Numerical simulation of coupled ocean-sea ice system
Yoshimasa Matsumura , Assistant Professor
Institute of Low Temperature Science / Graduate School of Environmental Science
High school : University Entrance Qualification Examination
Academic background : Department of Earth and Planetary Science, Graduate School of Science, The University of Tokyo
- Research areas
- Physical Oceanography
- Research keywords
- climate system, coupled ocean-sea ice system, numerical simulation, high-performance computing
What is your goal?
The oceans play an extremely important role in the global climate system. Areas with warm currents are warmer than others even at the same latitude, and phenomena such as El Niño, a local fluctuation of the sea surface temperature that impacts the weather and climate around the world, are well known. Sea ice also plays particularly important roles in the global climate.
Example of the results of a numerical simulation of the formation of Antarctic bottom water along the Antarctic coast
For example, along the coast of Antarctica, where a large amount of sea ice is formed every winter, the brine rejection caused by freezing forms cold and saline dense water masses that descend down the continental slope to the deepest layer of the world ocean. The oceanic thermohaline circulation that circulates the entire globe in a time scale of a few thousand years is maintained by such dense water formations caused by the freezing of sea ice. It has also been recognized that trace elements in seawater are absorbed inside sea ice when it freezes, then and suddenly released at the time and location of melting; such processes might play an essential role in the biogeochemistry and the ecosystem in the ocean. My research focuses on modeling this coupled ocean-sea ice system and performing numerical simulations using computers in order to investigate how this complex system impacts the global climate and material circulation.
What prompted you to start your present research?
I have long been interested in performing simulations of complex physical phenomena, which people could not perform in the past, by using the latest supercomputers. Numerical simulation is a very effective method to understand phenomena that are difficult to study through experimentation and direct observation. However, computers do not automatically simulate the phenomena taking place in the real world, of course. We have to build a model of the phenomenon of interest based on the fundamental laws of physics and chemistry, and express that model in the form of a program code that computers can process. At university I first studied the fundamentals of physics and information science. At the time I chose my major, an ambitious project was drawing interest: simulating the earth’s climate system by using a huge supercomputer, with incredible processing capacity, called the Earth Simulator. I was really attracted by the idea of being able to run my own program on the world's most advanced computer, so I started the study of climate dynamics in the Department of Earth and Planetary Sciences. Numerical modeling of the earth’s climate is a relatively new field, but in the sense of predicting the behavior of atmosphere by using computers, numerical weather forecasting has a long history, and many researchers have been establishing state-of-the-art atmosphere general circulation models. However, that is not the case for modeling the ocean and sea ice, another important component in the earth’s climate system, and there seem to be a lot of research themes that should be addressed. Because I was motivated to develop my own numerical model and run the simulation by myself from scratch, I soon became an oceanographer regardless of the fact that I had not been interested in the ocean in particular.
What equipment are you using, and how are you conducting your research?
Because my research style is centered on numerical simulations, I use computers. I can write program code on an ordinary PC, but actual simulations are performed on higher performance computers. I use a work station in my laboratory for debugging and testing until I can be convinced that the newly developed code is working appropriately as designed, and then I remotely log in to the supercomputer installed in the Information Initiative Center of the university to perform a large-scale simulation. Since a supercomputer is actually a cluster of many computers connected each other by high-speed network, it is important to coordinate a large number of computers so that calculations are efficiently performed in parallel to realize massive simulations. The technique of parallel computation that permits such effective use of supercomputers has become a research field called High Performance Computing (HPC). One remarkable feature of numerical simulation-based research is the ability to recover from a failure relatively easily compared to laboratory experiments or direct observation. In fact, simulations using newly-written programs are rarely successful at the first run. We have to be patient, with a long series of trial and error accompanied by incorporating a variety of physical models and numerical schemes and coordinating experimental setups. It is a moment of bliss when our own program code successfully simulate a new phenomenon at the end of a long period of debugging and trials. Although my research is based on numerical simulations, observational data is also essential in order to verify that our numerical program code correctly simulate the real ocean. Therefore, researchers specializing in simulations like me sometimes need to participate in an observation cruise to obtain data. Ideally, I should have written my program on board ship, but it was not possible because I always get seriously seasick.