Methodology
How every number on Newleaf is made: where the data comes from, how we clean it, how we project it to 2100, and what it cannot tell you.
At a glance
Newleaf brings together public data on people, food, land, wealth, energy, emissions, climate, health in one table: every indicator for every country, year by year, with projections up to 2100. Most of the data is collected through Our World in Data, which gathers and harmonises the statistics of the UN, the World Bank, the FAO and many other institutions.
Data collection
Every source is fetched from Our World in Data's chart and explorer pages, and every column becomes one indicator, with its unit. Three inputs come from elsewhere:
- Monthly surface temperature by country (Copernicus ERA5, also through Our World in Data), from which we derive the yearly average, the coldest and hottest months and their spread.
- Country boundaries from Natural Earth, which also give each country its continent and sub-region.
- Published factors from the scientific literature: lifecycle emissions of each source of electricity, the land, water and emissions footprint of foods, the carbon footprint of income groups, and the physical constants of the climate model.
Cleaning and harmonising
- The world is the sum of its countries. World values are recomputed from the countries in the database (sums for totals, population-weighted averages for rates), so a world line always matches the countries under it.
- Gaps. Missing years between two observations are interpolated linearly. A series that stops early is extended with its last yearly change, damped.
- Per person first. Totals are turned into values per person before they are projected, and back into totals afterwards: a country's emissions grow because each person emits more and because there are more people, and the two are kept apart.
- Regions. Countries are grouped in 10 regions: Europe, East Asia & Oceania, South & Central Africa, North America, Latin America, Middle East, North & West Africa, China, North East Eurasia, Indian Subcontinent.
- Income groups follow GDP per capita, with thresholds at 1,500, 5,000, 15,000 and 45,000 dollars a year.
Projections to 2100
Projections go year by year from 2026 to 2100. For some indicators the sources do not cover 2023–2025 yet: those years are estimated with the same method, so that every indicator reaches today. The projections are built in two passes, and the main indicators are linked to each other up to global warming:
1. Each indicator on its own
About a hundred indicators are projected country by country with the same rule: GDP per capita, energy and electricity by source, emissions by sector, land by type, food, and the income of the richest and the poorest. Each year's change is the sum of two terms:
- The country's own momentum: its average yearly growth over the previous five years, damped, with implausible jumps discarded and a maximum yearly rate per indicator.
- A common trend: a small yearly push, the same for every country, that fades linearly to zero by 2100, such as the expansion of solar and wind or the decline of coal.
That change is then multiplied by a saturation factor that slows growth as the indicator approaches its ceiling.
2. Closing the accounts
The second pass makes the indicators agree with each other, year by year:
- Population starts from the UN medium-variant birth, death and migration rates: each year's population is last year's plus births, minus deaths, plus net migration, with migration rebalanced so that it sums to zero worldwide. Age groups, fertility, life expectancy and median age follow the UN projections, adjusted when the population differs from the UN's.
- Totals are rebuilt from values per person times population, and summed into the world.
- Land cannot exceed the country: if urban, grazing, crop and forest land add up to more than 95% of its area, they are scaled down, and what is left is “other”.
- Electricity emissions are the generation of each source times its lifecycle emission factor. Emissions from land use follow the change in forest area: about 50,000 tonnes of CO₂ for every km² of forest lost, and as much absorbed for every km² regained.
- Incomes: the income of the poorest 90% and of the richest 1% are projected; the richest 10% get what is left of GDP. The wealth shares of the richest move with their income.
3. Three kinds of indicators
Not every indicator can be projected in the same way. Some have history only, because their past says little about their future, such as foreign investment or water quality surveys. Most are projected with the rule above: their own momentum, a common trend and a ceiling. Others follow or are bounded by reference projections: population, ages and life expectancy follow the UN, the emissions of each energy source use the IPCC's factors, and the trends of the energy sources are checked against DNV's Energy Transition Outlook.
What the projections leave out
The projections are not forecasts: they show where today's trends lead if they carry on. They do not include wars, pandemics, new policies or feedback loops nobody can foresee yet, such as the sudden thaw of permafrost. Each of these could move the results a lot, in either direction.
How much the method matters
A small test shows how sensitive a projection to 2100 is. Take the world's energy alone, 157,400 TWh in 2024, with no sources and no countries, and let its growth of the last five years, 1.20% a year, fade to zero by 2100:
| Starting growth | Energy in 2100 | Difference |
|---|---|---|
| Last 5 years, 1.20% a year | 246,000 TWh | — |
| 0.1 points faster, 1.30% | 256,000 TWh | +3.8% |
| 0.1 points slower, 1.10% | 237,000 TWh | −3.7% |
| Last 10 years, 1.26% a year | 252,000 TWh | +2.3% |
A tenth of a point a year is worth about 4% by 2100. The full model ends far lower, at about 135,000 TWh, because it adds up each source in each country, with efficiency, ceilings and the switch to electricity: the structure of the model matters more than the exact starting trend.
Compared with reference projections
| Newleaf | Reference | |
|---|---|---|
| Population in 2100 | 10.0 billion | 10.2 billion (UN, medium variant) |
| Global warming by 2100 | 2.6 °C | 2.7 °C (IPCC, intermediate scenario SSP2-4.5) |
| CO₂ in 2050 | 21 Gt | 22 Gt (DNV) |
| Peak of world energy | 2033, then −10% by 2060 | around 2040, then −4% by 2060 (DNV) |
Logistic curves
Nothing grows forever at the same rate, so every projected change is multiplied by how far the indicator still is from its ceiling K:
xt+1 = xt · (1 + gt · (1 − xt / K))
where g is the yearly growth described above. Far from the ceiling growth goes on as before; near it the curve flattens. This is the discrete form of the logistic curve. The ceiling depends on the indicator and on the country.
The climate model
Warming is computed from emissions with a chain of standard, simplified relationships, the same in the pipeline and in the Scenarios tool:
- Emissions. Fossil CO₂ is population × GDP per person × energy per dollar of GDP × CO₂ per unit of energy, plus CO₂ from forest loss.
- Concentration. CO₂ in the air is the pre-industrial 280 ppm plus a fixed fraction of cumulative emissions. Methane and nitrous oxide accumulate and decay, with lifetimes of about 12 and 109 years.
- Radiative forcing. For CO₂, 5.35 W/m² × ln(C / 280); for methane and nitrous oxide, proportional to the square root of the concentration above pre-industrial.
- Warming. The forcing of each gas times its sensitivity, summed over the three gases, calibrated on the observed warming of recent years. On top of it, a progressive feedback strengthens warming as the century goes on, in line with the latest UN projections: no correction this year, growing in a straight line to +15% in 2100. On Climate scenarios a lever moves it between 0 and 30%.
- Countries. Each country's contribution to warming is its share of cumulative CO₂; its local temperatures move by the same proportion as global warming.
Scenarios
The current trends scenario (the baseline) is the projection above: it is the scenario shown everywhere on the site.
The Scenarios tool goes further. Five indicators drive the climate chain: population, GDP per capita, energy per GDP, CO₂ per energy and forest share. Each of them has two levers:
- World trend shifts the yearly growth rate of every country by the same amount, compounded from today.
- Low/high income countries adds growth to the poorest countries and takes it from the richest, in proportion to their rank by income.
The map in the tool starts from each country's temperature anomaly in 2100 at current trends, and multiplies it by the ratio between global warming in your scenario and at current trends. All countries move by the same proportion: the levers act on the world's emissions, not on one country's climate.
The range of each lever is a choice of yearly rates around today's trends, wide enough to reach the pathways that the main studies consider plausible. Taking each of the five indicators at its current trend, its lowest and its highest, together with the climate feedback at 0, 15 and 30%, gives 3⁶ = 729 combinations: the grey bundle around the climate line is those 700+ runs of the climate chain, and the range at the top of the page holds 90% of them. They show how sensitive the result is to each assumption, not how likely it is.
Target scenarios
Five charts under the dashboards ask the opposite question: how much more than current trends the world would have to do, year by year from today to 2100, to reach a target.
- Mitigating global warming (Climate). The budget of a warming threshold (1.6 °C, 1.8 °C, 2.0 °C, 2.2 °C) is the cumulative CO₂ of current trends in the year its global warming reaches that threshold. Emissions are cut from current trends by r(x) = 200% · ln(1 + kx) / ln(1 + k): 0% today and 200% in 2100, so that emissions end as negative as current trends are positive. The shape k is solved for each threshold so that the CO₂ saved to 2100 is exactly what the budget requires: with k > 0 the cuts come early, with −1 < k < 0 late. The white dots mark the first year of net zero emissions.
- The next three charts share their targets: a cut of the world's yearly CO₂ emissions compared with current trends of 1, 2, 5, 10 billion tonnes, the same every year from today to 2100. The lines change over time only with what changes in current trends: the number of people, and the fossil fuels left to replace.
- Planting more trees (Land). The trees planted every year so that, from the time they grow up, they absorb 1, 2, 5, 10 billion tonnes of CO₂ a year. A tree counts after 10 years, and keeps absorbing in the following years; every year 0.25% of the trees are lost to deforestation, fires and disease (gross forest loss, about ten million hectares a year out of four billion in the FAO Forest Resources Assessment). A grown tree absorbs a different amount in each world region, from 12 kg of CO₂ a year in boreal forests to 28 kg in tropical ones, and the world figure is the average weighted by today's forest area. Trees absorb less as the CO₂ in the air grows, in inverse proportion to the concentration of current trends. So the first year plants all the trees the target needs, and the following years only those that replace the losses and make up for the lower absorption.
- Reducing meat consumption (Food). The grams of meat each person would eat less every week to cut 1, 2, 5, 10 billion tonnes of CO₂ a year: the cut divided by the CO₂ of a kilogram of meat and by the world population, never more than the meat of current trends. A kilogram of meat is the average of beef, pork, poultry and lamb, weighted by what the world eats today, with the greenhouse gases per kilogram of Poore & Nemecek (2018).
- Producing cleaner energy (Energy). Primary energy stays as in current trends, and renewables and nuclear replace fossil fuels, one terawatt-hour for one, until the CO₂ of the fossil fuels left out reaches 1, 2, 5, 10 billion tonnes a year. The CO₂ of a terawatt-hour of coal, oil and gas is the IPCC default emission factor, and the replaced fossil fuels have the mix of current trends in that year. Clean energy never goes above the total energy of current trends.
- Reducing wealth inequality (Wealth). For every year, one tax rate on the income of the world's richest 0.01%, with all the revenue given to the other 99%. As in Wealth distribution, the world is the stack of the income curves of its countries, so the rate that raises the GDP per capita of the 99% by 1, 2, 4, 8% is the increase divided by the ratio between the income of the richest 0.01% and that of the 99%. Where even a 100% rate is not enough, the line stops.
Models on the website
- Income distributions. For each country, income by percentile is rebuilt from GDP per capita, the income shares of the poorest 90%, the richest 10% and the richest 1%, and the share of people in extreme poverty: a log-normal curve up to the 90th percentile and a Pareto tail above it. Its average equals GDP per capita by construction. The world curve is built the same way from world values. The chart of income distribution over time compares decades with one recipe: without the poverty share, with each country's earliest income shares for the years before they start, and with the world as the sum of its countries.
