Seasonal Forecasting

The weather has such a profound impact on our daily lives—from how we dress to the cost of heating and cooling our homes—that it is not surprising that most Canadians have an almost obsessive curiosity about what the future will hold.

After all, Canada experiences some of the most variable weather on the planet from season to season—from scorching heat waves in summer to ice storms in winter. Accurate weather forecasts can make life easier, safer and more comfortable, and enable people in weather-sensitive industries—such as agriculture, shipping and recreation—to mitigate risks and take advantage of benefits.

Longer-term weather forecasts can be particularly beneficial because they provide greater advance notice of possible conditions. Firefighters may consider seasonal forecasts to position staff in areas susceptible to forest fires; merchants to stock up on items such as snowblowers, shovels and generators; farmers to determine the best time to plant their seeds; and commodities markets to trade futures in weather-dependent industries.

To address the need for this long-lead weather information, Environment Canada issues seasonal temperature and precipitation outlooks on the first of March, June, September and December each year. Unlike daily forecasts, which predict specific temperature values and the likelihood of precipitation, these long-lead forecasts describe the general character of the weather over the next 90 days and 3 to 12 months. The forecasts are depicted on the department's weather Web site in the form of colour-coded maps.

How it's done

Although Environment Canada began issuing seasonal forecasts more than a decade ago, it is only in the last seven years that they have been produced by automated computer models instead of subjective scientific analysis.

The process begins less than a week before the forecast is issued. The current conditions of the atmosphere and oceans are fed into two numerical forecasting models—one a global climate model used to predict climate change over periods of 20 to 100 years, and the other used for operational weather forecasting. Every day for six days, both models are run to simulate atmospheric changes over the next 90 days.

The results of the 12 runs are averaged out and compared to past climatology to create temperature and precipitation outlooks for the period. If all 12 runs say the same thing, the confidence or skill level of the outlook is said to be higher; if there is a lot of spread among them, it is lower.

How Accurate is it?

One way of verifying the skill of a forecast is to compare it with actual observations for the same period. To verify the models over a longer period of time, scientists at Environment Canada ran them to create seasonal outlooks for every season over the past 26 years. These were then compared to actual seasonal observations.

The results indicate a significant level of skill for the temperature forecasts for certain seasons and parts of the country—such as winter forecasting in western Canada—and room to improve in others. Precipitation forecasts, on the other hand, show little skill. Forecasts with skill levels of under 45 per cent are not considered much better than chance, since even a chance forecast is right 33 per cent of the time. Indicating skill level is an important way to tell users how accurate a forecast is likely to be.

For longer-term seasonal forecasts-that is, the three 90-day periods that will occur 3 to 12 months down the road, a statistical model is used that calculates future conditions based on the historic influence of sea-surface temperature, ice cover, snow cover, height of specific air-pressure levels, and other major forces.

Seasonal temperature outlooks are ranked as above, near or below normal, and precipitation outlooks as wetter, near normal, or drier. How much do temperature and precipitation have to differ to be rated as above or below the norm? Different thresholds have been determined to take into account various climate conditions across Canada. For temperatures, it may be as little as a few tenths of a degree Celsius in some coastal areas, due to the regulating effect of the oceans. On the other hand, in some parts of the Yukon it is more than 2°C in winter.

A look back

Seasonal forecasts made over the past year have shown remarkable accuracy when compared against observations. Last winter, while meteorological services in other countries predicted below-normal temperatures, Environment Canada's forecast predicted above-normal temperatures—and turned out to be 66 per cent accurate. And despite the fact that the previous 19 seasons brought above-normal temperatures, agreement among all 12 modelling runs resulted in a below-normal forecast for spring 2002 that was 81 per cent correct.

Room for Improvement

While seasonal temperature outlooks are gradually improving, precipitation has a long way to go. One reason is that temperature is continuous, while precipitation is sporadic—so a single event can have a significant impact on average rainfall for a particular period. Even on a daily basis, precipitation forecasting lacks accuracy due to the large number of interacting factors involved.

Despite these shortcomings, seasonal forecasting has made significant inroads for such a young science. With scientific observations and understanding of the atmosphere, oceans and environment constantly expanding, expectations are that our seasonal forecasting abilities will improve greatly over the next few years.