At least for a few days in advance, the meteorologists mean the change from high and low pressure areas, rain fronts and fine weather periods but quite good "under control" to have. The hit rate of the predictions for the next day is on average at "well over 90 percent", reported by the German Weather Service in Offenbach. However, detailed information on how this percentage has changed in the course of the past months, years and decades can not be elicited from the people of Offenbach. "You can not compare forecasts, as they were made a few years ago, with the methods available today, " press spokesman Meinhard Giebel points out. The numbers would therefore be "misleading".
The Berlin meteorology professor Horst Malberg and his staff have created a weather forecast for the city on the Spree every day for more than 20 years until the end of 1991. They used data provided by the German Weather Service - and evaluated the quality of their forecasts for precipitation, temperature, cloud cover and wind force on a scale of 0 to 100 percentage points.
The result: between 1971 and 1991, the hit rate of one to two day forecasts has barely improved. She climbed on the scale of the Berlin weather researchers from just under 84 to about 86 percent. "This shows how laboriously the atmosphere, despite all technical and scientific advances every single percentage point must be wrested, " says Malberg. He therefore considers the value of an average of far more than 90 percent officially stated by the German Weather Service to be unrealistically high. display
The Hamburg meteorologist Klaus Fraedrich, who has been working on predictive models for many years, sees a combination of classical computational models with novel methods - such as neural networks - as an opportunity to improve forecasts. Ensemble predictions are considered promising. "The goal of this method is to find out and state the error in each prediction, " explains Fraedrich. Instead of a single forecast, the computer calculates a large number of forecasts with slightly different initial values. The result is, for example, information about the probability with which it rains or storms in a certain place and day. This gives the meteorologists and users valuable information on the reliability of the forecasts at hand. Long-term forecasts for months or even whole seasons are high on the wish list of many weather researchers.
In Australia and North America, weather forecasts for several months have been published for many years. The meteorologists there can rely on a powerful ally: El Niño, which ensures a rhythm of several years for exceptionally high surface temperatures in the tropical Pacific. It has a very strong influence on the weather in the region around the Pacific Ocean - thus facilitating long-term forecasts.=== Ralf Butscher