- What if everyone was like each region. Each of these charts keeps the world as it is today and changes one thing. Emissions and warming use each region's amount per person, multiplied by the world's population. Electricity keeps the world's total and uses each region's mix of sources. Food keeps the world's kilograms and uses the composition of each region's diet. Land applies each region's yearly forest trend over the last 10 years to the world's forests, counting about 33,000 tonnes of CO₂ per square kilometre: the carbon of the FAO's Global Forest Resources Assessment 2020, without the soil.
- Age distributions. For each country, the number of people at every age from 0 to 100 is rebuilt from the births of each year since 1950: every generation is followed year by year with a mortality table matched to the country's life expectancy in that year, and the yearly net migrants are added by age, most of them young adults. Before 1950, births and life expectancy are extended backwards from the 1950s. The result is then scaled to match the UN population in five age groups (0–4, 5–14, 15–24, 25–64, 65+). Checked against the UN estimates by five-year age group, on average 1.5% of a country's population falls in the wrong group.
- Poverty lines in the SDG tracker. The share below $10 a day is the published one. The share below $5 a day is read between the published shares at $3 and $10, assuming incomes follow a lognormal curve between the two. The share below $20 a day comes from our model of each country's income distribution.
- Population growth. The tool follows the people of every age from one year to the next, without splitting women and men: each year they get one year older, some die, some arrive or leave, and the newborns join at age zero. It starts from this year's age distribution and, with the levers at zero, returns the UN projections exactly. Each lever is a percentage on those projections: zero this year, growing in a straight line up to its full value in 2100. Fertility scales the births, which also follow how many people are of childbearing age. Death rate raises or lowers the risk of dying at every age until the deaths per 1,000 people change by the chosen percentage, and life expectancy follows, within 40 to 100 years. Net migration scales the yearly flow, with arrivals and departures spread by age, most of them young adults. Fertility stays between 0.5 and 8 children per woman.
- Food. Production of fish and butter is derived from the FAO food balances (domestic supply minus imports plus exports); offal adds up the offal of seven animals. Animal protein per person comes from the FAO through Our World in Data and is projected like food consumption. Greenhouse gases, land and scarcity-weighted water per 1000 kcal come from Poore & Nemecek (2018); foods that the study does not cover take the value of the closest product. Macronutrients are typical values per kilogram of raw food from USDA FoodData Central. CO₂, land and water per kilogram of sweet potatoes, yams, coconuts and offal take those of the closest products of the same study, scaled to their calories; butter takes those of cheese, soybeans those of pulses, oats and rye their own, and fish the land and water of farmed fish. For foods without a land figure from the FAO, land is production times land per kilogram. Butter and offal come from milk and animals already counted, so they are left out of the sums of production. Food categories compared averages the foods of each category weighted by world consumption, with the energy from the macronutrients (4 kcal per gram of protein and carbohydrates, 9 of fat, 2 of fibre).
- Heatmaps by region and decade. Each cell of the heatmaps next to the target scenarios is a world region in a decade, from the 1960s (forest: the 1990s) to 2100. For population, GDP, forest, meat consumption, energy and CO₂ it is the average yearly growth from the first to the last year of the decade, of the sum of the countries that have both values. For the temperature anomaly it is the average over the years of the decade of the region's anomaly, itself the average of its countries weighted by their area. Colours are continuous: white for changes within ±0.1% a year, then through orange to red, or through light blue to blue, reaching the full colour at the 95th percentile of the grid.
- Temperature probability. The share of time and land above three temperatures chosen for each place: its hot months today, rounded to the degree, then one and two steps of its variability (also rounded, at least 1 °C), so that the thresholds make sense from the equator to the Arctic. It assumes temperatures spread normally around their yearly average with the country's variability: 1 − Φ((T − average) ÷ variability). The same share can be a heat that is widespread but rare or local but frequent. For the world, the regions and the European Union, the average of their countries weighted by land area.
- The tax simulator on Wealth distribution applies the chosen rates to these curves, bracket by bracket, and hands the revenue back to the chosen groups. It assumes that people and capital do not react to the tax.
- Live counters run between the yearly values of the database, at the pace of the current year.
