Method evaluation of GIS-based prediction tools for biodiversity : habitat suitability for birds in Stockholm, Sweden

Sammanfattning: Swedish cities are from a European point of view considered small, sparsely populated and green. Stockholm city has a great deal of its nature and older cultural landscape remaining, which is unusual in large metropolitan areas. During the last decades the spatial planning of urban environments has faced the challenge of including biodiversity concerns. This has proved to be difficult since there are no welldeveloped methods for quantifying and predicting the impacts of exploitation on biodiversity. As a result many green areas have been exploited and the flora and fauna are undergoing loss of habitats, fragmentation and alteration due to change in land use. There is an evident need to develop the planning and management methods for biodiversity in urban areas. Moreover, adequate methodologies for systematic and quantifiable predictions are needed. In this study landscape analyses have been carried out to predict the occurrence and suitable habitat in Stockholm municipality for sever birds: Lesser Spotted Woodpecker, Great Spotted Woodpecker, Green Woodpecker, Hawfinch, Nuthatch, Stock Dove and Tawny Owl. Two different prediction models have been used: an expert model and an empirical model. The basis for the study is a biotope map (Stockholm Municipality, 1999) and a species observation database (administrated by the Swedish Species Information Centre). The spatial analyses were conducted using GIS (ArcView 3.3 and ArcGIS 9.1). The most important conclusion is that it is possible to predict species distribution with both models. However, the quality and quantity of data is essential for predicting species occurrence. In this study the expert model is preferable since it is based on expert knowledge and a biotope map. The empirical model is based on occurrence data, a biotope map and software called Genetic Algorithm for Rule-set Production (GARP), which predict species distribution. The occurrence data has been gathered in an ad hoc manner. In this model the uncertainties in the occurrence data causes an overprediction mainly due to the bias in the data and the mismatch in the resolution of the biotope map and the occurrence data. The empirical model should consequently be used carefully and one should always consult with experts on different taxa and conservation biology. Which model to use depends on the available data and what knowledge there are in the certain organism group. In general it is easier to predict rare species than common ones that have a wider habitat criterion.