Loading field map
Climate science began with a simple question: what makes Earth's climate the way it is, and could it change? Over two centuries, researchers built a web of frameworks that transformed a single causal insight into a suite of models, proxy reconstructions, and statistical methods that must converge to inform policy. The history of climate science is not a linear march of discoveries but a series of frameworks that each addressed a limitation of earlier ones—sometimes by replacing, sometimes by coexisting with, and sometimes by providing infrastructure for others.
The first framework to offer a mechanistic explanation of Earth's climate was the Greenhouse Theory of Climate Change, proposed by Joseph Fourier in 1824 and later refined by John Tyndall and Svante Arrhenius. It identified that certain atmospheric gases—water vapor and carbon dioxide—trap outgoing infrared radiation, warming the planet. This was a causal theory: it explained why Earth is not a frozen ball and why changes in atmospheric composition could alter global temperature. Yet the Greenhouse Theory was qualitative; it could not describe the actual distribution of climates across the planet.
That descriptive gap was filled by the Empirical Climate Classification, most famously developed by Wladimir Köppen around 1900. Rather than explaining why climates differ, Köppen's system sorted the world into climate types (tropical, dry, temperate, continental, polar) based on temperature and precipitation thresholds. This framework coexisted with the Greenhouse Theory rather than replacing it: one provided mechanism, the other provided a map. The Köppen classification became the standard language for describing climate zones and remains in use today, though it has been refined and supplemented by more quantitative schemes.
A major puzzle for the Greenhouse Theory was the Ice Ages. Why had Earth cycled between glacial and interglacial periods? The Milankovitch Orbital Theory, developed by Milutin Milankovitch from 1911 onward, proposed that variations in Earth's orbit—eccentricity, obliquity, and precession—alter the seasonal distribution of solar radiation, driving glacial cycles. This framework addressed a question the Greenhouse Theory could not: natural, long-term climate change. For decades, Milankovitch's theory lacked precise paleoclimate data to confirm it, but it provided a testable hypothesis that later proxy reconstructions would validate.
Meanwhile, a different kind of question emerged: could human activity alter climate? In 1938, Guy Callendar compiled temperature records and atmospheric CO₂ measurements to argue that fossil fuel combustion was warming the planet. This Anthropogenic Climate Change framework revived and extended the Greenhouse Theory by linking it to industrial emissions. Unlike Milankovitch's natural cycles, Callendar's claim was about human perturbation. For much of the 20th century, this framework remained a minority view, awaiting more powerful tools to test it.
The mid-20th century brought a radical shift: the ability to simulate climate on computers. General Circulation Modeling (GCM) , pioneered by Norman Phillips in 1956, divided the atmosphere into a three-dimensional grid and solved the equations of fluid dynamics and thermodynamics. GCMs were the first frameworks that could integrate multiple physical processes—radiation, convection, advection—into a single predictive system. They did not replace the Greenhouse Theory but gave it computational muscle: now one could ask what happens to global temperature when CO₂ doubles.
Almost immediately, modelers confronted a fundamental limit. In 1963, Edward Lorenz discovered that simple atmospheric models exhibit sensitive dependence on initial conditions—the essence of Climate Dynamics and Chaos. This framework showed that weather prediction beyond about two weeks is inherently impossible, and that climate models must be run as ensembles (many simulations with slightly different starts) to capture the range of possible outcomes. Chaos did not invalidate GCMs; it redefined what they could promise. Instead of deterministic forecasts, climate modeling became a probabilistic enterprise.
To explore specific processes more efficiently, researchers developed simpler tools. Radiative-Convective Modeling, introduced in 1967 by Syukuro Manabe and Richard Wetherald, reduced the atmosphere to a single column and focused on the vertical balance of radiation and convection. This framework isolated the greenhouse effect's thermal structure without the computational cost of a full GCM. Energy Balance Models, emerging around 1968, went even further: they treated Earth as a single point or a few latitude bands, balancing incoming solar radiation against outgoing infrared radiation. These reduced-form models did not compete with GCMs; they formed a hierarchy. Energy Balance Models remain valuable today for rapid sensitivity experiments and scenario screening, because they can run thousands of simulations in seconds.
