BIRD SPECIES TURNOVER AND STOCHASTIC EXTINCTION IN WOODLAND FRAGMENTS

Year-to-year turnover in bird species composition was recorded across the whole size range (0.02-30 ha) of 146 woods studied. The mean numbe r of resident breeding species both lost and gained per wood between c onsecutive breeding seasons was 2 (range 0-8). No relationship was fou nd between this absolute turnover rate and woodland area, or any other of 24 predictor variables (describing woodland structure, isolation, connectedness and surrounding land use). Extinction and colonisation r ates (in terms of numbers of species lost and gained) were also unrela ted to woodland area. In all sizes of woods, the species most likely t o show local extinctions and colonisations were those with small popul ations within those woods, but the identity of the species concerned c hanged as woodland area increased. In the smallest woods, the majority of turnover involved common species, such as wren and dunnock, which occurred in only small numbers in these small woods. As woodland area increased, these species attained sufficient numbers to usually avoid stochastic extinction. The majority of turnover was then due to more s pecialist (and less numerous) woodland species, such as great-spotted woodpecker and marsh tit, which were usually lacking in small woods. I n Britain, much existing broadleaved woodland falls within the size ra nge studied. Thus the numbers of many bird species are liable to be sm all enough for yearly turnover in woodland bird communities to be appr eciable, and for the long-term persistence of individual species in pa rticular woods to depend on dispersal.