Weather: Why do seasonal forecasts seem so unreliable?
Tuesday, February 3, 2009
When forecasts change from one month to the next, Ontarians are entitled to ask why. The answer lies in the variables used in the Environment Canada climate models
by HENRY HENGEVELD
Recent seasonal forecasts issued by Environment Canada remind us that we should still be cautious in using them as a planning tool.
In the beginning of November 2008, forecasters predicted a 40 to 70 per cent probability that average temperatures for the following winter season (December through February) would be above normal for most of Ontario. By comparison, the likelihood of below normal temperatures was ranked at less than 20 per cent. They also suggested that the amount of precipitation for south and central Ontario would be above normal.
However, by Dec. 1, 2008, the forecast had changed significantly, suggesting that the winter season would experience near normal average temperatures across central and southern Ontario, below normal in northern Ontario (including the Thunder Bay region), and a residual pocket of above normal conditions in the eastern tip of the province. Precipitation projections remained above normal. The confidence levels given to these projections were rather disconcerting. In fact, the projections gave approximately equal chance that temperatures would average above normal, near normal or below normal. Not much better than a crap shoot!
These rather uncertain projections raise two prominent questions. How are such forecasts generated, and why do they appear to be so unreliable?
Environment Canada has been producing one- to three-month outlooks for temperature and precipitations anomalies for Canada since 1995.
These outlooks are produced by averaging the results of 40 simulations of future weather developments generated by four different weather and climate models.
All four models use the same current atmospheric data as inputs. However, three key global weather variables – sea surface temperatures, sea ice cover and snow cover – help determine how those conditions will change in subsequent months.
The four models differ in the way they treat these variables. Throughout the forecast period, all of the models keep sea surface temperatures fixed at the values last observed. One model uses average historical data for ice cover, while the other three allow the current ice conditions at the start of the forecast period to slowly move towards the long-term average for the season. All use differing methods of forecasting snow cover conditions.
By repeating the simulation 10 times with each model (each simulation with slightly different starting conditions) and examining how many of the resulting 40 simulations show above, near and below average temperature and precipitation conditions, the forecasters can obtain a measure of the likelihood, region by region, of the projected outlook.
Environment Canada also provides seasonal outlooks for three to nine months into the future. For these longer range projections, they use a statistical approach. The main variable used is global sea surface temperatures.
The statistical model used to develop the long-range outlooks is based on equations that describe how different patterns of sea surface temperature anomalies observed over the past 35 years have affected Canadian temperature and precipitation patterns for the 12 months following each observed anomaly.
The observed sea surface temperature anomalies for the most recent 12 months are then used as an input to this model to develop projections for temperature and precipitation anomalies for the next 12 months.
The skill in developing seasonal forecasts depends very much on how much the key variables used in the models differ from their long-term mean. The stronger the departure from normal, the greater the confidence in the projections.
During a strong El Nino event, for example, the unusually high surface temperatures over the eastern tropical Pacific Ocean cause a pronounced change in global pressure patterns and hence in the position of weather fronts and storm tracks. This usually results in much warmer than normal temperatures across Ontario, although other factors can always intervene to disrupt this pattern.
While total precipitation can still vary from El Nino to El Nino, the warmer temperatures also usually result in less snow accumulation on the ground. In this case, the skill in the seasonal forecast is quite good. During La Nina events, where eastern tropical Pacific Ocean surfaces are cooler than normal, winter temperatures in Eastern Canada usually cool significantly, with a corresponding increase in snow cover depth.
However, the effects on weather in the Great Lakes Basin are relatively modest. Under these circumstances, other determining factors that may be more difficult to predict can assume a greater role in seasonal projections for the region, and the skill decreases. In some years, when sea surface temperature anomalies are very weak – like this year – the skill is not much better than you would get from rolling the dice.
I often get asked why, when there can be such lack of confidence in seasonal forecasts, anyone would put any credence in climate projections for 30, 50 and even 100 years from now.
In response, I often compare the climate forecasting versus seasonal weather forecasting with the analogy of season-to-season climate prediction with day-to-day weather forecasting. While we may have little confidence in a projection for weather events one week from now, we can predict with confidence that average temperatures across Ontario next July will be some 12 to 14°C above the annual mean.
In the first case we are talking about week-to-week variations in weather caused by constantly changing positions of high and low pressure systems.
In the other, we are discussing seasonal climates dominated by predictable changes in our seasonal exposure to the sun's energy.
Likewise, seasonal weather is determined by often unpredictable variables while long term climate trends are influenced by global climate system forces that are much more predictable.
A final note. If you don't like the lack of confidence in Environment Canada's seasonal prediction, there is always the alternative provided by the Old Farmer's Almanac. This year, for southern Ontario, it proclaims that "winter will be slightly milder and noticeably drier than normal, with below-normal snowfall, except in southwest Ontario." BF
Henry Hengeveld is Emeritus Associate, Science Assessment and Integration Branch/ACSD/MSC, Environment Canada.