable.

What Mars still has to demonstrate

The figures assembled here mostly come from terrestrial systems, ISS operations, laboratory work, and mission architectures. They are not yet operating statistics from a Martian town. Early settlements will therefore need to measure their own water, oxygen, food, and energy flows, publish deviations from predicted balances, and revise margins from losses actually observed on Mars.

Water, oxygen, food and energy form a dependency graph. A greenhouse needs electricity and water; a power fault changes ventilation and cold storage; a water-quality event can affect both people and crops. Autonomy comes from preventing one fault from propagating through all vital resources, not from four independent stockpiles.

As the settlement grows, reserves become service-specific. Days of margin, seasonal demand, priorities and shedding rules must be explicit. A durable city knows which non-vital activity to curtail in order to preserve breathable air, potable water, minimum nutrition and medical capability.

Official sources and live resources

The references below document the underlying life-support, resource and technology evidence. They should be read for their measured scope and limitations rather than as endorsements of any settlement timetable.

NASA — Humans to Mars: living and working NASA — Mars Technology Development NASA — MOXIE instrument Open the complete Mars resources and live missions page

Official corporate pages describe the organization’s own plans and announced schedules. Public social-media feeds are dynamic and may include unverified third-party content.

Primary and institutional sources

Vital resources: foundations before the new sizing

Essential foundations already established

Energy is the common currency of every life-support loop

Water, oxygen, heating, agriculture, communications and industry look like separate problems until electricity disappears. Pumps stop moving water; fans stop carrying CO₂ to scrubbers; electrolysers stop making oxygen; greenhouses lose lighting and thermal control; workshops cannot make the part needed for repair. Electrical power is therefore not merely another subsystem. It sets the maximum rate at which almost every other loop can function.

In 2026 NASA states that nuclear fission has been selected as the primary initial Mars surface power-generation technology for human missions in the Moon to Mars architecture trade space. That does not make solar power irrelevant. It means the initial architecture values a firm source independent of sunlight and dust storms for vital loads, while solar, storage and other sources can supplement it and serve interruptible work.

A 4 / 20 / 100 / 1,000-person scenario for understanding firm power

Build a teaching model, not a NASA requirement. Assume a fixed common load of 50 kW for communications, computing, pumps, minimum workshop and thermal functions, plus 8 kW average per inhabitant combining daily life and early productive activity. Average power is P = 50 + 8N kilowatts, where N is population. Add a 35% design margin for maintenance, peaks and unavailable equipment: P_firm = 1.35 × P.

Four inhabitants give P = 82 kW and about 111 kW firm with margin. Twenty give 210 kW then 284 kW. One hundred give 850 kW then 1.15 MW. One thousand give 8.05 MW average and about 10.87 MW of firm capacity. A genuinely industrial town might demand much more; the point is scale. Moving from four people to a thousand means building a small industrial grid, not adding a few more panels.

Installed megawatts are not guaranteed megawatts

Solar output depends on time of day, season, latitude, orientation, temperature, atmospheric opacity and dust deposition. A reactor can provide a steadier base but still has outages, conversion hardware, radiators, controls and maintenance. Life support cares about the power that can be guaranteed in the selected adverse case after losses and failures.

Storage makes the scale visible. If the thousand-person example requires 10.87 MW firm and batteries were expected to cover only 12 hours, theoretical energy would be E = P × t = 10.87 MW × 12 h ≈ 130 MWh, before efficiency, depth-of-discharge, cold-temperature and ageing penalties. Forty-eight hours approaches 522 MWh. This does not rule out batteries; it explains why large settlements need firm generation, load shedding and the ability to postpone industrial work.

The microgrid must know how to break itself apart

A Martian grid should support islanding: a healthy district separates from a faulted one and maintains life-critical loads. Each island also needs black start. Loads are ranked: survival, safety, communications, essential production, then comfort and deferrable industry.

The hierarchy must be physically implemented and tested. A greenhouse can tolerate reduced lighting for a time; CO₂ removal in a crowded refuge cannot. Propellant production can pause during a peak; a thermal circulation pump may not. Grid intelligence is not keeping everything at 100%; it is deciding very quickly what must remain alive.

