Methodology

Introduction

carboncount turns an activity into an estimate of the CO2 equivalent it produced. You pick the activity, say how much of it there was, and it lands in a table alongside the emission factor used and the total. This page is the whole of the working: what the words mean, which number came from where, what was decided along the way, and where it is weak.

Everything here is an estimate. Emission factors are averages over a wide spread, so the total tells you the order of magnitude and the shape of your footprint, not a reading off a meter. Two people doing exactly the same things will genuinely have different footprints, and this cannot see the difference.

There is no server. The factors are compiled into a file at build time and the arithmetic happens in your browser, so nothing you type is sent anywhere or stored.

Definitions

Every term the rest of this page leans on. What the app does about each one is under Method.

CO2 equivalent (CO2e)

The unit every total on this site is in. Greenhouse gases other than carbon dioxide are converted into the amount of CO2 that would warm the climate as much over a hundred years, then added in. Methane and nitrous oxide dominate several food and fuel rows, so a CO2-only figure would understate them badly.

Emission factor

Kilograms of CO2e per unit of something: per km flown, per kg of beef, per kWh of electricity. A row is its factor times the amount you entered, and the factor is shown next to it so the arithmetic is visible rather than implied.

Activity based

An estimate built from a physical quantity multiplied by the emissions per unit of it. 420 km flown, 200 g of beef, 300 kWh of electricity. These are the accurate ones.

Spend based

An estimate built from what you paid, using the average emissions embodied in a euro spent in that part of the economy. Marked “by spend” in the table. These exist because for clothes, electronics and services there is no physical quantity anyone actually knows. They are much rougher: a cheap coat and an expensive coat of identical make will give different answers, and the answer will be wrong for both. You are never asked to choose between the two kinds of estimate. Each activity accepts only the units that make sense for it, so the method follows from what you picked.

Well-to-tank

The emissions of producing a fuel and getting it to where it is burnt, as distinct from burning it. DEFRA reports the two separately. Counting only the second is what a tailpipe figure does.

Radiative forcing

Aviation warms the climate by more than its CO2 alone, through contrails and nitrogen oxides emitted at altitude. DEFRA publishes a flight factor with and without that effect.

Great circle distance

The shortest path over the surface of the earth between two airports, which is what the airport picker measures. A real flight is longer: planes join airways, hold, and get vectored around traffic.

Scope 1, 2 and 3

The GHG Protocol's three buckets, which company carbon reports are organised into. Scope 1 is what you burn yourself, scope 2 is the electricity you buy, scope 3 is everything else in the chain on either side of you. They exist so that two companies accounting for the same supply chain do not both claim the same tonne. DEFRA labels every factor it publishes with the scope it belongs to. This app has no scopes, because a person has no reporting boundary and nobody is consolidating your total with anyone else's. What that changes is under Method.

Generation intensity, and full scope

Two different things an electricity factor can mean. Generation intensity is measured at the power station. A full-scope factor also counts what is lost getting the electricity down the wires and the well-to-tank of the fuel that made it. For the UK that is 0.13096 kg per kWh against 0.18436. Every other factor in this app is on the wider of the two. Unrelated to scopes 1, 2 and 3, which are about whose emission it is rather than how much of the chain is being measured.

Per vehicle and per passenger

Whether a transport factor prices the whole vehicle or one seat in it. DEFRA prices a car by the car; bus, train, taxi and ferry factors are already per passenger.

