Supercomputer storm blurs the lines between sim and reality
Saturday, August 15, 2015, 1:23 PM - There's movie special effects and then there's this: a new supercomputer visualization that's so real, it actually blurs the line between simulation and reality.
Simulating weather with computers is not an easy task.
Every hour of every day, Earth's atmosphere throws together countless air and water molecules, at different temperatures and speeds, all having their own specific effect on the outcome in producing the weather we see. While weather models are limited to kilometre-wide or larger scales to produce timely results for national and even regional forecasts - typically too small to pick up on a single storm cell - researchers can still tap into the power of supercomputers to delve down into individual storms.
This incredible simulation is one of their latest results:
This hyper-realistic visualization simulates the conditions that were in play that produced the EF5 wedge tornado near El Reno, OK, on May 24, 2011. This destructive tornado cut a swath over 100 kilometers long and as wide as 1.6 kilometres at some points along its track, killing nine people and injuring 181 others.
The simulation of this storm was designed by Leigh Orf, an Associate Researcher at the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin - Madison. Rather than using kilometre-wide scales, it was run by the Blue Waters supercomputer, at the University of Illinois' National Center for Supercomputing Applications (NCSA), using a resolution of just 30 metres.
When the computer had completed the simulation and delivered the roughly 100 terabytes of results, Orf passed them on to Rob Sisneros and David Bock, at the NCSA, where Bock used a custom rendering tool to add realistic lighting to the final visualization we see above.
"I call it 'sim-realistic,'" Bock told International Science Grid This Week (iSGTW).
According to Bock, this isn't just about producing something that's visually appealing.
"As soon as you see the storm, you recognize and connect aspects of the storm that you see in real life with the computational model," he told iSGTW. "If it's a thunderstorm, then it should look like a thunderstorm."
Orf agrees, and related this back to how it helps produce even better computer models for weather simulation.
"One of the ways to validate a numerical simulation is to visualize the model data in ways that match observations," Orf says. "If you can't get a simulated storm to look like a real storm, then there is something wrong with the model or the technique you are using. If you can get them to look similar, it lends much more credence to the simulation and its faithful representation of the atmosphere, the storm, and what goes on inside the storm."
The simulation was presented at the 2015 Extreme Science and Engineering Discovery Environment (XSEDE) conference that was recently held in St. Louis, MO, and was chosen as the XSEDE15 People's Choice award for best visualization.
WATCH BELOW: One year after, Jaclyn Whittal and Mark Robinson discuss the EF5 May 31, 2013 El Reno tornado - the widest tornado in US history