Featured Story Corner
Estimating the State of the Southern Ocean
—Matthew Mazloff, Massachusetts Institute of Technology
The state of the world's oceans, both in the past and present, must be understood in order to determine and predict the Earth's climate. With the goal of Estimating the Circulation and Climate of the Ocean, the ECCO consortium was founded in 1998. The ECCO group faces a computationally massive problem that is only feasible thanks to computing centers like the San Diego Supercomputing Center (SDSC). With the support of ECCO and SDSC, I am diagnosing and evaluating the state of the Southern Ocean. This remote and hostile ocean plays a crucial role in the Earth's climate, and is the least understood of all the world's oceans.
The ocean's great capacity to store heat and greenhouse gases gives it a vital role in climate change studies. Climatic trends are only one motivator for ocean study; the oceans also play a significant role in many other issues of human concern. Carrying heat, salt, nutrients, pollutants, and icebergs, ocean currents affect fisheries dynamics, shipping, offshore mining, and international policy. Sea level height change is of serious concern of coastal communities. Sea surface temperatures, which influence storm development, also influence global precipitation patterns and are capable of bringing seasonal droughts and floods. This brief list explains why there is a desire to determine the global ocean circulation.
Over the last two decades an international undertaking has led to the development of several ocean observation systems. Data collection methods utilized by these systems include ships, stationary moorings, satellites, and autonomous floats. The combination of increased observations and computing power has given scientists the unprecedented ability to estimate the state of the global oceans.
As a member of the ECCO consortium, I am working towards illuminating the climate and dynamics of the Southern Ocean. The Southern Ocean has been identified as a critical component to the Earth's climate system. This ocean absorbs more wind energy and atmospheric CO2 than any other ocean. It is the primary means of communication between the Atlantic, Indian, and Pacific Oceans. Despite its significance, the Southern Ocean remains inadequately explained largely owing to its extreme energy and complex topography.
My goal is to produce and analyze a state estimate of the world's oceans between Antarctica and latitude 24 South. The utility and realism of previous Southern Ocean state estimates is limited because of an inability to produce solutions with spatial resolution adequate to define the so-called Rossby radius of deformation, which is critical to producing ocean eddy (turbulence) fields of adequate realism. With DataStar and support from SDSC, I am able to begin to resolve this scale. The results thus far are promising. Sample output can be seen in Figure 1.
What I am doing, in essence, is using a modern ocean model to interpolate between ocean observations in order to gain a complete picture of the ocean state. Using a model developed at the Massachusetts Institute of Technology, the MITgcm, I can calculate the time-varying ocean state, which consists of velocity, pressure, temperature, salinity, and sea surface height. I then systematically change input parameters (e.g. atmospheric temperature or wind speed) within their prescribed uncertainty to bring the model state into consistency with ocean observations in an iterative process. An objective (cost) function made up of weighted model-data misfit terms allows quantification of the "goodness" of the model solution. Formally, I am minimizing the objective function in a weighted least squares optimization problem, while using Lagrange multipliers to constrain the solution to be a model solution. For a brief description of the actual procedure, which is known in the oceanographic community as the adjoint method, please see the paragraph appended below.
The ocean estimation and prediction problem is possibly the most computationally intensive problem in science. Thanks to oceanographers, there will be a consistent demand for computing centers to build larger and faster computers. Each year of the state estimate optimization I am carrying out requires comparing on the order of 1012 ocean model state terms to on the order of 109 ocean observations. A weighted least squares optimization problem with over a billion terms is a formidable challenge, yet SDSC has prided itself on building machines that are able to handle the required memory and communication speeds needed in such calculations. SDSC's DataStar supercomputer has made large-scale high-resolution state estimation a feasible exercise.
My work on DataStar has already provided an improved estimate of the Southern Ocean state for the year 2000. A short-term goal is to further improve this estimate and to extend it through the year 2003. The result will be shared with the general scientific community via the World Wide Web to maximize its benefit.
The ocean has a much longer memory than the atmosphere and ocean prediction will only be possible if we have a satisfactory initial state. My work is aimed at this initialization problem, but it is also a step toward the long-term goal of the ECCO consortium to provide (with the help of SDSC) an estimate of the global oceans from the last decade, to the present and beyond.
For more information please visit the ECCO website
This research is supported by the San Diego Supercomputer Center (SDSC), and by the National Ocean Partnership Program (NOPP) through the ECCO and ECCO-GODAE Consortia. Patrick Heimbach, Carl Wunsch, and the rest of the MIT-ECCO group are essential players in this project. Matthew Mazloff is reachable via email at firstname.lastname@example.org