Contents area

Background Information on Monthly Outlook

In order to provide a forecast on the average weather conditions for the coming weeks, all available measurement data from around the world are fed into complex meteorological models. With the help of model simulations, probability distributions can be worked out. This work requires high performance supercomputers.

Footer

Top bar Navigation

Swiss federal authoritiesSwiss federal authorities

The accuracy and quality of weather forecasting declines steeply as the forecast time period increases. A weather report for the following day is therefore much more reliable than a five-day forecast. This is due to the chaotic nature of the atmosphere: The fact that the smallest, unpredictable perturbations can quickly spread means that weather forecasts are restricted to around ten days ahead. In the light of these difficulties, how are predictions made for the coming weeks and months?

Predicting average weather condition trends

The answer lies in the fact that long-range forecasts do not predict individual weather events, but make predictions on trends in average weather conditions. Forecasting in this way removes the chaotic processes from the equation, thus allowing a longer forecasting period. At the same time, however, a whole range of other factors come into play, such as the soil moisture content and snow cover of the continents, and, above all, the oceanic conditions. This is particularly apparent in the case of the El Niño phenomenon, when abnormal temperatures in the equatorial Pacific Ocean cause increased precipitation in South America, and droughts in Australia. When these factors can be taken into consideration, it becomes possible to provide forecasts on weather trends over longer periods of time.

Model simulations on supercomputers

The long-range forecasts produced by MeteoSwiss are based on a coupled ocean-atmosphere-land model. This enables the evolution of conditions in the oceans and the atmosphere to be simulated with the help of complex comparisons. All available measurements from around the world are fed into these calculations at the starting point. These data are derived from satellites, buoys, aircraft as well as land-based stations. To estimate the uncertainty level of the forecast, numerous such model simulations are carried out. This enables the probability distributions for possible climate conditions to be quantified. The final step is the calibration of the forecasts with past measurements.

These types of vast and complex model simulations require an enormous amount of computing power in the form of high performance supercomputers. The long-range forecasts issued by MeteoSwiss are based on simulations that are carried out at the European Centre for Medium-Range Weather Forecasts (ECMWF). The ECMWF is jointly run by 34 countries including Switzerland, to ensure optimum pooling of the member states' individual resources for this highly complex and costly work.

Limited predictive ability of long-range forecasts

Even though significant advances in long-range forecasting have been made in recent years, the benefits of such forecasts are still limited in practice. The models are not equipped to reflect reality in all its complexity, but rely on a number of simplifications. In addition, not all regions of the earth are affected to the same extent by the framework conditions mentioned. There is always a possibility, therefore, that unforeseen, chaotic weather developments could be superimposed over the forecasted trend.

The models for long-term forecasts have only a coarse spatial resolution. Particularly in mountainous regions like Switzerland, this means that certain weather situations are only represented in a simplified way. An important example are the frequent high-pressure systems in winter. The large-scale weather situation is captured by the model, but not its specific effects in Switzerland with often lower temperatures in the lowlands and warmth in the highlands. The monthly outlook must therefore be interpreted carefully and always be understood as an average statement for an entire region.