AI Weather Forecasting - Exciting Innovation, But Human Expertise Endures
Alexis Thornton
Last yearForget everything you think you know about weather prediction—artificial intelligence is about to turn this game upside down. New AI forecasting models like GraphCast make waves by generating remarkably accurate predictions at breakneck speed.
When Hurricane Lee was brewing last year, GraphCast nailed its eastward trajectory a full three days before traditional models. Now that's a forecasting mic drop if I've ever seen one!
Of course, AI is far from perfect. These tools tend to play it safe by relying on familiar weather patterns from the past.
So they can miss the mark when confronted with unprecedented events outside historical norms like mega-storms juiced up by climate change. AI models also lack the fine-grained resolution to nail down hyper-local details like exactly which neighbourhoods will flood.
And let's not write off human forecasters just yet. AI tools aren't ready to fly solo regarding data collection, model interpretation, and anticipating novel threats. "The reports of my death are greatly exaggerated," says the weatherman, paraphrasing Mark Twain.