The Potential Distribution and Impact of Bracken in Upland Scotland: An Assessment using a GIS-based Niche Model

Bracken spread could have very selective impacts on the vegetation of upland Scotland (UK). Information on this would be of value in targeting appropriate management measures. However, it is difficult to establish accurately the current distribution of bracken, so modelling bracken spread using process-based models is not possible at present. An alternative modelling approach is described which uses knowledge of the climatic, soil and topographic limits to the existing bracken distribution in upland Scotland. This knowledge is formalized as a rule-based model. This is applied through a geographical information system (GIS) to identify the maximum potential bracken niche over the whole of Scotland with a nominal spatial resolution of 1 ha. The model is validated against field data from 14 sites around Scotland. Between 64–97% of the existing bracken lies within the predicted bracken niche. Locally low accuracies are ascribed to scale-effects derived from the input datasets, the nature of the ground data, as well as the universality of the model rules. The bracken niche is differentiated into two types: continuous and discontinuous. The total area of the niche is 6036 km 2 or 7.75% of the land area of Scotland with a broad 60:40 split between the continuous and discontinuous niche types, respectively. However, there is a strong regional bias in their distribution with the discontinuous niche type principally occurring in the west of Scotland. Overlaying the potential niche on the Land Cover of Scotland 1988 dataset shows that it impacts on 70% of the national area of ‘good rough grassland’ and 31% of the ‘poor rough grassland’ category. In addition, 16% of the national area of ‘heather moorland’ is impacted. There are strong regional variations in these impacts which may have implications for future bracken management. The limitations of the approach are discussed and the potential for linking it to dynamic models is identified.