by Bob Henson
David Randall. (Photo courtesy Colorado State University.)
A new NSF science and technology center based at Colorado State University is devoted to improving the way clouds are portrayed in global models, an issue that's bedeviled climate scientists for decades. The Center for Multiscale Modeling of Atmospheric Processes (CMMAP) was launched in July. Its initial funding of $19 million runs through 2011, with the possibility of another $20 million over the following five years. Around one-third of the NSF funding will go toward education and outreach related to clouds and climate (see below).
More than 40 onsite staff and students will occupy a new building for CMMAP set to be completed in 200on CSU's Foothills Campus. CMMAP's research will also involve 1cost-sharing partners and 10 other collaborating institutions.
"People have been struggling with this problem for 40 years," says CSU's David Randall, the center's director and principal investigator, referring to the difficulty of simulating clouds in climate models. "A lot of good work has been done, but we need a breakthrough."
As anyone who has ever gazed skyward knows, clouds come in a stupendous variety of sizes, shapes, and types. A sheet of stratus might stretch across several U.S. states, while a fair-weather cumulus might span less than a kilometer (0.6 mile). Most climate models are forced to summarize cloud behavior across grid boxes that typically extend at least 50 x 50 km (30 by 30 mi) horizontally and at least 100 meters (330 feet) vertically. As a model run unfolds—typically in time steps of five minutes to a half hour—moisture enters and leaves clouds, sunlight is reflected or absorbed, and so forth. But there isn't much room for detail within the grid boxes, given the limits of computing resources. Thus, important but complex phenomena such as overlapping cloud layers and the effects of turbulence on clouds are depicted only crudely.
Climate modelers realize that their current approximations, or parameterizations, of cloud behavior aren't good enough. That's especially true when the updrafts and downdrafts induced by turbulence produce cloud elements as small as 100–200 meters (330–660 feet) across. The slice-and-dice effects of turbulence on clouds are most dramatic in warm climates and in the atmosphere's lowest kilometer, the boundary layer.
"Turbulence interacts with everything else—clouds, rain, radiation, the surface—in complex ways that a paramaterization can only hope to treat in a cartoon-like fashion," says Christopher Bretherton (University of Washington), a CMMAP collaborator.
Those who take flights over cool subtropical oceans such as the northeast Pacific can often see the vivid contrast of small, bright white cloud cells against dark blue ocean. It's the signature look of marine stratocumulus clouds, huge yet thin sheets that produce a mackerel-sky effect. NCAR's Chin-Hoh Moeng, who will serve as deputy director of CMMAP, has spent years modeling this type of cloud. "Marine stratocumulus carry a lot of moisture from the ocean," says Moeng. They also reflect a great deal of sunlight, she adds, making them a key part of the global radiation budget. Yet marine stratocumulus are still not directly depicted in global climate models.
One model inside another
In order to treat the deficiencies of global models in handling various types of clouds, CMMAP will turn to a different tool: cloud-system resolving models (CSRMs). Several dozen of these have been developed since the 1970s to study local and regional cloud behavior in detail. Each CSRM covers a small enough area with sharp enough resolution (typically about 2 km, or 1.2 mi) to depict many cloud types in a fairly realistic way. Most CSRMs are two-dimensional models, extending upward and along a single horizontal axis usually oriented east-west or north-south.
Wojciech Grabowski and Chin-Hoh Moeng. (Photo by Carlye Calvin.)
NCAR's Wojciech Grabowski was heavily involved in developing CSRMs during the mid-1990s. "We always knew clouds were a big issue in climate modeling. At some point I got the idea of using CSRMs to study this problem," Grabowski recalls. Using data from 20 years of field projects, he began exploring the idea with colleagues, including Randall. The challenge was how to make CSRM-level detail compatible with a global model's limited resolution. The answer—at least for now—is to nestle a set of mini-CSRMs within the global model (see graphic). Although each CSRM covers only a fraction of a grid point, the technique provides modelers with a far clearer picture of local cloud behavior. That picture can then be extended, in various clever ways, across each grid point and thus across the globe.
This approach, called superparameterization, is at the root of the new center's work. Not everyone is convinced it will bear fruit. "The thesis is controversial," acknowledges Bretherton. Based on recent studies conducted at CSU, UW, and elsewhere, says Bretherton, "superparameterization seems to fix some of the long-standing problems of climate models but not others."
On the plus side, these new models develop showers and thunderstorms at the right time of day, whereas traditional global models tend to produce rain a bit too early. The superparameterized models also depict a more realistic Madden-Julian Oscillation, the 40-to-60-day cycle that pushes winds and rains eastward across the tropical Pacific. Yet it isn't clear that the new models do a better job overall with the yearly cycle of winds, rain, and clouds over the global tropics.
The biggest weakness of superparameterization may be that it doesn't yet go far enough. Even a modern CSRM is still too coarse to capture the localized eddies that shape marine stratocumulus and other boundary-layer clouds. This fact helps explain the startling results from two recent studies. One, led by Matthew Wyant (UW) and Marat Khairoutdinov (CSU), employed CSU's superparameterized version of the NCAR Community Atmosphere Model. The other, led by Hiroaki Miura of Japan's Frontier Research Center for Global Change, used the first global model that resolves clouds directly at each grid point. In both studies, there was an enhanced negative feedback from global cloudiness that made the climate more sluggish in responding to a forced rise in sea-surface temperature. If confirmed, the finding could have major implications for how the climate responds to human-produced greenhouse gases. However, the authors and other researchers agree that the poor depiction of low-level clouds in these early models means it's too soon to draw such conclusions.
