Yun-Bin Lin, Chianan University of Pharmacy and Science and Yu-Pin Lin, National Taiwan University.
The multivariate geostatistical method has an advantage over the traditional data reduction methods, e.g. principal component analysis (PCA) and correspondence analysis (CA), because it accounts for spatial correlations between variables at different scales. This study adopted the multivariate factorial kriging (MFK) method to carry out a multi-scale ordination in Taipei metropolis. Five eurytopic species selected from the community of urban birds were used in the ordination, and the ordination results were further interpreted by dense remote sensing data, including the land covers, the digital elevation model, and the normalized difference vegetation index. Spatial patterns of variation were recognized and mapped by the MFK at the regional scale, 16 km, and the local scale, 2 km. Comparisons between results derived from the MFK, PCA and CA, showed that the MFK can not only delineate the local-scale variation among species as done by the traditional PCA, but also apparently construct the spatial correlations among species at the regional scale as done by CA. The local-scale variation accounts for the larger proportion of the total variation and is affected by more factors than variation at the regional scale. Interpretation of the environmental data showed that the elevation heterogeneity is the geographical influence on the distribution of birds at the regional scale, while the meadow land-cover is the local influence in the highly developed part of the city. The ecotone between different forest stands and the edge between different land covers in the belt region in the city have both local and regional influences on birds. The physiological habits and characteristics of bird species were further discussed to support the environmental interpretations of the ordination results, and the evidence shows that the distribution patterns of species can be reasonably explained by their diet categories and body sizes.