Sometimes lifetime careers are spun from childhood events. Such was the case for Leigh Orf, Professor of Atmospheric Science at Central Michigan University, who lived through one of the worst tornadoes in recorded history. As Orf details in a recent blog, the supercell thunderstorm that hit New England on Oct. 3, 1979, spawned the deadly tornado that ran through the Western Massachusetts town of Feeding Hills, where Orf lived at the time. The tornado, which was quite rare for this part of the country, still ranks as the ninth most destructive in American history.
Orf goes on to explain that while most tornadoes lack the energy to cause much damage, a small percentage belong to the category of “long-track EF5 tornadoes,” considered the strongest class of tornado on the Enhanced Fujita scale, with sustained wind speeds above 200 miles per hour. These are the dreaded tornadoes that leave many miles of destruction in their wake. In recent years, such intense twisters are all too familiar to residents of Iowa, Missouri, Oklahoma, Mississippi and Alabama.
While weather forecasting has improved significantly in the last decade, tornado prediction is notably thorny. The ability to predict whether a thunderstorm will give rise to a tornado has so far eluded the best technology. But meteorologists continue to push back the barriers in an effort to better understand the dynamics of supercell thunderstorms, which when long-lived and characterized by a rotating updraft, are more likely to give rise to a long-track EF5.
To advance this goal, Orf and his colleagues – Robert Wilhelmson of the University of Illinois, Louis Wicker, National Severe Storms Laboratory and Bruce Lee and Catherine Finley of WindLogics – have been using the petascale “Blue Waters” supercomputer at the National Center for Supercomputing Applications (NCSA).
“Utilizing Blue Waters, we initialized the CM1 cloud model with the atmospheric conditions of the May 24, 2011, supercell that spawned a long-track EF5 tornado near El Reno, Okla.,” he writes. “Within this environment, a cloud was initiated by analytically forcing an updraft. This cloud grew into a supercell, and, more importantly, spawned an EF5 tornado that, over the course of its 65-mile path, produced winds near the ground of up to 300 miles per hour, which is also the strongest wind ever measured by Doppler weather radar in an observed tornado.”
Orf characterizes the simulation as a “milestone in severe storms meteorology.” The new level of fidelity achieved by the model was a results of the team’s addressing both meteorological and computational challenges.
“Getting the model to produce the ‘storm we wanted’ proved to be the biggest challenge, and it took dozens of attempts before we were able to simulate a supercell that produced a long-track EF5,” Orf continues. “On the computational side, the challenges primarily involved I/O. The amount of data produced by a simulation like this can be staggering – our simulation produced nearly 100 TB of data. I spent a significant amount of time trying different I/O approaches to maximize throughput during writes to Blue Water’s Lustre file system.”
After refining the model, the team rounded out its approach with middleware, visualization, analysis and conversion tools to model the data. The next step is to simulate multiple supercell storms, including ones that produce long-track EF5s and ones that do not, to highlight any distinguishing characteristics that can serve as an early warning system for this deadliest class of tornadoes.