A 10,000-Year Global Stochastic Tropical Cyclone Catalog with Wind-Dependent Track Transitions (WHITS)

Tropical cyclones are among the most powerful and destructive forces on Earth, capable of reshaping coastlines and causing billions of dollars in damage in a single season. For insurance companies, city planners, and engineers, the central challenge is not just knowing that these storms happen, but understanding exactly how often the most dangerous ones might strike a specific location. The problem is that the historical record of these storms is too short and too patchy to provide a clear answer. We have only a few decades of reliable data for many parts of the world, yet the most catastrophic events might only occur once every few hundred years. Relying solely on what we have seen so far leaves a dangerous gap in our knowledge, making it difficult to design buildings that can withstand rare, extreme winds or to price insurance policies that reflect the true risk.

To fill this gap, researchers have developed ways to create synthetic catalogs—massive libraries of simulated storms that follow the statistical rules of the real world but extend far beyond the limits of human observation. These simulations act as a bridge between the short history we have and the long future we need to plan for. The goal is to generate thousands of years of storm data that look and behave like real storms, preserving the complex shapes, speeds, and paths that nature produces, without introducing artificial errors that could skew risk calculations. This approach allows experts to ask questions about rare events that the historical record simply cannot answer.

A new study introduces a sophisticated tool called WHITS, a system designed to generate a ten-thousand-year global catalog of tropical cyclone tracks. Created by researchers at Columbia University and Arizona State University, this model moves beyond simple statistical guesses by building new storms from pieces of real, historical storms. Instead of trying to fit the data into a rigid mathematical formula, the system acts like a master seamstress, cutting and stitching together segments of actual storm paths to create entirely new, plausible journeys. The result is a vast library of storms that captures the messy, irregular reality of how these systems move, loop, and stall, providing a much clearer picture of risk for the entire globe.

The core innovation of WHITS lies in how it constructs these synthetic storms. The researchers started with the complete historical record of tropical cyclones from six major ocean basins, including the North Atlantic, the Western Pacific, and the South Indian Ocean. They broke these historical tracks into short, variable-length segments. When the computer needs to create a new storm, it picks a starting point and then begins to walk along a randomly chosen historical segment. As the simulated storm moves forward, the system decides whether to stay on that path or jump to a neighboring historical segment that looks similar. This decision is not random; it is guided by the local wind speed, the storm's age, its location, and how fast it is moving forward. By conditioning these choices on wind speed, the model ensures that the new storms maintain realistic intensity and smooth transitions, avoiding the jerky, unnatural jumps that plagued earlier versions of this technology.

This wind-focused approach is crucial because the damage a storm causes is deeply tied to its path and its speed. A storm that loops back on itself or stalls over a region can cause far more destruction than a storm that moves quickly in a straight line, even if their peak wind speeds are identical. Previous models often smoothed out these irregularities, effectively erasing the unique, chaotic shapes that real storms take. WHITS, however, preserves these irregular geometries. It allows for sharp turns, hairpin reversals, and complex loops, exactly as they appear in the historical record. The system does not force storms into neat, predictable patterns; instead, it lets them wander through the ocean in ways that are physically plausible, ensuring that the resulting catalog reflects the true variety of nature's behavior.

To make the data usable for engineers and insurers, the researchers added a final layer of refinement. Even with careful selection, jumping from one historical segment to another can sometimes create tiny, unrealistic gaps in position or sudden jumps in wind speed. The team developed a smoothing technique that gently blends these transition points, removing the digital artifacts while keeping the overall shape of the storm intact. This ensures that when the simulated storms are fed into models that predict storm surges or wind damage, the results are continuous and reliable, without spurious spikes that could lead to incorrect risk assessments.

The team tested their ten-thousand-year catalog against the best available historical data and against another widely used global storm model. The results showed that WHITS successfully reproduces the observed patterns of where storms travel and how often hurricane-force winds hit specific locations across all six major basins. The simulated storms matched the real world in their spatial density, showing the same high-risk corridors in the Caribbean, the coast of Mexico, and the waters east of the Philippines. More importantly, the model accurately captured the probability of a location experiencing hurricane-force winds in any given year, a key metric for long-term risk planning. The simulated fields were smoother than the historical records, which is exactly what is needed to estimate the likelihood of rare, extreme events with greater confidence.

Unlike other models that rely on complex physics equations or assume that storms follow specific mathematical distributions, WHITS relies on the empirical reality of what has already happened. By resampling real track segments, it inherits the natural relationships between a storm's shape, its intensity, and where it makes landfall. This approach avoids the bias that can creep in when models are forced to fit data into pre-defined shapes that may not exist in nature. The researchers found that a single set of rules could successfully generate realistic storms for six very different ocean basins, each with its own unique climate and storm behavior. This suggests that the underlying structure of the model is robust and capable of adapting to diverse environments without needing to be tuned separately for each region.

The resulting catalog is a massive resource for anyone concerned with climate risk. It provides a large, low-bias sample of physically plausible storms that can be used to estimate potential losses for insurance companies, to design buildings that can withstand extreme winds, and to plan for coastal resilience. Because the model generates ten thousand years of data, it offers a stable view of the tail end of the risk distribution—the rare, high-impact events that are too infrequent to appear in the short historical record but are critical for long-term safety. The researchers have made this dataset available for academic and non-commercial use, allowing scientists and planners to explore the full range of tropical cyclone behavior and better prepare for the storms of the future.

The work represents a significant step forward in how we understand and quantify the risk of tropical cyclones. By combining the statistical power of a massive synthetic catalog with the physical realism of historical track segments, the researchers have created a tool that respects the complexity of nature while providing the clarity needed for decision-making. The model does not claim to predict the future with certainty, but it offers a much more reliable way to imagine the possibilities, ensuring that our preparations for the next big storm are based on a comprehensive and realistic view of the risks we face.