- Billionaires. The number and wealth of billionaires come from published rankings; they calibrate the extreme tail of the distribution.
Country scores
On Countries, every country is scored against the world across seven themes. Each score sums up the main indicators of its theme, so that countries can be compared and ranked on several dimensions at once instead of on a single indicator taken on its own. The scores are indicative: they depend on which indicators are chosen and on which direction counts as better.
1. Each indicator
Every indicator is scored by how far the country is from the world today: 100 times the base-2 logarithm of the ratio between the two, so that twice the world value counts as +100 and half of it as −100. The score counts in favour of the country when the gap goes the right way: above the world for indicators to increase, below it for indicators to reduce. For gini coefficient and income share of the richest 1% the comparison is with the median country instead of the world, whose value is not that of a typical country, and the score is linear: 25% below the median counts +100, 25% above it −100. For life expectancy the score is linear too: 25% above the world counts +100, level with it 0, 25% below it −100. For forest area, yearly change the score compares the change over the last year with the world’s: the gap is the difference divided by the world’s change. Every indicator score is kept between −100 and +100, however large the gap.
2. Each theme
A theme score is the average of its indicator scores, kept between −100 and +100 and rescaled from 0 to 100: 50 means level with the world.
3. The total
The total score is the average of the seven theme scores, and needs at least 4 of them; a missing theme counts as level with the world.
The indicators
- People
- Fertility rate (higher is better); Life expectancy (higher is better); Death rate (lower is better).
- Food
- Undernourishment (lower is better); Food consumption per capita (higher is better); Meat consumption per capita (lower is better).
- Land
- Forest area, yearly change (higher is better); Forest land share (higher is better); Population density (lower is better).
- Wealth
- Median GDP (higher is better); Gini coefficient (lower is better); Income share of the richest 1% (lower is better).
- Energy
- Energy per capita (higher is better); Fossil fuels share of electricity (lower is better); Energy per dollar of GDP (lower is better).
- Emissions
- CO₂ per capita (lower is better); CO₂ per unit of energy (lower is better); Global warming impact per capita (lower is better).
- Health
- Air pollution (PM2.5) (lower is better); Child mortality (lower is better); Health spending per capita (higher is better).
Limitations
- Some recent years are estimates.
- For some indicators the sources have no data yet for 2023–2025: those years are estimated, not measured. From 2026 on every value is a projection, including the live counters on the homepage.
- Projections are trends, not forecasts.
- They extend each country's recent path with a growth rate that slows down and a ceiling it cannot pass. They do not model wars, pandemics, recessions, new technologies or policy changes, and they do not know about events after the latest data.
- Few feedbacks between indicators.
- Population, energy and land drive emissions, and emissions drive warming. Warming does not feed back into GDP, harvests, health or migration: climate damages are not part of the model.
- The climate model is simplified.
- It reproduces the observed warming but is far simpler than the models used by the IPCC: no aerosols, no ocean heat uptake, no carbon-cycle saturation, and only three gases. Country temperatures scale with global warming instead of being modelled region by region.
- Gaps are filled.
- Missing years inside a series are interpolated, and series that stop early are extended. Countries with patchy statistics therefore carry more modelled values than countries with complete ones.
- 168 countries, not all of them.
- Very small states and territories are left out because most indicators do not cover them. World totals are the sum of the countries in the database, so they can differ slightly from the world figures published by the sources.
- Income distributions are modelled.
- Curves of income by percentile are rebuilt from a few published shares and poverty rates; the richest fraction of a percent is extrapolated from the shape of the top of the distribution.
- Borders are not statements.
- Maps use Natural Earth boundaries, chosen for their coverage. They imply no position on disputed territories.
Sources and licence
Most of the data on Newleaf comes from Our World in Data. Data produced by Our World in Data is published under the Creative Commons BY 4.0 licence, which allows reuse with attribution. The rest is collated by Our World in Data from third-party providers, under their own terms: we cite each of them below and, under every chart, the provider of that chart's data and the processing it went through: by Our World in Data, and the projections and modelling by Newleaf.
