Image shows sea surface temperatures.
At Los Alamos, Supercomputers Tackle the Oceans
June / July 2005
Volume 3 Number 3
Scientists at Los Alamos National Laboratory, led by Project Leader Phil Jones use supercomputers to model our planet's oceans. Jones leads the Climate, Ocean and Sea Ice Modeling (COSIM) project. COSIM has succeeded in creating an accurate predictive model that can anticipate changes in ocean currents caused by changes in the weather. Also, Jones's ocean model has formed a key part of comprehensive global climate models. The ocean model forms a central piece of larger weather models because, as Stephen Lee, computer and computational sciences deputy division leader says, "The ocean drives the climate."
Computers solve some problems with ease, but other problems are extremely difficult for computers to tackle. Ocean currents present some of those difficult problems. A simple problem would be one where a mathematical calculation is done once, and the answer to that problem does not change any other calculation the computer has to perform. More typical computer models use the answers to sub-problems to get answers to new problems. These "normal" computer models finish one part of the problem, store the answer, and go on to solve the next problem. For example, a computer model of a boxer punching a bag could calculate the force applied by the boxer's fist and add that to the force applied by the swinging bag. Such models are "linear" and represent many of the predictive computer models of simple physical systems.
In contrast, ocean currents are "non-linear." Effects do not simply add together as they do in the boxer/punching bag example. Instead, currents, eddies, salinity, temperature and wind all interact with each other in complicated ways. A computer's linear way of thinking does not easily cope with complicated non-linear models. For this reason, realistic ocean simulations use supercomputers. Because simple calculations do not yield realistic answers, supercomputers run very large numbers of interacting calculations on small, simulated ocean pieces where small-scale models are realistic. Only when these small-scale models are combined do realistic ocean effects emerge from the larger model.
"It's a very complex non-linear physics problem," says Lee. "We have to model everything from the formation of eddies, the movement of currents, carbon sequestration, turbulence, the interaction of temperature and salinity and the formation of clouds. There is a lot of detail. A lot is going on." As Jones describes it, "For high-resolution modeling in particular the ocean in general has very small spatial scales and very long time scales. That means we need to simulate a very large number of ocean points over a very long time. To do that we need supercomputers."
The problem is complicated in particular by eddies. Eddies (whirlpools) spin opposite to the main current, and getting eddies right is the key to getting an overall ocean model right. "A lot of the kinetic energy in the ocean is governed by these eddies," says Jones. "They have a very strong role in surface currents of the ocean—€”things like the Gulf Stream. To get the Gulf Stream right, you have to solve the eddy structure."
LANL's supercomputer ocean model—€”the Parallel Ocean Program (POP)—€”gets eddies right. In POP, the whirlpools and backwaters that form on the ocean surface emerge naturally from the supercomputer simulations. These simulations closely resemble real ocean eddies, and when these eddies are combined in a global ocean mode, the resulting simulated ocean currents act in the same way sailors from the time of Columbus on have experienced.
Jones' team uses several supercomputers across the world, including the Earth Simulator in Japan (the world's third-fastest supercomputer) and LANL's own 256-processor supercomputer.
POP is currently seeing major use in two settings. First, POP forms the ocean portion of a comprehensive model of weather change used by the National Center for Atmospheric Research. Second, the U.S. Navy is using POP to predict ocean currents. "It definitely works," says Jones.
Jeffrey Stewart is a business development executive, technology
transfer, Los Alamos National Laboratory.
Copyright © 2012 | Innovation America