The coupled model intercomparison project (CMIP)

Gerald A. Meehl

National Center for Atmospheric Research, Boulder, Colorado

George J. Boer

Canadian Centre for Modeling and Analysis, Canada

Curt Covey

Program for Climate Model Diagnostics and Intercomparison, U.S.A.

Mojib Latif

Max Planck Institute for Meteorology, Germany

Ronald J. Stouffer

Geophysical Fluid Dynamics Laboratory, U.S.A.


CMIP is a project to study and intercompare climate simulations made with coupled ocean/atmosphere/cryosphere/land GCMs. There are two main phases (CMIP1 and CMIP2) which study: (1) the ability of models to simulate the present-day climate, and (2) model simulation of climate change due to an idealized change in forcing (a 1% per year CO2 increase). CMIP was initiated in December 1995 as part of the overall modelling activities within the World Climate Research Programme (WCRP). Since that time, present-day climate simulations from 19 models (CMIP1), and forced climate change results from 12 of these models (CMIP2) have been archived by the US Department of Energy Programme for Climate Model Diagnosis and Intercomparison (PCMDI). Virtually all global coupled modeling efforts in the world are represented in this data set held by PCMDI. The model results are available for analysis via "diagnostic subprojects" which concentrate on a particular aspect of climate and model behavior (and which attempt to entrain expertise outside of the modelling community in this analysis). There are currently 10 CMIP1 subprojects analyzing aspects of present-day climate simulations and 9 CMIP2 subprojects analyzing the CO2 increase experiments. Results from a number of subprojects were reported at the 1st CMIP Workshop held in Melbourne. Australia at the Bureau of Meteorology Research Centre from 14-15 October, 1998. Some recent advances in global coupled modeling related to CMIP were also reported. Presentations were based on preliminary (sometimes controversial) unpublished results. reflecting the personal views of the participants. Some of the results from the workshop are: 1) most observed processes are simulated in the global coupled models with varying degrees of fidelity, including the North Atlantic Oscillation and linkages to North Atlantic SSTs, the Antarctic Circumpolar Wave, El Nino-like events, monsoon interannual variability, etc.; 2) the amplitude of both high and low frequency surface temperature variability in some global coupled models is less than that observed, which may be partly due to omission of low frequency forcing agents in the models (e.g. solar, volcanos, etc.), and this has implications for climate change detection/attribution studies that rely on model results for estimates of natural low frequency variability; 3) in the CO2 climate change experiments, an El Nino-like pattern in the mean SST response such that there is greater mean surface warming in the eastern equatorial Pacific than the westren equatorial Pacific is simulated by a number of models but other models have little, or even a La Nina-like response; this complicates a simple understanding of climate change anomalies in the Pacific region; 4) though flux adjustment improves the simulation of present-day climate and produces a stable base state to enable very long term (1000 years and longer) integrations, it does not appear to have a major effect on model processes or model response to increasing CO2; and 5) recent multicentury integrations that produce a stable surface climate without flux adjustment (though still with some systematic simulation errors) signify a substantial increase in our understanding of the causes of climate drift and the need for flux adjustments. This has led to improvements in the model components and the initialization procedures used in these models.
Back to Publications List

To the Climate Analysis Section Home Page

Hongjun Zhang: