What makes a typhoon suddenly explode in strength? The answer may lie in tiny, invisible vortices swirling inside the storm. For the first time ever, Japan's Fugaku supercomputer has simulated an entire typhoon lifecycle at 100-meter resolution, revealing secrets that could transform how the world predicts deadly storms.

The World's First 100-Meter Typhoon Simulation

In January 2026, a research team from Tohoku University, the University of Tokyo, and Keio University announced a groundbreaking achievement in atmospheric science. Using Fugaku, Japan's flagship supercomputer housed at the RIKEN Center for Computational Science in Kobe, the researchers successfully simulated the complete lifecycle of a typhoon at an unprecedented 100-meter horizontal resolution.

The simulation tracked approximately four days of development, from a weak initial vortex all the way to a "super typhoon" with a central pressure around 920 hectopascals (hPa). The study, led by Associate Professor Junshi Ito of Tohoku University along with colleagues at the University of Tokyo and Keio University, was published in Geophysical Research Letters, a journal of the American Geophysical Union, on January 14, 2026.

Why 100-Meter Resolution Is a Game Changer

Current weather forecasting models typically operate at resolutions of about 2 kilometers, meaning the atmosphere is divided into grid cells roughly 2 km on each side. While this is sufficient for tracking a typhoon's general path and size, it is far too coarse to capture the complex, small-scale air currents swirling inside the storm.

A 100-meter resolution is 20 times finer, representing 400 times more data per area. Imagine going from a blurry satellite photo to a high-definition camera peering directly into the typhoon's inner workings. This approach, called "Large Eddy Simulation" (LES), directly computes larger turbulent eddies while modeling only the smallest ones. The sheer computational cost has previously made it impossible to apply LES to an entire typhoon lifecycle, until now.

Tiny Vortices with an Outsized Impact

The most striking discovery was the role of "mesovortices", small vortices with a radius of roughly 10 km that form near the eyewall. At 100-meter resolution, the researchers observed multiple mesovortices appearing, merging, and dissipating over time. Above these, rotating columns of powerful updrafts known as "vortical hot towers" rose into the atmosphere, surrounded by countless smaller eddies spinning in both clockwise and counterclockwise directions.

These small-scale vortices turned out to play a surprising role: they disrupted the airflow rushing toward the typhoon's center, weakening the average strength of updrafts and effectively delaying the onset of rapid intensification.

The numbers tell a compelling story. In the conventional 2 km simulation, rapid intensification began about 42 hours after initialization. In the 100-meter simulation, it did not begin until about 68 hours, a delay of roughly one full day. Crucially, though, the ultimate peak intensity (minimum central pressure around 920 hPa) was nearly identical in both cases. The "how strong" was the same; the "when" was dramatically different.

Why Predicting Rapid Intensification Matters

Rapid intensification (RI) is defined as an increase in maximum sustained winds of about 30 knots (roughly 15 meters per second) or more within 24 hours. Many of the most devastating tropical cyclones in recent history, from Typhoon Hagibis in 2019 to Hurricane Otis in 2023, underwent RI before making landfall, catching forecasters and communities off guard.

Typhoon Hagibis, which the current study used as a reference for its environmental conditions, saw its central pressure plummet to 915 hPa in just about 36 hours. It went on to cause severe flooding and landslides across Japan, resulting in over 100 deaths and billions of dollars in damages. If RI timing can be predicted more accurately, evacuation orders and disaster response can be initiated earlier, potentially saving many lives.

Fugaku's Power in the Global Supercomputer Race

The simulation was made possible by Fugaku's formidable computing capabilities. Covering a domain of 2,000 km × 2,000 km with 60 vertical layers at 100-meter resolution for four days demanded an astronomical volume of calculations.

Fugaku was jointly developed by RIKEN and Fujitsu and features Fujitsu's custom Arm-based A64FX processor across approximately 7.63 million cores. When it debuted in 2020, it topped the TOP500 supercomputer rankings across all four benchmarks, a historic sweep.

As of the November 2025 TOP500 list, the landscape has shifted. The United States' El Capitan leads at approximately 1.81 exaflops, followed by Frontier at 1.35 exaflops. Fugaku ranks 7th with 442 petaflops on the HPL benchmark. However, on the HPCG benchmark, which better reflects real-world scientific application performance, Fugaku holds a strong 2nd place globally at 16 petaflops, just behind El Capitan's 17.4 petaflops. This underscores Fugaku's continued strength in actual scientific computing, beyond raw theoretical speed.

Europe has also entered the exascale era, with Germany's JUPITER system at the Jülich Research Centre reaching 1.0 exaflops in late 2025, the first European system to cross this threshold. The global race in supercomputing power is accelerating, with nations investing heavily in climate prediction, AI, healthcare, and national security applications.

What's Next: Machine Learning and Beyond

The current study was intentionally designed as an "idealized experiment," excluding factors like vertical wind shear and the typhoon's translational movement to isolate the pure effects of small-scale turbulence. The research team has announced plans to incorporate these more realistic conditions in future simulations and to compare results with aircraft and satellite observations.

Perhaps most excitingly, the massive four-day LES dataset generated by this study is expected to serve as training data for machine learning-based typhoon intensity prediction models. It could also inform improvements to the computational schemes used in current operational weather forecasting models worldwide.

A World Where Storms Are Getting Stronger

Climate change is driving rising sea surface temperatures, fueling more powerful tropical cyclones around the globe. Category 5 storms have become more frequent, and the damages from extreme weather events continue to escalate. Accurately predicting when and where a storm will rapidly intensify is no longer just a scientific curiosity, it is a matter of life and death for coastal communities worldwide.

This world-first achievement by Fugaku represents a significant step toward better forecasting. The "secret of tiny vortices" discovered at 100-meter resolution may ultimately help build a stronger line of defense between hurricanes and the people in their path.

In Japan, typhoon season brings annual reminders of nature's power, and improving prediction accuracy has been a longstanding national priority. How does your country prepare for typhoons, hurricanes, or cyclones? What role do you think supercomputing should play in disaster prevention? We'd love to hear your perspective.

References