Thursday, January 29, 2015

The Joy Of Computer Models

 We just had a major snow storm go through New England. A good old nor'easter blizzard. Hurricane force winds. Lightening. The works. As people are prone to say lately, "it wasn't my first rodeo". In fact it's not even my second or third. Over the years, I have seen many blizzards and been through many years of heavy snow accumulation. 

Back in the day, it was illegal to broadcast a private weather forecast. All radio and TV forecasts were originated by the United States Weather Service. The TV stations would hire comely young women to deliver said forecasts. The forecasts were not terribly reliable. But that law giving government control of weather forecasting was changed. Then the TV stations hired trim young folks with degrees in meteorology, except for Al Roker, to deliver their own forecasts. It was still not terribly reliable.

Then came the glorious days of the computer and computer modeling. What could be better? Right? Well maybe, not so much. Computers don't just do these modeling things on their own. Someone has to write a program. That program writer decides what factors are important and what are not. Then they have to enter it. Test it. Debug it. And finally put it into beta testing. For an important program, it is a long and arduous process. And to be honest, some programmers are better than others.

But in the end, a computer model is a programmers idea of the relative importance of various factors that effect weather and the results those factors will create. There is not just one model that all meteorologists agree on. Our blizzard was tracked and forecasted by, probably, a dozen or more models. The forecasters look at the ones they consider most reliable and base their predictions on those.

Mother Nature has a way of dumping any or all of those models in the trash can. There are just too many variables involved in weather to get it right every time. There may come a time when computers are powerful enough to handle a task that complicated, but I wouldn't expect it any time soon.

I think that the forecasters did a formidably good job with this last storm. It moved off shore a little. That saved New York and New Jersey. They nailed it on New England. An A+ performance as far as I am concerned.

Now all of this having been said, modeling can only take you so far. Pretty accurate, two or three days. Sort of accurate, a week. Beyond that, a computer generated guess. But there are those that would have you believe that computer modeling can go far beyond that. They would have you believe that they can predict the climate for our whole planet out some thirty years or more. No. Not today. Not any time soon. And any time someone tries to tell me that the science is settled on any question, I know that they are fooling themselves or trying to fool me. The science is never settled on any question.

There is one thing that I know. And I know this for sure. Mother Nature has no tolerance for know it alls. And you can't fool Mother Nature.

No comments:

Post a Comment