Artificial Weatherman? AI Could Better Predict Hurricanes, Research Paper Says

Artificial Intelligence may be the key to more accurate and more timely hurricane predictions, China-based researchers have reported in the scientific journal Nature.

In two papers in the journal, the researchers, including scientists affiliated with the Chinese tech company Huawei, said that their AI method was an improvement over conventional, numerical weather prediction systems, known as NWP, now used by meteorologists around the world.

“Trained on 39 years of global data, our program, Pangu-Weather, obtains stronger deterministic forecast results on reanalysis data in all tested variables when compared with the world’s best NWP system,” Kaifeng Bi, Lingxi Xie and other researchers wrote.

NWP has been used for years and has improved, but is still considered slow and subject to approximations and inaccuracies, the paper noted. The AI approach also needs improving but holds “great potential.” AI will likely be integrated with traditional forecast methods to produce hybrid models, Lingxi Xie told The Daily Beast news outlet.

The researchers used a three-dimensional model, along with four decades of weather data and “reanalysis” to train neural networks to predict storm behavior and weather patterns, the Nature paper explained.

If proven to be successful, it’s possible that the AI system could assist in predicting last-minute movements in some hurricanes, which other studies have suggested are on the rise. Meteorologists and Florida officials were criticized last fall after Hurricane Ian took a sudden turn to the east after it was initially predicted to hit the Tampa area.

One of the Nature papers can be seen here. Two comments on the paper online questioned some of the methodology of the study.

Topics Catastrophe Natural Disasters InsurTech Data Driven Artificial Intelligence Hurricane