Incorrect forecast prompts BBC weatherman to apologize for thunderstorms

I'm sure it comes as a real shock to many that a weatherman got the weather wrong.

BBC forecasters were calling for a hot, sunny Sunday for London and the surrounding countryside, despite the UK Met office computer models showing the potential for thunderstorms that afternoon.

After the thunderstorms rolled through, weatherman Philip Avery made a public apology on the BBC News channel:

"There are thunderstorms which were not represented in our forecasts over the past couple of days or so," he said. "I have to say we can't even blame the computers, the computers actually wanted to put those thunderstorms in there but forecasters thought that it wasn't supported by enough evidence and so we went for the dry, hot option. Having said that, apologies to anyone who has had their next few hours ruined."

After forecasting weather for the past 12 years, I have to say that sometimes a forecaster does have to go on his or her gut feeling. The computer models aren't perfect. One of the first things you learn when you get into working with computerized weather models is their limitations. Some handle specific kinds of weather better than others. Some over-do it with the rain, while others don't put enough. You learn to look at several models at the same time, comparing them, in order to put together the best forecast you can.

[ Related: Enjoying your summer? Winter storms ahead could pack a punch ]

One of the biggest limitations forecasters can face is the lack of information. Forecasts along the west coast of the United States and Canada suffer because there aren't many weather stations out to sea, so there's not a lot of validated information to plug into the models. By the time a system has moved across the land for awhile, passing over weather stations so that we know what's actually happening, the computer forecast gets better and better, but its never perfect.

That's due to the limitations of a model's resolution. In order to compute the weather, the models essentially treat the atmosphere like blocks of information. The size of the blocks determines what kinds of weather it can pick up. If you want to run a computer model that takes the entire globe into account, or even just one hemisphere, those blocks have to be larger, or it will take days to run the model, instead of just hours.

A further complication is that weather models are typically run up to 4 times a day, so that you are putting in new information and getting updated forecasts, and sometimes different model runs can give you completely different forecasts. Other times, some weather feature can show up in one run, disappear for the next run, and then reappear for the run after that.

Sometimes, it really comes down to using your judgment.

[ Related: Canadians believe climate change is occurring as July 2012 becomes hottest month on record ]

I don't know what was going on in the BBC weather office over the weekend. I don't know what the UK Met office model was showing, or how long it was showing it. However, I do know how hard it is to put together weather forecasts, so I tip my hat to Philip Avery and the other forecasters at the BBC for taking the chance, and when it didn't work out, for stepping up and taking the heat for it.