1663
|
|
Predictions of Arctic Climate ChangeLos Alamos–based computer models project that by 2040 the Arctic Ocean may be ice-free for part of each year, bringing devastation to Arctic inhabitants. What makes this projection believable? And does it portend other abrupt changes in climate? Los Alamos–based computer models project that by 2040 the Arctic Ocean
may be ice-free for part of each year, bringing devastation to Arctic inhabitants. What makes this projection believable? And does it portend other abrupt
changes in climate? Photo credit: Jeanne M. Bowles
Abstract: Los Alamos computer models of ocean and sea ice circulation are enabling forecasts of Arctic sea ice melting and its possible impact on global climate change. Those forecasts project that, unless steps are taken now to reduce greenhouse gases, the Arctic Ocean will, by 2040, be ice-free for part of each year, bringing devastation to Arctic wildlife and altering ocean circulation and regional climate patterns. At a distance, the vast expanses of polar sea ice seem static and sublime, but they are actually in constant slow motion and internal change. Los Alamos researchers Elizabeth Hunke and Bill Lipscomb know that firsthand. Spending weeks onboard an icebreaker will teach you to read the signs. Ice floe extent, floe thickness, snow cover, and ratio of ice to open ocean—all change with location and season, and now with global warming. Hunke and Lipscomb are members of the Los Alamos effort, begun nearly 20 years ago, to apply the power of supercomputers to the problem of modeling climate change. Both have contributed for over a decade to accurate computer models of floating sea ice. Today their models are sought after by worldwide climate modeling groups who are trying to predict how rapid changes in the entire Arctic and parts of the Antarctic will impact Earth's climate. Sea Ice and ClimateSea ice plays a sensitive role in Earth's climate system for one simple reason: it is white and therefore has high albedo, or reflective power. While covering only 4 percent of Earth's surface, it accounts for perhaps 8 percent of the planet's total reflected energy, so it contributes significantly to the delicate balance between the energy delivered to Earth as sunlight and the energy radiated from Earth's surface as heat.  A typical icescape near the SHEBA field camp in the Arctic Ocean, May 1 1998. The snow ridges (sastrugi) are formed by wind action.
Photo credit: Mark Southwell, H.M.S. Endurance
Today, global warming is causing the sea ice to melt and give way to open ocean. This has caused the Arctic's albedo to decrease, and the increased absorption by the dark ocean surface causes more warming. Thus, melting begets melting in a feedback loop that amplifies the warming trend. In the Arctic, a second process is also at work. Ice frozen in the Arctic Ocean flows into the North Atlantic and melts, diluting the dense, salty water brought northward by the Gulf Stream and possibly disrupting the density-driven movement of global ocean currents. Understanding the possible strength and ramifications of these events requires a good sea ice model incorporated into to a full climate-system model. Tapping Los Alamos' Computing PowerGlobal warming was already a major concern in 1990, when the Department of Energy challenged climate researchers to ramp up their predictive powers. The department suggested adapting climate models to new, massively parallel supercomputers. 
Stephen Ackley and Elizabeth Hunke drilling a sea ice core in the Weddell Sea, Antarctica, in 1998. Photo credit: Mark Southwell, H.M.S. EnduranceThese computers, called "parallel" because they consist of hundreds of individual processors wired together and working simultaneously, had increased computer power by a factor of 10 (now a factor of 1000s). They should, it was argued, be able to handle more-realistic full-climate-system models that included atmosphere, land, ocean, and sea ice components. Of course, the component models would have to be rewritten to work in a parallel mode. Los Alamos had the computer power, having just acquired a Connection Machine from Thinking Machines, Inc., the first modern parallel computer. But it was a relative newcomer to climate modeling. Nonetheless, Bob Malone of Theoretical Division knew that the Lab's strength in fluid-flow modeling could make a big difference to ocean models. It was a two-part challenge—improve the physics and make the models run efficiently on the new parallel machine. By 1993 Malone and colleagues had developed the Parallel Ocean Program, or POP, the first model to accurately represent the path of the Gulf Stream. POP also set a new standard in computational efficiency and was soon chosen as the ocean component of one of the world's leading climate models, now known as the U.S. Community Climate System Model.  Bill Lipscomb’s height relative to an Arctic sea ice ridge.
