fluid dynamics equations

Fluid dynamics equations are a collection of physical laws that calculate how flowing substances like air or water behave. They work by precisely solving exact formulas to determine which direction the wind will blow or how pressure will change, similar to following a recipe with exact measurements.

The problem is that this "recipe" is extremely complex. To forecast the weather, you need to feed in today's data gathered from satellites and observation stations, then continuously solve these equations second by second. This used to require running supercomputers the size of a three-story building, and even small gaps or noise in the observation data risked throwing predictions badly off.

Recently, AI models have emerged that skip solving these equations from scratch. Instead, they learn patterns from decades of past weather data to predict what comes next. Rather than directly calculating physical laws, this approach essentially remembers how similar situations unfolded before, allowing forecasts to be made with far less computation.

The article explains that "past cyclone forecasting involved directly solving fluid dynamics equations using supercomputers the size of three-story buildings." It's easy to misread this as saying the method was outdated or wrong—but that's not the point. It was an accurate way to compute physical laws, but one with massive computational demands and vulnerability to gaps in observation data. AI models like WeatherNext instead took a different approach: rather than re-solving these equations, they learn patterns from historical data.