That is why climate scientists use two related but distinct measures: equilibrium climate sensitivity and transient climate response.
The difference is not a technical footnote. It separates the warming that unfolds over the coming decades from the larger response that remains locked into the climate system for centuries. In the language of climate policy, the distinction affects how quickly risks accumulate. In the language of physical geography, it describes where the heat is held, how fast it moves, and which feedbacks have had time to act.
The central question in equilibrium climate sensitivity vs transient climate response is therefore not which number is correct. Both measure real properties of the climate system. They answer different questions.
Two metrics, two clocks
The benchmark begins with a familiar reference point: the approximate pre-industrial atmospheric carbon dioxide concentration of 280 parts per million. Doubling that concentration brings it to roughly 560 parts per million. Climate models then estimate how the global mean surface temperature responds to the additional radiative forcing.
The word equilibrium is decisive. ECS describes the eventual warming after the atmosphere, land surface, ocean, and major climate feedbacks have had time to adjust toward thermal balance. That adjustment takes centuries, not a human planning cycle.
TCR is deliberately closer to the pace of industrial change. It is measured in idealized model experiments where carbon dioxide rises by 1% per year. At that rate, the concentration doubles in approximately 70 years. Scientists record the global mean surface warming around the point of doubling, while the deep ocean is still absorbing heat and the wider climate system remains out of equilibrium.
The two metrics can be set beside each other:
| Parameter | Equilibrium Climate Sensitivity (ECS) | Transient Climate Response (TCR) |
|---|---|---|
| Main question | How much does the surface eventually warm after CO₂ doubles? | How much has the surface warmed when CO₂ doubles during a gradual increase? |
| Timescale | Multi-century adjustment toward equilibrium | Approximately 70 years in the idealized 1% annual CO₂ experiment |
| Ocean state | Deep ocean and upper ocean have had time to approach thermal equilibrium | Deep ocean continues taking up substantial heat |
| IPCC AR6 likely range | 2.5°C to 4.0°C | 1.4°C to 2.2°C |
| IPCC AR6 best estimate | 3.0°C | 1.8°C |
| Why it matters | Long-term warming commitment and climate-system feedbacks | Near-term warming under continued forcing |
The gap between the ranges is not uncertainty in the ordinary sense. It is largely a consequence of heat distribution. The ocean does not respond as a single body. Surface waters exchange energy with the atmosphere relatively quickly, while the abyssal layers are connected to the surface through slower circulation, mixing, and water-mass formation.
TCR is the warming the surface reveals while the ocean is still taking the heat. ECS is the larger response that emerges after the climate system has had time to rearrange itself.
This distinction is essential when reading climate projections. A lower near-term temperature response does not mean that the underlying forcing has produced a small or temporary disturbance. It may mean that part of the energy has moved below the surface.
The transient response: a climate system in motion
The transient climate response definition is often reduced to a number measured at the moment atmospheric CO₂ doubles. The experiment behind it is more revealing than the number alone.
In the standard idealized run, carbon dioxide increases by 1% annually. The rise is smooth rather than abrupt, and it continues long enough for the concentration to double after about seven decades. At that point, the model reports the change in global mean surface temperature relative to the baseline.
The atmosphere responds rapidly to increased radiative forcing. Land surfaces warm, atmospheric circulation shifts, snow cover changes, and sea ice responds. But the surface is not the whole climate system. The ocean absorbs energy through contact with the atmosphere and redistributes it through currents and mixing. Heat can therefore enter the climate system without appearing immediately as an equivalent rise in global surface temperature.
This is why TCR is generally lower than ECS. During the transient experiment, the deep ocean acts as a brake on surface warming. It does not erase the energy imbalance. It changes where the energy resides and delays part of the temperature response.
That delay has a geographic expression. Tropical upper-ocean waters can warm while deeper layers remain cooler. High-latitude oceans may continue to absorb heat as circulation carries surface water downward. Sea ice loss can alter albedo, exposing darker ocean surfaces that absorb more sunlight. Over land, drying soils and vegetation stress can affect the partition between sensible heat and evaporation. These processes do not move in lockstep, and a global average conceals much of that unevenness.
TCR is therefore useful for understanding the warming trajectory relevant to the next several decades. It can inform assessments of heat extremes, shifting precipitation, glacier mass loss, wildfire weather, and ecosystem stress during a period when emissions and atmospheric concentrations are still changing.
But TCR should not be mistaken for a ceiling. The ocean’s uptake of heat postpones part of the surface response; it does not cancel it.
What climate models are actually calculating
The basic physics begins with radiative forcing: a change in the balance between energy entering the Earth system and energy leaving it. A rise in atmospheric carbon dioxide makes it more difficult for some outgoing infrared radiation to escape directly to space. The climate system warms until outgoing energy increases enough to restore balance, though feedbacks modify the size of that response.