From outpost to town, power becomes an institution

NASA and the Department of Energy are developing 40-kW-class fission surface-power technology for the Moon as a step toward broader exploration. Forty kilowatts is a technology milestone, not a Martian city power plant. A settlement will combine sources, maintain them, manage safe zones, converters, cables, spares and expansion over decades.

At hundreds of inhabitants, power also becomes governance expressed in hardware. Who may shed a factory? What reserve remains protected for hospital, communications and water production? How is industrial growth prevented from consuming the margin needed for a dust-storm emergency? These are policy questions only after engineers turn them into breakers, thresholds, control logic and real capacity.

Water and air connect directly to this network. Every litre mined, every kilogram of oxygen electrolysed and every kilogram of CO₂ removed requires thermodynamic work. Energy, water and ECLSS are not separate chapters of a colony; they are three views of the same matter-and-power system.

WaterMine ice, purify it, store it and recycle it.

OxygenSeparate it from CO₂ and water, then compress and store it.

FoodCombine stored staples with controlled-environment crops.

PowerDesign for dust, night, maintenance and peak industrial demand.

98% Recovery: where are the 2% — and should a settlement really target 100%?

The 98% demonstrated on ISS is impressive, but it must be interrogated. It is a recovery performance over defined collected and treated streams; it does not mean that 98% of every water flow in a Martian city, including agriculture and industry, can be recovered forever.

The unrecovered fraction can remain in concentrated brines and residues, stay trapped in filters, leave with wastes or arise from purges, leakage and maintenance. As more water is extracted from a residue, salts and contaminants become more concentrated, potentially increasing energy use, fouling and maintenance difficulty.

The engineering question is therefore a trade. If moving from 98% to 99.5% requires large extra mass, energy and consumables, it may be more robust to accept a small irreducible loss and replace it using locally extracted Martian water. The site now applies this rule to every impressive percentage: what exactly is measured, where are the losses, what does the final percentage cost, and what happens as the system ages?

Teaching example for the simplified 76-litres-per-day flow already shown for 20 people: at 98%, theoretical make-up is 1.52 litres/day. At 99%, it is 0.76 litre/day. The improvement is therefore only 0.76 litre/day for this defined stream. Before choosing 99%, compare that gain with the additional energy, membranes, filters, equipment and failure modes needed to obtain it.

This systems overview distinguishes demonstrated life-support performance from engineering extrapolations for a Mars settlement. Values that come from terrestrial or orbital systems are not treated as automatically transferable to Mars.

Life-support reality

A recovery percentage must be converted into make-up demand

A stated 98% recovery is useful only after defining the stream and the missing fraction. In a simplified 100-unit flow, 98 units return and 2 units require make-up. Over a long settlement lifetime, that small fraction becomes a supply requirement, so engineers must identify where it leaves the loop and whether recovering more would cost more mass, energy and maintenance than extracting local water.

Power must be budgeted as a firm service, not a nameplate number

Water processing, electrolysis, refrigeration, lighting and agriculture all draw from the same electrical system. A settlement therefore needs to know what power remains available during dust, maintenance, battery degradation and a failed generator. The design question is not only how many megawatts are installed, but how many kilowatts of critical service can be guaranteed hour after hour.

The colony does not need four resources; it needs one coupled system
NASA — Water recovery milestone on ISS

Water, oxygen, food and energy are often presented as separate chapters. On Mars they are a single network. Water is consumed by people, crops, hygiene, cooling and industrial chemistry. Oxygen is needed for breathing and potentially in far larger quantities as rocket oxidizer. Food production consumes water, nutrients, carbon dioxide, light and crew labor. Every pump, heater, compressor, lamp and fan consumes electrical power. Waste treatment can recover resources, but it also adds machinery that must be cleaned and repaired.

The engineering objective is not a perfectly closed ecological loop from day one. It is a controlled set of loops with known losses, adequate storage and the ability to bypass a failed processor. A settlement that recycles 98 percent of its water but has no replacement pump may be less resilient than one that recycles 90 percent and carries several months of reserve.

Life-support is also where the practical value of Mars becomes clear. Water ice, atmospheric carbon dioxide and mineral-rich regolith are not abstract “space resources”. They are the on-site inputs that can gradually transform a fragile outpost into a resilient settlement. For the broader strategic rationale, see Why go to Mars?.

