Climate
in a Hurry

The Whole Thing in One Page

Climate change is often presented as a contest between graphs, politics and frightening weather. The science is cleaner than that. Earth receives energy from the Sun and loses energy to space. If incoming and outgoing energy match over time, the planet's average temperature is stable. Change the rate at which heat escapes and the temperature must adjust until balance is restored.

Greenhouse gases change that escape. Sunlight passes through the atmosphere relatively easily, warms the surface, and the warm Earth emits infrared radiation. Molecules such as carbon dioxide, methane and water vapour absorb and re-emit some of that infrared energy. The natural greenhouse effect is why Earth is habitable. Add more long-lived greenhouse gas and the level from which the planet can efficiently radiate to space shifts upward into colder air. A colder emitter loses less energy. The climate then warms until the outgoing flow again matches the incoming sunlight.

Humans have changed the atmosphere. Burning fossil fuels, making cement, changing land use and agriculture have increased carbon dioxide, methane and nitrous oxide. Land and oceans absorb a large share of our carbon dioxide emissions, but not all. The remainder accumulates, making carbon dioxide a stock problem as well as an annual-emissions problem.

The expected fingerprints have appeared. The lower atmosphere and oceans have warmed. Glaciers and ice sheets have lost mass. Global sea level has risen. Heat extremes have become more frequent and intense across most land regions. The upper atmosphere has cooled while the lower atmosphere has warmed, a pattern expected from increased greenhouse gases and not from a brighter Sun. The oceans have taken up most of the excess heat, which is why ocean heat content is one of the clearest measures of the planetary imbalance.

Attribution does not rest on one thermometer line. Scientists compare observations with the responses expected from greenhouse gases, aerosols, volcanic eruptions, solar changes and internal variability. Models run with natural factors alone do not reproduce the modern warming. Add human influence and they do. The IPCC's conclusion is direct: human influence has unequivocally warmed the atmosphere, ocean and land.

Uncertainty remains, but it is not where public arguments often place it. We know the sign of greenhouse warming, the dominant cause of recent global warming and the broad direction of many impacts. The harder questions concern the size and timing of regional changes, cloud feedbacks, ice-sheet behaviour, compound extremes, adaptation limits and the exact emissions pathway societies will choose. Those uncertainties widen the range of possible outcomes. They do not create an equal chance that the basic mechanism is wrong.

The practical consequence is unusual. For carbon dioxide, warming depends mainly on cumulative emissions. To stop adding further CO2-driven warming, global net carbon dioxide emissions must fall to roughly zero. Cutting methane can slow near-term warming because methane is shorter-lived. Mitigation changes how far the climate moves. Adaptation changes how much harm a given climate causes.

Climate science therefore ends with a distinction rather than a slogan. Physics tells us how forcing changes temperature. Observations tell us what has already changed. Attribution tells us why. Models map conditional futures. Risk science asks who and what sits in harm's way. Those findings constrain the choices available to societies, but they do not decide how costs, benefits, speed or fairness should be weighed. That final step is political and ethical, not a hidden output of a climate model.

That is the book.

Why You Should Care

In 1958, Charles David Keeling began measuring carbon dioxide from a station high on Mauna Loa in Hawaii. The location was chosen because air arriving over thousands of kilometres of Pacific Ocean could give a clean view of the background atmosphere. The instrument recorded a saw-toothed line. Carbon dioxide fell each northern spring and summer as plants took it up, then rose in autumn and winter. Underneath that annual breathing, the whole line climbed.

That curve is now one of the most recognisable records in science because it catches the climate problem in a single image: a natural cycle with a human trend laid across it. The seasonal wiggle is large enough to prove the biosphere is active. The rising baseline proves that the biosphere and oceans are not removing carbon dioxide as fast as humans are adding it.

Little of the core argument depends on trusting a forecast. Laboratory spectroscopy establishes that greenhouse gases absorb infrared radiation. Atmospheric concentrations are measured directly. Satellites measure incoming sunlight and outgoing heat. Ocean heat, glaciers, sea level, snow and heat extremes add independent evidence. Models matter, but the case does not begin with one.

That matters because climate is a subject in which several different questions are routinely fused. Is the planet warming? What is causing it? How much more warming will a given amount of greenhouse gas produce? What will that warming do in a particular place? Which responses are technically possible? Which are affordable, fair or politically acceptable? A person can answer one correctly and still disagree about the others. Keeping the questions separate is the quickest way to think clearly.

You should also care because climate change shifts the background conditions under which many systems operate. A hotter baseline changes heatwaves. Warmer air can hold more water vapour, raising the potential for heavy rainfall where weather dynamics allow it. Warmer oceans expand and sea level rises. Carbon dioxide also changes ocean chemistry. Farmers, engineers, insurers, public-health systems and ecosystems then face altered combinations of heat, water, fire and slow change.

The effects are uneven. Some cold-related harms decline. Rainfall can rise in one place and fall in another. A degree of global warming is not a degree everywhere, and a global average does not tell you who is exposed or vulnerable. Climate risk cannot be read from temperature alone.

The latest observations make the scale concrete. The World Meteorological Organization assessed 2025 as the second or third warmest year in the instrumental record, about 1.43°C above the 1850 to 1900 average, after the record warmth of 2024. One unusually warm year above 1.5°C does not mean the Paris Agreement's long-term temperature threshold has formally been crossed, because that target concerns sustained warming over decades. It does show how close the background climate has moved to a level that recently sounded distant.

There is one final reason to care, and it is more useful than fear. Climate change has a stopping rule. Carbon dioxide does not make temperatures rise forever after emissions cease. The best current understanding is that CO2-driven global warming stabilises approximately when global net carbon dioxide emissions reach zero, though sea level and other parts of the system continue changing for much longer. That gives the problem an unusual shape. Every tonne avoided reduces the eventual warming commitment. Every fraction of a degree avoided reduces some risks. Delay matters because carbon accumulates, but there is no scientific point at which further effort becomes meaningless.

The climate problem is therefore neither a hoax that vanishes under inspection nor an apocalypse with a fixed date. It is a physical perturbation whose scale depends heavily on choices. The useful position is not optimism or pessimism. It is understanding the machine well enough to know which levers are real.

The Core Ideas

Climate Is an Energy Balance

Put Earth in the dark and it would cool. Put it in sunlight and it warms. The average temperature settles where energy absorbed from the Sun is balanced by energy radiated back to space. Climate begins with that accounting identity.

The Sun sends mostly visible and near-visible radiation. Some is reflected by clouds, aerosols, ice, snow and bright surfaces. The rest is absorbed by the atmosphere and surface. A warm Earth then emits infrared radiation. The hotter an object is, the more thermal radiation it emits, so warming itself provides the stabilising response that eventually restores balance after a disturbance.

If Earth had no atmosphere capable of absorbing infrared radiation, its surface would be far colder. The atmosphere is partly transparent to incoming sunlight but interacts strongly with outgoing infrared at particular wavelengths. Water vapour, carbon dioxide, methane, nitrous oxide and some other gases absorb and emit infrared radiation because of their molecular structure. This is measured spectroscopy, not a climate hypothesis invented from temperature records.

The common blanket analogy helps only briefly. A blanket reduces heat loss mainly by limiting convection. Greenhouse gases work through radiation. A useful teaching model is that adding greenhouse gas shifts the effective altitude from which infrared energy escapes to space upward into colder air. Colder air emits less infrared energy, so Earth temporarily loses less energy than it absorbs and the surface and lower atmosphere warm until outgoing radiation rises enough to restore balance. The real atmosphere is not a single emitting layer: different wavelengths escape from different heights, and convection, clouds and water vapour interact with radiative transfer. The altitude picture compresses that spectrally complex process without changing the underlying energy accounting.

This change in the planetary energy budget is called radiative forcing. Greenhouse gases provide positive forcing. Some aerosols provide negative forcing by reflecting sunlight and altering clouds. A large volcanic eruption can cool the climate temporarily by injecting reflective particles into the stratosphere. Changes in the Sun can alter incoming energy. The climate responds to the sum of these disturbances, filtered through feedbacks and internal variability.

The distinction between forcing and feedback prevents several common errors. Carbon dioxide added by burning fossil fuels is a forcing because it is imposed on the climate system from outside the fast climate response. Water vapour is mainly a feedback because warmer air can contain more water vapour, and water vapour itself is a greenhouse gas. Warming therefore increases atmospheric water vapour, which amplifies the initial warming. Treating water vapour as a rival explanation to carbon dioxide reverses the causal sequence.

Earth does not respond instantly because it contains enormous heat reservoirs. Land temperatures can change quickly. The upper ocean changes more slowly. The deep ocean can take centuries to approach a new state. Ice sheets can take longer still. The energy-balance idea therefore contains both the cause of warming and the reason the response unfolds over time.

The first Core Idea is the one to keep underneath everything else. Climate is not primarily a story about weather records, political targets or computer simulations. It is a story about an energy imbalance. The rest of the science asks what created that imbalance, how large it is, where the energy goes and what follows.

