Numerical Models

Numerical  Forecasts Satellite Surface Obs Forecasts


About Numerical Weather Models

Both Environment Canada and the U.S. National Weather Service (along with many other countries) maintain a set of national numerical weather prediction models. These models typically have a range of valid times that go out as far as 16 days into the future, but are usually not particularly reliable until about a week in advance. Reliability increases as the target date approaches, and by two or three days out, the models accuracy is usually good enough for planning a final eclipse-viewing site.

Model range is usually related to model resolution — the level of detail in the calculations — and is often expressed as the separation between grid points used in the calculations. Long-range models have relatively coarse resolution, while shorter-range models calculate on a much smaller grid. The U.S. GFS (Global Forecast System) model, which goes out for 16 days, runs with a resolution of about 13 km for the first 10 days and then at 25 km for the balance of the “run.” The NAM (North American Model) has a resolution of 12 km and a run of 3.5 days. The Canadian GDPS and RDPS models have resolutions similar to the GFS and NAM, respectively, but use a different suite of numerical algorithms to calculate the meteorological parameters.

Very high-resolution models (the Canadian HRDPS and the U.S. HRRR) have resolutions of 3 km or less, but shorter runs (48 hours for the HRDPS and 21 hours for the HRRR). To compensate, these models run much more frequently (every hour for the HRRR) than their longer-range cousins.

College of DuPage Numerical Models Map display of output from U.S. long- and short-range models
UQAM Weather Centre NWPS Map display of numerical output from US and Canadian models Graphical display of several U.S. and Canadian model outputs at user-selected locations
Skippy Sky Astronomy Weather Map display of weather elements over North America from the U.S. GFS model
Clear Outside Global 7-day cloud & weather forecast at user-selected locations
Table 1: Chose a link from the list above to explore computer-produced forecasts for your observing site.

How to Use the Model Outputs

The links provided in Table 1 above will lead you to sites where the model outputs – temperature, wind, cloud, precipitation – can be viewed. The problem you will encounter is “how reliable are the predictions?”

There are no guarantees, so you will have to do your homework.  Here are some clues:

  • Does a single model forecast the same cloud pattern from one run to another for eclipse day? When the patterns begin to repeat from one day to the next, it is a sign that the model is “locking in” on the pattern.
  • Do different models forecast the same cloud pattern (Figure 1)? The various models use different (though similar) algorithms to forecast cloudiness. Some are tuned to convective cloud formation, others to larger-scale systems. Environment Canada uses a different formulation than the National Weather Service. Don’t be surprised if you see a lot of difference even three or four days out; just hold your breath and wait until the models converge. There will almost always be some lingering differences, but the areas of similarity will be more reliable forecasts
Figure 1: Comparison of 30-hour forecasts of cloud cover from the NAM and GFS models. These images are taken from the College of Dupage site.

Ensemble Forecasts

Figure 2: Ensemble temperature and dewpoint forecasts from

Modellers know that that numerical predictions aren’t entirely reliable, often because the initial condtions that go into the models are incompletely measured around the globe. To help evaluate this uncertainty, the models are run many times with slightly different initial conditions and the differing predictions amalgamated into an “ensemble” forecast that shows the range of possible outcomes (Figure 2). All of the links in the table offer ensemble forecasts, but the easiest to follow might be those at, where they are presented in graphical format. These ensembles, while somewhat discouraging in the extent of their ranges, will give you a measure of the reliability of the forecast for eclipse day. A small range means a high degree of stability in the forecast.

Climatology and weather for celestial events