A simplified sensitivity relationship is often expressed as the ratio between equilibrium warming and the forcing associated with doubled CO₂. In practical model experiments, however, climate sensitivity is not calculated from one isolated atmospheric property. It emerges from interactions among:
- water vapour, which amplifies warming because warmer air can hold more moisture;
- snow and sea ice, whose loss reduces the reflection of incoming sunlight;
- clouds, which can either reflect sunlight or trap outgoing infrared radiation depending on their altitude, structure, and location;
- lapse-rate changes, describing how temperature varies with height through the atmosphere;
- ocean heat uptake, which shapes the pace at which surface temperature rises;
- aerosols, which can cool the climate by reflecting sunlight or modifying clouds.
The uncertainty is concentrated in these feedbacks rather than in the fact that carbon dioxide absorbs infrared radiation. Clouds remain particularly consequential because their effects vary across short distances and timescales. A model may reproduce broad circulation patterns yet still represent low clouds, convective cloud systems, or cloud responses to warming imperfectly.
That is one reason the phrase ECS vs TCR climate models needs care. Models do not simply contain two competing sensitivity settings. They simulate a climate system whose feedbacks operate at different speeds and whose representation differs among model generations.
ECS and the long horizon of thermal equilibrium
Equilibrium climate sensitivity asks what happens after the climate system has had time to settle following a sustained doubling of atmospheric CO₂. In the classic definition, the oceans and major fast feedbacks have approached equilibrium with the altered atmospheric composition.
That horizon is long because the ocean is stratified. The upper ocean exchanges heat with the atmosphere, but the deep ocean is separated by density structure and circulation patterns that slow the transfer. The climate system can remain energetically out of balance even after the surface has experienced decades of warming.
ECS also allows more feedbacks to express themselves. Snow and sea ice may retreat further. Atmospheric moisture patterns can reorganize. Vegetation and land-surface properties may shift. Some slow components of the Earth system, including large ice sheets and long-lived carbon-cycle processes, may continue changing beyond the standard multi-century definition of ECS. Their exact contribution is not fully captured by the conventional metric.
This matters because ECS is not a forecast of the temperature reached on a particular date. It is a measure of the eventual surface response to a defined forcing level. It provides a way to compare the strength of the climate system’s feedbacks across models and observational constraints.
The IPCC Sixth Assessment Report assessed the likely range of ECS at 2.5°C to 4.0°C, with a best estimate of 3.0°C. For TCR, the assessed likely range is 1.4°C to 2.2°C, with a best estimate of 1.8°C. The report also gives a very likely TCR range of 1.2°C to 2.4°C.
These ranges are not promises, and they are not symmetrical descriptions of every possible future. They express assessed probabilities under the evidence available, including instrumental observations, paleoclimate records, physical understanding, and climate-model behaviour.
The narrowing of the ECS range in AR6 was scientifically important. Earlier assessments gave a wider likely range. The newer range does not mean that uncertainty has vanished; it means that several lines of evidence now place stronger constraints on the most plausible sensitivity values.
Why the deep ocean creates the gap
The difference between ECS and TCR is sometimes described as a delay. That is accurate, but incomplete. The deep ocean is not merely waiting for the atmosphere to catch up. It is actively redistributing energy through a vast, layered circulation system.
At the surface, warming can be moderated when heat passes into seawater. The mixed layer responds on relatively short timescales, while deeper water masses absorb energy more slowly. This produces a vertical temperature structure in which the surface response is smaller than the eventual equilibrium response for the same atmospheric forcing.
The ocean also transports heat horizontally. Currents move warm and cold water across basins, connecting tropical, temperate, and polar regions. The result is not a uniform global warming field but a patchwork of regional responses. Some ocean areas warm rapidly; others remain comparatively cool at the surface while accumulating heat below.
The ecological consequences follow the same geography. A reef, estuary, kelp forest, or polar marine food web responds to local temperature, oxygen, stratification, acidity, and circulation—not to the global mean alone. TCR can help describe the pace of broad warming, but local ecosystems experience the combination of transient surface conditions and longer-lived ocean changes.
The delay can also complicate public interpretation. If surface warming slows temporarily because more energy is entering the ocean, the underlying forcing has not weakened in proportion. A short interval of reduced surface warming cannot be read as evidence that the climate system has returned to balance. The energy is still moving through the system.
This is where the physical meaning of climate sensitivity becomes more useful than a single headline number. The atmosphere, upper ocean, deep ocean, cryosphere, and biosphere are coupled, but they do not share one clock.