The sea ice component, however, remained a stumbling block, and Malone suggested that John Dukowicz, a POP developer, team up with newcomer Hunke to develop a fresh approach. "Dukowicz and Hunke were as immune as the rest of us to the 'this is the way we've always done it' mindset, "Malone says proudly, "and they completely rethought the traditional methods for calculating sea ice dynamics." How Arctic Ice Floes FlowDriven by fierce winds and ocean currents, Arctic sea ice advances across the North Pole from Siberia to northern Greenland in a slow-motion walk known as the Transpolar Drift. About 10 percent annually flows into the North Atlantic through the Fram Strait, between Greenland and Svalbard, Norway. Some sea ice gets trapped by the huge Beaufort Gyre, north of Canada and Alaska, and circulates for years or piles up near the Canadian coast. That older, thicker ice gradually feeds into the Transpolar Drift, completing the turnover of Arctic sea ice about once every 10 years. Within the two major drift systems, individual ice floes crash together and deform into ridges. Under tension, they crack and split into many pieces. Sunlight reflects from their surface but also penetrates to warm their interiors and the surrounding ocean, causing summer's melting. Winter's cold, turbulent winds cause freezing. No computer model can account for all the large and small motions and thermal changes of the yearly cycle of melting and freezing—there simply isn't enough computer power. So the important processes must be abstracted—idealized—and converted into equations. As their biggest abstraction, sea ice models treat the Arctic's millions of square miles of individual floes and ice packs as a continuous flat layer of viscous, plastic material (like tough putty or very dense molasses) that slows down as it stiffens. This two-dimensional layer, sandwiched between atmosphere and ocean, is divided into a grid of computational cells, about 60 miles on a side. When forces push more ice into a given cell and create thick ridges, the increase in thickness makes the ice more rigid in that cell. The traditional treatment of sea ice as a viscous-plastic sheet imitates reality very well, except where the sheet encounters an obstacle and must stop. In that area the modeled rigidity of the ice pack increases and spreads, and much computational work is required before disturbances in this region die down. Thus, a barrier encountered in one area dramatically slows the entire calculation. That's a disaster when the goal is to simulate decades of sea ice changes and their global consequences. Breaking the Computational BarrierHunke and Dukowicz broke through the computational barrier by introducing a clever numerical scheme. "When a portion of the material begins slowing down, the energy of its forward motion is converted to elastic waves that, like sound waves in a tuning fork, cause the material to effectively 'vibrate' in place without deforming, while the neighboring ice floes move past," explains Dukowicz, who initiated the scheme.  The two main sea ice drift systems in the Arctic the Beaufort Gyre and the Transpolar Drift.
The new elastic viscous-plastic scheme mimics the real behavior of the ice, affecting only the local region that encounters the barrier. And it is fast and efficient on modern parallel supercomputers. Similarly, Lipscomb, a newcomer in 1998, worked to model thermal processes more realistically. He improved the model representation of how brine pockets affect heat flow, melting, and freezing within each computational cell. Also in 1998 Lipscomb traveled to the Arctic as part of the international SHEBA (Surface Heat Budget of the Arctic) project. There he learned that sea ice in the central Arctic had thinned from about 10 feet to 6 feet or less in the 20 years since the previous such project. Anomalies associated with strong winds were part of the explanation, but then in 1999, newly released submarine sonar measurements showed that overall Arctic sea ice thickness had decreased about 40 percent from the 1970s to the 1990s. Since sea ice thinning must affect heat flow between ocean and atmosphere, Lipscomb added to the thermal model the ability to represent variable sea ice thickness within a computational cell.  An aerial view of Arctic sea ice. Leads (large cracks tens to hundreds of yards across) and smaller cracks form as sea ice moves.