While models simulated the future, another tradition reconstructed the past. The Paleoclimate Proxy Reconstruction School, growing from the 1960s onward, used ice cores, tree rings, ocean sediments, and other natural archives to infer past temperatures, CO₂ levels, and ice volumes. This framework provided the empirical backbone for climate science. Proxy data confirmed Milankovitch's orbital theory by showing that glacial cycles matched orbital variations. More critically, ice cores from Antarctica revealed that CO₂ and temperature had risen and fallen together over hundreds of thousands of years, strengthening the case for the Greenhouse Theory.
Proxy reconstructions also constrained a key quantity: Climate Sensitivity—the equilibrium global warming from a doubling of CO₂. The Climate Sensitivity Framework, formalized in the late 1970s, asked: how much will Earth warm? Early estimates ranged from 1.5°C to 4.5°C, a range that persisted for decades. Alongside sensitivity came the Radiative Forcing Framework, also crystallized around 1979. Radiative forcing measures the change in Earth's energy balance caused by a factor (CO₂, aerosols, solar variability) in watts per square meter. Together, these two frameworks became the core of the Intergovernmental Panel on Climate Change (IPCC) assessment methodology. Radiative Forcing provides the input; Climate Sensitivity provides the response. They were jointly developed and jointly adopted: the IPCC's first report in 1990 used both to project future warming.
By the 1990s, GCMs had grown to include oceans, sea ice, and land surfaces. The next step was to couple them with biogeochemical cycles—carbon, nitrogen, sulfur—to create Earth System Modeling. This framework, emerging around 1990, absorbed and extended GCMs by treating the Earth as a single interacting system. An Earth System Model (ESM) can simulate not just the physical climate but also how ecosystems respond to CO₂ and how those responses feed back on the climate. ESMs are the most comprehensive tools available, but they inherit the uncertainties of their components: cloud processes, aerosol interactions, and carbon cycle dynamics remain major sources of spread.
As models improved, a new question became tractable: can we detect a human influence on climate and attribute it to specific causes? The Climate Detection and Attribution framework, formalized in the mid-1990s, uses statistical fingerprinting to compare observed climate changes with the responses simulated by models under different forcings (natural vs. anthropogenic). Detection asks whether the observed change is larger than natural variability; attribution asks what fraction of that change is due to human activities. This framework depends on Earth System Model outputs and Radiative Forcing inputs: without accurate forcing histories and reliable model responses, attribution would be impossible. By the IPCC's Fifth Assessment Report (2013), detection and attribution had been extended from global mean temperature to continental-scale warming, ocean heat content, and even individual extreme events.
The frameworks that dominate contemporary climate science are Earth System Modeling, Climate Detection and Attribution, the Climate Sensitivity Framework, and the Radiative Forcing Framework. They agree on the fundamentals: Earth is warming, human activities—especially CO₂ emissions—are the dominant cause, and continued emissions will produce further warming. The basic physics of the greenhouse effect is not in dispute.
Yet significant disagreements remain. The Climate Sensitivity Framework still wrestles with a wide range: the IPCC's Sixth Assessment Report (2021) gave a likely range of 2.5°C to 4.0°C for equilibrium climate sensitivity, but some studies suggest values below 2°C or above 5°C are possible. The source of this spread lies in cloud feedbacks, which different models parameterize differently. Radiative Forcing faces its own uncertainty: aerosols—sulfates, black carbon, dust—have a cooling effect that is poorly quantified, making it hard to know how much of the observed warming has been masked by pollution. Climate Detection and Attribution has successfully attributed global temperature rise, but attributing individual extreme events (heatwaves, floods, hurricanes) to climate change remains contested. Some events show a clear human fingerprint; others are dominated by natural variability. The National Academies of Sciences, Engineering, and Medicine has called for more research on event attribution methods.
Earth System Models serve as the infrastructure for all these debates. They are the primary tools for projecting future climate under different emission scenarios, and their outputs feed into detection and attribution studies. But they are not monolithic: different modeling centers produce ESMs with different parameterizations, leading to a spread in projections. The community manages this spread through multi-model ensembles (e.g., CMIP6), which treat model diversity as a source of information rather than a flaw. The leading frameworks today do not compete; they depend on each other. Radiative Forcing quantifies the drivers, Climate Sensitivity translates forcing into warming, Earth System Models simulate the full response, and Detection and Attribution tests whether those simulations match reality. The history of climate science is the story of these frameworks learning to work together.