This also explains why the ocean matters so much. The atmosphere contains little heat compared with the ocean, so surface air can jump around while the total system continues gaining energy. A hot or cool year can therefore distract from the more stable question: is Earth, taken as a whole, retaining more energy than it loses?

Carbon Dioxide Is a Stock, Not a Tap

Imagine a bath with the drain open. Water level depends on the difference between inflow and outflow, not on whether the drain exists. The carbon cycle works the same way, with one complication: it has several drains operating on different timescales.

Carbon moves naturally among atmosphere, ocean, vegetation, soils and rocks. Plants take up carbon dioxide through photosynthesis. Respiration and decay return much of it. The surface ocean exchanges carbon dioxide with the air. Ocean circulation moves carbon into deeper water. Weathering and sediment processes act more slowly. Before industrialisation, these enormous flows were roughly balanced over the timescales relevant to recent climate.

Humans added a new flow by taking carbon stored underground for millions of years and oxidising it rapidly. Land-use change added another by releasing carbon from vegetation and soils while weakening some sinks. The natural carbon cycle did not stop. It absorbed a large fraction of the added carbon. That uptake is why atmospheric carbon dioxide has risen by less than it would have if every emitted molecule had stayed in the air.

But the atmospheric concentration still climbed from roughly 280 parts per million before industrialisation to more than 430 parts per million at Mauna Loa during the seasonal peak in 2026. The oceans have absorbed carbon and become more acidic. Land ecosystems have, on balance, absorbed carbon too. Those sinks are a service provided by the Earth system, but calling them a solution is misleading. The water level is rising because the tap is running faster than the drains.

This gives carbon dioxide a property that methane does not share to the same degree: cumulative emissions matter. A pulse of carbon dioxide is redistributed among atmosphere, ocean and land on several timescales. There is no single atmospheric lifetime after which it disappears. Some of the perturbation is removed relatively quickly, some persists for centuries, and a smaller fraction affects the system for much longer.

That persistence links total human carbon dioxide emissions closely to total CO2-driven warming. The relationship is not perfectly exact, but over policy-relevant ranges it is close enough to support the idea of a carbon budget. If a temperature limit is the goal, there is a finite amount of additional carbon dioxide that can be emitted for a chosen probability of staying near that limit. The budget depends on climate response, non-CO2 gases, Earth-system feedbacks and how the target is defined, so it should never be treated as a precise countdown clock. Its logic is sound even when its exact size is uncertain.

Methane behaves differently. It is a powerful greenhouse gas molecule for molecule, but atmospheric methane is chemically removed on a timescale of roughly a decade. Sustained methane emissions maintain elevated concentrations. Cut those emissions sharply and the methane concentration can decline, reducing its warming contribution relatively quickly. Nitrous oxide is longer-lived. Aerosols can last days to years, depending on type and altitude. Different pollutants therefore require different stopping rules.

The stock-flow distinction explains why stabilising annual carbon dioxide emissions at a high level does not stabilise climate. If emissions remain positive, atmospheric carbon dioxide continues to accumulate, though the fraction remaining in the atmosphere can vary. To stop adding further CO2-driven warming, net global carbon dioxide emissions need to fall to roughly zero. That is not a political slogan inserted into physics. It is the consequence of treating carbon as a stock.

The Warming Is Measured in More Than Air

If the only evidence for climate change were a surface thermometer record, scepticism would be easier. Thermometers move, cities grow around stations, instruments change, and the record requires careful correction. Climate science is stronger because the same energy imbalance appears in many independent parts of the system.

Global surface temperature records combine measurements from land stations and ocean observations. Several research groups use different methods and make different choices about gaps, urbanisation, sea-surface measurement changes and baseline periods. Their precise annual numbers differ slightly. Their long-term result does not: the world has warmed markedly since the late nineteenth century, with the fastest sustained warming in recent decades.

The ocean is the larger ledger. Water has a high heat capacity and the oceans cover most of the planet. More than 90 per cent of the excess energy accumulated in the climate system has gone into the ocean. Thousands of autonomous Argo floats now profile temperature and salinity through the upper ocean, while earlier ship and instrument records extend the picture backwards. Ocean heat content has risen to record levels. This matters because a warm year can be moved around by El Niño or La Niña, but a persistent increase in ocean heat reveals a planetary energy imbalance beneath short-term surface noise.

Ice tells the same story in a different language. Mountain glaciers have retreated across most regions. Greenland and Antarctica have lost ice mass. Arctic sea ice has declined strongly in late summer since satellite monitoring began. Snow cover has changed. These indicators have different measurement systems and different regional complications, yet they respond consistently to a warming climate.

Sea level combines several processes. Warmer seawater expands. Melting land ice adds water to the ocean. Changes in land water storage can push the total up or down. Tide gauges show a long rise over the twentieth century; satellites have measured global mean sea level since the early 1990s with far better spatial coverage. The rate has increased over the satellite era. A millimetre of global average sea-level rise sounds trivial until it is multiplied across the area of the oceans and added to storm surge, tides and coastal subsidence.

The atmosphere has fingerprints too. The troposphere has warmed while the stratosphere has cooled. That vertical pattern is expected when greenhouse gases increase because added carbon dioxide warms the lower atmosphere while enhancing radiative cooling higher up. A simple increase in solar output would not produce the same pattern. Nights have generally warmed faster than they would under a pure solar explanation. The spectrum of outgoing infrared radiation measured from space also shows greenhouse-gas effects.

Then come biological and seasonal observations: earlier spring events in many regions, shifts in species ranges, changes in growing seasons and coral bleaching during marine heat. None of these is a perfect thermometer. Together they form a consistent physical picture.

The discipline's strongest evidence is therefore not one graph but convergence. Independent instruments, different physical variables and different research groups point to the same imbalance. The reasonable argument is about exact rates, regional detail and causes of individual events. The claim that there is no warming signal requires dismissing thermometers, oceans, ice, sea level, satellites and basic radiation physics at once.

Attribution Is a Fingerprint Test

Knowing that the world is warming does not by itself tell you why. Climate has changed before humans burned fossil fuels at industrial scale. Volcanoes, solar changes, orbital cycles, ocean variability and random fluctuations all affect climate. Attribution is the work of separating those influences.

The most basic test is magnitude. Solar observations do not show an increase capable of explaining recent warming. Large volcanic eruptions tend to cool the planet for a few years rather than produce a century-scale warming trend. Orbital changes operate over much longer timescales and are not pointing towards rapid modern warming. Internal variability can redistribute heat between ocean and atmosphere and create years or decades that are warmer or cooler than the trend, but it cannot create energy from nowhere. A persistent increase in the total heat content of the climate system requires a net forcing.

The stronger test is pattern. Different causes leave different fingerprints across height, latitude, season and physical variable. Greenhouse forcing warms the troposphere and cools the stratosphere. Solar forcing would tend to warm both. Greenhouse warming is larger over land than ocean and especially strong in the Arctic because of feedbacks. Aerosol pollution has concentrated cooling effects in regions where emissions are high and can alter cloud properties. Volcanic aerosols produce identifiable temporary cooling after major eruptions.

Climate models are useful here because they allow counterfactual experiments that cannot be run on the real planet. Researchers run ensembles with natural forcings only, with human forcings, or with both. The natural-only simulations reproduce much of the earlier variation but not the strong late twentieth and early twenty-first century warming. Simulations including human greenhouse gases and aerosols reproduce the observed broad trend and spatial pattern far better.

This is sometimes caricatured as models being tuned to produce the answer. Models do contain parameters for processes too small or complex to calculate directly at global scale, especially clouds and turbulence. But attribution does not rely on fitting one curve. It draws on physical forcing estimates, observed fingerprints, paleoclimate evidence, energy-budget measurements and models with many independently tested components. The central result survives changes in method.

The IPCC's strongest wording reflects that convergence: human influence has unequivocally warmed the atmosphere, ocean and land. The word does not mean every flood, drought or warm afternoon is caused solely by greenhouse gases. It means the net global warming since the nineteenth century cannot be explained without human influence, and the human contribution is the dominant cause of the observed warming over recent decades.

Event attribution is a narrower task. Instead of asking whether climate change caused an event in the sense of creating it from nothing, researchers ask how human-caused climate change altered its probability or intensity. A heatwave may have been possible in the old climate but become far more likely or hotter in the new one. Heavy rainfall may intensify because a warmer atmosphere contains more moisture, while the precise storm path still depends on weather. For some hazards, such as certain droughts or tropical cyclone characteristics, regional dynamics make attribution harder.

The language matters. Climate change loads the dice, but the dice analogy can also mislead because the dice themselves keep changing and different hazards respond differently. Attribution is better understood as a fingerprint test conducted on a moving distribution. It asks whether the observed pattern and changed odds match the physics expected from known forcings.

Feedbacks Decide How Much Warming You Get

Doubling carbon dioxide creates a calculable radiative forcing. The difficult question is how strongly the rest of the climate system responds. That is the job of feedbacks.