Water: the first local resource to industrialize

Robotic data show that subsurface water ice is present in parts of the Martian mid-latitudes, including regions where it may lie within roughly a meter of the surface. That does not mean settlers can simply melt clean blocks. Ice may be mixed with regolith, salts and dust. Its depth and concentration vary. Excavation equipment must work in cold, abrasive conditions and transfer material without releasing large amounts of vapor into the thin atmosphere.

A practical chain could involve enclosed excavation, insulated transport, thermal extraction, vapor capture, filtration, mineral removal and storage. The water system would then split into quality grades. Potable water requires the strictest treatment. Hygiene, cooling and industrial processes can often use lower grades. Fire reserves and radiation-shielding tanks can double as strategic storage.

Location decisions should therefore use delivered water cost, not just ice presence. That cost includes excavation energy, heater efficiency, contamination, equipment wear, distance from the habitat and the labor needed to maintain the plant.

Recycling water without creating a single point of failure

Space stations already recover water from humidity and urine, and future deep-space systems aim for higher closure. A Mars settlement would recycle cabin humidity, hygiene water and processed wastewater. Yet biological growth and industrial operations create streams more complex than those of a small spacecraft. Soil substitutes, nutrient solutions, cleaning chemicals and trace contaminants must be monitored continuously.

The safest design separates loops where necessary. Drinking water should not depend on one biological reactor. Crop water can tolerate different treatment pathways. Emergency tanks should be isolated from daily circulation. Sensors must be backed by manual sampling because a failed sensor can be as dangerous as contaminated water.

Oxygen: demonstrated on Mars, not yet industrialized

NASA’s MOXIE instrument operated sixteen times and produced oxygen from the carbon-dioxide-rich Martian atmosphere. It proved the central chemical principle under real Martian conditions. A colony-scale plant, however, would need to ingest and filter large volumes of thin, dusty air; compress it; heat an electrolysis stack; separate gases; reject heat; and store oxygen safely for years.

Breathing oxygen is only one demand. A crew of several people consumes a manageable daily quantity compared with the oxidizer required to launch a large vehicle from Mars. This is why mission architectures often connect oxygen production to propellant strategy. If return flight depends on local oxidizer, the plant must run long before the crew arrives and fill verified tanks with substantial margin.

Oxygen can also be produced by electrolyzing water, which yields hydrogen as well. The best architecture may combine atmospheric carbon-dioxide processing with water electrolysis, using each pathway according to power availability, propellant chemistry and equipment condition.

Food: stored calories first, biological production second

The first crews will not live entirely from Martian agriculture. Stored food provides predictable nutrition with far less power and hardware than crop production. Fresh crops nevertheless matter for vitamins, menu variety, morale and partial air and water regeneration. NASA’s space-crop research is explicitly aimed at sustainable production systems for deep-space exploration, but the technology remains an active field rather than a solved farm package.

Crops would grow in controlled chambers, not in untreated Martian soil. The chambers must regulate pressure, temperature, humidity, carbon dioxide, light, nutrients and pathogens. Martian regolith contains no organic ecosystem equivalent to agricultural soil and can contain chemically troublesome compounds. Processed local minerals may eventually contribute structural media, but nutrients and microbial management require deliberate systems.

A sensible food strategy develops in layers. Early missions carry all staple calories and grow small quantities of leafy vegetables and herbs. Later farms add compact crops with high edible yield, predictable cycles and useful nutritional profiles. Only a larger settlement can justify extensive staple production, oil crops and animal protein systems. Even then, cellular agriculture, fermentation and insect or aquaculture systems may use resources more efficiently than conventional livestock.

Waste is inventory in the wrong form

Human waste, crop residues, carbon dioxide and packaging all contain recoverable material. ESA’s MELiSSA program studies a regenerative loop intended to recover food, water and oxygen from waste streams using linked biological and physical processes. The long-term potential is significant, but biological loops introduce their own vulnerabilities: contamination, population instability, slow recovery and the difficulty of diagnosing invisible changes.

For that reason, early settlements should combine biological and physicochemical methods. Drying, oxidation, filtration, distillation, nutrient recovery and controlled bioreactors can be arranged as modular processes. Modules can be isolated, sterilized and restarted. A single elegant ecosystem that cannot be bypassed would be too fragile.

