The Human Cost Calculator

Every building emits carbon, and carbon costs lives. If you are planning one, this estimates how many, and shows which of your decisions change that number.

Project
How much you expect the world to do

This setting is an assumption about what everyone else does, not about this building. The same tonne kills more people in a world that fails to act, because the mortality response to temperature steepens sharply above 2 °C. 2.4° is Bressler's optimal pathway, which assumes full global decarbonisation by 2050. 4.1° is the DICE baseline, and the default here. Alt is the same baseline with the damage function specified differently. Upper is the top of the published range. If you believe the 2.4° pathway, the grid factor in the model below should probably fall on the same schedule; the two are set separately so you can see what each one does on its own.
Start from a typical building, then change anything

0.00
Statistical deaths

What it costs

Four steps, from the carbon to the people. Each figure is traceable to a single published paper, listed at the foot of the page.

  1. 01

    This project releases 0 tonnes of CO₂e over its life.

    Per 0
    Upfront A1–A50
    Same as the lifetime emissions of0
  2. 02

    The mortality cost: 0 statistical deaths.

    Bressler (2021) puts one death at 4,434 tonnes of CO₂e on the baseline pathway. That figure counts deaths from extreme temperature alone. Flood, fire, crop failure, disease and conflict all sit outside it.

  3. 03

    The financial damage: $0.

    At the mortality-inclusive social cost of carbon of US$258 a tonne. Bressler's published range runs from a small net benefit to US$545, which would put this project at up to $0. Rennert et al. (2022) reach US$185 independently, with a 5 to 95 per cent range of US$44 to US$413. Kikstra et al. (2021) find that letting economic damage persist instead of dissipating raises the estimate by an order of magnitude, which would put this project nearer $0. Carleton et al. (2022) value the mortality risk of unmitigated warming at roughly 3.2 per cent of global GDP by 2100. Every one of these figures leaves out damage categories: biodiversity loss, conflict, migration, and the tipping points none of the models resolve. The real number is very likely higher than any of them.

  4. 04

    Where does the damage land?

    The poorest half of the world emits 12 per cent of global emissions and carries around 75 per cent of the relative income losses. The richest tenth emits close to half and loses about 3 per cent.

    High income nations

    Absorbed as adaptation

    The share of this damage falling on wealthy countries is largely paid for: seawalls, cooling, reinforced infrastructure, insurance. It registers as expenditure.

    Low income nations

    Absorbed as mortality

    Where there is no capacity to adapt, the same damage lands on health systems, water and food supply. On the Preston curve, falling income tracks falling life expectancy, so economic loss converts into deaths that the temperature figure never counts.

The intersectional toll

Heat is the only cause the headline figure counts. These are two of the things it leaves out.

Food, water and conflict 0 further statistical deaths

Pearce and Parncutt (2023) reviewed 180 studies and found the literature converging on one death per 1,000 tonnes of fossil carbon, which is 3,664 tonnes of CO₂. Set against Bressler's temperature-only figure, the difference is the toll from crop failure, water scarcity, disease and climate-driven conflict.

Order-of-magnitude estimate. The authors put the true value between a tenth and ten times this. The two studies use different methods, so the difference is a comparison of two studies, not a breakdown of one.

Species 0 buildings like this commit one species to extinction

A tonne of CO₂e commits 2.3 × 10⁻⁷ species to eventual extinction. This project accounts for 0 of a species. The factor does not change with the emissions scenario. A tonne emitted today does the same damage as one emitted in 1950.

Global Change Biology 30:e17037 (2024). Committed extinctions, not immediate ones: the species persists for some time and then does not.

The ice it removes

Both forms are drawn to the same scale, by area. The small one is one floor plate of the building. The large one is the summer sea ice its emissions take off the Arctic, permanently. The field is what is lost, not what is gained.

Sea ice removed0m², September minimum
Its own floor plate0m², one floor of it
Ratio0times its own footprint

Notz and Stroeve, Science 354:747 (2016). Three square metres of September sea-ice area per tonne of CO₂, plus or minus 0.3, for emissions from fossil fuels and cement.

When the carbon is spent

One bar per year of the study period, coloured by whichever stage dominates it.

20282088

Change any of it

Every figure above comes from the model below. Open a section to change the assumptions, or switch tab to enter a figure you already have.

Changing the use sets the structure, facade, services, fitout and energy standard ordinarily built for that typology. Each of those can then be changed independently. Nothing is fixed behind the selection.
Gross floor area, measured to the outside face of external walls and including circulation and plant. The benchmark schemes each define area differently, which is one reason the comparison against them is indicative.
Structural quantities per square metre increase with height through longer load paths, larger columns, deeper foundations and, above approximately twenty levels, transfer structures. The model applies a stepped factor.
Design life determines how the upfront carbon is distributed and how many replacement cycles occur within the study period. Sixty years is the common assumption. Actual service lives are frequently shorter, particularly in commercial stock, and the upfront carbon is spent regardless of how long the building stands.
Sets where the study period begins on the grid decarbonisation curve. A building finished in 2040 spends its early operational years on a cleaner grid than one finished in 2028, so completing later reduces operational carbon while leaving upfront carbon unchanged.
Not a reduction against business as usual. The Australian Reduction Roadmap works down from the planetary boundary for climate change: a global safe operating space of 2.51 Gt CO₂e a year, Australia's 0.33 per cent Paris share of it, three per cent of that for new housing, divided by the floor area the country builds. The answer is 6.63 kg CO₂e per square metre per year by 2028, against a current figure of 461.8, a reduction of 98.57 per cent. Every other scheme in the field beside this one sets targets relative to a conventional reference building, so a project can sit comfortably inside one and remain two orders of magnitude outside the budget. The verdict at the foot of the results is made against this number. The Roadmap states 6.63 as an average and specifically for housing, and notes that different building types need differentiated limits, since a detached house has more reduction potential than an office.
Context, not the verdict. Four schemes that set numerical limits. GBCA is the Australian roadmap: the Green Building Council puts 500 kg per square metre at low embodied carbon for a commercial building and 1,000 at high, and by 2030 every Green Star building must show a 40 per cent cut against a conventional reference, which against a concrete framed Australian office of 520 kg per square metre puts the target near 310. Those bands are published for commercial buildings, so housing and schools are read against the same numbers here for want of an Australian equivalent. RIBA 2030 is a voluntary UK target covering upfront carbon. RE2020 is French regulation capping embodied carbon over fifty years. Denmark's BR18 regulates whole life carbon per square metre per year. Each is compared only against the modules it covers, with the excluded tenancy fitout reported separately beneath. Floor area definitions differ between the three, so read them as indicative, not as a compliance check. This model has a broader scope than any of them.
Structure is usually the largest single share of upfront carbon and the most difficult element to revisit once the frame is procured. The range between mass timber and heavy concrete is approximately threefold.
Applies only where the structure contains concrete. The majority of concrete's carbon is in the Portland clinker. Supplementary cementitious materials such as slag and fly ash displace a proportion of it without altering member sizes. The principal trade off is slower early strength gain. Reinforcement does not vary in this model.
0%
Retention avoids both the manufacture of a new frame and the removal of the existing one. Retained structure is charged at eight per cent of new construction to allow for investigation, strengthening and repair.
A unitised curtain wall carries approximately three times the carbon of masonry or lightweight construction. Aluminium and its supporting steelwork account for the majority.
Plant, ductwork, cabling, risers and lifts. Services are typically renewed two or three times within a sixty year life, so the initial installation is usually the smaller share of the total.
Partitions, ceilings, joinery, floor finishes and loose fit. The shortest lived element of the building.
1.00
Manufacturing carbon varies between countries, largely with the energy used to make the product. The default scales the energy sensitive share of A1 to A3 by this country's grid intensity relative to the world average of 0.473 kg per kWh. The remainder does not move, since cement calcination is a chemical reaction and is the same everywhere. It cannot account for the origin of individual products. If you have Environmental Product Declarations, use those figures and set this to 1.00.
Mass is generally more significant than distance. Heavy local materials usually outperform lighter imported ones. The exception is specialist facade and finishes shipped internationally.
Material ordered and not installed carries its full upfront carbon. Coordinating to standard sheet and member sizes is among the lowest cost reductions available. Site energy adds a further 15 kg per square metre.
Commercial leases typically run five to ten years. Make good provisions require the tenancy to be returned to base condition at the end of each term, and depreciation schedules favour replacement over retention. Across a sixty year life this can amount to nine complete interiors, which may exceed the carbon of the structure containing them. The lease term sets these cycles, not the condition of the materials, and none of it appears in the design documents.
Sealed glazing units and their gaskets are commonly renewed at around forty years. Where the design life exceeds the facade life, the difference is a second facade.
Plant and equipment are typically renewed at twenty to thirty years. Risers, primary ductwork and containment usually survive, so replacement is charged here at eighty five per cent of the original installation.
Refrigerants are small in mass and significant in effect. R-410A has a global warming potential approximately two thousand times that of carbon dioxide, and all systems leak over their service life. R-32, propane and carbon dioxide plant reduces this contribution substantially with no change to the plan.
A shortcut that writes a figure into the field beside it. Editing that figure directly switches this to custom.
Delivered energy at the meter, before on site generation. Modelled figures commonly fall below measured performance. Metered data from a comparable building is preferable to a preset.
Charged at 0.20 kg per kWh, covering combustion and upstream fugitive emissions. Unlike electricity this figure does not fall over the study period, so gas becomes a larger share of operational carbon each year the grid cleans up.
Set by the country selected above. These figures are indicative, decline annually, and are calculated differently between jurisdictions. Confirm the current national factor before publication.
Operational carbon declines as the grid decarbonises. Upfront carbon is spent at practical completion and does not. Adjusting this date shows how the balance between the two shifts.
Photovoltaic arrays carry their own embodied carbon. The model assumes 1,250 kg per installed kilowatt, one replacement at year thirty, an annual yield of 1,350 kWh per kWp and degradation of half a per cent per year.
Demolition, transport and disposal. A small share of the total, though it determines what is available to the next project.
Module D falls outside the assessment boundary and is reported separately rather than deducted. Danish regulation takes the same position. The higher rates need mechanical connections, bolted or screwed, not welded, glued or cast in, together with a material passport and accessible fixings. Steel frames and demountable mass timber can recover a substantial share on this basis. Recovery claims made at design stage remain projections about demolition practice several decades ahead.

