Here's a little insider information about weather forecasting. Hopefully this puts into perspective how good computer models are, and why humans still have a role despite the increased power of computers.

Example 1: Computers rule, humans help

Colorado had a powder day on Monday afternoon and Tuesday morning. Four days ahead of time (on Thursday, 1/13) I pointed out that Monday could be a good day. The computer models had pointed to Monday as "the day" and as a meteorologist that knows the local mountains, all parameters seemed to be good. I sounded the "alarm", and the powder came as expected. This is an incredible feat of technology -- four to five days ahead of time we could point to certain ski areas and certain times and say "Hey, there will be a lot of powder at this time!". Amazing, and something that we could not do 25+ years ago, mostly because computing power has increased exponentially, which helps the forecasts to improve accordingly.

Example 2: Computers stink, humans rule

The image below shows a computer model forecast of snow from Wednesday through midday Friday. Notice that up to 40" is predicted near Steamboat in north-central Colorado. However, this will not happen, and it takes a human to "correct" the computer model. The weather pattern during this time is very moist (good for snow) but also very warm with temperatures at 10,000ft just below freezing (bad for snow). I know from experience that this weather pattern will bring snow, but nowhere near the amounts the models predict. Because there is so much moisture, the models go gung-ho on predicting snow. But they miss the fact that temperatures will limit the amount of snow. So, thanks to the computer model for predicting that it will snow, but a local human forecaster must override the model's forecast.



For more on the difference between computer model forecasts and hand-crafted forecasts, see this short video (1m 48s).