Habitat models of bird species' distribution: an aid to the management of coastal grazing marshes.
Coastal grazing marshes comprise an important habitat for wetland biota but are threatened by agricultural intensification and conversion to arable farmland. In Britain, the Environmentally Sensitive Area (ESA) scheme addresses these problems by providing financial incentives to farmers to retain their grazing marshes, and to follow conservation management prescriptions. A modelling approach was used to aid the development of management prescriptions for ground-nesting birds in the North Kent Marshes ESA. This ESA contains the largest area of coastal grazing marsh remaining in England and Wales (c. 6500 ha) and supports nationally important breeding populations of lapwing Vanellus vanellus and redshank Tringa totanus. Counts of ground-nesting birds, and assessments of sward structure, surface topography and wetness, landscape structure and sources of human disturbance were made in 1995 and again in 1996, on 19 land-holdings with a combined area of c. 3000 ha. The land-holdings varied from nature reserves at one extreme to an intensive dairy farm at the other. Models of relationship between the presence or absence of ground-nesting birds and the grazing marsh habitat in each of c. 430 marshes were constructed using a generalized linear mixed modelling (GLMM) method. This is an extension to the conventional logistic regression approach, in which a random term is used to model differences in the proportion of marshes occupied on different land-holdings. The combined species models predicted that the probability of marshes being occupied by at least one ground-nesting species increased concomitantly with the complexity of the grass sward and surface topography but decreased in the presence of hedgerows, roads and power lines. Models were also prepared for each of the 10 most widespread species, including lapwing and redshank. Their composition differed between species. Variables describing the sward were included in models for five species: heterogeneity of sward height tended to be more mportant than mean sward height. Surface topography and wetness were important for waders and wildfowl but not for other species. Effects of boundaries, proximity to roads and power lines were included in some models and were negative in all cases. Binomial GLMMs are useful for investigating habitat factors that affect the distribution of birds at two nested spatial scales, in this case fields (marshes) grouped within farms. Models of the type presented in this paper provide a framework for targeting of conservation management prescriptions for ground-nesting birds at the field scale on the North Kent Marshes ESA and on lowland wet grassland elsewhere in Europe.