An Overview of CMIP5 and the Experiment Design
K. E. Taylor, R. J. Stouffer & G. A. Meehl
Observation and Context
Climate researchers rely on standardized model comparisons to understand global climate change and natural variability. Prior to the fifth phase of the Climate Model Intercomparison Project (CMIP5), significant gaps existed in evaluating critical climate feedbacks, cloud behaviors, and decadal predictability. Earlier model phases, such as CMIP3, lacked high-spatial-resolution representations, interactive biogeochemical carbon-cycle couplings, and easily accessible data for researchers outside of standard physical climate modeling (such as those studying socio-economic impacts and adaptation). To address these issues in time for the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5), the scientific community needed a more detailed and coordinated modeling framework.
Hypothesis
If global climate modeling centers perform a standardized, tiered suite of coordinated simulations — combining long-term projections, near-term initialized predictions, and interactive biogeochemical cycles — then the international scientific community can isolate the physical mechanisms driving model differences, improve near-term predictive skill, and establish a more robust consensus on climate variability and feedbacks.
Experiment and Methodology
Over 20 global modeling groups coordinated to run more than 50 different climate models. The experiments were divided into a structured “core” and “tiered” design across two primary categories:
- Long-Term Century-Scale Simulations: Models projected climate responses to changing atmospheric composition and land cover from the mid-nineteenth century to the year 2100 and beyond. Researchers utilized standard Atmosphere-Ocean Global Climate Models (AOGCMs) and newer Earth System Models (ESMs). Unlike standard models, ESMs featured interactive biogeochemical components that closed the carbon cycle by calculating CO₂ concentrations directly from simulated emissions. Forcing was driven by the new Representative Concentration Pathways (RCPs), representing different mitigation scenarios like RCP4.5 and the high-emission RCP8.5 (reaching about 8.5 W/m² of radiative forcing by 2100).
- Near-Term Decadal Predictions: These exploratory 10-to-30-year integrations were initialized with actual observed ocean temperatures and sea ice states starting from 1960 to the present. This allowed models to attempt to track the actual unforced trajectory of climate evolution rather than relying on randomized unforced variations.
- Idealized Diagnostics: Models ran specialized diagnostic tests, such as abrupt quadrupling of CO₂ and 1% per year CO₂ increases, to calculate climate sensitivities and isolate cloud-radiation feedbacks.
Results and Data
CMIP5 successfully generated an unprecedented multimodel dataset exceeding 3 petabytes of archived output, which is nearly 100 times larger than the CMIP3 archive. Spatial resolution dramatically improved, with roughly half of the atmospheric models achieving a latitudinal resolution finer than 1.3 degrees (compared to only one model in CMIP3).
The catalog of saved variables expanded extensively to include data for the atmosphere (60 variables), ocean (77), land surface and carbon cycle (58), ocean biogeochemistry (74), sea ice (38), land ice (14), and clouds (100). Crucially, data delivery was decentralized through the Earth System Grid Federation (ESGF), a global network of data nodes that allowed users to easily search, filter, and access standardized datasets from any participating center. This unified archive significantly reduced sharing delays for researchers across all three IPCC working groups.
Conclusion and Climate Impact
The CMIP5 experiment design proved that a highly structured, collaborative framework can successfully standardize global climate modeling. While near-term decadal forecasting remains exploratory and faces hurdles like model drift and the need for complex bias corrections, the coordinated tier structure allows scientists to directly compare diverse model behaviors.
The climate impact of this project is substantial: CMIP5 provides a robust scientific foundation for global environmental policy. By delivering high-resolution, downscaled datasets and multi-century projections, it empowers society to assess regional-scale climate impacts, model local vegetation and water changes, and design effective national adaptation and mitigation strategies.
Citation
Taylor, K. E., Stouffer, R. J., & Meehl, G. A. (2012). An overview of CMIP5 and the experiment design. Bulletin of the American Meteorological Society, 93(4), 485-498.