The CMIP6 “hot model” problem
The sixth phase of the Coupled Model Intercomparison Project, known as CMIP6, included models with a wide spread of raw ECS values. Some Earth system models produced ECS values above 5°C, and the reported raw model range extended from 1.8°C to 5.6°C.
Those results attracted attention because they raised the possibility of much stronger warming from a doubling of atmospheric CO₂. But a raw multi-model range is not the same as the IPCC’s assessed likely range for the real climate system.
Several CMIP6 models with very high ECS were linked primarily to stronger short-wavelength cloud feedbacks. In these models, changes in cloud behaviour amplify warming more than observational evidence currently supports. That does not make the models useless. They can still contribute to research on circulation, regional climate, extremes, and process interactions. But their high sensitivity values should not simply be averaged into a statement that the scientific consensus expects more than 5°C of equilibrium warming.
The distinction is between model output and assessed evidence. The IPCC AR6 evaluated model results alongside observations, paleoclimate information, and physical constraints. Its likely ECS range of 2.5°C to 4.0°C reflects that synthesis, with a best estimate of 3.0°C.
A model ensemble is not a democratic vote in which every simulation receives equal authority. Models are experiments built with different parameterizations and structural choices. Some reproduce particular observations better than others; some contain known biases relevant to the variable under discussion. Interpretation requires examining why a model produces a given sensitivity and whether that mechanism is consistent with the wider evidence.
A high-sensitivity model is a signal to investigate a feedback, not a shortcut to the upper end of the climate forecast.
The same caution applies in the opposite direction. A model with low ECS does not prove that climate risks are modest. Lower sensitivity can coexist with strong regional changes, rapid extremes, sea-level rise, ecological disruption, or dangerous warming committed over longer timescales.
What the metrics mean for policy
TCR and ECS serve different policy questions.
TCR is closely connected to the rate of warming under continued emissions. That makes it relevant to near-term adaptation: infrastructure design, heat-health planning, water management, agricultural calendars, wildfire risk, and ecosystem conservation. Communities experience the coming decades before they experience a fully equilibrated climate response.
ECS is more relevant to the long-term consequences of cumulative forcing and the warming commitment associated with a sustained atmospheric CO₂ concentration. It helps frame the stakes of stabilizing concentrations, removing carbon dioxide, and assessing climate states that may persist long after emissions decline.
Neither metric can substitute for an emissions pathway. A sensitivity range does not tell us whether a particular concentration will be reached, how quickly it will be reached, or how regional climate will respond. Those questions require emissions scenarios, carbon-cycle modelling, ocean circulation, atmospheric dynamics, and impact assessments.
Nor can either metric be translated directly into a local temperature estimate. Global mean surface temperature is a diagnostic of the planetary energy balance. It does not describe the temperature of a specific watershed, estuary, mountain range, or city. Regional warming can exceed or fall below the global mean, and precipitation, drought, snowpack, and extreme heat may change in ways that are not proportional to the global average.
Still, the two metrics establish a necessary physical boundary around policy debate. A transiently moderated surface response is not evidence that the final climate response will remain small. Conversely, a high raw model estimate is not automatically the most credible estimate of the Earth system’s equilibrium sensitivity.
Reading climate sensitivity without losing the timescale
The most reliable way to read climate sensitivity estimates is to keep four questions in view:
1. What is the forcing experiment?
ECS and TCR are defined using different idealized setups. A result from a 1% annual CO₂ increase cannot be interpreted as though it were an equilibrium experiment.
2. What part of the climate system has had time to respond?
TCR captures a moving system in which deep-ocean heat uptake suppresses part of the surface warming. ECS includes a much longer adjustment.
3. Is the number a raw model result or an assessed range?
CMIP6 output spans values broader than the IPCC AR6 likely ranges. The reason for that difference is scientific evaluation, not editorial preference.
4. Is the claim about global mean temperature or local impact?
A global sensitivity metric is not a regional forecast. Ecosystems respond to local thermal stress, hydrology, chemistry, disturbance, and habitat change.
The physical world does not deliver warming in neat reporting intervals. Heat enters through the atmosphere, reaches the upper ocean, descends into deeper layers, alters circulation, and interacts with ice, clouds, soils, and living systems. ECS and TCR are attempts to measure different stages of that process.
TCR tells us how much warming is visible while the climate system is still in motion. ECS tells us how much more the surface may warm after the slower components have adjusted. The difference is not a contradiction between climate metrics. It is the signature of a planet whose atmosphere reacts quickly, whose oceans store heat for centuries, and whose ecological consequences unfold across several clocks at once.
That is the broader climate implication: delaying surface warming is not the same as removing the forcing. The heat that does not appear immediately at the surface has not left the system. It has moved—through water, circulation, feedbacks, and time.