Menu Close

Explainer: the models that help us predict climate change

Good climate models can predict and help us respond to weather events. David Dashwood/Flickr, CC BY-SA

What will the weather be like next week, next season, or by the end of the century? In the absence of a second Earth to use in an experiment, global weather and climate model simulations are the only tools we have to answer these questions.

Having access to this information is vital for the community, government and industries to make informed decisions – this includes sectors like tourism, natural resource management, agriculture and emergency services to name a few.

Weather and climate may never be completely predictable, but the science has now come far enough for us to be more confident when it comes to knowing whether it will rain this afternoon and for projecting what Australia’s climate may look like many decades in the future.

We’re also getting better at predicting the next season or two, so that we can be more prepared to respond to the extremes in weather like cyclones, heatwaves and flooding rains that already impact Australian communities.

Climate modelling from CSIRO on Vimeo.

Looking ahead

General Circulation Models (also referred to as global climate models) are built using mathematical representations of the dynamic Earth system. Their fundamentals are based on the laws of physics including conservation of mass, energy and momentum. These models represent, in three dimensions, the large-scale circulations of the atmosphere and ocean, such as the progression of high and low pressure systems and large-scale oceanic currents. Models also include the cryosphere (snow and sea ice) as well as the land surface.

Climate models help us to understand our present weather and climate, and also allow us to consider plausible future scenarios of how the climate might change. They generate simulations to tell us what happened or what might happen under a range of different scenarios, such as for greenhouse gas concentrations .

Although models used for weather prediction and climate applications share the same fundamentals, they are a little different.

Weather models are run at higher spatial “resolution”, and incorporate the very latest set of satellite and ground measurements using advanced data assimilation methods. This defines the starting point from which the model predicts the evolution of weather events over the next week or so.

Climate models do not seek to forecast the exact “weather” on a particular day months or years ahead (which is impossible), but rather predict the “statistics” of the weather (i.e. the “climate”), such as the average conditions, over a season or trends over decades.

Climate happens on a regional scale. Marcus Greig/AAP

While General Circulation Models simulate large scale Earth system processes, there are some processes, such as cloud formation and rainfall, which occur at small scales and make changes in the Earth system difficult to predict perfectly.

Despite these challenges, continuous improvement of models (e.g., higher resolution, better representation of physical processes and improved use of data particularly from satellites) over the past three decades has improved our ability to predict weather and to make climate projections.

There are now over 40 global climate models run around the world. These modelling groups use a common set of greenhouse gas and aerosol scenarios, called Representative Concentration Pathways. This co-ordinated approach permits ready comparison of projections across the many thousands of model simulations for which data are available.

Likewise, weather prediction centres verify daily weather forecasts using internationally defined metrics that enable ready comparison of predictions made by the centres.

The Representative Concentration Pathways fall into three categories:

  • high: greenhouse gas emissions continue to rise over the 21st century without abatement, with a decline in aerosols

  • intermediate: greenhouse gas emissions peak then decline

  • low: greenhouse gas emissions peak quickly and then decline rapidly to very low values (a strong mitigation case).

No matter which model or greenhouse gas scenario we use, a substantial and robust warming signal is evident in the projections of future climate, larger for high emission scenarios. Models also project differences in the timing and magnitude of warming and a range of changes in rainfall and other elements.

So rather than one single climate future, we need to consider a range of possible futures.

Which models are best?

All climate models go through rigorous evaluation to determine the extent to which they can represent daily weather, and past and current climate.

There are many tests carried out to assess a climate model’s performance. For example, scientists can assess how well the model simulates historical climate (such as average Australian rainfall over the last 20 years), or the model’s ability to represent or predict specific features such as monsoon onset, El Niño, or the paths of tropical cyclones.

Researchers investigating the effects of future climate change might decide to select a subset of models based on performance. However, selecting the “best” model or subset of models depends on what performance measure you use.

For example, recent evaluation of climate models for Australian conditions showed that there is no fixed “subset” of climate models that can represent all important aspects of climate better than simply using the full set of available models.

Climate projections often come with a measure of confidence, based on physical understanding, robustness of model projections, and consistency of projections with observed trends or past changes. The performance of climate models with respect to past climate is a critical factor in establishing our level of confidence in future projected changes. The confidence ratings provided for Australia’s latest projections are a novel and useful feature of assessing the range of projected changes in Australia’s future climate.

Australia’s world leading climate model

Time series for Australian average temperature for 1910–2090 as simulated in CMIP5 models, relative to the 1950–2005 mean. Bureau of Meteorology observations are shown in thick brown and a series from a typical model (ACCESS1-0) are shown into the future in light purple. The shading represents the spread amongst all the models for the historical period (grey shading) and future period (purple-high emissions; blue – intermediate; yellow – low emissions). For further details on projections see Chapter 7 of the NRM Tech Report: ( Climate Change in Australia

Australia’s own climate model, the Australian Community Climate and Earth System Simulator, or ACCESS, is consistently shown by national and international groups to be among the top performing models across a range of climate features important to Australia.

ACCESS was developed jointly by the Bureau of Meteorology and CSIRO through their research partnership, The Centre for Australian Weather and Climate Research. It was developed in collaboration with Australian Universities and the UK Met Office with support from the Department of Environment. ACCESS is specifically designed to be used for both weather prediction and climate simulation.

In “weather mode” ACCESS is used by the Bureau of Meteorology to provide Australia’s weather forecasts. Thanks to ACCESS, the Bureau’s four day forecast is now as accurate as the three day forecast was just ten years ago. Comparisons with forecasts from overseas operational centres show ACCESS to be among the top performing models in the world.

The “climate” version of ACCESS was used to generate climate projections that were submitted by Australia to recent co-ordinated international climate change experiments and in support of the recent 5th Assessment Report of the Intergovernmental Panel on Climate Change (IPCC).

ACCESS will continue to be developed and improved, encompassing and modelling the Earth’s component systems with greater detail and precision.

This is the final article in a series on climate change in Australia, to coincide with the launch of new climate websites by CSIRO. Read more:

Want to write?

Write an article and join a growing community of more than 185,600 academics and researchers from 4,982 institutions.

Register now