There is no published record of the assumptions used in practice for design life, replacement cycles or retention rates. Contributed models build that record and allow the tool to show how a project compares against others.

sendscountry, building use, floor area rounded to two significant figures, storeys, design life, structure, cement, retention, facade, services, fitout, replacement cycles, refrigerant, energy use, grid factor, PV, and the calculated result
never sendsthe project name, your location, your address, any file, or anything else you have typed

For use when you already have a figure from a life cycle assessment. Confirm which modules it covers before entering it. An upfront A1 to A5 figure and a whole life figure differ by a factor of two or more, and reports do not always state which is quoted.

Method, and what this number is not

The mortality figure

R. Daniel Bressler, The mortality cost of carbon, Nature Communications 12:4467 (2021). The DICE integrated assessment model is extended with a climate to mortality damage function, finding that a tonne of carbon dioxide emitted in 2020 causes 2.26 × 10⁻⁴ excess deaths between 2020 and 2100 on the baseline pathway. That is one death per 4,434 tonnes.

The four pathway settings are figures published in that paper: the central baseline estimate, two alternative specifications of the damage function, and the upper bound of its reported range.

Why the worse cases are worse

The mortality response to temperature is convex. Below 2 °C the paper projects roughly 100,000 excess deaths a year. Above 2 °C the curve steepens sharply, reaching over four million a year at 4 °C. A tonne emitted now is therefore far more lethal in a world that fails to act than in one that does.

On the baseline pathway the paper projects 83 million cumulative excess deaths by 2100. Following the optimal path avoids 74 million of them.

What the number is not

Temperature related mortality only. Flood, fire, storm, crop failure, water scarcity, disease vectors, conflict and displacement all fall outside it. The figure should be read as a lower bound.

The bottom of the published range is negative, indicating a small net benefit. The top is three times the central figure. The four settings span the central estimate, two alternative specifications, and the upper bound.

The deaths are statistical and global, and no individual can be identified. They fall disproportionately on hot, low income regions that contributed least to the emissions.

Does the building data travel?

In part. The mortality conversion is global. The lifecycle module structure is EN 15978 and international. Grid factors are national and vary by a factor of twenty five, so they matter most and are handled directly.

Material intensities are the principal source of uncertainty. Cement calcination is a chemical reaction and is approximately constant everywhere, but steel varies substantially with the production route and aluminium with the electricity used in smelting. The materials factor is an approximation of this. Absolute figures should be treated as indicative; comparisons between two options within the same setting are considerably more reliable.

Floor area definitions differ between the benchmark schemes. RE2020 uses net internal area, BR18 uses gross heated area, RIBA uses gross internal area.

The cascade

Direct mortality: Bressler, The mortality cost of carbon, Nature Communications 12:4467 (2021). One death per 4,434 t CO₂e, temperature only.

Systemic mortality: Pearce and Parncutt, Quantifying Global Greenhouse Gas Emissions in Human Deaths to Guide Energy Policy, Energies 16:6074 (2023). A review of 180 studies converging on one death per 1,000 t of fossil carbon, which is 3,664 t CO₂. The authors state this is an order-of-magnitude estimate and the true value could be a tenth or ten times it. The difference between that figure and Bressler's compares two studies with different methods. It is not a decomposition of either.

