This past June, the U.S. National Weather Service began using an updated version of its primary weather forecast model, the Global Forecast System. According to the government press release, “this is the first major upgrade in almost 40 years to the model’s dynamical core, which is a key model component that computes wind and air pressure for successful numerical weather prediction.”
The reason that this is important is that it should lead to increased accuracy in weather forecasts covering the next 15 days, not only from the National Weather Service, but from media meteorologists and from weather companies like AccuWeather and The Weather Channel.
Back when I started forecasting the weather in the 1960s, we had computer forecast models, but they were not very accurate and producing weather forecasts still had a large artistic and intuitive component in addition to science and pattern recognition.
The accuracy of these computer forecast models suffered from two primary deficiencies:
- They were dependent on the accuracy and completeness of current weather condition reports, which were sorely lacking. Today, we have a better network of ground-based observation stations, Doppler radar, and satellite observations, which has helped to make the assessment of the initial state of the atmosphere more accurate and of higher resolution.
- Computers were much slower and had much less memory than today’s supercomputers, which meant that many simplifications had to be made to both the resolution of the data and the representation of atmospheric physics (the “dynamical core”) in order to have the computer forecasts generated in a timely manner.
Over the ensuing decades, as computers grew in speed, memory, and other capabilities, the forecasts generated by computer models became much more accurate, to the point where it became almost impossible for even the best forecast meteorologist to consistently improve the forecasts that they generated.
This allowed the creation of the algorithms and artificial intelligence that generate the forecasts that you receive on your phone and computer. All of those forecasts, with their pinpoint localization and comprehensive detail, are generated automatically with little, if any, human intervention.
The keys to the generation of these forecasts are the forecast models from the National Weather Service and international government agencies such as the European Centre for Medium-Range Weather Forecasts, which generates the “ECMWF” model often considered to be the most accurate, on average.
These forecast models generate the forecast data that are ingested by companies such as AccuWeather. Then, through the use of artificial intelligence, this information is enhanced and adjusted to be made even more accurate for its eventual role as the basis for the electronic forecasts that you likely utilize.
What an Updated System Means For You
A couple of years ago, the National Weather Service acquired much more powerful computers. Since then, they have worked diligently to take full advantage of their enhanced capabilities by improving their forecast models.
With this improved model now being used operationally, all weather forecasts should be more accurate, as all forecasts depend on this model to some extent. The primary increase in accuracy will be in the longer range of the model forecasts, where errors sometimes grew larger over time. There likely will not be a noticeable improvement in forecasts covering the next day or two, but forecasts 5 to 15 days in the future should be significantly more accurate.
Weather from The Old Farmer’s Almanac
Of course, we have our own methods here at the Almanac. Read about how we predict the weather here.
For long-range weather predictions for this fall 2019 through summer 2020, pick up a copy of The 2020 Old Farmer’s Almanac! Here’s where to find the 2020 edition near you.