Evolution of General Circulation Models (GCMs)
Origins and structure General Circulation Models (GCMs) emerged in the 1960s and 1970s as early digital tools to simulate atmospheric behavior. Based on first-principles physics (the Navier-Stokes equations for fluid motion, thermodynamic laws, and radiative transfer models) initial GCMs were highly simplified, often one- or two-layer atmospheric simulations with no coupling to oceans, land, or ice.
Complexity expansion By the 1990s and 2000s, GCMs had evolved into Earth System Models (ESMs), incorporating modules for oceans, sea ice, land surface dynamics, carbon cycling, aerosols, and biogeochemical feedbacks. Cloud microphysics, dust transport, volcanic aerosols, and even socio-economic coupling (via IAMs: Integrated Assessment Models) were layered in. While model scope expanded, so too did computational demands and parametric uncertainty.
The CMIP Framework The Coupled Model Intercomparison Project (CMIP) was launched by the World Climate Research Programme (WCRP) to standardize simulations across institutions. Each generation (CMIP3, CMIP5, CMIP6) defined common scenarios, forcings, and output diagnostics. CMIP results serve as the empirical basis for IPCC assessments and downstream policy modeling, including carbon budgets and adaptation targets. CMIP7, currently under development, aims to improve cloud physics, regional resolution, and scenario diversity.
Forecasts vs. Projections vs. Scenarios
Forecasts Forecasts are short-term (days to seasons) and initialized using real-time observational data. They are governed by initial conditions, bounded by observational uncertainty, and generally verifiable within their temporal window (e.g., ENSO forecasting, seasonal precipitation).
Projections Projections are conditional simulations of climate responses to externally imposed forcings, such as rising CO₂ concentrations. Unlike forecasts, projections do not predict what will happen but what could happen under specified assumptions; they are “if-then” exercises, not testable predictions.
Scenarios Scenarios represent long-range socio-economic narratives embedded with assumptions about population growth, energy mix, land use, and policy frameworks. The IPCC’s Shared Socioeconomic Pathways (SSPs) (e.g., SSP1-1.9 (sustainability) vs. SSP5-8.5 (fossil-fueled development)) are common inputs to GCMs, not outputs. Their plausibility varies widely.
Best practice The correct application of climate models involves multi-model ensembles and scenario sampling, not reliance on single pathways or models. Yet public communication and policy implementation routinely fail this standard.
CMIP5 and CMIP6 as Political Instruments
Scientific infrastructure, political function CMIP5 and CMIP6 outputs form the backbone of IPCC Assessment Reports (AR5 and AR6, respectively) and serve as inputs for regulatory regimes (e.g., UNFCCC frameworks, EU Taxonomy, TCFD/NGFS disclosures). However, their deployment is rarely neutral. High-end scenarios like RCP8.5 and SSP5-8.5 (originally intended as worst-case bounds) are routinely portrayed as likely or “business as usual” in media and policy documents.
Inflation of risk narratives Although global emissions have not tracked SSP5-8.5 since 2010, and energy trends diverge from its fossil-use assumptions, this scenario continues to dominate risk assessments, driving policy justifications for net-zero mandates and decarbonization incentives. Academic critiques (Hausfather & Peters, 2020) note this distortion but it persists.
The feedback loop A model-scenario-policy loop has emerged: models shape IPCC reports; reports influence regulation; regulation drives investment mandates; those mandates reinforce the salience of high-risk model scenarios, regardless of plausibility. This circularity embeds speculative science into systemic economic structures.
Structural Uncertainty and Model Dependence
Structural uncertainty Different models encode different structural assumptions about climate processes, including convection schemes, cloud parameterization, and land surface interactions. These differences yield divergent outputs even under identical forcing scenarios.
Parameterization and ensemble spread Processes like cloud formation, aerosol indirect effects, and ocean mixing are computationally unresolved and thus parameterized, approximated using empirical or semi-empirical coefficients. This creates wide variation in outcomes. For many variables (e.g., regional precipitation), the spread across models is greater than the forced signal itself.
Model interdependence Despite the appearance of independence, many GCMs are structurally related, sharing legacy code, parameterization libraries, or institutional lineages (e.g., GISS, Hadley, MPI). This introduces the “model democracy problem”: ensemble diversity is narrower than it appears, exaggerating apparent consensus.
The “Consensus” Illusion and Suppression of Scientific Dissent
The Cook et al. Distortion The famous “97% consensus” figure comes from Cook et al. (2013), which reviewed 11,944 climate paper abstracts and found that only 41 (0.3%) explicitly stated humans were the primary cause of warming. The rest either made no causality statement or referred to anthropogenic contribution broadly.
Rhetorical leverage, scientific cost This figure has been used extensively in media, ESG reports, and public policy to imply unanimity on not just warming attribution, but also on model accuracy, sensitivity, and urgency. In reality, a broad majority of scientists agree climate is changing and humans contribute, but disagreement is strong on sensitivity estimates, scenario realism, and mitigation policy.
Institutional pressures and conformity incentives Funding, tenure, publication, and professional advancement often depend on adherence to IPCC-aligned framing. Research that challenges high-sensitivity models or explores natural variability is often marginalized or rejected. This creates a self-reinforcing echo chamber and inhibits pluralism in methodological development.
Term | Definition | Example/Use Case |
Forecast | Short-term, initial-condition-based prediction | Seasonal rainfall projection |
Projection | Long-term scenario-dependent simulation | Global temperature by 2100 under SSP2-4.5 |
Scenario | Socio-economic narrative with energy and policy assumptions | SPP1-1.9 (”sustainability”) vs. SSP5-8.5 (”fossil”) |
Ensemble | Multi-model collection of runs capturing uncertainty | CMIP6 ensemble spread for sea level rise |
Structural Uncertainty | Divergence caused by model architecture or parameterization | Cloud feedback variation among GCMs |
"Consensus” | Claimed unanimity used rhetorically rather than scientifically | "97% of scientists agree…” |
Current (May 2025) Developments and Implications
CMIP7 underway Expected to improve regional resolution, cloud representation, and incorporate “storyline” scenario construction to reflect narrative diversity rather than rigid pathways.
Growing recognition of limits Drafts of the forthcoming IPCC AR7 (expected 2027) acknowledge systematic model over-prediction in high-emissions scenarios and reinforce the need for transparent communication of scenario plausibility.
Shifts in policy frameworks Due to increasing criticism from economists, modelers, and ESG practitioners, regulatory agencies and financial institutions are beginning to reevaluate:
- The use of RCP8.5 in capital stress testing
- The reliability of model-based transition risk frameworks
- The legitimacy of model-driven mandates within climate finance