Social cost of carbon: Bressler gives US$258 per tonne once mortality is included, with a published range from a small net benefit to US$545. Rennert et al., Nature 610:687 (2022), reach US$185 independently. Carleton et al., Quarterly Journal of Economics 137:2037 (2022), value the mortality risk of unmitigated warming at roughly 3.2 per cent of global GDP by 2100, with an interquartile range of −5.4 to 9.1 per cent.

Sea ice: Notz and Stroeve, Science 354:747 (2016). Three square metres of September sea-ice area per tonne, plus or minus 0.3, for CO₂ from fossil fuels and cement.

Species: Urban, Accelerating extinction risk from climate change, Science 348:571 (2015). A synthesis of 131 studies giving 2.8 per cent of species at present warming, 5.2 at 2 °C, 8.5 at 3 °C and 16 at 4.3 °C. The four points are plotted as published, with nothing interpolated between them. Urban's 2024 update in Science, across 485 studies, finds extinctions accelerate above 1.5 °C and puts the highest emissions scenario near a third of species.

Human climate niche: Lenton et al., Quantifying the human cost of global warming, Nature Sustainability 7:1237 (2023). Roughly 140 million more people exposed to dangerous heat per 0.1 °C. The per tonne figure used here converts that with the IPCC transient climate response to cumulative emissions of about 0.45 °C per trillion tonnes, which is our chain rather than theirs. Tol, Climatic Change 177 (2024), disputes the underlying method and argues the affected population is in the hundreds of millions.

Inequality: Chancel, Nature Sustainability 5:931 (2022) and the Climate Inequality Report 2023. The bottom half of the world emits 12 per cent of the total and carries about 75 per cent of relative income losses; the top tenth emits 48 per cent and loses about 3 per cent. The income to life expectancy relationship is the Preston curve, Preston (1975).

The limit

Australian Reduction Roadmap, version 1.4, January 2026. University of Melbourne (Robert Crawford, James Helal, Dan Hill, André Stephan), TERROIR, UTS, QUT and Beyond EPiC, after the Danish original by EFFEKT, Artelia and CEBRA. Licensed CC BY-NC-SA 4.0.

It works down rather than across: the planetary boundary for climate change gives a global safe operating space of 2.51 Gt CO₂e a year (Petersen et al., 2022) against 53.9 Gt actually emitted. Australia's Paris share of 0.33 per cent gives a national budget of 8.23 Mt a year against 574 Mt emitted, seventy times over. New housing takes three per cent of that. Divided by the floor area Australia builds, the limit is 6.63 kg CO₂e per square metre per year by 2028, against a current 461.8, a reduction of 98.57 per cent.

Its metric counts initial embodied carbon in full in the year it is spent and annualises only the recurrent and operational parts. That is a deliberate departure from the Danish approach, on the grounds that construction emissions reach the atmosphere during construction and cannot be spread across fifty years. It is why the numbers here are so much larger than a whole-life-divided-by-fifty figure.

The Roadmap gives 6.63 as an average and specifically for housing, and states that building types need differentiated limits since a house has more reduction potential than an office. Read it as the order of the gap rather than a compliance threshold for every use.

As a rough check on scale, a conventional Australian detached house comes out of this model at around 466 kg CO₂e per square metre per year, alongside the Roadmap's 461.8. The two are not measuring quite the same thing: theirs is an average new dwelling of 186 m², blended from detached houses at 232 m² and apartments at 137 m², while this is a single detached house of whatever area you enter. Different data, different building mix. Read the closeness as reassuring about the order of magnitude and nothing more. Both draw on the EPiC Database at the University of Melbourne, so they are not fully independent either.

Sources

Grid factors: Ember Global Electricity Review for most countries, DEFRA for the United Kingdom, EPA eGRID for the United States, and the National Greenhouse Accounts for the Australian states. World average 0.473 kg per kWh.

Benchmarks: GBCA Climate Positive Roadmap and Green Star upfront carbon credits, with the 500 and 1,000 kg per square metre bands from the Green Building Council of Australia (2023) and the conventional Australian office figure of 520 kg per square metre from Towards net zero embodied carbon, Sustainable Cities and Society (2024). Also RIBA 2030 Climate Challenge, France RE2020 Ic construction caps, and Denmark BR18 whole life limits.

Infrastructure Australia (2024) puts upfront embodied carbon at 7 per cent of Australia's national emissions, and finds a 23 per cent cut achievable by 2026 with strategies already available.

Building intensities: indicative values calibrated for early stage comparison between options. They do not substitute for a life cycle assessment to EN 15978 using measured quantities.

Coefficients

Statistical deaths0.00
Whole life0 t