|Although studies of changes in the timing of passerine spring migration
are numerous, less is known about timing of their autumn departure.
We present phenological data on 22 species based on mist-netted birds
caught on the Baltic island of Christiansø during autumn migration
between 1976 and 1997. We used least square and quantile regression
to test for changes in departure time of the last individuals and the date
at which 5%, 50% and 95% of the season total remaining to be trapped
(corrected for catch effort). For the entire group of 22 migrants there
was no overall change in median departure (–0.0426 days year–1, P =
0.40). Testing the 12 species for which the entire migration period was
included (thus excluding many long-distance migrants), we found an
overall earlier departure (–0.18 days year–1, P = 0.007). Short-distance
migrants tended to show earlier departure, while long-distance migrants
only showed advanced departure amongst the last individuals to leave.
The magnitude of change in departure time was correlated with migration
distance, longer distance migrants showing the least change.
Overall, timing of autumn migration was more variable between species
than the unidirectional change reported in studies of spring arrival.|