Scientific background of extended range forecasts

The accuracy and quality of weather forecasts decreases as the predictions stretch further into the future. Everyone knows from their own experience that weather forecasts for the next day are much more reliable than, say, a five-days forecast. This is due to the chaotic nature of the atmosphere: Smallest and unpredictable perturbations can grow quickly, limiting the predictability of weather to approximately 10 days. How should it then be possible to do a forecast for the week after next? Or even further into the future, as in our seasonal forecasts?

The answer to resolve this seeming paradox is that long-range forecasts do not aim at forecasting specific weather events, but rather focus on weekly means. With this "trick", the prediction range can be extended significantly. By taking weekly averages, the predictions are less sensitive to chaotic perturbations, while a range of other factors, such as soil moisture, snow cover and especially the state of the oceans, gain in importance. If such factors are considered appropriately, then it is also possible to extract information on the expected longer term weather trend.

For these reasons, extended range forecasts at MeteoSwiss are based on a coupled ocean-atmosphere-land model, which calculates the evolution of oceans and atmosphere by solving complex equations. These calculations are initialized with data obtained from measurements all over the world, including satellites, buoys, airplanes and surface stations. To quantify the uncertainties of the prediction, a whole ?swarm? of such model-simulations is carried out. That way, the range and probability of possible climate states can be quantified. Finally, the predictions are calibrated with past measurements.

Even though there has been a lot of progress in the development of extended range prediction systems, the quality of such forecasts is in practice still limited. The models cannot reproduce the full complexity of reality. They are based on numerous simplifications. Moreover, not all regions in the world are to the same extent affected by the boundary conditions mentioned above. Thus, it can always happen that unpredictable and chaotic weather events distort the expected mean tendency. 

Back to current weekly climate outlook