Energy: the resource that determines all the others

Power demand on Mars is not limited to lighting and computers. Ice extraction requires excavation and heat. Oxygen production requires compression and high-temperature electrochemistry. Habitats require continuous circulation, heating and carbon-dioxide removal. Greenhouses may require artificial light. Workshops, laboratories and medical systems create variable peaks.

Solar arrays are attractive because they are modular and have extensive operational heritage. Mars, however, receives less sunlight than Earth and experiences dust deposition and storms that can reduce output. Arrays require cleaning strategies, storage and enough excess capacity to survive seasonal lows. Fission surface power can provide continuous output independent of sunlight; NASA is developing systems first for lunar use with extension toward Mars. A robust settlement would likely value diversity more than ideological purity: distributed solar, storage, fission baseload and emergency generators serving separated microgrids.

Storage is not passive

Large tanks and batteries are often treated as secondary hardware. In reality, storage decides whether variable production can support continuous life. Water tanks buffer extraction outages. Oxygen tanks separate production from consumption. Food stores bridge crop failures. Batteries or other storage bridge night and transient demand. Thermal storage can preserve heat from industrial processes.

Inventory must be physically distributed. A fire, impact or pressure failure should not remove all reserve of one resource. Digital records should track quantity, quality, location and expiration, while local gauges allow verification when software or networks fail.

The governing design rule: fail gradually

Life-support systems should not switch from normal to fatal in one step. Loss of one greenhouse should reduce fresh food, not total calories. Loss of one oxygen unit should consume reserve while another unit increases output. Loss of a power source should shut down non-essential industry before cabin circulation. This hierarchy requires explicit load shedding, protected emergency buses and crew procedures tested under realistic delay and isolation.

The first Martian settlement will not be self-sufficient in the romantic sense. It can nevertheless become resilient by knowing every flow, keeping strategic reserves, diversifying production and designing every critical loop for inspection and repair.

Three simple calculations that make the coupled system visible

1 — Water reserve

2,400 L usable at a net demand of 80 L/day gives 2,400 ÷ 80 = 30 theoretical days without production.

2 — Oxygen production

If one line produces 20 kg/day and the scenario demand is 16.8 kg/day, nominal surplus is 20 − 16.8 = 3.2 kg/day. That surplus is not a reserve: it disappears if the line stops.

3 — Available power

Four 40 kW units would represent 160 kW of nameplate power if all could operate at that level. A real design must then account for maintenance, outages, losses and margins.

These calculations are deliberately pedagogical. They show the key questions: how much is stored, how fast can we produce, and how long can we endure if one function disappears?

Industrial priority: secure critical dependencies before pursuing total autonomy

Measure and control. Reliable sensors, analysis and metrology.
Secure power and thermal control. Without them, other loops slow or stop.
Secure water. Local extraction, treatment, recycling, separated storage and emergency purification.
Secure oxygen and atmosphere. Multiple production or reserve paths, leak detection and compartment isolation.
Expand local food. Gradually, because biological systems require time, light, nutrients and biosafety.
Recycle materials. Turn wastes and effluents into reusable inventories.

This is a priority logic, not a universal blueprint. Site conditions, population, transport capability and accepted risk can change the detailed order.

The real system bottleneck: energy available when it is needed

Mining ice, pumping fluids, electrolyzing water, compressing oxygen, heating a greenhouse and recycling waste all require energy. Two systems with the same maximum nameplate power can provide very different services if one is intermittent and the other dispatchable.

NASA is developing 40-kilowatt-class fission surface power concepts for continuous off-world operation. This illustrates an essential distinction: installed power is not automatically firm power available at every hour. A real architecture must include generation, storage, margins, maintenance, conversion losses and load-shedding priorities.

During a crisis, the relevant power figure is the firm capacity still available to air, water, thermal control, communications and shelter after the failed generation or storage element has been isolated.

Three reserve horizons: immediate, operational and strategic

Exact durations must be sized for the mission and risk model, but it is useful to separate several decision horizons.

Immediate reserve absorbs a short failure or planned maintenance. Operational reserve provides days to diagnose, repair and test. Strategic reserve protects against a longer outage, unavailable spare or missed resupply.

These labels do not imply universal values such as exactly 24 hours, 7 days or 30 days. Their purpose is to connect stored quantity to reaction time. A tank becomes a resilience tool, not merely a container.

A vital resource must be designed as a chain, not as a machine

For water, the full chain may be: prospecting → extraction → melting → filtration → purification → storage → distribution → use → collection → treatment → quality control → return to storage. Every arrow is a possible failure point.

