Multivariate analysis and self-organizing mapping applied to analysis of nest-site selection in Black-tailed Gulls

Abstract

The factors affecting nest-site selection and breeding success of Black-tailed Gulls ( Larus crassirostris) were studied in Hongdo Island in Korea during the breeding seasons in 2002 and 2003. Two analyzing methods, Principal Component Analysis (PCA) and Self-Organizing Map (SOM) – an unsupervised learning method in artificial neural networks, were applied to multivariable datasets characterizing nest-sites of the gulls. Both methods provided insights on the major trends in nest-site selection by Black-tailed Gulls. PCA showed that the variables regarding the “wall” effect such as rock cover and nest-wall (positively), and the nearest distance between neighbors (negatively) were related to breeding success of Black-tailed Gulls. SOM confirmed ordination of the sample sites by PCA and efficiently classified nest-sites according to environmental condition for breeding. Grouping based on the “wall” effect on PCA was more finely revealed in subdivision on SOM regarding the variables of slope and the nearest distance between neighbors. The use of techniques in ecological informatics such as SOM would be an efficient tool in analyzing data for breeding behavior of birds.