Flooding the system - improved flood simulation technology

Materials World magazine
,
1 Mar 2012

Noah saw it coming. Before the floods came, he built an ark. As it bobbed on the rising sea, everyone else drowned. If only the forsaken had a computer simulation to hand; they could have made straight for the Mountains of Ararat, and waited for the waters to subside.

Nowadays, we can predict the occurrence of floods. Three dimensional computer simulations map the progress of natural disasters, but do so slowly. One hour of flood simulation can take an hour of central processing unit time. A Norwegian research team at SINTEF ICT in Oslo claims to have developed a shallow water simulator that is up to 400 times faster than existing computing methods. The key to the technology is the use of sophisticated graphics cards.

Lead Researcher André Rigland Brodtkorb explains, ‘The graphics card makes the simulation run far faster than it would on a standard CPU. We apply these existing algorithms onto the graphics card to make it run in a new way, to make it run more efficiently. And then we solve real world problems with the simulator.’

The addition of graphics cards enables the team to process more calculations simultaneously. The result, they say, is a higher quality result. ‘With tsunami warnings, you could pick 21 scenarios that are representative of extreme tsunamis you would expect to see,’ he says. ‘You then average these results to see where the tsunami will hit and which areas will be affected. By running things faster, you can run more realisations and more prototype tsunamis to compare against. Once the tsunami warning is held, you start running computations.’

The team has produced a working simulation of the Malpasset Dam collapse in France in 1959. The ensuing flooding killed 423 people after 55bln litres of water hurtled through nearby Fréjus. The researchers could recreate the flood once they knew where the dam broke and various measure points along the valley. ‘For a simple dam break simulation, you just need to know what the water elevation at each spatial point is,’ Brodtkorb notes.

‘It would take a matter of minutes or a couple of hours, depending on how long you would like to run the simulation for and how many scenarios you would like to check. For a CPU, it would take days and weeks for similar types of simulations.’

At present the SINTEF team is working with UK-based company Halcrow to bring the technology to end-users for use in artificial dams and flood management systems. They are also exploring its use in carbon sequestration and ocean currents:

‘These simulations could aid in oil spill mitigation or in search and rescue operations. Once you have the ocean current [measurements] that this type of simulation will give you, you can then get an accurate description of where things will propagate.’