Oxygen has its own chain: feedstock, production, purification, compression, storage, distribution, sensors and reserves. Food has another: seed or stored food, water, light, nutrients, growth, harvest, processing, storage, distribution and waste recycling.

The design question is therefore always double: what happens if one step stops? and what alternative path can maintain a minimum service?

Stock, flow and consumption: three numbers that must not be confused

A settlement can own a large amount of water and still face a shortage tomorrow morning. Three different quantities govern resources: stock, flow rate and consumption.

Stock answers “how much do we have now?” Flow rate answers “how much can we produce or process per hour?” Consumption answers “how much do we use over a given time?”

A huge stock does not permanently compensate for a treatment plant that is too slow. A fast plant does not protect against failure if there is no buffer stock. A robust system combines production capacity, reserves and the ability to reduce demand.

Teaching example

If a usable reserve contains 2,400 L and net consumption is 80 L/day with no production, the theoretical endurance is 2,400 ÷ 80 = 30 days, provided demand stays near 80 L/day and the full stock remains usable.

Where would humans actually live on Mars?

A realistic guide to Mars habitats: pressure shells, radiation shielding, regolith cover, lava tubes, dust control, interior design, repair and emergency zoning.

Who should be selected for the first permanent settlement on Mars?

A realistic framework for selecting the first Mars settlers: overlapping skills, medical risk, psychology, diversity, team compatibility, legitimacy and replacement planning.

How would the first Mars colony govern itself?

How a Mars settlement could govern emergencies, scarce resources, crime, work, contracts and local autonomy while remaining legally connected to Earth.

Explore the books behind the broader Mars project

This resource chapter stays focused on the coupled flows that keep a settlement alive: water, oxygen, food, and energy. Wider political, social, and architectural consequences are handled elsewhere so that the calculations here can remain explicit and auditable.

Frequently asked questions

Is there enough water on Mars for a colony?

Mars contains abundant water ice, but usable supply depends on site, depth, concentration, contamination, excavation energy and processing reliability.

Has oxygen been made on Mars?

Yes. NASA’s MOXIE technology demonstration produced oxygen from atmospheric carbon dioxide. Colony-scale continuous production has not yet been demonstrated.

Could people grow food directly in Martian soil?

Not safely without processing and controlled conditions. Crops need pressure, water, nutrients, light and protection from contaminants; untreated regolith is not equivalent to fertile soil.

Is nuclear power necessary on Mars?

Not necessarily for every architecture, but continuous fission power can reduce dependence on sunlight and storage. A resilient settlement may combine fission and solar rather than rely on one source.

2026 Source update — 98% water recovery does not mean “no losses”

Two NASA figures are often quoted for water recycling, but they do not describe exactly the same system boundary. In 2023, after the Brine Processor Assembly was incorporated, NASA reported a demonstration of about 98% total water recovery aboard the ISS; the agency noted that overall recovery had previously been about 93–94%. A NASA ECLSS reference updated in 2025, meanwhile, describes the Water Recovery System as recovering about 90% of station water. Those figures should therefore never be copied as if they were a universal constant: the subsystem, configuration, denominator, and test period must be stated.

If an architecture reaches 98%, the pedagogical arithmetic is straightforward: out of 100 litres entering the chosen accounting boundary, 98 litres return as recovered water and 2 litres are not recovered within that boundary. NASA does not publish one universal breakdown saying that those 2% always correspond to fixed amounts of evaporation, purge, or leakage. Loss paths depend on hardware, operations, and what is included in the accounting boundary. A Mars architecture should therefore instrument losses explicitly: water retained in residues or brines, maintenance and purge flows, humidity that is not captured, water discarded with some waste streams, leakage, EVA losses, and water rejected because it fails quality limits.

The engineering consequence matters: a higher recovery fraction is useful only if quality, reliability, maintainability, and energy demand remain acceptable. NASA’s recovery chain uses multiple filtration and catalytic oxidation stages, checks water quality with sensors, and reprocesses water that does not meet standards. Cabin air follows the same closed-loop logic: the Air Revitalization System removes carbon dioxide and trace contaminants, while the Oxygen Generation System electrolyzes water to make breathing oxygen. Some hydrogen can then be sent to a Sabatier-type carbon-dioxide reduction process to recover water.

On Mars, closing the recycling loop is not enough: the settlement also needs a local make-up source. NASA’s Advanced Mars Water Acquisition System, updated in 2026, studies recovery and purification of water from Martian soils for life support, oxygen and fuel production, food production, and radiation shielding. NASA’s SWIM maps also identify regions where subsurface water ice has been detected, because access to ice can directly influence the choice of a human landing and settlement site.

Engineering check: whenever an efficiency or recovery figure appears, ask: “efficiency of what, measured across which boundary, for how long, and where exactly does the remainder go?”

Primary sources: NASA — 98% water recovery milestone · NASA — ECLSS reference · NASA TechPort — Advanced Mars Water Acquisition System · NASA — SWIM water-ice map.

Go further in the books

Arcadia — technical companionThe technical architecture behind the first Martian city.

I Walked on Mars — four novelsExperience the settlement and growth of Arcadia from inside.

Author’s Visual Vision — four illustrated booksThe illustrated visual companion collection to the novels.

Surface power is the first resource

Without stable electricity, life-support loops do not merely become less comfortable; they progressively stop being loops. Surface power therefore has to be read as a continuity-of-life problem.

Baseline fission. NASA architecture work selects fission as the primary surface-generation technology for initial Mars missions; that provides a continuous backbone, not unlimited power.

Complementary solar. Photovoltaics can still diversify sources, charge storage, serve remote outposts, and reduce dependence on a single generator.

PMAD and electrical buses. Generation and loads are separated by conversion, switching, protection, and distribution; a bus fault can disable otherwise healthy generators.

Black start. After a complete blackout, some equipment must restart without assuming that the network it is supposed to restore is already alive.

Priority loads. Ventilation, thermal control, safety communications, and air processing cannot be shed on the same terms as a workshop or a noncritical greenhouse.

When vital resources tip into crisis

The night the grid collapses

A distribution fault cuts power to two districts. The scenario follows load shedding, battery endurance, ECLSS priority, converter restart, and the decision to keep a zone occupied or evacuate it.

Seven days without the main water processor

Contamination forces the main processor offline. The case tests the real value of buffer storage, the backup train, usage restrictions, and the laboratory’s ability to qualify water before return to service.

Greenhouse lost, population fed

A crop disease removes part of production for two cycles. The town combines reserves, reduced menu variety, short-cycle crops, and energy rationing without confusing an agricultural crisis with immediate loss of life.

Slow leak in an inhabited district

Pressure drifts over several hours. The goal is to detect the leak before a critical threshold, isolate the volume, move residents, and restore gases without exhausting strategic reserves.

Vital resources should be managed as coupled inventories with different recovery times. Water can often be buffered for days or weeks; electrical power may need balancing second by second; food combines daily consumption with harvest cycles measured in weeks or months; oxygen can be produced continuously yet also stored for emergencies. Treating all four resources with the same “days of autonomy” metric hides these dynamics. Operators need an inventory view that shows current stock, production capacity, degraded production, expected demand, and time to the next recoverable state for each resource.

Coupling can improve efficiency but also concentrate risk. Electricity drives water treatment, oxygen generation, pumps, lighting, refrigeration, and food production. A power shortage can therefore appear as several separate resource shortages hours or days later. Conversely, stored water or oxygen can buy time during electrical repairs. The architecture should make these substitutions explicit: which loads can be shed, which production can be deferred, what stock is consumed during the delay, and what minimum restart power is required. This converts “self-sufficiency” from a percentage into a time-dependent operational model.

Agriculture adds another delay that industrial systems often do not have. A failed crop cannot be replaced instantly by increasing power to a machine. Seed stock, growth stage, crop diversity, stored food, nutrient solution, lighting, water quality, and plant disease all determine recovery time. A settlement that obtains a large fraction of calories locally may therefore require more food reserve, not less, during the transition to mature agriculture because it has introduced new biological failure modes that are slow to recover.

Reserve policy should also account for restart losses. A greenhouse, water processor, or chemical reactor may consume extra energy and material during cleaning, warm-up, sterilization, or requalification after shutdown. A buffer sized only for steady-state consumption can therefore expire just as the failed system returns to service. Recovery budgets should include the transient cost of restarting the process, not merely the time spent waiting for repair.