Start with the stabilising response. A warmer Earth emits more infrared radiation to space. This Planck response acts like a thermostat and prevents a finite greenhouse forcing from producing unlimited warming. If that were the only response, climate sensitivity would be relatively low.

Then the amplifiers arrive. Warmer air contains more water vapour, which is a greenhouse gas. That increases warming. Snow and sea ice melt, exposing darker land or ocean that absorbs more sunlight. That increases warming. Changes in atmospheric temperature with height alter how efficiently heat escapes. Clouds can either cool by reflecting sunlight or warm by trapping infrared radiation, and changes in cloud amount, altitude and type can alter the balance.

Cloud feedback has long been one of the largest sources of uncertainty in climate sensitivity. Better observations and models have narrowed the plausible range, but clouds remain difficult because they are small-scale, dynamic and connected to circulation. The key public misunderstanding is to treat this uncertainty as if feedbacks might plausibly cancel the greenhouse effect. The evidence instead points to a net positive feedback overall.

The standard summary measure is equilibrium climate sensitivity, the long-term global surface warming after atmospheric carbon dioxide doubles and the climate approaches a new equilibrium while fast feedbacks operate. IPCC AR6 assessed a best estimate of about 3°C, with a likely range of 2.5°C to 4°C. That range is much narrower than the broad uncertainty carried through earlier assessments.

There is also transient climate response, which asks how much warming occurs around the time carbon dioxide doubles under a steadily rising concentration. It is lower than equilibrium sensitivity because the deep ocean has not yet caught up. Human decisions unfold in this transient climate, while the deep ocean, ice sheets and sea level continue responding long after a particular emissions decision has been made.

Feedback language becomes more complicated when carbon-cycle and ice-sheet changes are added. Warming can weaken some land and ocean carbon sinks, leaving a larger fraction of future emissions in the atmosphere. Thawing permafrost can release carbon dioxide and methane. Large ice sheets can cross thresholds that commit them to long-term loss. These Earth-system feedbacks do not all operate on the same timescale, and they should not be bundled into a single dramatic number.

This is where tipping points enter. A tipping element is a component that may undergo large, self-reinforcing change after a critical level is crossed. Candidates include parts of the Greenland and Antarctic ice sheets, coral reefs, permafrost and large circulation or ecosystem systems. Evidence and confidence differ sharply among them. Some changes are gradual over the range we can observe. Some may have thresholds. Crossing a threshold does not usually mean the whole planet changes overnight. It can mean a process becomes difficult or impossible to reverse on human timescales.

Sensitivity is therefore feedback arithmetic. The direct greenhouse forcing is well understood. The uncertainty lies mainly in how strongly water vapour, clouds, ice, circulation and slower Earth-system components amplify or reshape the response. That uncertainty is important because it changes the warming expected from a given emissions path. It is not a blank space where any answer is equally plausible.

Models Are Conditional Machines, Not Crystal Balls

A climate model is a set of physical equations applied to a three-dimensional grid representing atmosphere, ocean, land and ice. It calculates conservation of energy, mass and momentum, radiation, fluid motion and many chemical and biological processes. Where the grid cannot resolve a process such as individual clouds, the model represents its average effects through parameterisations informed by observations and smaller-scale theory.

That sounds less magical once you see what models are for. Weather prediction asks for the exact state of the atmosphere days ahead. Climate projection asks for the statistics of weather under changed boundary conditions over decades. A model can fail to tell you whether it will rain in London on 15 June 2050 and still estimate that warmer air raises the probability of heat extremes over southern England. The tasks are different.

Models are tested against the present, the historical record, volcanic eruptions, seasonal cycles, ocean circulation, paleoclimates and patterns they were not individually tuned to reproduce. No model gets everything right. Some run too warm or cool in particular regions. Clouds, rainfall and regional circulation remain harder than global temperature. Model ensembles are useful because disagreement among models reveals part of the structural uncertainty rather than hiding it.

Future projections add another uncertainty that science cannot eliminate: human choice. A model can estimate the climate response to a specified greenhouse-gas pathway. It cannot know how many power stations, forests, cars, factories or policies will exist in 2070. That is why modern assessments use scenarios. A scenario is not a prediction that claims to know society's future. It is a conditional statement: if emissions and concentrations follow something like this path, the climate response is likely to fall in this range.

Three kinds of uncertainty should therefore be kept separate. Internal variability is the climate's own chaotic movement, which matters strongly over years and regions. Model uncertainty concerns how the physical system responds. Scenario uncertainty concerns what humans emit. In the near term, internal variability can dominate local trends. Over longer periods, emissions choices become increasingly important for global temperature and many related changes.

The stronger reason to trust climate projections is that models are not alone. Simple energy-balance calculations, paleoclimate evidence, observed responses to past forcings, modern satellite measurements and complex models converge on similar sensitivity ranges. A climate model adds geography, circulation and interactions to a physical argument that exists without it.

The correct posture is neither worship nor dismissal. Treat models as conditional machines. Ask what was assumed, which variable is robust, how wide the ensemble spread is, whether independent evidence supports the result and whether the decision would change across the plausible range. A model is most valuable when it turns uncertainty into a structured set of consequences rather than a single number pretending to be fate.

Resolution also matters. A global model can represent a mountain range while missing the exact valley where rainfall decides a reservoir's future. Regional downscaling can add detail, but detail is not automatically accuracy. The closer a decision gets to one catchment, crop or city block, the more local observations and process knowledge must sit beside the global model. That scale test prevents false precision from masquerading as better science.

Risk Is the Product of Physics and Exposure

A hotter climate does not translate mechanically into a fixed amount of harm. Risk appears when a physical hazard meets people, ecosystems or assets that are exposed and vulnerable. That distinction is where climate science meets the human world.

A heatwave is a hazard. The risk depends on who experiences it, whether homes can cool, whether people work outdoors, how old or ill they are, whether electricity and health services function, and whether the city has trees, shade and night-time relief. A coastal storm becomes more damaging when sea level is higher, but the final loss also depends on buildings, defences, evacuation, wealth, insurance and recovery capacity. The same physical event can be inconvenience in one place and catastrophe in another.

Climate change alters hazards in several ways. Heat extremes become more frequent and intense as the baseline warms. Heavy precipitation intensifies in many regions because warmer air can contain more water vapour, though circulation determines where storms form and travel. Agricultural drought reflects rainfall, evaporation, soils and crops, so its response is regionally complex. Tropical cyclones are not expected to become more numerous everywhere, but the strongest storms can carry more rainfall and a larger share may reach high intensity in a warmer climate. Fire weather rises in many regions where heat and dryness align, but ignition, vegetation and land management remain crucial.

Slow changes matter too. Sea-level rise raises the starting point for every coastal flood. Ocean warming and acidification stress marine ecosystems. Glacier loss changes seasonal water storage in mountain regions. Shifts in heat and rainfall affect crop yields, labour capacity, disease ecology and infrastructure. Some impacts can be adapted to. Others become harder or more expensive as warming increases.

Adaptation means reducing harm under the climate that occurs. It includes heat-health plans, flood protection, water storage, drought-tolerant crops, building standards, early-warning systems, ecosystem restoration, insurance design and, in some places, relocation. Good adaptation is specific to a risk. Air conditioning can reduce heat mortality while increasing electricity demand. A sea wall can protect one coast while shifting erosion elsewhere. Irrigation can protect crops until water itself becomes scarce. Adaptation has limits.

Mitigation does a different job. It reduces the greenhouse-gas forcing that creates future climate change. For carbon dioxide, that ultimately means bringing net global emissions to roughly zero. Rapid cuts in methane can reduce near-term warming. Protecting and expanding carbon sinks can help, but land is finite and stored biological carbon can burn or decay. Carbon dioxide removal may be needed to balance residual emissions from hard-to-eliminate sources and, in some pathways, to reduce atmospheric CO2 after an overshoot. Its scale, cost, permanence and ecological effects matter.

The causal loop closes here. The book began with a planetary energy imbalance created by forcing. Because carbon dioxide accumulates, as long as net CO2 emissions remain substantially above zero the stock continues to rise and adds further warming. The same physics that explains the first extra fraction of a degree supplies the stopping rule for the last: halt the growth of the long-lived forcing.

Risk then turns that physical endpoint into a practical one. Mitigation limits how far the hazard shifts. Adaptation reduces exposure and vulnerability to the change that cannot be avoided. Uncertainty strengthens the case for treating them together because the tails of the distribution contain costly outcomes. The science does not choose a carbon price, technology mix, burden-sharing formula or fairness rule. It does identify which interventions change greenhouse forcing, which reduce exposure or vulnerability, and which claims about the physical system are incompatible with the evidence. That boundary is important: climate science can narrow the decision space without pretending to settle the values inside it.

