Vilhelm: A Novel AI-based Global Weather Forecasting System for the Prediction of Atmospheric Hazards

Cyclones are known as one of the biggest hazards to life and property. They can lead to a number of different hazards such as storm surges, floods, very high wind speeds, tornadoes and lighting. The combination of these hazards greatly increases the potential for loss of life and damage of properties. California experienced a devastating and deadly "bomb cyclone" on March 21st (2023), causing widespread destruction and claiming at least five lives. Many people remained without electricity until Thursday morning (March 23rd). Whereas it is hard for models to predict the formation and trajectory of bomb cyclones, Jua’s model, Vilhelm, predicted the formation of the cyclone about 40 hours before it hits California. In addition, it was offering a high level of accuracy in predicting the storm's trajectory 24 hours in advance. Furthermore, Vilhelm's predictions regarding the start of cyclone's dissipation on 22nd March proved to be accurate.

During our analysis on the cyclone, we selected wind speed and pressure as suitable indicators to forecast the formation and trajectory of the cyclone. At 00:00 on March 20th, Vilhelm forecast wind speeds at the 850 hPa and 1000 hPa pressure levels that correctly identified the formation, extension and movement of the cyclone that would hit California on March 21st. In addition, the mean sea level pressure forecasts of the Vilhelm model on 20th March (00:00) approved these findings at the atmospheric layers and therefore the physical consistency of the AI based system. 

Given that the Vilhelm model offered an accurate forecast for the formation, extension, dissipation and crucially, the trajectory of the cyclone, there is a strong potential for it to be used as a global “Atmospheric Hazards Warning System”.