Weather forecasts are famously imperfect, but the science behind them is far cleverer than we usually give it credit for.
In this week's podcast, Professor Luke O’Neill explores how weather forecasting works, and why floods remain one of the hardest things to predict. Luke isn’t a meteorologist — although he did briefly consider it in college — but too much physics put him off. Still, he’s an ideal guide to the basics because the weather is really about a few core ideas behaving badly.
At its heart, forecasting comes down to temperature, air pressure, humidity, sunlight, and the way air flows like a fluid. Air moves from high to low pressure, dragging wind and weather systems with it. Add water vapour into the mix and things get interesting very quickly.
People have been trying to predict the weather for hundreds of years, using almanacs, folklore, and observation. It was never perfect, but it mattered hugely to farmers and sailors. Rain, in particular, remains tricky. Moist air rises, cools and condenses into clouds — but rain doesn’t just appear. It needs tiny particles like dust, sea salt, or pollen to form droplets, and those microscopic details are hard to pin down.
Today’s forecasts rely on satellites, radar, weather balloons, and ground stations, all feeding data into powerful computer models. Those models keep improving, and artificial intelligence is now helping to sharpen predictions.
Flooding is even more complicated. It’s not just about how much rain falls, but how fast it falls, how long it lasts, and where it lands. Soil type, vegetation, evaporation, and urban concrete all matter. Forests and wetlands act like sponges, while cities can make flooding worse — something Ireland knows well after decades of building on flood plains.
Some countries lead the world in flood modelling, but nowhere reliably predicts flash floods. Luke argues that weather is a brilliant way to teach science, and that we already know how to reduce flood risk. The challenge now is acting on that knowledge and getting on with it.
