- UCAR Home
- About Us
- For Staff
June 4, 2014 | Twice as much flour and water should give you twice as much bread. Earth’s atmosphere doesn’t respond to greenhouse gases quite that simply. But researchers are finding new ways to work with some aspects of climate change that are surprisingly linear.
This work could point the way to saving time and money in future climate research while producing a richer set of studies—and a richer range of information for policy makers—on various climate change scenarios and how society might mitigate and/or adapt to the changes.
Several dozen experts gathered at NCAR in April to discuss “pattern scaling,” one of several statistical approaches that allow scientists to interpolate between two highly detailed simulations and get simpler but still-useful results that fill in the gap between them. Pattern scaling is designed to map out how climate might evolve if emission pathways differ from those that have been studied intensively through full simulations with global climate models.
Part of the challenge in preparing for climate change is that it can take thousands of hours of costly supercomputer time to produce a single projection of how climate might unfold over the next century. Moreover, nobody knows the extent to which nations and industries might act to stem the rise of greenhouse gases.
Because of this, it takes multiple simulations—with emissions allowed to plummet in some runs and to grow unchecked in others—in order to provide a broad range of possible trajectories for global climate.
For example, the modeling carried out for the recent Fifth Assessment Report from the Intergovernmental Panel on Climate Change involved four different “representative concentration pathways.” Each of these RCPs specify the amount of carbon dioxide in the atmosphere expected by a certain point (such as the year 2100) (see graphic).
Pattern scaling aims to fill in the gaps between these simulations. The technique could assist policy makers who want to know the impacts of a different emission level, such as something in between two of the four existing RCPs. It’s also widely used in other types of research on climate impacts.
Pattern scaling appears most effective in projecting changes at large regional scales, where scientists expect a climate model to be most skillful. For example, it’s well established that emitting more greenhouse gases will tend to make subtropical areas drier and subpolar regions wetter. The models also show polar regions warming more quickly than lower latitudes, a trend already being observed. These changes should be roughly proportional to how much global average temperature rises, which in turn is a function of how much total greenhouse gas is emitted. The latter relationship is easily simulated with very simple and inexpensive models. Pattern scaling allows those results to be translated into projections of future climate change across large regions in a computationally efficient way.
Not all phenomena lend themselves to the pattern-scaling approach. It doesn’t capture the effects of internal (natural) climate variability, which can temporarily mask the signal of human-caused changes. As an example, much of the southeastern United States has seen little temperature rise over the last century. Climate models didn’t predict this regional “warming hole,” since they’re not designed to track the ups and downs of internal variability. It’s also difficult to use pattern scaling to determine how quickly the warming hole will diminish, as it’s eventually expected to do. Pattern scaling also has its limits when warming pushes climate past a major threshold, such as the melting of Arctic summertime sea ice.
Participants came out of the workshop agreeing that it’s time for a more thorough evaluation of pattern scaling. They want to determine where and how pattern scaling might be used more extensively. For example, it could play a role in research leading up to the next major assessment from the Intergovernmental Panel on Climate Change.
“Pattern scaling can’t completely replace the need for comprehensive Earth system model runs, but perhaps we can rely on it more than we think we can now,” says NCAR scientist Brian O’Neill, who helped coordinate the workshop.
Workshop website: "Pattern Scaling, Climate Model Emulators, and their Application to the New Scenario Process," April 23-25, 2014