A probability forecast does not only provide a prediction of future weather events, but is also a measure of how likely the events are.
By counting the number of weather forecasts with similar output values, a weather forecast ensemble is created. If the different forecasts within the ensemble indicate similar weather events, it is more probable that those events will actually take place. If the different forecasts in the ensemble show a wider spread of weather events, the future weather is more uncertain. In other words, an ensemble can provide a measure of the uncertainty of the weather forecast.
Probability forecasts provide added value
The production of wind power is of course dependent on weather conditions, one of the important factors being ice on turbine blades. The Swedish Energy Agency is funding a project in which SMHI, Uppsala University and WeatherTech Scandinavia are developing the ability of forecast models to calculate icing rates and the concomitant loss of production from wind turbines. This development has continued for several years and the first successful tests of the ensemble forecasts have recently been carried out.
"Probability forecasting provides the wind power industry with a better basis for production planning. Two decisive factors in how much energy a wind turbine can produce are the weather and predictions of icing conditions. Delivering a probability assessment for the weather forecast has high added value for wind energy producers," says Esbjörn Olsson.
The latest developments in probability forecasts for the wind power industry were announced at the Winter Wind conference in Åre at the beginning of February.
Test of probability forecasts for local weather
For some time now, SMHI has been testing a new way of presenting probability forecasts for the general public. In SMHI's weather service available in Swedish at smhi.se there is a beta version that displays several options for weather developments based on local forecasts; what weather situation is most likely and what is least likely.