For many years I worked with high end computational science and engineering project, so I have an enduring appreciation for state of the art scientific computation.
Dylan Keon, Cherri M. Pancake, and Harry Yeh of Oregon State University have and interesting article in November IEEE Computer, “Protecting Our Shorelines: Modeling the Effects of Tsunamis and Storm Waves“.
In principle, computational modeling of the ocean is very similar to modeling the atmosphere. (There an article in the same issue about Numerical Weather Prediction.) Both are mostly about modeling energy flows and the resulting physics, especially storms. Most of the physics is understood, but high resolution models are gigantic, testing or exceeding our computing capacity.
In recent decades we have improved weather prediction quite a bit, with large ensembles of large computers. But Keon et al. are modeling oceans, particular events such as tsunamis and storm surges, is not as far along. Why?
One very important reason is that, for most purposes, the atmosphere can be modeled as a uniform grid, with no obstructions. While there are some landforms that matter (large mountain ranges), most of the planet is ocean and the mountains only affect the lowest parts of the atmosphere. Air and energy flow in all directions.
The oceans, on the other hand, inhabit non-uniform basins, of greatly varying depths and surrounded by coastlines. And coastlines are complex, indeed, fractal geometries in 3D. Water does not flow freely everywhere, which makes the modeling very complicated, and also requires good data about the basin itself.
Keon and colleagues would like to be able to model tsunamis and extreme waves, to be able to give warnings akin to hurricane warnings. These phenomena affect the shores, and also depend on the geometry of the coast—the messy boundaries of the ocean, the most difficult challenges.
As they discuss, even if you have an idea of the magnitude of a wave, predicting its effects on shore is very difficult. It depends on the land that it encounters, and possibly other factors (such as tides or recent rainfall or seasonal ground cover). Given the rarity, we don’t have a lot of history to help predict.
Think about the complexity of modeling the details of rocks, dunes, and marshes along kilometer after kilometer of coast line, to the detailed resolution of 1m or less. It’s no wonder that it is difficult to know what will happen if a wave arrives.
In addition, details of the actual wave that arrives at shore is difficult to model. Even assuming that there are good measurements of the wave at sea, working out how it will rise as it enters shallow water and up to the shore is difficult. In practice, this means that the estimates are so broad as to be difficult to act on. The difference between a 1 meter or a 3 meter or a 5 meter wave are considerable!
Finally, of course, time is short. While hurricanes brew up over many days, a tsunami crosses the ocean in half a day, and will hit nearby shores in an hour or less. Even if we had precise predictions, there is barely time to evacuate. And if the predictions are uncertain, they may be worse than useless, they may actually lull people into mistaken belief that there is no danger.
Keon et al sketch some of the current efforts to develop both real time prediction and risk assessment models using High Performance Computing. As they say, this is “compelling for scientific as well as practical and humanitarian reasons”. p. 31
By the way, the November issue of IEEE Computer has several articles about cutting edge science that is built on high performance computing, from planetary scale to sub-atomic.
(Ask your friendly local librarian to help you get a copy of the full article)
- Keon, Dylan, Cherri M. Pancake, and Harry Yeh, Protecting Our Shorelines: Modeling the Effects of Tsunamis and Storm Waves. Computer, 48 (11):23-32, 2015.