1. Identifying suitable restoration sites for a scarce subarctic willow (Salix arbuscula) using different information sources and methods.
- Author
-
Pakeman, Robin J. and Torvell, Lynne
- Subjects
- *
WILLOWS , *NATIONAL parks & reserves , *HABITATS , *LINEAR statistical models - Abstract
Background: Subarctic willow scrub is an extremely restricted habitat within the UK. Restoration efforts are now targeted at expanding current populations and establishing new stands. Aim: Selecting suitable target sites for restoration has to be cost-effective. Efficacy can be ensured through the development of habitat suitability models with high predictive ability but these have to be efficient in using only a small part of the resources of a restoration project. Location: The study was concentrated on the Lochan na Lairige catchment at Ben Lawers National Nature Reserve, Perthshire, Scotland. Additional data were collected from three other sites in the Southern Highlands; Fireach na Moine, Meall Ghaordie and Meall na Samhna. Methods: This research compared the merits of using vegetation or environmental parameters, and a combination of both, to build habitat suitability models for a scarce species of upland Britain, Salix arbuscula. In addition, models derived from three different statistical methods (generalised linear models, linear discriminant analysis and classification and regression trees) were compared. Results: Models based on either vegetation and environmental parameters were very similar in predictive ability, though there was only moderate agreement on which sites were suitable for restoration. The combination of both sets did not improve predictive ability, and hence only one set of parameters is necessary to develop a cost-effective method of identifying suitable replanting sites for S. arbuscula. The models developed by generalised linear modelling were consistently and significantly higher in predictive ability than those developed using the other methods, but there were considerable differences in which sites were highlighted as suitable for restoration between the different methods. Conclusions: Predictive models for suitable restoration sites for S. arbuscula could be assembled cost-effectively from just one parameter set, with a GLM approach providing the most effective statistical approach. However, as there was only moderate consistency between modelling approaches a model averaging approach could be profitably employed in site selection. [ABSTRACT FROM AUTHOR]
- Published
- 2008
- Full Text
- View/download PDF