PAOC Spotlights

Digital Computing for the Atmosphere, Oceans and Climate

Wed May 8th, 2013
Helen Hill

Chris Hill's work centers on how digital computing technology in its broadest sense applies to Earth science and especially the ocean. 

Computing in general has a long and proud history at MIT. An iconic image from the 1930's shows Vannevar Bush next to an analog difference engine built to integrate differential equations that could not be solved analytically. It is one of the many historical MIT images that epitomizes the mens et manus spirit William Barton Rogers first espoused in Boston in the 1860s. Bush was aided in his work by an MIT student, Claude Shannon. At the same time as wrestling with Bush's mechanical computer, Shannon authored a masters thesis articulating a mapping between Boolean logic and electrical circuits. Arguably, the ideas in that thesis underpin much of modern digital computing. Thanks in part to Shannon's thesis work, it ultimately became possible to integrate forward differential equations that describe fluid, chemical and biological systems expressed not as mechanical gears but rather as functions and modules in a digital computer program. Going back to the 1950s, previous generations of researchers at MIT, including Jules Charney, Norm Philips and Ed Lorenz, used such digital capabilities to explore computational solutions to the equations of motion for a stratified fluid on a rotating sphere. Moving forward technologically, in the 1980's, Danny Hillis (an MIT doctoral student of Claude Shannon) was an important trail blazer in another computing revolution, around building computers out of thousands of networked processors with very close coupling.

Fast forward to 2013 and these historical threads are now firmly joined. We can now deploy digital computer technology at scales that were unimaginable not that long ago. We can formulate digital computer programs that employ literally trillions of transistors working in concert to compute quantities of interest. We can use this technology to ingest data from sensors at unprecedented rates, to explore the behavior of sets of billions of discrete equations that we are unable to comprehend otherwise and to analyze and visualize large volumes of digital information. Today we use computational models to examine how basic ideas and theories play out in the context of virtual approximations of real-world complexity. A recent example is work that researcher Oliver Jahn is working on with Chris Hill to understand better the role of the ocean in atmospheric CO2 budget variations. A number of recent satellite missions in Europe, Japan and the US have focused on measuring atmospheric column inventories of CO2 with a goal of mapping sources and sinks of CO2 on an ongoing basis. This is information that can potentially inform our basic understanding of planetary dynamics as well as contributing to thinking and debate around planetary stewardship and global sustainability. To fully interpret the satellite observations estimates of air-sea fluxes of CO2 and their spatio-temporal variability and uncertainty are needed. While it is well known that the uptake and out gassing of CO2 by the ocean varies spatially and temporally, the magnitude of the variation is not well characterized. Jahn and Hill, along with colleagues Holger Brix and Dimitris Menemenlis at JPL are using innovative computational capabilities to examine these questions. Their work relies on computational tools developed by multiple generations of researchers in the ocean group EAPS. These tools are used to combine observations and theoretical understanding to create a baseline estimate of the ocean carbon budget for the last ten years. Then a second set of computational tools is used. These have been developed by researchers at Argonne National Laboratory in close collaboration with EAPS ocean researchers Chris Hill, Patrick Heimbach and Carl Wunsch. This second set of tools can reformulate a compute program that numerically calculates a high-dimensional function into one that calculates the sensitivity of that function with respect to other variables. In the case of the Jahn and Hill's current work, this tool allows them to compute maps that show quantitatively how sensitive air-sea exchange of CO2 is to aspects of ocean dynamics. Using computation in this way can help see how and where different processes and pieces of the Earth system interact. It seems a far cry from the work of Vannevar Bush on integrating trajectory equations using mechanical wheels, but it owes much to that work and work that followed.

Chris Hill also maintains a watchful eye over the so-called M.I.T. General Circulation Model (or MITgcm) project. A technological underpinning of several efforts in EAPS including the ECCO and Darwin projects, the MITgcm is a remarkable collaborative effort. It is a software initiative that has been effectively growing within EAPS for nearly 20 years. Over that time it has had numerous contributors and many more users. Today it is one of the most widely used university based global ocean models. Its key strength is its diverse capabilities and community of applications. The scientific literature includes applications to exoplanets, to gas giants and to brown dwarf stars in addition to ocean and atmosphere problems closer to home. It has also been the origin of stunning visual representations of the richness of ocean behavior. This has included forays into the media worlds with visualizations that garnered more than one million youtube viewers (even earning the model a New York Times editorial) as well as numerous cameo roles throughout the recent NOVA Earth from Space special. Like many successful infrastructures, its role can sometimes be forgotten. It would be remiss not to acknowledge this piece of infrastructure that ocean researchers in EAPS, together with a broad collection of associated collaborators, voluntarily contribute to the field. It is, after all, another in a line of examples that show how supporting computational research can pay back many times over, not just through direct research results but also through the furtherance of technology that can benefit many other research endeavors.

Links

Chris Hill specializes in the application of large-scale computation to all aspects of understanding Earth and planetary systems. He is a founding developer of the M.I.T. General Circulation Model (MITgcm), a numerical simulation tool used for a wide range of basic science and applied studies in planetary fluid dynamics. Hill  also helped launch the Earth System Modeling Framework (ESMF), a major open standard for creating multicomponent models of Earth system processes. In both activities he has been active in developing prognostic model and model-data synthesis tools.

Among various projects, Hill co-leads the research, education and outreach committee of the Massachusetts Green High Performance Computing Center (MGHPCC). The MGHPCC is a multi-university enterprise in Massachusetts, dedicated to enhancing large-scale computing infrastructure in the area.

Recent News

Modeling buoyant surface plumes MITgcm News

Construction begins on high-performance computing center MIT News