The improved thermal model fit smoothly with the Hunke-Dukowicz dynamics model, and together they became known as CICE (pronounced "sea ice"). CICE can be downloaded from the Web, and users find it well documented and easy to run. "Our users send us questions when they get funky or suspicious results, and sometimes that leads to improvements in the model. Also, because many people have applied it to different problems, the program is quite trustworthy. I hate to call it 'bug-free,' but 'almost' is OK," says Hunke wryly. Is CICE a Crystal Ball?CICE's accuracy has been assessed by comparing "hindcasts" (simulations of past sea ice changes) with observations.  The U.S. Community Climate System Model forecasts a rapid decline in average September (minimum) sea ice extent. White areas have at least 50 percent ice coverage. The model projections for 2010–2019 (left) and 2040–2049 (right) show the decline relative to the observed ice extent averaged for the 1990s (red curve).
For example, when CICE is used along with POP in the U.S. Community Climate System Model, it accurately reproduces satellite observations of ice pack movements over the cycle of seasons in both the Arctic and Antarctic—including current variations in the expanse of Arctic sea ice from about 2.5 million square miles at summer's end to 6 million square miles at the end of winter. In addition, when major components of CICE are used in both the Community Climate System Model and the United Kingdom's Hadley Centre Climate Model, those two come much closer than other models to reproducing the dramatic decline in summer Arctic sea ice from 1957 to the present. In view of these hindcasting successes, CICE might well be a crystal ball for predicting the Arctic's future. For a scenario in which greenhouse gas emissions follow a middle-of-the-road projection, the Community Climate System Model, with CICE and POP components, forecasts that Arctic summer ice could decline precipitously in coming decades and almost disappear by 2040.  New Focus—The Greenland Ice Sheet - How fast will it take for the ice covering
Greenland to melt? "That’s the great unanswered
question we’ll try to answer next," says Phil Jones, present leader of climate modeling at Los Alamos.
Large outlet glaciers in Greenland have accelerated their movement to the coast and thinned during the past decade. Scientists are concerned that the entire Greenland ice sheet could decay within a few centuries, instead of over several millennia, as previously believed.The melting would raise sea level by about 20 feet. Melting of the West Antarctic ice sheet could contribute another 15 feet. Existing ice sheet models, which track the movement of ice on land, are too crude to predict these changes, in particular, the initiation of rapid sliding at the base of the ice sheet. Los Alamos plans to develop a more accurate model to account for processes that enable accelerated sliding, for example, ice melt reaching and lubricating the base of the ice sheet. The model will then be coupled to full-climate-system models to study the feedbacks that might result as the ice sheet retreats. Could the ice sheet's shrinking trigger local climate changes that further affect the rate of retreat? Could the retreat alter the ocean circulation? What role do ice sheets play in transitions between interglacial and ice age climates? Modeling is the key to answering these questions. [The figure shows Dr. Alberto Behar of the California Institute of Technology, Jet Propulsion Laboratory, standing beside a moulin, or a large vertical shaft. Surface meltwater flows into the moulin and plunges deep into the Greenland ice sheet.]
When the initial conditions of a computer run are varied to represent natural climate variation, the pattern of sea ice decline shows one or two brief periods of abrupt change (30 percent or more), flanked by slower changes. Interestingly, these abrupt events were triggered by unanticipated influxes of warm water from neighboring oceans into the Arctic Ocean. The complete disappearance of Arctic sea ice for part of the year would destroy the habitat of polar bears and other species dependent on floating ice. In addition, the associated Arctic warming could trigger the onset of other major changes, including accelerated melting of the mile-thick ice sheet covering Greenland. The melting of so much ice from a land mass would cause a rapid rise in sea level. (Sea ice already displaces its weight in water and does not raise the sea level when it melts.) Arctic sea ice melting may be a warning for the whole planet: if greenhouse gases double or triple by the end of this century, global warming could wreak havoc on our only home in the universe. It seems climate modeling has come of age just in the nick of time, allowing humanity to project into the future the consequences of its actions and to develop and evaluate strategies for both adapting to and mitigating the changes that are already in motion.
References: 
|