For 21 years, diurnal migration of 46 bird species was observed under standardized conditions at Randecker Maar, SW Germany. The data are examined as to whether changes in median dates of passage of these species can be detected. For this purpose the data were analysed separately for three periods, i. e. 1970-76, 1977-83 and 1984-90. In 1970 to 1983, according to meteorological data the weather-related onset of autumn was delayed by five days in northern Germany and by six days in southern Germany. During the last observation period, 19 (67 %) of the 28 passerine short-distance migrants passed on average up to 10 days later compared to the first observation period. In 14 of these species the median dates changed continuously. Only 5 species migrated earlier than before- The majority of the short-distance migrants, therefore, seems to have adapted to the later onset of autumn. Currently there seem to be no plausible reasons for delayed migration other than climatic changes. The median dates in Non-Passerines were less clear-cut. Among short-distance migrants, delayed migration was found in 5, advanced migration in 4 species and among long-distance migrants, delayed migration was noted in 4 species, advanced migration in 3 species. Short-distance migrants may gain advantages from a longer stay on their Central European breeding grounds. Populations departing late should tend to become resident, thus advancing the median date for the fractions of population still migrating. Even advanced migration may thus be explained by global warming. Long-distance migrants, however, may only occasionally profit from a longer stay on their breeding grounds. Their arrival south of the Sahara coinciding with the short monsoon rainfalls probably has greater selective significance, Deviations in arrival times caused by climatic changes may thus lead to catastrophies as is currently discussed in the case of Tawny Pipit. The influence of global warming on bird migration and the changes in migration patterns ensuing from it were to be expected after laboratory results. Here, this influence is shown in a long-term study of free-living populations of a large number of species.
