Climate models are less reliable in predicting climate extremes such as floods
Policymakers must find flexible plans to ensure water security despite the uncertainties of climate change, says climate scientist Mark New.
Much of the developing world is dependent on rainfall, snow or ice for its natural water supplies, so researchers and policymakers are concerned about how climate change will affect water security.
Long-term water availability is largely determined by the average conditions in a river basin, but it is often extreme events that cause acute water insecurity. Importantly, water security is not limited to scarcity — excess water can be devastating too, as the recent flooding in Pakistan shows.
As a result, assessing how climate change affects water security requires information on changes to both the mean and the variance of climate, especially rainfall, on a range of timescales, from hours to decades.
Uncertainty in predictions
Uncertainty in projections from global climate models fall into three categories. 'Initial condition uncertainty' relates to the chaotic nature of weather and climate, where a lack of information about the current climate means we cannot — even with a perfect model — precisely predict the climate. Running multiple simulations, identical except in their initial conditions, can help quantify this uncertainty.
'Model uncertainty' arises from imperfections in climate models, which are, after all, computerised simplifications of the real climate system. We can get some measure of this uncertainty by looking at the spread of results from different models, or by running the same model several times with different configurations for key processes such as precipitation.
'Forcing uncertainty' relates to doubts about the level of future emissions, and how much of them will end up in land or ocean sinks, or in the atmosphere. The lack of a concrete global emissions policy means this uncertainty remains large.
All global climate models project a global average increase in rainfall as the climate warms. Most models agree that rainfall will increase over mid-latitudes and the wet tropics, and decrease over the dry tropics and sub-tropics.
But projections at a regional scale are more diverse. Even where models agree on the direction of rainfall change, there are considerable differences in magnitude.
Climate models tend to be less reliable in representing and predicting climate extremes. For example, some models fail to adequately represent the dynamics of the El Niño–Southern Oscillation, which is essential for predicting drought, or processes that can lead to intense flooding events, such as tropical storms and heavy monsoon rainfall.
At the finer spatial scales required for on-the-ground assessments, interpreting output from global climate models is particularly tricky. Their coarse resolution means they often miss processes such as topographic forcing and land-surface feedbacks that are important for local climate.
A large 'downscaling' industry has emerged to address this weakness, but the results from different approaches are extremely variable.
Improvements to models over the coming years may help to overcome some of these limitations. But most climate scientists expect the spread of uncertainty to widen before it starts to narrow.
So, for the foreseeable future, users of climate model projections will have to deal with large uncertainties.
Ensuring water security
We can expect areas that are currently arid or semi-arid to get drier, and average rainfall intensity is likely to increase almost everywhere. But when looking at specific cases, with some inevitable uncertainty in future climate, policymakers planning to ensure water security must employ strategies that account for this uncertainty.
A relatively easy first stage is 'climate screening' — assessing whether the local water security situation is sensitive to climate. It may be that climate change is not an issue, for example where water scarcity is driven by political or socioeconomic factors such conflict, poor infrastructure or pollution, or where a problem results from existing climate vulnerabilities.
Where climate is a factor, data from climate models can be used to explore where the situation sits relative to the 'envelope of uncertainty' — the range of possible future climates projected by climate models.
If, after this screening, climate change turns out to potentially important, more sophisticated analyses may be needed. These include techniques based on decision theory and risk analysis that identify the options that are most robust to the uncertainty.
Crucially, outputs from climate models must be carefully interpreted, as some models may be less reliable than others in particular geographical domains. This requires climate scientists to be involved in the analysis and planning process, rather than simply providing data to end-users.
In many instances, policymakers can help local communities cope with today's water shortages by building resilience to existing climate vulnerabilities.
For longer-term initiatives, such as transboundary treaties or large infrastructure developments, including dams, the key is to avoid getting locked into situations that can be exacerbated by climate change.
This means choosing the easy, 'no regrets' options first, and ensuring that flexibility is built into any strategic plans or agreements.
Policymakers must not allow uncertainty about climate change to be an excuse for inaction — governments, businesses and individuals make decisions based on uncertain information all the time, and climate change should not be any different.
Mark New is a Reader in climate science at the University of Oxford and Tyndall Centre for Climate Change Research, United Kingdom.
All SciDev.Net material is free to reproduce providing that the source and author are appropriately credited. For further details see Creative Commons.