Improving the Accuracy of Canada’s Weather Forecasts

The Global Environmental Multi-scale (GEM) Model

Forecasting the weather always stands as one of the most challenging tasks for any computer system. Meteorological researchers and computer scientists are always anticipating the next, more powerful supercomputer in order to expand their calculations – attempting to simulate more accurately one of the most complex processes on Earth. It was in the 1950s that the first electronic computers enabled the operational use of Numerical Weather Prediction (NWP) on a systematic basis. Since that time, standard measures of the quality of weather forecasting from the Meteorological Service of Canada's MSC Canadian Meteorological Centre in Dorval show dramatic progress. For example, a 5-day forecast in the year 2003 is as accurate as a 36-hour (1.5 day) forecast was in 1963. As the accuracy and range of forecasts increases, so does their value to the general public and to the 1/7^th of Canada's GDP (Gross Domestic Product) that is weather sensitive.

Environment Canada's Super Computer The front end of Environment Canada's supercomputer which is housed at the Canadian Meteorological Centre (CMC) in Dorval, PQ.The CMC is the main location for the Numerical Prediction Research Division and other components of the Meteorological Research Branch, and other Meteorological Service of Canada employees.

There is also another way to increase computing capacity. That is through the software that is used. The more efficient the programming is, the more calculations can be done within the time limits imposed by meteorologists and clients.

Researchers at the MSC made a large intellectual investment in the early 1990s to develop the Global Environmental Multi-scale model (GEM) which became operational in 1997. The heart of this numerical model is the dynamical core, which provides the computational foundation for describing the conservation of energy and momentum. All the other aspects of the model, such as predicting radiation, cloud formation, land/surface exchange etc., plug into this flexible core. The GEM is unique in the world in its ability to run with fixed or variable spatial and temporal resolution from the global to local scales - all with the same dynamical core.

The GEM dynamical core is so efficient and flexible that it now also powers several different forecast programs. The model is run twice daily at 24km horizontal resolution, with sophisticated data assimilation techniques that take advantage of real-time surface and satellite meteorological data from around the world, producing 48-hour forecasts for all of Canada. GEM is run at a coarser resolution of 100km to produce daily forecasts out to ten days for the entire globe. In somewhat of a scientific tour-de-force, GEM is also used for local and convective scale modelling out to 30 hours to simulate hurricane, severe squall line, and heavy precipitation weather.

The next test for GEM will be to extend its time horizon to produce seasonal forecasts that will be more accurate than those from our existing multi-model ensemble prediction system. Replacing the current multi-model system with a single GEM-based system will result in significant cost savings.

One of the biggest challenges facing meteorological services around the world is merging weather, ocean, climate, hydrology and air quality models into integrated environmental prediction models. Such integrated models can link atmospheric change to ecosystems, human health and the economy. This is another area where GEM holds great promise. Research is currently underway to develop these GEM-based unified modelling platforms for environmental prediction of the impact of human activity on the environment and the environment's impact on the health, safety, security and businesses of Canadians. GEM is an excellent example of doing more with less.