Hey everybody, and welcome to this week’s Weekend Weather Word. I know it’s been a while since I published the last edition, but real life just has this talent for throwing a spanner in the works. For the last two editions, I concentrated on two of the basic building blocks of weather systems worldwide. Today, I’m going to take a more theoretical approach. If you’re a regular reader of this blog, and there’s no reason you shouldn’t be, you’ll have read the words “Weather model” more than once. These models are crucial to modern meteorology, in fact, they are the very foundation upon which it is built. But what exactly are these weather models? Well, if you’re thinking about dangerously skinny girls standing around in skimpy clothes in all kinds of weather, you better think again.
A weather model is basically a complex computer program that creates numerical predictions about the movement of air masses in the atmosphere. These predictions are created by horrendously complex algorithms that are regularly fed with data from weather stations all over the coverage area of the model. Sounds simple, doesn’t it? Well, unfortunately, it is anything but simple. You see, given the chaotic nature of the atmosphere, specifically of the troposphere, the lowest part of the atmosphere were 99.9% of weather takes place, there is no one formula to predict the weather. What’s more, the models are victims of the data that is being fed into them, “Garbage in, garbage out” as some veteran programmers like to say. This is exacerbated by an ever decreasing number of weather stations, while the models themselves grow more complex, requiring ever larger data centers to run them. Furthermore, there is no one single model that has proven to be accurate enough to serve as a global gold standard in weather forecasting. Add national pride, inter-service rivalry between military branches, and even competition between different universities to the mix, and you’ve got a pretty big mess on your hand. Oh, and there are now private forecasting companies as well, with their own models of course. Have I totally confused you yet?
Pretty chaotic, isn’t it? Well, many of these competing models only work for a certain region, the ICON model operated by the German weather service for example is focussed on central Europe, while many US models are understandably designed to provide accurate forecasts for the lower 48, Alaska, and Hawai’i. The number of large scale global models is rather limited, with only 4 being reliable enough to be of use in my forecasts on this blog. I’ll be focussing on these models below.
Global Forecast System (GFS)
The Global Forecast System is operated by the US National Weather Service (NWS). It calculates forecasts for up to 16 days into the future, although in my experience, accuracy pretty much goes down the drain after around 7 days. It works with a resolution of 13 kilometers, meaning that it calculates weather conditions for grid points spaced at 13 kilometer intervals. This resolution drops to 27 kilometers for forecasts beyond 10 days. Because the GFS is created by the NWS which is a US Government Agency, its forecast products are in the Public Domain, making it one of the most easily available models.
It is not without its flaws, though. In my experience, GFS is a “slow” and “wet” model, meaning that weather forecasts produced by this model predict a slower movement of weather systems, as well as rather more rainfall than eventually comes to pass. This is not much of an issue for hobby forecasters like me, however this data is also used by the National Hurricane Center in Florida to predict the path of hurricanes and recommend evacuations accordingly. While part of the problem here may be related to a lack of supercomputer resources at the NWS, work has been underway since 2010 on a successor model using the wonderful designation “Flow Following Finite Volume Icosahedral Model” or FIM for short.
Global Environmental Multiscale Model (GEM)
Moving north from the US to the civilised part of North America, the GEM model is operated by the Canadian Meteorolgical Center. It produces forecasts for up t0 10 days into the future. Like the GFS, GEM is a pretty low-resolution model, although it does predict a wide range of weather parameters. I generally don’t like GEM that much, namely because the forecasts are generally made in 12-24 hour steps, rather than the three hour resolution of the GFS model. Still, it is a good place to go to when looking for a second opinion.
Navy Global Environmental Model (NAVGEM)
To put it bluntly, the NAVGEM model is a clone of the Canadian GEM model operated by the US Navy’s Fleet Numerical Meteorology and Oceanography Center. I touched on this model back in one of the first Weekend Weather Words. It produces weather forecasts for up to 8 days ahead, which are updated 4 times a day. This is generally speaking a pretty good model to keep an eye on, as it is quite accurate. That makes sense, after all the US Navy is dependent on high quality weather forecasts, and no admiral wants to willingly sail his task group into a category five hurricane.
European Center for Medium-range Weather Forecasts (ECMWF)
This alphabet soup is actually not a weather model per se, but the name of an independent European forecasting center. Based in Reading in the UK, this institution is set up to produce accurate forecasts in the medium term, meaning up to 15 days ahead. It is jointly run by 22 nations. The model produced by the ECMWF is actually called the Integrated Forecast Model, however it is rarely referred to as that, mostly being referred to as the European Model over in the US, while I prefer just to refer to the ECMWF model as “Reading”. But while the nomenclature may be up for debate, the accuracy and track record certainly is not. From my experience, Reading produces the only medium term forecasts that are even remotely reliable.
Of course, there are a number of other models out there, many of them very specific to certain regions, like the Dutch HARMONIE model, and I certainly call upon those when needed. However, weather models can only take you so far. As important as computers are in weather forecasting these days, nothing beats a look at the sky, or doing local measurements at regular intervals. Meteorology is still as much an artform as it is a science.