How It Actually Works

Sunlight in, heat out

Every climate story begins at the top of the atmosphere. Averaged over the whole sphere and over time, Earth receives about 340 watts of solar power per square metre. Roughly three tenths is reflected back to space by clouds, aerosols, ice, snow and the surface. The rest is absorbed. To avoid warming indefinitely, the planet must lose an equal amount of energy.

That simple balance hides a layered atmosphere. The surface absorbs sunlight and emits infrared radiation upward. Some wavelengths pass directly to space through atmospheric windows. Others are absorbed by greenhouse gases and clouds, then emitted in all directions. The atmosphere itself radiates. The effective radiation to space comes from a range of heights, depending on wavelength and cloud conditions.

Add carbon dioxide and the atmosphere becomes more opaque to infrared radiation in parts of the spectrum. The energy that used to escape from a lower level now escapes, on average, from a higher one. Because the troposphere generally becomes colder with altitude, that higher emitting region initially sends less heat to space. The top of the atmosphere is out of balance. Energy accumulates, warming the lower atmosphere, surface and ocean until outgoing radiation rises enough to match incoming sunlight again.

One reason old objections persist is that carbon dioxide absorbs strongly in some infrared bands already. If those bands are saturated near the surface, why should more CO2 matter? Because the relevant question is not whether one layer absorbs nearly all radiation at a particular wavelength. It is where in the atmosphere the radiation that eventually escapes to space is emitted. Additional CO2 shifts that effective emission upward. The absorption spectrum also has wings rather than a single black line, and pressure, temperature and overlapping gases shape the result. Saturation near the ground does not mean zero additional forcing.

The carbon leaves the ground

For most of human history, atmospheric carbon dioxide varied within a relatively narrow range compared with the industrial rise. Ice cores trap ancient air bubbles and extend direct samples of past atmospheres back hundreds of thousands of years. They show glacial-interglacial cycles in which carbon dioxide moved broadly with temperature, ice and ocean circulation.

The carbon cycle then distributes the added CO2. Plants respond to carbon dioxide, temperature, water and nutrients. Some regions and ecosystems take up more carbon; others release it during drought, fire or land clearing. The ocean absorbs carbon dioxide where surface chemistry and circulation allow it, then transports some into deeper layers. Absorption changes carbonate chemistry, lowering pH even though seawater remains alkaline overall. This is ocean acidification: a movement towards greater acidity, not a claim that the ocean has become an acid.

The share of emissions taken up by land and ocean varies from year to year, especially with El Niño, drought and fire. Over recent decades, those natural sinks have removed roughly half of anthropogenic carbon dioxide emissions, with the atmosphere retaining the rest. That fraction should not be treated as a permanent law. As the planet warms and the ocean accumulates carbon, sink behaviour can change.

The accounting nevertheless closes well enough to make the source clear. Human emissions are larger than the observed annual atmospheric increase. Natural sinks explain the difference. A natural source cannot explain a rising atmosphere while the land-ocean system as a whole is taking carbon out of the air.

From a curve to a climate signal

Keeling's Mauna Loa record gave climate science a precise atmospheric trend. Temperature required a more difficult global reconstruction. Land thermometers are unevenly distributed. Ocean measurements changed from buckets hauled onto ships to engine-room intakes, buoys and satellites. Stations moved and instruments changed. Urban heat islands affect local temperatures. A global record therefore needs homogenisation and statistical treatment.

Different groups solve these problems differently, which is useful. NASA, NOAA, the UK's Met Office and University of East Anglia, Berkeley Earth and others produce independent global temperature analyses. They disagree by small amounts in individual years and sparse regions. They agree on the long rise. The agreement is not surprising because most of the raw measurements overlap, but methodological diversity helps expose sensitivity to processing choices.

The planet's heat inventory provides a harder-to-fake cross-check. The ocean dominates heat uptake. Argo floats, deployed globally since the 2000s, repeatedly descend and rise through the upper two kilometres of ocean, measuring temperature and salinity. Earlier records are less complete and require more correction, but the multi-decadal increase in ocean heat is clear. WMO reported record ocean heat content again in 2025.

Sea level is another integral measure. Thermal expansion and land-ice loss both add to it. Tide gauges reveal local relative sea level, which also includes land movement. Satellite altimeters estimate global mean sea level. The global rate has accelerated over the satellite record, while local outcomes differ because of ocean circulation, gravity changes, subsidence and uplift.

The cryosphere responds visibly. Glaciers are losing mass in most monitored regions. Greenland and Antarctica are losing ice overall. Arctic sea ice has declined strongly in extent and thickness, especially in late summer. Antarctic sea ice is more variable and has behaved differently from the Arctic, a reminder that two icy poles do not share one mechanism.

Natural variability rides on the trend

The climate system has internal oscillations because atmosphere and ocean exchange heat and momentum. El Niño and La Niña are the most familiar. During El Niño, warm surface water spreads eastward across the tropical Pacific and more heat is released from ocean to atmosphere, often lifting global surface temperature. La Niña tends to do the reverse. Neither creates or destroys the long-term planetary energy imbalance. They move heat around and alter where rainfall and temperature anomalies appear.

Other modes influence the North Atlantic, Southern Ocean, monsoons and regional weather. A decade can therefore warm faster or slower than the forced trend. This is why cherry-picking a start year can manufacture a supposed pause or acceleration. Climate is a distribution observed over enough time and space to separate the signal from chaotic movement.

Volcanoes produce another visible interruption. The 1991 eruption of Mount Pinatubo injected sulphur dioxide into the stratosphere, where sulphate aerosols reflected sunlight and cooled global surface temperature for several years. The event became a natural experiment for models and observations. It showed both that radiative forcing can move global temperature and that the climate rebounds when a short-lived forcing fades.

The Sun varies too. Satellites measure total solar irradiance and show an approximately eleven-year cycle. Solar activity contributed to climate changes before the industrial era, but the measured modern trend does not provide the forcing required for recent warming. The observed vertical fingerprint, with a cooling stratosphere over the long term, also points away from increased solar output as the main cause.

The fingerprints line up

By the late twentieth century, the scientific question had moved from whether carbon dioxide could warm the planet to whether the human signal had emerged from variability. Attribution developed as a formal discipline using observed patterns, forcing estimates and ensembles of model runs.

Suppose greenhouse gases, aerosols, solar changes and volcanoes each produce a different expected spatial and vertical pattern. Researchers can ask how strongly those fingerprints appear in the observed climate. They can also run models with subsets of forcings. Natural factors alone fail to reproduce the post-1950 warming. Human greenhouse gases alone tend to warm more than observed in some periods because human-produced aerosols offset part of that forcing. Combine the major human and natural influences and the broad observed history emerges.

This does not mean every detail is captured. Aerosol forcing has been difficult to estimate because particles differ chemically and affect clouds. Regional rainfall trends are noisier than global temperature. Internal variability can produce mismatches over a decade. Attribution works because the global and large-scale patterns remain much larger than those uncertainties.

The result is stronger than a statement that humans contributed some warming. IPCC AR6 assessed the best estimate of total human-caused warming from 1850 to 1900 to 2010 to 2019 as about 1.07°C, while natural solar and volcanic drivers contributed close to zero net warming over that period. Human greenhouse gases caused more warming than the observed net change, with aerosols and other human influences offsetting part of it.

That offset creates an awkward feature of air-pollution control. Sulphate aerosols from fossil-fuel combustion damage health and reflect sunlight. Cleaning them from the atmosphere is desirable for health, but it removes some cooling. If greenhouse-gas emissions remain high while aerosol pollution falls, part of the greenhouse warming that had been masked can appear. This is not an argument for preserving dirty air. It is an argument for cutting the warming pollutants at the same time.

Sensitivity turns forcing into temperature

Once the cause is established, the next question is scale. How much warming follows a given forcing?

A doubling of atmospheric carbon dioxide relative to pre-industrial levels creates a radiative forcing of roughly 3.7 watts per square metre before the climate has fully responded. The direct surface warming would be modest if nothing else changed. Feedbacks magnify it.

Water vapour feedback is strong and positive. A warmer atmosphere can contain more water vapour, and relative humidity remains roughly stable on large scales, so atmospheric water vapour increases. Lapse-rate feedback describes changes in the vertical temperature profile and partly offsets water-vapour amplification globally. Surface-albedo feedback is positive because loss of snow and ice exposes darker surfaces. Cloud feedback is assessed as positive overall but carries substantial uncertainty.

Evidence for climate sensitivity comes from more than models. Scientists use the instrumental period, the response to volcanic eruptions, paleoclimate states such as the last glacial maximum, and physical understanding of feedbacks. Each line has different weaknesses. Their combination tightened the AR6 estimate.

The equilibrium sensitivity range matters, but it should not be confused with a forecast for 2100. Actual twenty-first century warming depends on emissions, non-CO2 forcing, ocean heat uptake and transient response. A highly sensitive climate warms more for the same emissions, yet lower emissions can still produce less warming than high emissions in a lower-sensitivity world. Human choice and physical response multiply rather than compete.

