Hurricane formation has long remained shrouded in mystery, with meteorologists unable to discern exactly what forces cause the devastating storms (also known as tropical cyclones) to materialize. Now, researchers at Florida State University (FSU) have used HPC to uncover the troubling truth: even the slightest changes in atmospheric conditions can feasibly trigger hurricanes to form.
“The motivation for this research was that we still don’t have a universal theoretical understanding of exactly how tropical cyclones form, and to be able to forecast that storm-by-storm, it would help us to have that more solidly taken care of,” said Jacob Carstens, a researcher at FSU who conducted the research alongside atmospheric scientist Allison Wang.
The duo began with minimalist numerical models in order to explore the disturbances required to initiate hurricane formation. They applied uniform conditions across the models, then added random temperature fluctuations to stir the pot. Then, they examined how cloud formations were affected.
“We’re trying to go as bare-bones as possible, looking at just how clouds organize without any of these external factors to form a tropical cyclone more efficiently,” Carstens said. “It’s a way we can look at the tropical cyclones themselves rather than the surrounding environment’s impact on them.”
What the researchers found was less than encouraging: in simulations of latitudes between 10 and 20 degrees (around the equator, including storm-prone regions such as southern India, Indonesia and Haiti), hurricanes formed every single time – including when no disturbances were artificially introduced. In the stable scenario, a vortex emerged and, a few simulated days later, evolved into a hurricane.
To run the simulations, the researchers turned to two different HPC resources. First was the Cheyenne supercomputer at the NSF-supported National Center for Atmospheric Research (NCAR). Cheyenne is equipped with 4,032 Intel Xeon Broadwell-based nodes, an aggregate 313 TB of memory and a Mellanox EDR InfiniBand interconnect. At 4.8 Linpack petaflops, Cheyenne placed 44th on the most recent Top500 list of the world’s most powerful publicly ranked supercomputers. Additional computing was conducted on FSU’s own HPC cluster, based in its Research Computing Center. The FSU cluster comprises 776 nodes and, according to FSU, delivers 393 teraflops.
The researchers published their findings as “Tropical Cyclogenesis From Self‐Aggregated Convection in Numerical Simulations of Rotating Radiative‐Convective Equilibrium” in the April 2020 issue of the Journal of Advances in Modeling Earth Systems.
“Convection can evidently organize from isolated cotton-ball clouds to a full-blown hurricane without any help from favorable large-scale conditions,” said Eric DeWeaver, a program director in NSF’s Division of Atmospheric and Geospace Sciences. “Sometimes people talk about the ‘butterfly effect,’ the idea that a butterfly can cause a storm by flapping its wings, but here we don’t even need the butterfly.”