A study by the University of Reading has revealed that even modern artificial intelligence (AI) weather models would have struggled to accurately predict the weather during the D-Day landings in June 1944. The research, published in the journal Weather, used AI models to create 50 different forecasts of the storm that impacted the Allies' invasion of Normandy. The researchers assessed how confident the AI model was in its predictions of the weather that Group Captain James Stagg had to forecast on June 5 and 6. On June 5, the original planned invasion day, the model correctly identified the storm’s position and the poor weather over Normandy, matching Stagg’s analysis. However, for June 6, the model predicted the storm would move toward Norway, bringing calmer conditions, while the actual weather was more dangerous than anticipated. Professor Andrew Charlton-Perez, the lead author of the study, noted that Group Captain James Stagg faced one of the most challenging tasks in history, as an incorrect forecast could have jeopardized the invasion and the outcome of World War II. The study suggests that if Stagg had had access to AI forecasts, he might have still chosen to delay the invasion, as the AI model indicated a real risk of dangerous winds on June 5, which would have justified a one-day delay. However, the AI incorrectly predicted that June 6 would be much calmer and clearer than it actually was, leaving soldiers to face more severe weather conditions. Stagg believed the conditions on June 6 were acceptable, leading General Eisenhower to issue the famous order, “OK, we’ll go.” The AI model also incorrectly predicted that the storm would clear away from Britain too quickly on June 6, suggesting calm and clear conditions. In reality, some of the invading forces encountered high winds and stormy seas. The model still showed a 30% chance that the wind would exceed the safety limit for landing craft on June 6, a risk that was overshadowed by the model’s overall reassuring picture. Additionally, the model predicted clearer skies than actually occurred on June 6, contributing to a false sense of security about the weather conditions. Weather records from 1944 were limited, so the researchers used newly discovered paper weather logs from the Faroe Islands, obtained from the Met Office archive. These islands, located near the storm’s path, provided readings that captured the storm at a point missed by other records. Converting these old paper logs into digital data allowed for a more detailed analysis of the storm’s strength, suggesting it may have been slightly stronger than previously thought. Digitizing more historic weather logs from archives across Europe could allow scientists to test storms like this one with greater accuracy in the future.