"One of the things we have to do is to improve the way that boundary-layer clouds are incorporated in these models," says Randall. "I'm very confident that we can make it better." In one approach, Bretherton is working with Khairoutdinov and UW's Peter Blossey to add a third player to the modeling team: a large-eddy simulation (LES) model with 100-m (330-ft) resolution. LES models will be added at each grid point, where they will interact with the CRSM and the global model.
What are NSF's Science and Technology Centers?
NSF recently funded six new Science and Technology Centers, or STCs (two in 2005 and four in 2006, including CMMAP) as part of its fifth round of STC funding since 1989. According to NSF, the centers are designed to support "innovative research and education projects of national importance." Knowledge transfer is another focus.
Each STC is funded for five years, with the option of renewal for another five. Home institutions agree to arrange cost-sharing that totals 30% of a center's budget. Forty STCs have been supported to date, with 17 of those currently receiving NSF support. Many of the 23 other centers whose 10-year NSF support has expired continue to operate with funding from other sources. These include the Center for Analysis and Prediction of Storms (University of Oklahoma) and the Center for Clouds, Chemistry and Climate (Scripps Institution of Oceanography/University of California, San Diego).
Along with innovations like these, the new center will also explore better ways to carry out more traditional types of global modeling. "Conventional parameterizations will still be needed for very long simulations," notes Randall in a 2003 overview paper for the Bulletin of the American Meteorological Society. After all, superparamerization doesn't come cheap: the embedded modeling approach can increase the cost of a typical climate simulation by a factor of 100 or more. One benefit of the extra detail, though, is that it enables a more direct comparison with high-quality data streaming in from current and future satellites, including the just-launched CloudSat platform (see sidebar below).
At NCAR, William Collins is eager to see how CMMAP expands on the legacy of a previous NSF cloud-climate center at the Scripps Institution of Oceanography (see right). Since Collins is a lead player in the Community Climate System Model as well as in CMMAP, he hopes to serve as a bridge between the new center and the larger modeling community. While he doesn't see superparamerization eclipsing more conventional techniques anytime soon, it's an important step forward, he says. "This is exploratory research. For those of us in the business of bread-and-butter climate simulations, I view it as a very useful exercise."
Teaching and communicating
As with NSF's other Science and Technology Centers, CMMAP plans to foster diversity and education on a variety of fronts. The center and its partners will generate a set of improved curricula in Earth system science at the middle and high-school levels, along with a teacher training course and internships aimed at K-12 teachers. CSU's Little Shop of Physics program is working with CMMAP to develop engaging methods of teaching about clouds and climate. "We'll look at questions like ‘Why are clouds white?' and ‘Why do clouds stay up?'—very interesting questions that have to do with the basic science of air, water, energy, and light," says Brian Jones, director of the Little Shop of Physics. Meanwhile, the public will learn about CMMAP science through the UCAR-based Windows to the Universe site as well as other Web vehicles.
As part of its education strategy at the undergraduate and graduate level, CMMAP will leverage existing programs that increase the pool of scientists from underrepresented groups. These include the Colorado Alliance for Minority Participation and, at UCAR, Significant Opportunities in Atmospheric Research and Science (SOARS), which will offer at least two protégés per year the chance to work on CMMAP science. "We are pleased to work with CMMAP and to build on a history of collaboration between SOARS and CSU's atmospheric science department," says SOARS director Rajul Pandya. "It is especially exciting to see a project like this begin with a commitment to broadening participation in atmospheric sciences."
CMMAP will also launch two entirely new vehicles to spread the word about the past, present, and future of climate modeling. One is an all-electronic, peer-reviewed, nonprofit journal, tentatively titled the Journal of Global Environmental Modeling (GEM). It will be offered online at no cost, with funding provided through page charges and agency funding. Global modeling results are now spread across journals in meteorology, oceanography, and math and physics, says Randall, adding that nothing comparable to GEM currently exists.
The other new vehicle is a book designed to capture the history of global modeling. CMMAP plans to solicit chapters from scientists heavily involved in cloud parameterization, while also including transcripts of interviews with others who embarked on climate modeling in the 1950s and 1960s at such institutions as NCAR, NOAA's Geophysical Fluid Dynamics Laboratory, Los Alamos National Laboratory, and the University of California, Los Angeles. The time for such a book is ripe, says Randall: "The pioneers of atmospheric general circulation modeling are now in or near retirement."
In a global climate model (left), each grid cell might represent an area as large as 2.8° latitude by 2.8° longitude, or about 300 x 250 km in mid-latitudes, with more than a dozen vertical layers. In the CMMAP strategy, the clouds within each grid cell are based on data from sets of small cloud-system resolving models (CSRMs, shown in red at left). A single cell might contain a row of 64 one-column CSRM models (right), each depicting clouds over a 4 x 4 km area (2.5 x 2.5 mi). Yellow arrows show where model elements extend farther than shown. (Illustration by Mike Shibao, based on imagery from CMMAP.)