Coupled Model Intercomparison Project

• 1. Introduction

  ›

  The Coupled Model Intercomparison Project, known as CMIP, is a collaborative framework that scientists around the world use to improve knowledge of climate change. It was organized in 1995 by the Working Group on Coupled Modelling, part of the World Climate Research Programme. What does it mean to "couple" a climate model? And why does coordinating dozens of research groups across more than a dozen countries actually matter for understanding Earth's future? Those are the questions this documentary will answer.

  At its core, CMIP works because climate is not one system. It is the atmosphere talking to the oceans, the oceans influencing ice sheets, ice sheets changing land surfaces, and land surfaces feeding back into the atmosphere. A coupled model is a computer-based model that links all of those subsystems so they can interact during a simulation. Before frameworks like CMIP, research groups ran their own models in their own ways, making direct comparison almost impossible. The project changed that.

• 2. Building the Framework

  ›

  Lawrence Livermore National Laboratory, home to the Program for Climate Model Diagnosis and Intercomparison (PCMDI), has been the operational backbone of CMIP since its earliest phases. PCMDI helped the Working Group on Coupled Modelling determine the scope of each phase, maintained the project's database, and participated in data analysis. That infrastructure role made it possible for modeling centers around the world to contribute on equal footing.

  CMIP has collected model output from two main types of simulations: pre-industrial "control runs", which establish a climate baseline without human influence, and simulations in which atmospheric CO2 rises by 1 percent per year. Both types give researchers a controlled setting for comparing how different models respond to the same forcing. Some phases went further by incorporating more realistic historical scenarios, paleoclimate reconstructions, and projections of future conditions, under the banner of the 20th Century Climate in Coupled Models effort, known as 20C3M.

• 3. Early Phases Take Shape

  ›

  The response to the first CMIP announcement exceeded expectations. Up to 18 global coupled models participated in the initial data collection, representing most of the international groups that had built global coupled general circulation models at the time. That broad uptake validated the idea that a shared comparison framework was something the climate science community genuinely wanted.

  At a September 1996 meeting of the CLIVAR numerical experimentation group, held in Victoria, Canada, scientists agreed on the design for CMIP2. The second phase would focus on integrations lasting 80 years, with CO2 increasing at a compound rate of 1 percent per year, reaching doubled concentrations at around year 70. That specific design gave every participating group a common benchmark. A model's sensitivity to a CO2 doubling is one of the most fundamental questions in climate science, and CMIP2 gave researchers a shared lens for examining it.

• 4. CMIP3 and the IPCC

  ›

  During 2005 and 2006, PCMDI coordinated and stored a large collection of climate model outputs covering simulations of past, present, and future climate scenarios. The timing was deliberate. Climate scientists were preparing the IPCC Fourth Assessment Report, and CMIP3 was designed to support that work, including groups outside the major modeling centers who needed access to high-quality simulation data.

  For CMIP3, a list of 20 different experiments was proposed. PCMDI kept documentation of all the global climate models involved. That documentation function is easy to overlook but hard to overstate: without a clear record of which model was run under which conditions, comparing results across groups becomes unreliable. The archive PCMDI maintained for CMIP3 set a standard that later phases would build on and extend.

• 5. CMIP5 Raises the Bar

  ›

  CMIP5 ran from 2010 to 2014 and introduced a major advance in how simulation runs were described. Previous phases recorded the outputs; CMIP5 also recorded extensive metadata describing the scientific, technical, and numerical specifics of every model run. The METAFOR project created an exhaustive schema to capture those details, which was archived alongside the simulation data itself.

  The scientific goals for CMIP5 were explicitly tied to questions left open by the IPCC Fourth Assessment Report process. The CMIP5 experimental protocol was endorsed by the 12th Session of the WCRP Working Group on Coupled Modelling, signaling broad institutional support. Results from CMIP5 fed directly into the IPCC Fifth Assessment Report, making the phase one of the most consequential rounds of coordinated climate modeling the field had seen to that point.

• 6. CMIP6 and the DECK

  ›

  Planning for Phase 6 began in 2013, and an overview of its design was published in 2016. By 2018, CMIP6 had endorsed 23 Model Intercomparison Projects involving 33 modeling groups spread across 16 countries. That scale of participation required a new organizational architecture.

  The structural innovation at the heart of CMIP6 is a framework called CMIP Diagnostic, Evaluation and Characterization of Klima, shortened to DECK. The word "klima" is Greek for "climate." DECK provides a common baseline of experiments that all participating models must run, while Endorsed MIPs allow groups to address a wider range of specific scientific questions built on top of that shared foundation. The structure is designed to persist into future phases.

  CMIP6 also introduced standardized forcing datasets coordinated through a system called input4MIPS. Those datasets cover historical emissions of short-lived climate forcers, biomass burning, land-use change, greenhouse gas concentrations, ozone, aerosol optical properties, solar forcing, and sea surface temperatures, among others. For future projections, CMIP6 replaced the Representative Concentration Pathways used in prior phases with Shared Socioeconomic Pathways. The deadline for papers contributing to the IPCC Sixth Assessment Report Working Group I was set for early 2020.

• 7. What Comes Next

  ›

  CMIP7 has been designed with a different release philosophy than any previous phase. Rather than delivering all results in a single large batch, the project announced a more continuous release approach, with so-called fast track experiments that can address specific requirements as they arise, alongside the standard DECK releases and a growing number of community MIPs.

  First data from CMIP7 is expected by the end of 2025. That timeline places the next round of coordinated climate model output directly in the window when scientists will be assembling the evidence base for future international assessments. The Atmospheric Model Intercomparison Project, AMIP, remains a related effort running in parallel, focused on global coupled ocean-atmosphere general circulation models.