Weather swings wildly within days, but climate is the long-term statistical state. In 1967 Manabe built a one-dimensional column model of the atmosphere, calculating radiation and convection together and showing a ~2 °C rise for doubled CO2. About a decade later Hasselmann treated fast, chaotic weather as “noise” and linked it to slowly varying climate with stochastic equations. He also devised fingerprint techniques to separate natural and human signals, demonstrating that observed warming stems from human emissions. Their work now underpins IPCC assessments and global mitigation policy.

Manabe’s model combined radiative-convective equilibrium with a fixed-relative-humidity assumption to numerically obtain vertical temperature profiles up to the stratosphere. The outcome—lower-tropospheric warming coupled with stratospheric cooling—provided a greenhouse signature distinct from solar-forcing heating. Hasselmann mapped the climate system to a Langevin equation driven by white noise via a Lagrangian integral, opening a framework that jointly estimates climate sensitivity and internal variability spectra. His maximum-likelihood detection method projects observed spatial patterns onto modeled signals, enabling statistical detection & attribution of CO2 forcing. These ideas feed directly into CMIP design and Bayesian estimates of equilibrium climate sensitivity.

Manabe & Wetherald (1967) solved two-stream radiative transfer and moist-adiabatic convective adjustment in a 1-D RCM; their 2×CO2 sensitivity of 2.3 K remains a robust lower bound across later multilayer GCMs. Hasselmann (1976, 1979) introduced temporal scale separation, describing the climate response with a reddened spectrum S(ω)=σ²/(1+τ²ω²); using a Fokker–Planck treatment, he derived Green-function solutions for forced heating interacting with internal noise. The fingerprint method employs EOF projection and optimal detection, weighting model-ensemble variance via a covariance matrix to disentangle contributions of multiple forcings. This statistical-mechanical framework underlies modern detection & attribution, emergent constraints, and parameter assimilation in GCMs.