What one degree does to a distribution

Global mean temperature is an average of land, ocean, tropics, poles, day and night. Its importance lies in what happens to distributions underneath it.

Take heat. Shift the average temperature upward and an event that used to sit in the far hot tail becomes less rare. The distribution may also change shape, and soil moisture can amplify heat over land. This is why the increase in extreme heat is one of the clearest consequences of warming.

Heavy rainfall follows another physical relation. Warmer air can hold about 7 per cent more water vapour per degree Celsius near typical surface temperatures if relative humidity stays similar. That does not mean rainfall everywhere rises by 7 per cent per degree. Storm dynamics, moisture supply and circulation matter. It does mean storms that can access the extra moisture have greater potential to produce intense rainfall.

Drought is more complicated because it has several meanings. Meteorological drought concerns low precipitation. Agricultural drought concerns soil moisture. Hydrological drought concerns rivers, reservoirs and groundwater. Warming increases evaporative demand in many places, so soil can dry even without a large fall in rainfall. Regional precipitation changes can either reinforce or offset that effect.

Tropical cyclones illustrate why simple claims fail. Their total global number does not need to rise for risk to increase. Warmer oceans and air support heavier rainfall. Sea-level rise makes storm surge reach farther inland. Evidence also supports an increase in the proportion of intense tropical cyclones with warming, while basin-scale frequency trends remain noisy and affected by observation changes.

Wildfire risk also separates hazard from outcome. Hotter, drier conditions can increase fire weather and dry fuels, but ignition, vegetation, invasive species, suppression history and land management decide whether a landscape burns. Climate is part of the causal chain, not the whole chain.

Impacts are unequal by construction

The physical system creates hazards. Societies create much of the exposure and vulnerability. Two coastal settlements can face the same sea-level rise and radically different risk because one has defences, strong buildings, warnings and finance while the other has fragile housing on low ground. Agriculture, health, water and ecosystems show the same pattern: climate pressure interacts with local conditions, other stresses and capacity to respond.

Risk often compounds. Heat and drought can occur together. A cyclone can destroy power just when heat and humidity make cooling essential. Drought can reduce hydropower while increasing electricity demand. A crop shock can interact with conflict, trade restrictions or high debt. Climate change changes the odds inside systems that already have weaknesses.

Some adaptation is straightforward: cool roofs, shade, flood warnings, drainage, water efficiency, resilient crops, building codes. Some requires major infrastructure or relocation. Some ecosystems cannot be engineered into their old state once temperature or chemistry moves too far. The effectiveness of adaptation therefore declines for some risks as warming rises.

Mitigation changes the forcing

Mitigation is often discussed as a list of technologies, but climate science gives it a cleaner structure. Reduce sources of greenhouse gases, increase durable removals where appropriate, and cut short-lived warming agents that can lower near-term forcing.

Carbon dioxide dominates long-term warming because of its scale and persistence. Fossil-fuel combustion and industrial processes remain the largest source. Land-use change contributes additional CO2. Methane comes from fossil-fuel systems, agriculture, waste and natural sources. Nitrous oxide is strongly linked to agriculture and fertiliser use. Fluorinated gases contribute smaller shares but can be potent.

The queue gives Energy, Coal and Oil their own books, so the machinery of power generation does not belong here. For climate, the important fact is that technology and demand alter emissions pathways. Low-carbon electricity, electrification, efficiency, land management and changes in industrial processes can all reduce greenhouse-gas forcing. Their engineering and economic trade-offs belong elsewhere.

Net zero does not mean every smokestack vanishes. It means remaining anthropogenic carbon dioxide emissions are balanced by anthropogenic removals, globally. The smaller the residual emissions, the smaller the removal burden. Biological removals can be reversed by fire, disease or land-use change. Geological storage can be more durable but requires capture, transport, injection and monitoring. Direct air capture is physically possible but energy-intensive and expensive at present. None of these facts erases the value of emissions cuts.

Methane allows a different lever. Because it is shorter-lived, reducing methane emissions can lower its atmospheric concentration and warming contribution within decades. This is especially useful for slowing near-term warming, but methane cuts cannot substitute for net-zero carbon dioxide if long-term temperature stabilisation is the goal.

Adaptation changes the damage

Even under strong mitigation, some further warming and many climate effects are unavoidable because of past emissions and system inertia. Adaptation is therefore not an admission that mitigation failed. It is the second half of climate risk management.

A useful adaptation begins by naming the failure mode. Heat kills through physiology and through infrastructure failure, so warnings, cooling, labour rules, urban design and reliable electricity can reduce risk. Flood risk depends on water level, drainage, buildings and land use, so protection can combine barriers, retention areas, building standards and retreat. Drought adaptation can include storage, efficient irrigation, crop choice, groundwater management and demand rules.

Maladaptation occurs when a response reduces one risk while creating another. Air conditioning powered by high-emissions electricity can increase future forcing. Hard sea defences can encourage more development behind them or shift erosion. Irrigation can deplete groundwater. Insurance can preserve incentives to rebuild repeatedly in exposed locations. Adaptation therefore requires feedback thinking too.

There are hard and soft limits. A hard limit means no feasible adaptation can avoid intolerable risk under given conditions, for example where an ecosystem loses the climatic conditions it requires. A soft limit means options exist in principle but finance, governance, technology or social constraints prevent their use. The distinction matters because one calls for more capability and the other for less hazard.

The decision under uncertainty

Climate decisions cannot wait for every uncertainty to disappear because emissions and infrastructure choices are being made while the uncertainty persists. The correct question is therefore not whether scientists know the exact temperature in every region in 2080. It is whether uncertainty changes the ranking of available actions.

Some uncertainties cut both ways. Higher climate sensitivity makes a given emissions path worse; lower sensitivity makes it better. Faster ice-sheet loss raises coastal risk; slower loss buys time. If the response is irreversible or expensive to reverse, uncertainty can increase the value of avoiding the high-risk tail rather than justify inaction.

Other uncertainties concern technology costs, political cooperation, economic growth and social preferences. These are not climate-model errors. They are uncertainties about the human system. Keeping them separate prevents science from being blamed for questions it never claimed to answer.

The modern climate problem can now be stated without drama. Humans changed atmospheric composition. That created positive radiative forcing. The Earth system warmed and stored excess heat, mostly in the ocean. Feedbacks amplified the response. The warmer baseline changed hazards. Exposure and vulnerability converted those hazards into unequal risks. Future emissions determine additional forcing; adaptation determines part of the resulting harm.

How we know

The physical account rests on several independent forms of evidence. Laboratory and satellite spectroscopy establish greenhouse-gas absorption and changes in outgoing radiation. Atmospheric measurements, carbon isotopes, oxygen decline and emissions inventories identify the source of the carbon dioxide rise. Surface records, ocean heat, sea level, glaciers, ice sheets and satellite observations document a warming system. Detection and attribution studies test observed spatial and vertical fingerprints against greenhouse gases, aerosols, solar variability, volcanoes and internal variability. Paleoclimate evidence, the instrumental record, energy-budget constraints and climate models jointly inform sensitivity.

Important uncertainties remain in cloud feedback, regional precipitation, ice-sheet dynamics, carbon-cycle responses, compound extremes and adaptation limits. Models resolve large-scale circulation better than local weather and cannot predict future policy choices. Annual global temperature also contains natural variability, so a single year above a threshold is not the same as a sustained multi-decadal crossing. None of these uncertainties overturns the central attribution of recent global warming to human influence.

What People Get Wrong

“Climate has always changed, so this warming is natural”

The first half is true. Climate has changed throughout Earth's history. The second half does not follow.

Past climate change gives scientists mechanisms to test. Orbital cycles pace ice ages over tens of thousands of years. Large volcanic eruptions cool the planet temporarily. Solar output changes. Continents move. Greenhouse-gas concentrations rise and fall through geological and biological processes. The fact that several mechanisms exist means the modern change has to be attributed, not assumed.

Modern observations identify the forcing. Atmospheric carbon dioxide has risen sharply from fossil-fuel and land-use emissions. The isotopic composition of atmospheric carbon shifts as expected from fossil carbon. The troposphere warms while the stratosphere cools. Solar measurements do not show a trend capable of producing the observed warming. Natural-only climate simulations fail to reproduce the modern rise.

Past climate change therefore strengthens rather than weakens the greenhouse case. Ice cores and other paleoclimate records show that greenhouse gases and temperature are tightly linked through forcing and feedback. Sometimes temperature changes first and carbon dioxide follows as a feedback, especially during orbital transitions. That does not imply CO2 cannot also initiate warming when humans add it first. A substance can be both cause and feedback depending on what starts the sequence.

The correction matters because “natural versus human” is not a philosophical choice. It is a causal question with fingerprints. Modern warming contains the fingerprints expected from human greenhouse forcing.

“CO2 is only a tiny part of the atmosphere, so it cannot matter much”

Concentration alone does not determine physical importance. Ozone is present in trace amounts and blocks damaging ultraviolet radiation. Tiny quantities of pigment can change the colour of water. A gas matters because of what wavelengths it absorbs, where it sits in the atmosphere and how its concentration changes the energy budget.

