So exactly how does a computer make a weather forecast? Basically, a computer model has two parts: weather observations - the starting point for the forecast - and mathematical equations that describe how the current weather patterns will evolve in the future. Some of these equations may even be familiar to you - for example, Isaac Newton's famous F=ma, force equals mass times acceleration, is used to predict the wind.

To link the observations and the equations, meteorologists overlay a grid of points onto the geographical area of interest. A grid is also stacked up vertically to create a 3D array of points. At each grid point, the computer takes the weather observations and "solves" the equations for pressure, temperature, wind, and moisture to produce forecasts a few minutes ahead. Those forecast values are then used to repeat the process for the next time interval, producing a forecast a few more minutes into the future, and so on. It takes about an hour and trillions of calculations for a supercomputer to make a two-day forecast for North America.

Theoretically, you can get more detailed forecasts by increasing the number of grid points. But the more grid points there are, the more powerful the computer has to be, and even lightning-fast supercomputers have practical limits. As a result, there will never be a perfect computer weather forecast. Never.