Operational Model Forecasts

Numerical Weather Prediction Products

What are these products?

The products that are made available on this site originate from the operational runs of the GEM Regional Deterministic Prediction System (RDPS) model and the GEM Global Deterministic Prediction System (GDPS) model. Traditionally, model forecast output has been processed into raster graphics as four-panel charts that can be printed out on large sheets of paper. In the course of the forecast process, meteorologists often like to annotate these charts, compare them with previous ones, and with the output from other models.

The oldest and most familiar of these products is the classic four-panel chart, which depicts several of the traditional parameters used in general meteorology. More specialized charts are also made available, depicting model forecasts of fields such as air turbulence, aircraft icing, or parameters used in forecasting seasonal severe weather and ocean waves.

What is Numerical Weather Prediction?

Numerical Weather Prediction (NWP) is the forecasting of weather elements through the use of numerical models, such as CMC's Deterministic Prediction System models. All NWP models are based on the following idea:

If we know enough about the state of the atmosphere at the present time, and we know the physical laws that govern the atmosphere, then we can code these laws into a sufficiently fast computer and do a virtual "fast-forward" to see what the atmosphere's evolution is going to be, through the next few hours or days.

Much like a digital camera contains an array of pixels instead of a smooth, continuous picture, the model performs its calculations on a three-dimensional grid. You could also think of a huge 3D spreadsheet where each cell holds the forecast for a particular point in the atmosphere. For instance, the GEM Global Deterministic Prediction System (GDPS) model operates on cells that are each 0.3 degree of latitude by 0.45 degree of longitude. Time-wise, the evolution of the weather is also calculated in steps. The current GDPS model "sees" time in increments of 15 minutes.

Uses of NWP

Model forecasts have become an indispensable source of information in virtually every aspect of weather forecasting. The level of detail in modern models allows for a wide variety of products and forecast fields to be delivered, for use not only in general meteorology, but also in specialized areas such as aviation and air quality. Advances in data visualization and delivery methods show great promise for the users of meteorological products and the practitioners of the science. However, managing the mass of forecast data created by the models is fast becoming a science of its own. Output from modern NWP models requires post-processing to make it intelligible and, most importantly, expert human interpretation in order to assess its meaning, qualities, and possible flaws.

Limitations

While NWP is the greatest success achieved by the science of meteorology, its application can still be said to be only partially effective. There are three main reasons for this:

  1. One has to face the problem of creating a sufficiently accurate picture of the state of the atmosphere at the outset of the forecast process. Errors introduced at the beginning of the forecast will propagate and amplify at each forecast interval, gradually eroding its accuracy and usefulness.
  2. To move from a theoretical understanding of the weather to computer code that can generate a weather forecast, modelers will inevitably have to make some approximations.
    1. The first kind of approximation is the model's resolution, or its ability to capture smaller details.
    2. The second kind of approximation, called parametrization, becomes necessary to take into account the large-scale effects of phenomena that are too small to be picked-up at the model's resolution. For instance, individual thunderstorms are too small to be forecast by the model; yet in order to be useful the model must still produce a good approximation of the effects of thunderstorms on large-scale precipitation and temperature patterns.
  3. To be successful, the model must integrate an understanding of many different phenomena and their interactions: how the wind blows; how heat is received from the sun and transformed by the oceans, the ground, the air, and the clouds; how water vapour condensates into clouds and how droplets of water turn to rain, ice and snow; how friction near the ground mixes the lower layers of air. Thus, errors in handling one type of phenomenon can contaminate other parts of the model, or amplify errors in other model sub-systems.