Supercell thunderstorms produce the most violent tornadoes. Accurately forecasting such tornadoes to alert the public in advance remains a significant challenge, partly due to our incomplete knowledge of how supercells and tornadoes work. This grant supported research into exploring the most violent supercell thunderstorms through the use of highly sophisticated numerical models run on powerful supercomputers.

In order to achieve proposal goals, a significant amount of computer programming focusing on data reduction was done. Data reduction was necessary for this project to succeed, as traditional methods for saving data frequently, necessary in order to achieve proposal goals, would quickly overwhelm supercomputing systems. This reduction in data was done primarily through the use of lossy data compression. We used a form of lossy data compression (called ZFP) that reduces data by a factor of 10 to 100 times, degrading the original data only slightly and not enough to affect research results. Once ZFP was written into the code that writes and reads data, simulation data was saved at extremely high frequency (as frequently as five times per second) where it was then post-processed, visualized and analyzed. Visualizations and animations play significant roles in our analysis workflow, as many important flow features associated with tornado formation are very small and very fast.

Two groups of simulations were conducted under this grant. The 24 May 2011 El Reno, OK environment (where a supercell producing a long-track EF5 tornado occurred) was used for the bulk of simulations. Simulations run at 30 and 10 meter grid-spacing simulations were done using a sounding from this environment to initialize the model. Simulations reveal features such as the streamwise vorticity current (SVC), discovered by the PI's team in simulation data, play key roles in modulating the strength of the low-level mesocyclone of the supercell, where intensification of this updraft often immediately precedes tornadogenesis in simulations. Temporal averaging of 30 meter data indicated a rapid drop in pressure centered around 2 km above ground precedes tornadogenesis, and that this pressure drop is associated with a strengthening SVC. However, the details of tornadogenesis involve many small-scale vortex mergers that are lost in temporal averaging. These mergers occur primarily between cyclonic vortices along cold pool boundaries beneath the low-level mesocyclone. Both 30 and 10 meter simulations exhibit vortex mergers along the storm's cold pool boundary, and in both simulations, a rearward progression of vortices (in a storm-relative sense) halts beneath the mesocyclone, where additional vortices merge into a stationary vortex that rapidly reaches tornado strength near the ground.

The tornado formation phase of the 10 meter El Reno simulation consists of three clearly identifiable vortex mergers that occur to the single vortex that is becomes the tornado.  Each merger results in a strengthening vortex, with a third merger coiling/zipping two strong vortices together to form a strong two-celled tornado. The results of this simulation highlight the critical importance of resolving vortex-vortex interactions in supercell simulations designed to study tornadogenesis.

An idealized environment was used to study how above anvil cirrus plumes (AACPs) form.  AACPs are stratospheric ice clouds that often form on top of strong thunderstorms, and in 3/4 of supercells exhibiting AACPs, AACP formation precedes severe weather (large hail and/or strong winds) at the surface by an average of 30 minutes. AACP formation mechanism was identified in 50 meter simulations conducted on the Frontera supercomputer. These simulations were conducted on a regular mesh - grid spacing everywhere, including well above the top of the cloud, was a constant 50 meters. Traditionally, a vertical stretched mesh focusing resolution near the ground is used in cloud modeling, but this approach does not resolve the top of the cloud. Our simulations revealed a process occurring in AACP forming supercells where a thin (200 or so meters deep), intense (exceeding 90 m/s), primarily horizontally-oriented jet of air kicks off a hydraulic jump. The hydraulic jump in the air flow results in an upward explosion of ice and water vapor into the lower to middle stratosphere that streams downwind of the storm's overshooting top in a storm-relative sense. Identifying the physical mechanism behind AACP formation is the first step in identifying any mechanisms that may link AACPs to severe weather at the Earth's surface, and also exploring the role of AACPs in hydrating the stratosphere.

Last Modified: 06/15/2023
Modified by: Leigh Orf