The Impossibility Of Long Range Weather Forecasting

The Impossibility Of Long Range Weather Forecasting

Sometimes I feel sorry for the local weatherman. He or she has to get in front of the television camera every day and give forecasts up to a week in advance knowing full well that those forecasts are apt to be wrong. Here’s why.

Weather forecasting is based on computer modeling and these computer models are based on two things; the physics and the data. Each model that the forecaster can choose from is based on different physics assumptions, sometimes very different.

For example, will a storm just entering the west coast of the United States travel across the continent without any intensification and exit the east coast three days later? Or, will it slowly intensify as the days go on and exit the east coast as a major noreaster? Each computer model handles the situation differently and the forecaster must choose which one to use.

Quite often the atmosphere is in a state which is a combination of several different physical models, not just one. In this case, it is not at all unusual for the computer models to predict wildly differing forecasts for three to seven days into the future. When this happens, the best the forecaster can do is to put all the options out on the table. Of course, this is not what the public wants to hear and a single forecast must be made. When the inevitable happens and the forecast is wrong the weatherman must indeed have a thick skin if he is to withstand the barrage of jokes and insults which will surely come his way.

As if that is not enough, here is the real problem. These forecasting models are based, of course, on the data that is input into the computer program. This data is a compilation of current weather observations and this data is VERY incomplete. As an example, suppose two observations are fifty miles apart. The observers diligently compile the relevant weather statistics and send them off to the national weather service where the data is then input into each computer model.

Here is the problem. In the fifty miles seperating the observers, there might be some weather phenomena that is going, through no fault of anyone, unreported. This may be something very small but as the program extrapolates out days and weeks ahead, that little something may, and usually does, have a huge effect on the weather that will be seen in the forecast area.

It is known as the “butterfly effect”( a butterfly flapping his wings halfway aroung the world will eventually have an effect on the weather seen locally), and it is the dirty little secret of weather forecasting. Until we are able to collect data in a close to continuous fashion between observation points there is very little chance that our weather forecast a week ahead will be correct and nobody sees this possibility occuring in the near future.

So the next time you see or hear your local weather forecaster speaking confidently about next week’s forecast, know that deep inside he or she is grimacing and taking that forecast with a whole bushel full of salt. And please, hold all those weather jokes!

Leave a Reply