How to better forecast once-in-a-millennium weather events
For all that day-to-day weather forecasts have improved, it remains a challenge to forecast events that might happen once in 1,000 years—like the deadliest heat waves.
Forecasting extreme weather events, like those that occur once in a millennium, remains a challenge for traditional supercomputer-based climate models. These models are time-consuming and energy-intensive, making it difficult to predict rare occurrences such as deadly heat waves. Artificial intelligence (AI) has improved short-term weather forecasts but often struggles with predicting rare, extreme events that weren't part of their training data.
A team of researchers from the United States and France, led by Pedram Hassanzadeh of the University of Chicago, has developed a hybrid method called AI+RES to tackle this issue. This method combines the efficiency of AI with the trustworthiness of traditional models, allowing for faster and more accurate predictions of rare events using fewer resources.
The AI+RES method specifically enhances the rare event sampling (RES) technique, which is useful for short-duration extreme events. By integrating AI's predictive capabilities, the method effectively identifies conditions that lead to short, rapidly developing extremes, such as weeklong heat waves. In tests using 50,000 simulations, the AI+RES method achieved results equivalent to traditional models with only one-hundredth of the simulations.
This hybrid approach has the potential to significantly reduce computational costs for forecasting rare heat and precipitation events, providing valuable information to decision-makers for climate adaptation and mitigation planning.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
