Early in The Long Winter, Pa clearly pays attention to the signs for the coming winter. He checks the thickness of muskrat dens and notes the patterns of birds flying south. He talks to his fellow townsmen. Of course, Pa also listens to an old Indian predict that every 7th winter is severe and every 3rd of those is especially severe (that’s a subject that will earn its own post another day). He didn’t just trust one thing. He took in information from all kinds of sources to give Ma his prediction that the coming winter would be a hard one. And if you think Laura lays it on thick in the book, it’s even more extensive in the manuscript.
Those of us who make weather and climate predictions now aren’t all that different. Instead of muskrats, birds, and folklore, we turn to computer models and observations around the globe to make our predictions. But we definitely consult more than one source. In fact, when it comes to our forecast models, we want as many as possible. We call it an “ensemble” approach, to take as many models as possible and see what range of solutions they give us. Forecast models are run in the U.S. (mainly from the National Centers for Environmental Prediction, a part of the National Weather Service [NWS]… but also from the military and other federal research labs and universities), Europe (the European Center for Medium Range Forecasts [ECMWF] and the U.K. Met Office), and Canada, for the most part.
Sometimes, they agree with each other about an event pretty well. When they do agree, we feel a bit more confident about the forecast. When our maps start to look like spaghetti with all of the different possible solutions — and yes, we literally call those “spaghetti plots” — then we know that confidence is low and there is a wider range of possibilities.
One thing that has caught some fire in the news lately is the success of the “European model” (ECMWF) compared to our federally funded American models. The ECMWF pegged the track for Hurricane Sandy days before the American models settled on the same track as a most likely solution, and several news outlets made a pretty big deal about it (for example, USAToday, Politico, and Popular Mechanics, just to name a few). As a forecaster, I can tell you that I’m more comfortable with the ECMWF than most others on most days, especially at longer ranges (4 days out and onward). It does outperform our homegrown models, on the average. Like every model, it does sometime get it wrong, too. It just happens less often.
What makes the European model (I’ll call it the ECMWF here) so much better? One reason is that it runs at finer resolution for more days out in time — meaning, kind of like a TV or computer monitor, the picture is better when there are more pixels covering the same amount of space. So it’s basically like an HD TV compared to an old tube TV, especially from about 8 days out to about 14 days. Another reason is that it takes in more data from real observations to start each model run, and it does it using a method that seems to be better. And the reasons the Europeans can do these things and NWS can’t is because they have better computers to run all these high-powered calculations.
What’s up with us here in the U.S.? Why can’t we keep up with the Europeans?
In a word…money. Federal funding for NWS in general, including our forecast models, simply isn’t keeping up with technological advances that would be needed to improve the models. We can’t afford to upgrade our computers to improve how we run our models. We can’t afford the research to know what improvements would be helpful to forecasters. Heck, we are barely able to keep the lights on in NWS with our current budgets, and we’re running on significant staffing and equipment shortages. The Europeans can afford it because they charge for their model information, while we Americans let everyone have equal access to it for free. The Europeans also don’t hire forecasters to provide your weather forecasts and warnings for free, with a network of offices around the country to keep in touch with what’s going on locally. They are focused much more intensely on pure modeling.
Is that better? I’m not sure. I think it’s nice that there’s an NWS office pretty close to everyone in the country, and we have a much bigger patch of the globe to cover. What we need, really, is enough money to support both. A forecast model isn’t much good if someone isn’t there to interpret it and make decisions from it. A forecaster can’t do as good of a job if the tools in our toolbox aren’t sharp. We need both the models and the forecasters to give the best forecasts to our American citizens.
How do we get better models in our National Weather Service? Well, since I’m a federal employee, I really can’t tell you what to do, even though I’m on my own time and my own blog and my own equipment now. I’ll leave it to your imagination to think about what to do from here, keeping in mind who it is that makes the decisions about our budgets in a federal agency.