Carbon dioxide is measured in parts per million because its concentration is small compared with nitrogen and oxygen. Nitrogen and oxygen make up most of the atmosphere but interact weakly with infrared radiation at the wavelengths relevant to Earth's surface temperature. Carbon dioxide has molecular vibrational modes that absorb infrared strongly.

The natural greenhouse effect depends heavily on water vapour and clouds, but water vapour responds quickly to temperature. Add carbon dioxide and the initial radiative forcing warms the air; warmer air supports more water vapour; the added water vapour amplifies the warming. Remove much of the CO2 and the planet cools, reducing water vapour as well. The gases do different jobs in the feedback system.

The right question is therefore not what percentage of the atmosphere CO2 occupies. It is how changing that concentration alters outgoing infrared radiation. Spectroscopy and satellite observations answer that directly.

“Scientists rely on computer models because the observations are weak”

Models are indispensable for projecting conditional futures and separating forcings, but the observed climate signal does not depend on them.

Atmospheric carbon dioxide is measured directly. Infrared absorption is measured in laboratories and from satellites. Surface temperature is measured by thermometers and ocean instruments. Ocean heat is measured through ship observations and floats. Sea level is measured by tide gauges and satellites. Ice mass is measured using field observations and satellite gravimetry. These are observations before they become model inputs.

Models then answer questions observations alone cannot. What would the twentieth century have looked like with natural forcings but no human greenhouse gases? What range of regional rainfall change follows a high-emissions pathway? How does ocean circulation alter the timing of warming? Those are counterfactual or future questions, so some model is unavoidable.

Models are strongest where they converge with observations, simple theory and paleoclimate constraints. That convergence is far stronger for human-caused global warming than for precise local rainfall decades ahead. Rejecting climate science as “just models” discards the instruments that already show the planet's energy imbalance.

“One cold winter disproves global warming”

Weather is a particular state of the atmosphere. Climate is the statistical distribution of weather over longer periods. A warming climate can still produce cold days, cold snaps and snowy winters.

Raise the average of a distribution and the cold tail does not vanish instantly. It becomes less common while the hot tail becomes more common. Regional circulation can still deliver Arctic air into populated mid-latitudes. Snowfall can remain heavy where air is cold enough, and a warmer atmosphere can contain more moisture. Local experience therefore remains noisy even while the global distribution shifts.

The reverse mistake also occurs. One heatwave does not, by itself, prove the whole climate trend. Event attribution asks how climate change altered the event's probability or intensity compared with a counterfactual world without the human forcing. For heat extremes the signal is often strong. For other events the answer is more conditional.

This boundary protects both Weather in a Hurry and this book. Weather explains why next Tuesday develops as it does. Climate explains why the odds from which Tuesdays are drawn are changing.

“If climate sensitivity is uncertain, scientists do not know how serious the problem is”

Uncertainty is not ignorance. It is a probability range constrained by evidence.

The direct radiative effect of doubling carbon dioxide is well established. The larger uncertainty comes from feedbacks, especially clouds and some slower Earth-system responses. IPCC AR6 assessed equilibrium climate sensitivity at a best estimate near 3°C, with a likely range of 2.5°C to 4°C. Values outside that range are not impossible, but they are less supported by the combined evidence.

For decisions, the range matters because high sensitivity produces more warming and risk for the same emissions. But the existence of a range does not make emissions irrelevant. Across the plausible range, lower cumulative carbon dioxide emissions produce less long-term warming than higher cumulative emissions.

Many engineering and financial decisions are made under wider uncertainty than this. Nobody argues that an uncertain fire probability proves a building cannot burn. The rational question is how sensitive the decision is to the range and what the expensive tail looks like.

Climate uncertainty is often treated backwards. Uncertainty about severity can strengthen the case for risk management where outcomes are irreversible rather than justify assuming the mildest case.

“Net zero means zero emissions everywhere”

Net zero carbon dioxide means anthropogenic CO2 emissions are balanced by anthropogenic CO2 removals over a stated system and period. It does not require every source to reach literal zero.

Some emissions are easier to eliminate than others. Electricity can be produced without direct carbon dioxide emissions. Some industrial process emissions, aviation, agriculture and dispersed sources are harder. A net-zero system can retain residual emissions if durable removals balance them.

That does not make removals a licence to continue all current emissions. Removal capacity is limited by cost, energy, land, permanence and scale. Forest carbon can be reversed by fire or clearing. Soil carbon can saturate. Direct air capture requires substantial energy and infrastructure. Geological storage requires monitoring. The more residual emissions remain, the larger and harder the balancing system becomes.

There is also a separate concept of net zero greenhouse-gas emissions, which balances warming and cooling effects across gases using an accounting framework. Because methane and carbon dioxide have different lifetimes, simple tonne-for-tonne comparisons can obscure temperature effects.

The correction matters because the physical stopping rule concerns the stock of long-lived carbon dioxide. Net zero is an accounting description of how to stop adding to that stock, not a claim that all human activity becomes emission-free.

“Once 1.5°C is exceeded, there is no point doing anything”

A temperature target can be missed without turning physics into a cliff.

Climate risks generally rise with additional warming. Some increase gradually; some rise faster after thresholds; some changes can become irreversible on human timescales. None of that creates a universal boundary at which 1.6°C and 3°C become equivalent.

The Paris Agreement's 1.5°C goal concerns limiting long-term warming, not whether one calendar year temporarily exceeds 1.5°C above the nineteenth-century baseline. Natural variability can push an individual year above or below the underlying trend. WMO's recent annual observations show how close the world is to sustained 1.5°C warming, but the policy target is assessed over longer periods.

If sustained 1.5°C is exceeded, lower emissions still reduce the peak and eventual warming. Mitigation can limit overshoot. Carbon dioxide removal could, in principle, lower atmospheric CO2 later, though large-scale removal carries cost and feasibility limits. Adaptation can still reduce harm at every warming level.

The correction matters because fatalism and denial make the same physical mistake. Both pretend choices no longer affect outcomes. Cumulative emissions ensure that choices continue to matter until net CO2 emissions reach zero, and adaptation continues to matter after that.

Use It

Separate mechanism, observation, attribution and projection

When you hear a climate claim, first ask what kind of claim it is.

A mechanism claim says how a process works: carbon dioxide absorbs infrared radiation. An observation claim says what has been measured: ocean heat content has risen. An attribution claim assigns causes: human greenhouse gases are the dominant cause of recent global warming. A projection claim is conditional: under a specified emissions pathway, future warming is expected to fall within a range.

These categories have different evidence and different uncertainty. Confusing them creates bad arguments. A model error in regional rainfall does not erase laboratory spectroscopy. A cold year does not falsify the long-term attribution. A robust attribution does not tell you exactly which emissions policy to choose.

Use the distinction anywhere evidence and forecasts are mixed. Ask which layer is under dispute before deciding whether the whole claim rises or falls. A newspaper headline that says “scientists predict” may be reporting an observation, an attribution study, a scenario result or an economic assumption. Those are not interchangeable claims.

Look for the stock and the flow

Climate becomes easier once you ask whether the variable accumulates.

Carbon dioxide is a stock problem because emissions add to an atmospheric perturbation that natural sinks remove only partly and over several timescales. Annual emissions are the flow. Temperature stabilisation therefore requires the net flow into the long-lived stock to fall to roughly zero.

Methane is different because chemical removal is much faster. Aerosols are different again. This is why “a tonne is a tonne” can fail across pollutants and why cutting different gases produces different temperature responses over time.

The lens transfers beyond climate. Debt, reservoirs, inventories, infectious prevalence and pension liabilities all behave differently when a flow accumulates into a stock. Before reacting to the latest annual number, ask what level has already built up and what rate would stabilise it.

Read extremes as changed odds, not isolated anecdotes

A flood, wildfire or heatwave can be used badly by both sides of a climate argument. One event does not establish a global trend, and a global trend does not mean climate change created every event from nothing.

Instead ask how the probability distribution changed. Was the event made more likely, more intense or longer-lasting? Is the mechanism clear? Is the attribution robust for this hazard and region? What role did exposure, land management, drainage, ignition or building quality play?

This lens avoids two symmetrical errors: declaring every disaster proof of climate change and declaring every non-climate contribution evidence against it. Causation can be multiple. Climate can increase the hazard while human development determines the loss.

Ask what uncertainty would change the decision

A range is useful only if it affects what you do.

Suppose a coastal asset lasts eighty years. The exact sea-level rise by 2100 is uncertain. If every plausible range requires raising electrical equipment and improving drainage, those measures are robust. If only the high-end range justifies a large sea wall, the decision depends on risk tolerance, reversibility and the cost of waiting.

Climate decisions often benefit from this robustness test. Do not demand one precise future number when several futures lead to the same sensible action. Conversely, do not hide a choice behind “the science” when different values, discount rates or risk tolerances would produce different policies from the same evidence.

This is the boundary between scientific uncertainty and political judgement. Science can narrow the range. It cannot decide how much risk a society should accept. A good decision record should therefore state both the evidence range and the value judgement used to act within it. That makes disagreement inspectable rather than rhetorical.

Distinguish stopping the cause from reducing the harm

Mitigation and adaptation answer different questions.

Mitigation asks how to reduce the future forcing by cutting emissions or increasing removals. Adaptation asks how to reduce damage from the climate that occurs. A flood barrier can protect a city without changing global temperature. Closing a coal plant can reduce future warming without protecting a particular street from next winter's flood.

The distinction reveals weak proposals. An adaptation plan that assumes hazards can keep rising indefinitely without limits is incomplete. A mitigation plan that ignores present-day heat, water and coastal risk is incomplete. Strong climate strategy does both because one acts on cause and the other on consequence.

Follow the denominator

Many climate arguments depend on a denominator that disappears in rhetoric.

“Renewables doubled” means something different if they doubled from 2 per cent to 4 per cent than from 30 per cent to 60 per cent. “A country cut emissions” can mean territorial emissions fell while imported emissions rose. “A disaster cost more” can reflect greater hazard, greater exposure, inflation or all three. “Deaths fell” can reflect adaptation even while heat hazard increased.

Ask: per person, per unit of energy, per unit of GDP, per square kilometre, relative to which baseline, and measured where? Climate is full of large numbers. Denominators turn them into information.

The limits

Climate science cannot tell you the single correct policy. It can estimate physical responses, risks and the consequences of emissions pathways. It cannot derive a fair burden-sharing rule from radiation physics. Questions about who pays, how quickly capital should turn over, which landscapes should host infrastructure, how much present consumption should be traded for future risk reduction, and how responsibilities should be divided are political and ethical as well as technical.

Models also remain weaker at some scales and processes. Regional precipitation, clouds, ice-sheet dynamics, compound extremes and local ecosystem responses carry larger uncertainty than the sign of global greenhouse warming. Economic damage estimates add another layer of assumptions about growth, adaptation, discounting and non-market loss. Treating one integrated damage number as a physical constant is a category error.

There are limits to adaptation. Wealth and technology can reduce many risks, but they cannot guarantee preservation of every coastline, glacier, coral reef or climate-sensitive ecosystem. Nor does every mitigation technology scale without land, mineral, institutional or social constraints. A technically possible pathway is not the same as a politically delivered one.

The final limit is psychological. Climate invites both certainty theatre and despair. The evidence supports neither. The mechanism is strong enough to say human emissions are warming the planet. The future is open enough that emissions and adaptation choices still materially change outcomes. Holding both facts at once is harder than choosing a tribe, and more useful.

The one thing to keep

Keep the balance sheet.

Sunlight comes in. Heat must go out. Greenhouse gases change that exchange. Carbon dioxide accumulates because human additions exceed the rate at which natural sinks remove the perturbation. The planet stores the resulting excess energy, mostly in the ocean, until warming restores radiative balance. Feedbacks decide how much warming is required. Geography and vulnerability decide where the damage lands.

That sequence lets you test almost any climate claim without memorising a debate. Ask what changes the energy budget. Ask whether the pollutant accumulates. Ask what has been observed independently. Ask whether the proposed cause leaves the right fingerprint. Ask which uncertainty matters to the decision. Ask whether a response changes the forcing or only the damage.

The climate problem is large because the atmosphere is shared and the carbon stock is cumulative. It is tractable because the physics supplies a stopping rule. Net carbon dioxide emissions near zero stop adding further CO2-driven warming. Every avoided tonne lowers the eventual forcing compared with emitting it. Every fraction of a degree changes some risks. Adaptation can reduce harm even while the physical system continues responding.

You do not need to believe in climate change as an identity. You need to follow the accounting. The atmosphere does not vote, exaggerate or compromise. It responds to molecules, radiation and time.

Terms

Albedo. The fraction of incoming sunlight reflected by a surface or the planet as a whole. Snow and ice have high albedo; open ocean is darker. Changes in albedo can amplify warming or cooling.

Anthropogenic. Caused by human activity. In climate science it distinguishes human influences such as greenhouse-gas emissions and aerosols from natural influences such as volcanoes and solar variability.

Attribution. The assessment of the causes of an observed change. Climate attribution compares expected fingerprints and model counterfactuals to estimate the contribution of human and natural drivers.

Carbon budget. The cumulative amount of carbon dioxide that can be emitted while retaining a stated probability of limiting warming to a chosen level. It is an estimate with uncertainty, not a precise countdown clock.

Carbon cycle. The movement of carbon among atmosphere, ocean, vegetation, soils and rocks. Human emissions add carbon to this natural exchange faster than sinks remove the added perturbation.

Carbon dioxide removal. Human activity that takes CO2 from the atmosphere and stores it durably, through biological, chemical or engineered methods. Removal is distinct from avoiding an emission in the first place.

Climate. The statistical character of weather over time, including averages, variability and extremes. Climate can change even though individual days remain noisy.

Climate sensitivity. The amount of global warming produced by a specified forcing, commonly expressed as the long-term warming after carbon dioxide doubles.

Cloud feedback. The change in Earth's energy balance caused by warming-induced changes in cloud amount, altitude, type or optical properties. It is positive overall in current assessments but remains an important uncertainty.

CO2 equivalent. A way to express the climate effect of different greenhouse gases in a common unit using a chosen metric and time horizon. Useful for accounting, but capable of obscuring differences in gas lifetimes.

Equilibrium climate sensitivity. The long-term global mean surface warming after atmospheric CO2 doubles and the climate approaches equilibrium with fast feedbacks operating. AR6 gives a best estimate near 3°C.

Exposure. The presence of people, ecosystems or assets in places that can be affected by a climate hazard. Risk can rise because exposure rises even if the hazard itself does not.

Feedback. A response that alters the original climate change. Water vapour and ice-albedo feedbacks amplify warming; the increase in infrared emission from a warmer Earth stabilises it.

Forcing. A change in Earth's energy balance imposed on the climate system, such as increased greenhouse gases, volcanic aerosols or a change in solar output.

Global mean surface temperature. An estimate of average temperature change across land and ocean. It is an index of the climate system, not the temperature experienced everywhere.

Greenhouse effect. Warming caused because gases and clouds absorb and emit infrared radiation, changing the altitude and temperature from which Earth effectively loses heat to space.

Greenhouse gas. A gas that absorbs and emits infrared radiation at relevant wavelengths. Important examples include water vapour, carbon dioxide, methane, nitrous oxide and fluorinated gases.

Hazard. A physical event or trend capable of causing harm, such as extreme heat, flood, drought or sea-level rise. Hazard becomes risk through exposure and vulnerability.

Internal variability. Climate fluctuations generated within the coupled atmosphere-ocean system without a change in external forcing. El Niño and La Niña are major examples.

Maladaptation. An adaptation response that increases vulnerability, shifts risk elsewhere or worsens future climate risk, such as protection that encourages more development in an exposed zone.

Methane. A potent greenhouse gas with an atmospheric lifetime of roughly a decade. Cutting sustained methane emissions can reduce its warming contribution faster than comparable cuts in long-lived CO2 affect the carbon stock.

Mitigation. Action that reduces the causes of climate change by cutting greenhouse-gas sources or increasing removals. It changes future forcing rather than merely protecting against impacts.

Net zero CO2. A state in which anthropogenic carbon dioxide emissions are balanced by anthropogenic removals. Reaching global net zero CO2 approximately stabilises CO2-driven global temperature.

Ocean acidification. The decline in seawater pH caused mainly by uptake of anthropogenic carbon dioxide. The ocean remains alkaline, but its carbonate chemistry shifts in ways that affect marine organisms.

Radiative forcing. The change in the planetary energy budget caused by a driver before the full surface-temperature response. Positive forcing warms; negative forcing cools.

Representative concentration pathway and SSP scenario. Structured assumptions used to explore future greenhouse-gas concentrations, emissions and socioeconomic conditions. They are conditional futures, not forecasts that claim to know what society will do.

Risk. The potential for adverse consequences arising from interactions among climate hazards, exposure and vulnerability. The same hazard can create radically different risk in different societies.

Tipping element. A part of the Earth system that may undergo large, self-reinforcing change after a threshold is crossed. Confidence, threshold location and timescale differ among proposed elements.

Transient climate response. The global warming around the time carbon dioxide doubles under a steadily rising concentration. It is lower than equilibrium sensitivity because the deep ocean has not fully warmed.

Vulnerability. The propensity of people, ecosystems or assets to be harmed by a hazard, shaped by sensitivity, resources, institutions, health, infrastructure and capacity to adapt. It explains why equal hazards can produce unequal losses.