Where each source comes from

Method

• The arithmetic is the GHG Protocol's, the boundaries are not. Activity data times a published emission factor, in CO2e on a hundred-year basis, physical measurements preferred over spend wherever one exists, and every source, assumption and departure written down: that is the GHG Protocol's calculation approach, and DEFRA publishes these factors for company reporting under that standard. What this app does not do is keep the scopes apart. A company reports the electricity it buys as scope 2 and the losses in the wires as scope 3; here they are one number, because a person has no reporting boundary to divide along. The result is a consumption footprint rather than an inventory. It is complete, but no row in it can be lifted straight into a corporate report.
• Well-to-tank is included. Both of DEFRA's rows are counted, which makes most numbers roughly a fifth higher than a calculator that counts only the tailpipe. For a company those two rows fall in different scopes, which is why a corporate figure for the same litre of fuel is the smaller of the two.
• Flights include radiative forcing. DEFRA publishes both figures and this uses the higher one. In company reporting the uplift is optional and commonly left off, so a flight costs more here than the same flight in an audited inventory. The question this app is answering is what the flight did to the climate, not what has to be declared.
• Flight distance is not the map distance. DEFRA handles indirect routing with a flat uplift, 95 km, which is added to every route on top of the great circle distance.
• Flights use a curve, not DEFRA's bands. DEFRA bands flights relative to the UK, which is no use between two other countries. This app used to remap those bands onto distance, with the domestic factor under 785 km and the international one above. The two differ by a factor of 2.09, so a flight of 786 km cost less than half one of 784 km. Now there is one continuous curve: 74.4657 kg per passenger spread over the route, plus 0.100898 kg for every km of it. Short flights still cost more per km, because the climb is paid once however far you go, but nothing jumps.
• The curve is a fit, not a physical model. It passes through DEFRA's domestic factor and its non-UK international factor at assumed stage lengths of 460 and 3000 km, which DEFRA does not publish. Its fixed term works out at 74.4657 kg per passenger, far more than a climb actually costs, because it is absorbing every reason short flights are worse: smaller aircraft and emptier ones as well as the climb. Moving the assumed distances across their plausible range moves a mid-haul answer by about 5%. The fit is mine, not DEFRA's.
• Below 460 km the curve stops. It is held flat at DEFRA's domestic figure rather than continued, because continuing it produces numbers higher than anything DEFRA publishes and there is no evidence down there to check them against.
• Cabin premiums grow with distance. DEFRA prices short-haul business at 1.5 times economy and long-haul business at 2.9, because a short-haul business seat is an economy seat with the middle blocked and a long-haul one is a bed. Those are interpolated rather than switched. Premium economy and first have no short-haul DEFRA row at all, since airlines do not sell them on short flights, so below the long-haul distance they are priced as the economy and business seats they actually are, and the row says so.
• Country grids are scaled to match. Ember publishes electricity intensity as generation at the power station, so a raw Ember number would measure something narrower than the full-scope number it replaces. Each country is multiplied by 1.40776, which is the ratio between those two scopes in DEFRA's UK rows. Real transmission losses vary by country, so this is approximate, but it approximates the right quantity. The multiplication is mine, not Ember's. The UK is not taken from Ember at all: DEFRA publishes it directly at 0.18436 kg per kWh.
• Electric cars are not zero. DEFRA scores them zero at the exhaust and puts the real emissions under electricity. This app adds those back, so an electric car shows what charging it cost.
• Water counts both ends of the pipe. DEFRA's two water figures are easy to misread. “Water supply” is not raw water: it is abstraction, treating it to drinking standard, and pumping it to your tap. “Water treatment” is the sewage works at the other end. This app used to count only the first, which was not “untreated water” but “no sewerage”, about 53% of what a metered bill covers. Both are counted now. Water that never reaches a drain, in the garden or a leak, is about half what this shows.
• A hotel with no country is a placeholder. DEFRA publishes 39 countries and a further 16 as zero, meaning no data. Picking nothing, or picking one of those, gives the median of the published ones, 32.1 kg a night. It used to give the UK figure, which is seventh lowest of the thirty-nine and flattered everyone who did not pick.
• Cars are per vehicle, public transport is per passenger. The form asks how many of you were in the car and divides. Bus, train, taxi and ferry factors are never divided again.
• A route is one way unless you say otherwise. The airport picker has a return box. When it is ticked the distance doubles but the factor does not move: two flights of 600 km are two short hops, not one long haul, and the curve is asked about one leg.
• Currency is converted at a fixed rate, set 1 July 2026. A live rate would mean the same receipt gave a different answer tomorrow.
• The tree line is not offsetting. One tree absorbs about 60 kg of CO2 a year, so the comparison says how long trees would need to take your total back out of the air. It does not say the total has been cancelled, because it has not. Planting the tree is a separate act with its own long and uncertain timeline.
• A row that cannot be resolved is shown, not dropped. If the data no longer has a factor for something you entered, the row says so and stays out of the total. A total that quietly went down would be worse than a row you can see.

Attribution and licensing

• UK Government GHG conversion factors 2026 (Department for Energy Security and Net Zero (DESNZ) and Defra, Open Government Licence v3.0)
• Reducing food's environmental impacts through producers and consumers (Poore & Nemecek 2018, Science, via Our World in Data, CC BY 4.0)
• Carbon intensity of electricity generation (Ember, via Our World in Data, CC BY 4.0)
• OurAirports (OurAirports, Public domain)
• Greenhouse Gas Equivalencies Calculator (US Environmental Protection Agency, US Government work, free to use)
• EXIOBASE v3 spend-based factors (EXIOBASE consortium, CC BY-SA-NC (non-commercial use only))

Contains public sector information licensed under the Open Government Licence v3.0.

The code is MIT. The compiled factor data is not: it carries the licences of the sources above. EXIOBASE v3.9 is CC BY-SA-NC, so carboncount stays non-commercial for as long as the spend-based factors are in use.

Fonts are served from this site rather than a third-party CDN, so loading the page tells nobody else that you did.

Limitations

• The spend factors are the weakest data in the app. They are hand-entered sector averages at the right order of magnitude rather than an automated extraction from EXIOBASE, and replacing them is the top open task.
• This is not a GHG Protocol inventory and does not claim to be. There is no organisational boundary, no split into scopes 1, 2 and 3, no market-based electricity figure for anyone on a renewable tariff, no base year and nothing verified by a third party. Those are what make one company's report comparable with another's, and none of them are available to one person counting a flight.
• The hundred-year warming potentials underneath these factors are not all of one vintage. DEFRA has moved most of its rows onto the IPCC's fifth assessment values, but bioenergy is still on the fourth, some refrigerants are on the sixth, and the hotel figures arrive as CO2e with no gas breakdown at all. Mixing vintages is normal practice and the effect is small next to everything else on this list, but the total is not built on one consistent set of them.
• DEFRA is a UK dataset. Transport factors reflect the UK vehicle fleet and UK load factors, which are not the world's. Electricity and hotel rows can be set to another country; the rest cannot yet.
• The scaling that puts country grids on the same scope as everything else uses the UK's ratio of transmission losses and fuel supply to generation. Those losses genuinely differ by country, from around seven percent to around nineteen, so a country's number carries that error on top of Ember's own.
• DEFRA publishes a hotel figure for only 39 countries. Everywhere else, and every count where no country is picked, gets the median of those: a placeholder standing in for a range that runs from 4.7 kg a night to 152. Picking your country matters more here than anywhere else in the app.
• The flight curve rests on two distances DEFRA does not publish, and below 460 km it is not a curve at all, just DEFRA's domestic figure held flat. Short flights are the part of this app with the least evidence under them.
• Food factors are global averages farm to retail. They do not include cooking, transport to your kitchen, or the emissions of what you threw away.
• The tree comparison is in CO2 against totals in CO2e, and it is the one number in the app typed in by hand rather than fetched from its publisher, so unlike every other figure here it cannot fail the build if the publisher moves it (entered 30 July 2026).
• Nothing you enter is saved. Closing the tab clears the count, which is what the printable receipt is for.