Wednesday, March 6, 2013

A Case for Better Weather Models



I am disappointed. Though I knew the weather on the East Coast would be different from what I was used to when I moved to Maryland, I hadn’t quite realized just how different it would be. I had expected fewer tornadoes, and maybe even fewer thunderstorms, but I wasn’t really prepared for such mildness in the winter. Having grown up in the Midwest, a winter with less than five inches of snow was really no winter at all. When my facebook and twitter feeds were flooded with information about snow in Boston, Kansas City, Lubbock, Long Island, and various other places across the country where I did not have the pleasure of being, I often responded with sentiments of a sad and/or jealous nature. Naturally, you might imagine my joy and hopeful anticipation when it was said earlier this week that the DC-Baltimore area would be blanketed on Wednesday, give or take a day, with a nice, thick, layer of snow.

Well, the weather gnomes had different ideas. As if to mock us for cancelling school, closing government offices, and advising people to hunker down and stay home whilst the skies unleashed their fury and hurled thousands, nay trillions!, of snow things down upon our weary heads, it became obvious by 1:00 PM EST that DC’s “Snowquester” might not even involve much snow at all. Sure, some areas west of the city saw copious amounts of Grade A snowman making, heart attack inducing snow, the DC-Baltimore area was dealt some early morning slush and surface temperatures that, wish as I might, refused to dip below freezing.

I’d had my doubts. The temperature yesterday topped out yesterday in the upper 40s, the temperature on Thursday was also forecast to be above 40F, and the rational part of my mind was saying “There’s no way that we’re going to get much of any snow with this...” A coworker of mine even left work on Tuesday doubting we’d get as much as 5 millimeters (because he’s French). The emotional part of my mind, however, was hoping against hope that somehow the cold would prevail and we’d get at least 5 or so inches (because I’m American) of snowfall to soften the harshness of such a mild, uneventful DC winter. My hopes were of course buoyed by forecasts from weather models, like the NAM, which was spitting out forecasts on Tuesday that suggested the DC-Baltimore area would receive a substantial amount of snowfall. 

These forecasts, as we now know, were actually pretty wrong as far as the DC area was concerned. They didn’t take into account the possibility of the surface temperature just not being cold enough for any appreciable snow. And the mistake was a costly one. A large portion of the DC-Baltimore area essentially shut down in preparation for this excessive accumulation of snow that didn’t end up accumulating. In doing a brief post-mortem of the event, The *gulp*Weather Channel mentioned that the model to perform the best for this Snowquester snowstorm in the DC area was, in fact, the ECMWF’s model, while American models lagged behind with varying degrees of wrong-ness. Granted, no model forecast is ever going to be 100% correct, 100% of the time (or even any percent of the time). It’s just not possible. In order to be useful, weather models have to approximate atmospheric processes, simply because atmospheric processes cannot be PRECISELY represented by equations, and because these processes are so complex and contain so many variables that it would be impossible to model them with absolute accuracy. And despite the snowfall forecast for the DC area being pretty much a complete bust, the models did do a decent job of telling us that there would be a storm and that some regions in Maryland, Virginia, and West Virginia would get a substantial amount of wintery precipitation. 

I can’t help but think of the irony of this storm. Just earlier this week, the nation’s Congress refused to put aside their frivolous differences and work (for once this decade) on some sort of compromise to avoid the major budget cuts outlined in The Sequestration, and now a poor forecast has hit the nation’s capitol and likely cost a substantial amount of money. What’s morbidly funny is the fact that these kinds of losses will only be avoided if the National Weather Service and our weather models can be improved, and if the weather data from satellites and terrestrial-based sources that we rely on to make our forecasts aren’t taken away from us. We busted. We busted because our short range, high resolution models need more development, and because the computing sources at the disposal of our National Weather Service probably aren’t currently good enough to even run better models. ECMWF arguably has the best weather model out there, and it’s arguably improving at a pace our operational forecasters don’t have the resources to match.

Americans: Isn’t it tragic that our weather, which is amongst the most volatile and diverse weather in the world, is better modeled by the Europeans than it is by any American-made model? Isn’t it tragic that we have the knowledge and talent in this country to vastly improve our National Weather Service’s forecast skill, but that we’re hard pressed to find an opportunity to do it?