Go Deeper

Andrew E. Dessler, Introduction to Modern Climate Change, 3rd edition

Cambridge University Press, 2021. Start here for the science with enough mathematics to make the mechanisms explicit without turning the subject into atmospheric physics training. Dessler covers observations, energy balance, the carbon cycle, forcing, feedbacks, attribution, projections, impacts and policy in a compact sequence. The equations are elementary and can be skipped on a first pass without losing the argument. It is especially good on the distinction between forcing and feedback, and on why policy conclusions require assumptions beyond the physical science.

Intergovernmental Panel on Climate Change, Climate Change 2021: The Physical Science Basis

Cambridge University Press, 2021. This is the main assessment behind the physical claims in this book. Do not read it front to back. Read the Summary for Policymakers, Technical Summary and the relevant Frequently Asked Questions, then use chapters as reference. Its strength is synthesis: hundreds of authors assess thousands of studies and state both confidence and uncertainty explicitly. The summaries are dense, but they let you see which claims are assessed as high confidence and which remain structurally uncertain. Use the interactive atlas for regional questions, but resist reading one model cell as a local forecast.

Spencer R. Weart, The Discovery of Global Warming, revised and expanded edition

Harvard University Press, 2008, with the author's online history subsequently updated. Read this for how the science developed. It shows that modern climate understanding did not arrive from one model or political campaign but from spectroscopy, oceanography, ice cores, radiation measurements, computing and repeated argument across disciplines. The historical detail is especially useful for seeing old objections in their original scientific context. It is also a corrective to the idea that climate science appeared fully formed in the late twentieth century.

Raymond T. Pierrehumbert, Principles of Planetary Climate

Cambridge University Press, 2010. This is the demanding option. It develops radiation, thermodynamics, atmospheric structure and climate from first principles and applies them across Earth and other planets. The mathematics is substantial. Read it when you want to understand why the greenhouse effect behaves as it does rather than accepting a verbal summary. Its planetary comparisons are useful because they show which parts of climate physics are general and which depend on Earth's particular atmosphere, oceans and orbit. It is the book to open when a verbal analogy starts feeling too easy and you want the physics underneath it.

Notes and Sources

The Whole Thing in One Page and Why You Should Care

The central physical account follows IPCC Working Group I, especially the Summary for Policymakers, Technical Summary and chapters on Earth's energy budget, the carbon cycle, attribution and future climate. IPCC AR6 states that human influence has unequivocally warmed the atmosphere, ocean and land. Its assessed warming for 2011 to 2020 relative to 1850 to 1900 is about 1.1°C.

The current annual climate state was checked against the World Meteorological Organization's State of the Global Climate 2025, published 23 March 2026. WMO assesses 2025 as the second or third warmest year on record at about 1.43°C above the 1850 to 1900 average and reports record ocean heat content. Annual temperature above 1.5°C is distinguished from the Paris Agreement's long-term threshold.

NOAA Global Monitoring Laboratory's Mauna Loa record was checked on 11 August 2026. NOAA reports a June 2026 monthly mean of 431.44 ppm, compared with 429.61 ppm in June 2025. The manuscript uses “above 430 ppm” rather than treating one monthly value as a permanent atmospheric level.

Core Idea 1: energy balance and greenhouse effect

The radiative mechanism, forcing framework and feedback terminology follow IPCC WGI AR6 and Andrew Dessler's Introduction to Modern Climate Change. The effective-emission-height explanation is a reader-facing simplification of line-by-line radiative transfer, not a claim that all wavelengths escape from one physical altitude.

Historical references to Fourier, Tyndall and Arrhenius were checked against Spencer Weart's history of climate science and standard histories cited in IPCC assessments. The manuscript avoids claiming that any one of them possessed the modern greenhouse model in full.

Core Idea 2: carbon cycle and cumulative emissions

The carbon-cycle account follows IPCC WGI AR6 and the Global Carbon Project. The Global Carbon Budget 2025 projects fossil CO2 emissions of 38.1 GtCO2 in 2025, a record high, and provides current estimates for atmospheric growth and land and ocean sinks. The narrative avoids locking the airborne fraction to a fixed percentage.

The close relationship between cumulative CO2 emissions and warming, and the implication that global net-zero CO2 approximately halts further CO2-driven warming, follows IPCC WGI and the AR6 Synthesis Report. Residual Earth-system responses, especially sea-level rise, continue after temperature stabilisation.

Core Idea 3: observations

Global temperature, ocean heat, sea-level, glacier and ice-sheet evidence follows IPCC WGI AR6, WMO 2025 and the major observational datasets assessed by those reports. The manuscript uses ocean heat as an independent indicator of planetary energy imbalance and does not treat any single surface-temperature dataset as uniquely authoritative.

Core Idea 4: attribution

Detection and attribution claims follow IPCC WGI AR6. The estimate that human influence caused about 1.07°C of warming in 2010 to 2019 relative to 1850 to 1900, with natural drivers near zero net contribution over that period, is drawn from the Working Group I Summary for Policymakers. Human greenhouse-gas warming is partly offset by cooling from human aerosols.

Event-attribution language is deliberately probabilistic. Confidence is generally strongest for heat extremes and varies for precipitation, drought, tropical cyclones and wildfire conditions by region and metric.

Core Idea 5: feedbacks and sensitivity

Equilibrium climate sensitivity is taken from IPCC WGI AR6: best estimate 3°C, likely range 2.5°C to 4°C. Feedback descriptions follow the WGI chapter on Earth's energy budget, feedbacks and climate sensitivity. Tipping elements are presented as heterogeneous risks with different confidence, thresholds and timescales rather than one global cliff.

Core Idea 6: models and scenarios

The description of climate models, parameterisation, ensembles, internal variability and scenario uncertainty follows IPCC WGI AR6. The historical model-performance point is consistent with peer-reviewed retrospective evaluations that compare modelled warming against observations after accounting for the difference between projected and realised forcing. The body keeps the claim qualitative.

Core Idea 7: impacts, risk, mitigation and adaptation

Risk framing as hazard, exposure and vulnerability follows IPCC Working Group II. Statements on heat, heavy precipitation, drought, tropical cyclones, wildfire conditions, ecosystems, agriculture and adaptation are scoped to assessed levels of confidence rather than universalised.

Mitigation claims follow IPCC Working Group III and the AR6 Synthesis Report. The manuscript leaves detailed energy technology, coal and oil histories to neighbouring titles and keeps only the mechanisms needed to explain emissions pathways, net zero and removals.

UNEP's Emissions Gap Report 2025 was checked for current pathway context. It estimates warming of roughly 2.3°C to 2.5°C this century if the 2025 round of nationally determined contributions is fully implemented, and about 2.8°C under current policies. These are scenario-dependent policy projections, so the body does not present them as a forecast of what will occur.

UNEP's Adaptation Gap Report 2025 estimates developing-country adaptation finance needs of about US$310 billion to US$365 billion per year by 2035, compared with US$26 billion in international public adaptation finance in 2023. The figures inform the production record but were omitted from the narrative because the book owns climate science rather than climate-finance policy.

What People Get Wrong

The seven corrections draw on the same IPCC physical science, observations and risk assessments. The water-vapour correction distinguishes fast feedback from imposed forcing. The model correction separates observed evidence from counterfactual and projection uses. The weather correction preserves the boundary with Weather in a Hurry. The net-zero correction distinguishes physical temperature stabilisation from accounting claims. The 1.5°C correction follows WMO and IPCC treatment of annual variability versus long-term warming levels.

Bibliography

Assessments, observations and original data

Global Carbon Project. Global Carbon Budget 2025. 2025.

Intergovernmental Panel on Climate Change. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Edited by Valérie Masson-Delmotte et al. Cambridge: Cambridge University Press, 2021.

Intergovernmental Panel on Climate Change. Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Edited by Hans-Otto Pörtner et al. Cambridge: Cambridge University Press, 2022.

Intergovernmental Panel on Climate Change. Climate Change 2022: Mitigation of Climate Change. Contribution of Working Group III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Edited by Priyadarshi R. Shukla et al. Cambridge: Cambridge University Press, 2022.

Intergovernmental Panel on Climate Change. Climate Change 2023: Synthesis Report. Core Writing Team, Hoesung Lee and José Romero, eds. Geneva: IPCC, 2023.

National Oceanic and Atmospheric Administration, Global Monitoring Laboratory. Trends in Atmospheric Carbon Dioxide. Boulder, Colorado. Data checked 11 August 2026.

United Nations Environment Programme. Adaptation Gap Report 2025: Running on Empty. Nairobi: UNEP, 2025.

United Nations Environment Programme. Emissions Gap Report 2025: Off Target. Nairobi: UNEP, 2025.

World Meteorological Organization. State of the Global Climate 2025. Geneva: WMO, 2026.

Modern works

Dessler, Andrew E. Introduction to Modern Climate Change. 3rd ed. Cambridge: Cambridge University Press, 2021.

Pierrehumbert, Raymond T. Principles of Planetary Climate. Cambridge: Cambridge University Press, 2010.

Weart, Spencer R. The Discovery of Global Warming. Revised and expanded ed. Cambridge, MA: Harvard University Press, 2008.