Social Channels


Additional Options

Date Range

Receive a Daily or Weekly summary of the most important articles direct to your inbox, just enter your email below. By entering your email address you agree for your data to be handled in accordance with our Privacy Policy.

Carbon Brief Staff

24.12.2013 | 1:40pm
SciencePredicting the future: The challenge of regional climate projection
SCIENCE | December 24. 2013. 13:40
Predicting the future: The challenge of regional climate projection

Mat Collins is professor and joint Met Office Chair in Climate at the College of Engineering, Mathematics and Physical Sciences, University of Exeter.

Niels Bohr, the famous Danish Physicist, once remarked ‘Prediction is very difficult, especially when it’s about the future’.

This quote is sometimes used to suggest that making predictions of future climate change is an impossible task. But, of course, we can predict the future. Meteorological services around the world do it every day, with ever increasing accuracy.

To predict future changes in climate, scientists use climate models. We feed in assumptions about future levels of greenhouses gases, then run the models forward in time and diagnose the output. We usually speak of “projections” to indicate that our predictions are not definitive, they are conditional on those economic, social and technological assumptions about future greenhouse gas levels.

All of this activity takes place in modelling centres around the world. A recent revolution in the field has been to collect the output from experiments performed at different modelling centres into a central repository, making the data available to a larger community of climate researchers. This is called the Coupled Model Intercomparison Project (CMIP).

A changing climate

With a changing climate, there’s demand for information about the changes we can expect at scales that might affect particular populations, ecosystems and economies. So how confident can we be about changes projected for a given region?

One problem is that different models represent the complex climate system in slightly different ways, so can produce different projections. Even at the global scale, the level of warming from now until the end of the century can be three degrees Celsius different, depending on the model used. That’s one reason our future climate projections are uncertain.

That science (and life) can be uncertain is nothing new, of course. We generally deal with uncertainties by expressing them as odds or probabilities. Perhaps the natural thing to do would be to compute probabilities from the CMIP model projections by counting the number of models that exceed two degrees of warming by 2100, for example.

If 20 per cent of the models exceed two degrees Celsius for a certain scenario, does that mean that there is a 20 per cent chance of that threshold being crossed? Unfortunately, it’s not that simple. Models are not independent. They use approximations and have flaws that are common to all models.

So how would we make an assessment of the statistical likelihood of future climate change, such as those in the Intergovernmental Panel on Climate Change’s 5th Assessment Report (AR5)?

Estimating earth’s sensitivity

Let’s take global mean temperature. Uncertainties in projections for a given emissions scenario are mainly driven by uncertainties about the sensitivity of the climate system. This results from the combined effects of different processes that can amplify or dampen warming, known as feedbacks.

The Transient Climate Response ( TCR) is defined as the global mean temperature change at the time of doubling of carbon dioxide above preindustrial levels. It’s a compact way of characterising the sensitivity of the climate relevant over the next few decades.

Fortunately, there are ways to estimate the TCR that are independent of each other, and ways that combine information from both models and observations of the climate system. Different studies produce slightly different estimates (for both the mean and the uncertainty). But because there are multiple lines of evidence, we felt confident enough in the AR5 to make an expert assessment of the TCR – that it’s likely (more than 66 per cent probability) to be in the range 1.0°C to 2.5°C and extremely unlikely (less than 5 per cent) to be greater than 3°C.

Because we know the TCRs from the CMIP models, we can then relate their spread of projections to the assessed TCR range.

Region by region

So what can we say about regional climate change? Firstly, we need to think about what we mean by regional, as it means different things to different people. Here we are talking about regions the size of a large country or sub-continent.

One way to project changes on this sort of scale is to “zoom in” on the output from a global climate model. You can see an example of this in the figure below, it’s the projected change in rainfall in the South Asian region this century derived from global climate models.


Changes in June-September rainfall in in the South Asian region in 2081-2100, expressed as a percentage increase or decrease compared to 1986-2005. Hatching indicates where changes are not significant. Source: KNMI Climate Change Atlas.

You can see that the level of detail in the map is not particularly fine. Each square of the model grid represents 2.5 degrees longitude by 2.5 degrees latitude. That’s about the average for a typical global climate model. Depending on where you are in the world, it could be equivalent to an area about the size of the Republic of Ireland.

The map shows an intensification of the South Asian Monsoon rainfall in a warmer world. There are multiple factors that could affect this. The monsoon flow, which comes across the Indian Ocean before turning north and depositing its moisture-laden air is projected to weaken under climate change. But this is counteracted by a general increase in the available moisture as a result of warming. Overall, the moisture increase wins out and the monsoon rains intensify.

Those are the underlying mechanisms the global model can represent. Can we use it to put a figure on how much South Asian Monsoon rainfall will increase? This is where we run into difficulties.

We know that there are multiple factors that might affect how much rainfall the models project; the contrast between ocean and land-temperatures, local recycling of moisture between the atmosphere and the soil, the remote influence of factors such as El Niño, the poor representation of the Western Ghat mountains in these coarse resolution models, to name a few.

We do not have a large number of studies that attempt to quantify all these factors together in the same way that we have studies that quantify global temperature change. At best, we can be confident in saying that the monsoon rainfall will increase, but the crucial information of how much is still beyond our grasp.

Moving forward

How can the situation be improved? I think better coordination between studies that look at climate change in a particular region could help. Authors should be encouraged to try to say something about projections, not just to investigate the processes involved in bringing about changes.

Also, we need more research into developing generic ways of making regional projections that take into account what we know about the dynamics of regional climate, and which combine models with observations.

Professor Mat Collins specialises in quantifying uncertainty in climate projections and is a Co-ordinating Lead Author of Chapter 12 of the Intergovernmental Panel on Climate Change (IPCC) 5th Assessment Report. Follow @mat_collins on Twitter.


Expert analysis directly to your inbox.

Get a Daily or Weekly round-up of all the important articles and papers selected by Carbon Brief by email. By entering your email address you agree for your data to be handled in accordance with our Privacy Policy.


Expert analysis directly to your inbox.

Get a Daily or Weekly round-up of all the important articles and papers selected by Carbon Brief by email. By entering your email address you agree for your data to be handled in accordance with our Privacy Policy.