AI Aids in Building Hurricane Resistance Research
Researchers at the National Institute of Standards and Technology have reportedly created a new method of digitally simulating hurricanes with the potential to develop improved guidelines for building designs.
The data was devised using 100 years of hurricane data alongside modern artificial intelligence techniques, demonstrating that simulations can accurately represent the trajectory and wind speeds of a collection of actual storms. The study results were recently published in Artificial Intelligence for Earth Systems.
Hurricane Model Method
According to the NIST, state and local laws that regulate building design and construction codes have designers use standardized maps to find the level of wind their structure must handle based on its location and its relative importance.
These wind speeds are reportedly derived from scores of hypothetical hurricanes simulated by computer models, which are themselves based on real-life hurricane records.
The researchers reportedly developed the maps by simulating the inner workings of hurricanes, influenced by parameters such as sea surface temperatures and the Earth’s surface roughness. However, NIST notes, the requisite data on these specific factors is not always readily available.
“Imagine you had a second Earth, or a thousand Earths, where you could observe hurricanes for 100 years and see where they hit on the coast, how intense they are. Those simulated storms, if they behave like real hurricanes, can be used to create the data in the maps almost directly,” said NIST mathematical statistician Adam Pintar, a study co-author.
To advance this process, NIST postdoctoral researcher Rikhi Bose, together with Pintar and NIST Fellow Emil Simiu, used new AI-based tools and years of hurricane records to “tackle the issue from a different angle.”
Pintar said that, rather than having their model mathematically build a storm from the ground up, the team taught it to mimic actual hurricane data with machine learning. With enough quality information to study, machine-learning algorithms can reportedly construct models based on patterns they uncover within datasets that other methods may miss and then simulate specific behaviors, such as the wind strength and movement of a hurricane.
The study material came from the National Hurricane Center’s Atlantic Hurricane Database (HURDAT2), which contains information like coordinates of travel paths and windspeeds for hurricanes going back more than a century.
Splitting data on more than 1,500 storms into sets for training and testing their model, the researchers challenged this with concurrently simulating the trajectory and wind of historical storms it had not seen before. As a result, the model scored highly.
“It performs very well. Depending on where you're looking at along the coast, it would be quite difficult to identify a simulated hurricane from a real one, honestly,” Pintar said.
Additionally, they used the model to simulate sets of 100 years’ worth of hypothetical storms in a matter of seconds. The authors report that they saw a large degree of overlap with the general behavior of the HURDAT2 storms, suggesting that their model could rapidly produce collections of realistic storms.
However, the team states there were some discrepancies, such as in regions where HURDAT2 data was sparse and the model generated less realistic storms.
“Hurricanes are not as frequent in, say, Boston as in Miami, for example. The less data you have, the larger the uncertainty of your predictions,” Simiu said.
For the next steps, the team reportedly plans to use simulated hurricanes to develop coastal maps of extreme wind speeds as well as quantify uncertainty in those estimated speeds. Since the model’s understanding of storms is limited to historical data for now, it cannot simulate the effects that climate change will have on storms of the future.
NIST says that within the next several years they aim to produce and propose new maps for inclusion in building standards and codes.