Thursday, January 30, 2014

Oh, the weather outside is frightful...

On two occasions I have been asked [by members of Parliament]: 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.- Charles Babbage

As you may have heard, it snowed down here in the South. Birmingham and Atlanta were paralyzed, and most stuff just shut down. That part is normal. Southern cities do not invest heavily in snow removal equipment because it wouldn't be used more than a day or two every couple of years. The normal reaction to snow is to simply close everything down. Everyone stays home, or leaves work or school early, and watches the snow hit the ground from the comfort of their respective living rooms. The next day it melts and things are back to normal.

Well, it didn't work out that way this time. There were two problems. First, the temperature dropped precipitously as the sleet and/or snow fell. Normally, some snow falls, hits the relatively warm ground and promptly melts. In this case though, as the icy stuff hit the warm ground, it melted then got covered by more stuff that didn't melt as the temperatures fell. The ice then got covered by snow, which is about the worst case scenario for winter driving. Yet even that wouldn't have been a huge deal had it not been for the second problem: The weather guessers screwed up big time.

To begin with, winter weather warnings were issued early on (some as early as Sunday evening) that said the bad stuff would arrive around 10 AM and would affect Alabama south of I-85 and US 80. Basically, that's everything south of Montgomery. Birmingham, on the other hand, might get a “dusting” of snow. Now, if I were still doing my ridiculous commute to work in Birmingham, I probably would have considered going into work because both Birmingham and my area (slightly north of Montgomery) were outside the nasty stuff. Fortunately, I am retired now, so I didn't have to make a decision. Thank goodness.

When I got up Tuesday morning, it obvious something had gone very wrong with the predictions. First of all, it was sleeting, almost four hours before the precipitation was supposed to start. Second, looking at the radar, it was obvious that it was snowing like the dickens up in North Alabama, including Birmingham, where it wasn't supposed to. Because people up there believed that this was the “dusting” they were supposed to get, they went to work and school. By the time they realized just how wrong the weather services had been, the roads were a slick mess. But, the schools and businesses decided to tell everyone to make a break for it, which of course ended up with a major traffic catastrophe.

So what happened? Well, it seems that weather models were wrong. To begin with, the storm was farther north than the models had predicted. Then there was this from a National Weather Service person: “ 'The models we use generally show storms coming in slower than they do, so we backed up the predicted time that storms are expected to hit.' ” According to [Mark] Linhares, the NWS forecast backed up Tuesday's storm arrival by several hours, but was still off by a couple of hours.”

Now chew on that for a moment. What he's saying is that the computer models the National Weather Service uses are wrong and haven't been corrected. That is a frightening statement. Yet we see evidence of this all through hurricane season, when the models predict storm paths that are radically different from one another, and which change from one hour to the next. If you ever want to see why the path predictions are so weird, go to a weather site that shows all the models. A couple of years ago, at least one model showed a hurricane making it all the way to Wisconsin. One would think they would have dumped that model, but my suspicion is that on some other occasion it had been accurate, or at least had been more reasonable.

Computer modeling is notoriously difficult, especially when it's about real-world systems. It's one thing to predict collision interactions in a particle accelerator or how galaxies interact gravitational. It's quite another when people's lives are affected by predictions of where a hurricane will strike or how much snow is going to fall. Businesses and governments make decisions based on these predictions, and the consequences of being off by a “couple of hours” can be very costly in terms of money and in terms of lives.

So, when the Weather Service admits the model is wrong and admits that they didn't fudge it right by “a couple of hours” (in some areas it was more like 3 or 4 hours), it's time for someone to evaluate the modeling that's being done. Maybe it's time to dump the models all together and go back to actually having people read the surface maps. I suspect that forcing these people to actually analyze the weather patterns with their heads rather than almost solely with computers and fudge factors might give us better forecasts.

Face it. It's unlikely it could be much worse.

No comments: