BIG DATA

Oceans cover nearly three-quarters of the Earth’s surface and are constantly in motion. This vastness and dynamism makes it difficult to get a sense of the world’s oceans as a whole, and to understand how they’re changing.

In 1964, when Carl Wunsch, professor of physical oceanography at the Massachusetts Institute of Technology, first went to sea, oceanographers faced old and incomplete data and models that didn’t match the reality of ocean currents at all.

“For 25 years, I’ve been trying to solve the problem of observing the global ocean as a whole, on the grounds that you can’t understand the ocean’s climate implications without being able to perceive and understand the whole system,” Wunsch said. “Recently, observational systems have been developed that allow one to observe the oceans at least as fast as they’re changing.”

New and improved monitoring devices — from floating drifters to satellites to instrumented and radio-equipped elephant seals — have made massive amounts of data available to researchers. However, these diverse streams of information are a challenge in their own right.

“Having in some sense solved the problem of how to observe such a difficult system, the question became, what to do with all the data?” Wunsch asked. “You have different kinds of data that appear at irregular temporal and spatial intervals, and you need to synthesize these into a picture of what the ocean is doing day by day. This led to my heavy use of computational facilities.”

Using the Ranger supercomputer at the Texas Advanced Computing Center (TACC) — as well as high-performance computing systems at NASA, NOAA, NCAR and other systems on NSF’s TeraGrid — Wunsch and his colleagues from the ECCO (Estimating the Circulation and Climate of the Ocean) Consortium have developed increasingly subtle methods to coordinate the billions of pieces of observational data now available. By synthesizing diverse remotely-sensed and in-situ observations with known dynamics and thermodynamics, ECCO creates a working ocean model that can characterize the entirety of the ocean at any given time.

“I’m trying to understand what the ocean is doing and why it’s doing it, with the hope, one day, of being able to say something about what it might look like in the future,” Wunsch said. “But that’s a much more difficult problem.”

One large chunk of the ocean that requires special scrutiny is the Southern Ocean — the waters surrounding Antarctica. Complex topography, extremely energetic currents, and a lack of historic observations make the dynamics of these remote waters particularly difficult to understand and explain why so little is known about the Southern Ocean. Yet this body of water, which is the main communication link between the other major ocean basins, is a vital component of the Earth's climate.

Under the guidance of Wunsch, Matt Mazloff, now a post-doctoral researcher at Scripps Institution of Oceanography at UC San Diego, is using Ranger to develop state-of-the-art assessments of the Southern Ocean. His regional model — which incorporates emerging observations and increases the resolution of the ECCO model from one latitudinal degree (about 100 kilometers per grid space) to one-sixth of a degree (16 km.) — has led to unexpected discoveries.

“A main feature of the Southern Ocean is the Antarctic circumpolar current, which is probably the biggest current system on earth. It was originally thought to be one massive current, but as research evolved and we started analyzing satellite data, we realized that it’s not one massive current but a whole system of currents — in some places as many as nine streams, sometimes converging to two,” Mazloff explained. “But these currents are small by ocean scales, so you couldn’t resolve them with the global models. If you wanted to see what the Southern Ocean was doing daily, you needed to resolve these fast narrow streams of current with a high-resolution model. That’s where Ranger came in.”

Mazloff’s high-resolution simulations have used nearly two million computing hours on Ranger, and three million hours on Datastar at the San Diego Computing Center, where the project was begun, to create increasingly accurate models of the circulation and conditions of Southern Ocean from 2005 (when observations in the region reached critical mass) to 2007. Over the next year, he will be updating the model to bring it to the present date.

Mazloff’s simulations reveal the complexity and energetic nature of the Southern Ocean system, and is guiding scientists working on the International Polar Year project, a collaborative international research effort that ended in March 2009.

“Observationalists trying to figure out how to do their process studies have been interested in using the product that we’re developing on Ranger because they recognize that it’s important to see these small scale features,” Mazloff said.

This active interest in computational models represents a landmark for oceanographic simulations, which for a long time were not widely used by observational researchers. “Now that the modelers and the computer and numerical theoreticians have put this model together, it’s starting to get used by the observationalists as it was intended,” said Mazloff. “This wouldn’t have happened without high-performance computing.”

Mazloff’s model is assisting researchers who are looking to quantify the mixing of properties (e.g. heat, nutrients, pollutants) in the Southern Ocean, which has significant consequences for climate and policy. For instance, the research team from the DIMES project (Diapycnal and Isopycnal Mixing Experiment in the Southern Ocean, a US/UK field program aimed at measuring mixing in the Southern Ocean), plans to drop a tracer, or dye, at about 1000 meters deep, and then go back every year to monitor where the dye has moved and in what concentrations — information that will help determine how oceans blend.

Researchers used the output from Mazloff’s state estimate to decide where to drop the tracer to best study this process; they will continue to use the model output to locate and track the tracer in the coming years.

“Historically, this study would’ve been carried out by the observational side,” Mazloff explained, "but now they’re using these model syntheses performed on high-performance computing systems to guide their study and to do the post-analysis.”

Other uses for Mazloff’s model include studies of the carbon cycle — exploring how much CO2 is taken up by the Southern Ocean — and geodesic research, studying Earth’s gravitational field, where ocean conditions play an important role.

Most valuable, however, is the fact that Mazloff’s model provides a baseline description of the current state of the Southern Ocean, which has until now been unknown. Unlike the atmosphere, which has been systematically studied for hundreds of years, researchers have only been trying to comprehensively understand the oceans for half a decade. Add to that the much smaller scales of oceanic weather, complex topography, and longer “memory” of the oceans, and it’s only with the emergence of petascale systems like Ranger that computational oceanography is coming into its own.

“I think people are surprised when they hear that we don’t know what our oceans are doing,” Mazloff said. “It’s only recently that we actually started to observe this huge part of our planet. So this is our best estimate of what the oceans look like on any given day in the modern time.”

Oceans are important in and of themselves, but increasingly scientists are realizing the integral role they play in global climate change. This has led many researchers to try to use the ECCO models to predict the future state of the oceans. However, according to Wunsch, too much remains unknown about the present to look too far ahead, as some scientists have tried to do.

“Oceanographers have never been in the prediction business, because we don’t know enough,” Wunsch explained. “If you don’t know today’s weather, you can’t predict the weather for tomorrow or next week. We’re still in the process of determining the equivalent of today’s weather in the oceans, and then deciding if there is predictive skill.”

Wunsch is quick to note that uncertainty doesn’t equal inaction. Policy makers need to make decisions about climate change knowing that the uncertainty is not going to diminish any time soon.

“That’s a very unpalatable message,” Wunsch said. “However, we can say a great deal about the possibilities and the threats. We can say there’s a possibility that the current rate of sea level rise will accelerate and if it does, we’re going to have a real problem and maybe we should be thinking now about what we’re going to do about it.”

Long-range study, increased observation, and ever-improving ocean models, may eventually enable scientists to predict distant changes in the oceans and their impact on our climate, and with the help of systems like Ranger, we may be able to learn enough about the oceans to change course before it’s too late.

******************************************************************************************************************************

Funding for the ECCO Consortium has come principally from the National Ocean Partnership Program and NASA. To learn more about this research, visit the ECCO consortium homepage or explore the International Polar Year project.

Aaron Dubrow
Texas Advanced Computing Center
Science and Technology Writer