Taming the Ocean at MIT

Through this article, I would like to share my research experience and activities of the MSEAS-Multidisciplinary Simulation and Assimilation Systems research group belonging to one of the most prestigious universities, The Massachusetts Institute of Technology. The laws of physics rule our every action and the oceanic ecosystems are no different. Now, as we know that the oceans are bounded by the laws of physics, so can we delve deeper and devise new means to explore, develop and utilize the possibilities?

It was summer 2014 and I received an opportunity to learn from an extremely enthusiastic group of researchers led by Prof. Pierre Lermusiaux at the Department of Mechanical Engineering, Massachusetts Institute of Technology. Oceans are unpredictable, and this is what motivates Prof. Lermusiaux towards all of his research. Conditions at sea have long influenced human activities, from exploration to commerce, fisheries, tourism and even wars. Combining the power of ‘Computational Fluid Mechanics’ with probabilistic modelling, he develops models and assimilation schemes to better predict ocean behavior for a wide range of applications — from planning the most efficient paths for underwater robots, to predicting phytoplankton blooms.

Path planning for robots has been vastly studied in literature, but Prof. Lermusiaux’s research is quite unique. He along with his team of graduate students develops algorithms introducing the uncertainty of the ocean to achieve time and energy optimal path planning for Autonomous Underwater Vehicles (AUVs). The surrounding medium of travel for an AUV is the ocean, a highly dynamic and multi-scaled system with a considerable variability in both time and space. Hence utilizing the dynamic ocean currents to ones benefit, is a great challenge. AUVs in certain cases may choose to ride an ocean current, in order to save energy. However on other hand, there may be forbidden regions due to safety, hazardous conditions, security or naval considerations which the AUV is required to avoid during its course. Making the AUVs intelligent enough to take such decisions, is what the novel algorithms developed by MSEAS do. Such AUVs can be used for seafloor mapping, commercial exploration, military reconnaissance and coastline protection or maybe finding the supposedly drowned Malaysian Airline MH370, who knows!

Next major focus area of MSEAS is Data Assimilation (DA). The ocean physics involves a multitude of phenomena occurring on multiple scales, from molecular dissipation process scales to tsunamis. Oceanic processes cover a wide range of space scales, from about 1 mm to 10,000 km, and of time scales ranging from about 1 s to 100 years and even more (Lermusiaux, 2006). Hence, a reliable modelling scheme becomes an increasing difficult task to accomplish due to practical simplifications, inexact representations or parameterizations and numerical limitations. Or if we try to be totally dependent on ocean data measurements made from ships, aircrafts, underwater vehicles or platforms at sea etc., then the problem demanding consideration is that, such data are limited in time and space. So, there is an inevitable need to combine these various sources of data and different dynamical models, and where finally DA comes to the rescue. Even with all these, the measurements from the oceans have high uncertainties which needs to be taken care of.

My own project dealt with modelling of Nutrients, Phytoplankton and Zooplankton (NPZ) in idealized ocean banks, called Coupled Ocean Biological-Physical Dynamics. There exist mathematical models which can model NPZ systems based on the amount of sunlight available, death rate, regeneration, etc. But these are highly dependent on the values of parameters involved or the complexity of the model. Moreover, the uncertainty of the physics involved (i.e. initial conditions of our simulations or Reynolds number of the flow, etc.) also affects the evolution of NPZ variables over time. To deal with these kinds of problems, we employ stochastically dynamic models. These models can deal with the uncertain initial conditions represented by probability distribution functions, can evolve them over time, and update them based on observations to reduce the uncertainty in the system. For example, we may start with a number of biological models with different parameters, run our simulations, and then use real life observations from the oceans of the NPZ concentrations, to make Bayesian updates and finally tell which model or parameters best describe the real life situation. These kind of studies help us better predict processes like upwelling and hence tell our fisherman where to find fishes.

Learning from this group and contributing to this cause was one of the phenomenal turnovers in my life. Prof. Lermusiaux very actively collaborate with researchers from different institutions all over the globe. And when asked about his take on international internships, he said “Such exchanges are fundamental for the understanding of nations and their people, allowing to make scientific and engineering contributions without borders”. I would also hereby like to thank S. N. Bose Scholars Program 2014, for funding my internship and for all the good memories, I had with this really awesome research group. As I headed back home they are still cultivating ideas for the future, making our lives better and more secured.

For more information, please visit mseas.mit.edu

References

Lermusiaux P. F. J., Uncertainty estimation and prediction for interdisciplinary ocean dynamics, Journal of Computational Physics 217 (2006) 176–199

---

---