CoECSS engaged in international CMIP5 experiments
by Sophie Lewis
Centre of Excellence researchers are playing an important role in an international project contributing to the Coupled Model Intercomparison Project Phase 5 (CMIP5), which consists of over 20 modelling groups worldwide participating in a standard set of climate model experiments.
The objective of the project is to use standardised model experiments across various groups to assess how accurately climate models simulate the recent past, to provide projections of future climatic changes over the next century or so and to investigate various complex climate processes, such as the role of clouds in the climate system.
As part of this, Centre of Excellence researchers are producing historical climate simulations using the Australian Community Climate and Earth System Simulator (ACCESS) version 1.3 with support from the Centre for Australian Weather and Climate Research (CAWCR) and the National Computational Infrastructure (NCI).
One of the highest priority CMIP5 experiments in which the Centre is involved is a simulation of the historical period that aims to reproduce the evolution of the climate from 1850 to 2005 as accurately as possible.
Each model is run with standard forcings to replicate climatic changes occurring over this period. These forcings include changing atmospheric composition (including changing carbon dioxide), volcanic influences, solar influences and changes in aerosols.
In addition to this historical simulation with all forcings, there are a series of CMIP5 detection and attribution experiments.
For example, in one detection and attribution experiment the model is forced by historical changes in greenhouse gases but no natural variations (such as volcanic eruptions) are imposed on the model.
In another, and complementary experiment, the model is forced by natural variations but anthropogenic changes in atmospheric composition are ignored.
The object of these experiments is to detect changes in climate and then to attribute the detected changes to a particular climate forcing.
All the modelling groups involved in the CMIP5 project generally provide more than one contribution, or ensemble member, for each experiment.
To achieve consistent and comparable results, the ensemble members run under identical experiment conditions. The only difference is in the way in which each experiment has been initialised.
The different initial conditions for each ensemble member leads to slightly different climate trajectories. Through this variation, an overall envelope of modelled climates is generated that encompass the full spread of model differences.
By doing this over the full series of model simulations the statistical confidence in the model results is increased and the internal variability of each model is more clearly revealed.
CAWCR conducted the first ACCESS1.3 historical ensemble as part of the CMIP5 project and the Centre of Excellence has recently completed two further ensembles. The results of these historical simulations are available to the Australian research community through NCI as pre-published CMIP5 fields.
The model output is undergoing rigorous quality checking and the highest priority climate variables are being uploaded to the Earth Systems Grid for general use.
The comparison between observed temperatures and the simulations of the ACCESS1.3 Climate Model
At a global scale, the model results from ACCESS1.3 for the historical simulation with all forcings included agree closely with the other models contributed to CMIP5 and also with observational records of climate from 1850 to 2005 (see figure above).
Comparing these modelled surface air temperatures to observation provides a preliminary evaluation of the performance of each model. Many other climate variables are also produced during each simulation and these will be included in many analyses of CMIP5 models.
The suite of historical ACCESS model simulations with all forcings has been completed. The Centre has now started a suite of detection and attribution experiments using greenhouse gas forcings only.
Each ensemble member takes approximately three months to complete and an update will follow soon.
For further information on ACCESS1.3 CMIP5 simulations and available output, go to the CSIRO ACCESS wiki.