How we cite
“Provider (year) – with processing by Our World in Data and projections by Newleaf”. For the data collected through Our World in Data: Our World in Data, ourworldindata.org, licensed under CC BY 4.0.
On Windows, emoji and flags are drawn with Twemoji (graphics licensed under CC BY 4.0, font under the SIL Open Font License).
People
- United Nations, World Population Prospects — population, births, deaths, migration, fertility, life expectancy, median age, age groups
- UN DESA, International Migrant Stock — immigrants and emigrants
Wealth
- World Bank — GDP, GDP per capita, access to electricity and clean cooking
- World Bank, Poverty and Inequality Platform — people by income threshold
- World Inequality Database — income and wealth shares of the richest
- Penn World Table — working hours and labour productivity
- IMF, OECD, SIPRI, UNESCO — government revenue and spending by function
- WHO/UNICEF Joint Monitoring Programme — clean water and sanitation
- ITU — internet use
Food
- FAO — crop and livestock production, food supply, land used by crop, undernourishment
- WHO — overweight adults
- UN Inter-agency Group for Child Mortality Estimation — child mortality
Land
- HYDE, History Database of the Global Environment — long-term land use
- FAO — forest area, land area, agricultural land
Energy
- Energy Institute, Statistical Review of World Energy — primary energy by source
- Ember — electricity generation by source
- U.S. Energy Information Administration — energy use per capita
- IEA, Global EV Outlook — electric cars
Greenhouse gases
- Global Carbon Budget — CO₂ emissions, cumulative emissions, CO₂ per unit of energy
- Climate Watch — CO₂ by sector
- Jones et al. — methane and nitrous oxide emissions, and each country's contribution to warming
Climate
- NOAA and ice-core records (EPICA, Law Dome) — CO₂, methane and nitrous oxide concentrations
- Copernicus Climate Change Service, ERA5 — monthly surface temperature, temperature anomalies, precipitation
- Met Office Hadley Centre — global and sea surface temperature anomalies
- National Snow and Ice Data Center — Arctic and Antarctic sea ice
- FAO AQUASTAT — freshwater withdrawals
Published factors
- Poore & Nemecek (2018), Science — greenhouse gases, land and scarcity-weighted water per kilogram and per 1000 kcal of food
- Gephart et al. (2021), Nature — greenhouse gases of farmed and wild fish
- USDA FoodData Central — macronutrients per kilogram of food
- IPCC (2014), Fifth Assessment Report, and Turconi et al. (2013) — lifecycle emissions per terawatt-hour of electricity
- Lovering et al. (2022), PLOS ONE — land use per terawatt-hour of electricity
- Macknick et al. (2012), Environmental Research Letters — water use per terawatt-hour of electricity
- IRENA (2024) and Lazard (2024) — cost per terawatt-hour of electricity
- Markandya & Wilkinson (2007) and Sovacool et al. (2016), through Our World in Data — deaths per terawatt-hour of electricity
- IPCC (2006), Guidelines for National Greenhouse Gas Inventories — combustion emissions of coal, oil and gas
- Oxfam and Stockholm Environment Institute, Emissions Inequality Dashboard (2025) — CO₂ per person by global income group
- Canadian Centre for Policy Alternatives (2008) and Li et al. (2021), Sustainability — how land and water footprints grow with income
- FAO, Global Forest Resources Assessment — gross forest loss
Maps
- Natural Earth — country boundaries (public domain)
Newleaf is independent and is not affiliated with or endorsed by Our World in Data or any of the providers above. Errors in processing, projections and modelling are ours. If you spot one, please tell us.
Newleaf is free for everyone.
No paywall, no account, no ads. If you find Newleaf useful, a donation helps to keep it accessible for whoever is standing up for our planet — thank you.