Finding Profiles of Forest Nutrition by Clustering of the Self-Organizing Map

Understanding the nutritional states and profiles of tree species is important for monitoring the well-being of forests. Data from foliar surveys are available, but there is still need to better understand the underlying nutritional mechanisms in trees. In this paper, the nutrient concentrations of pine and spruce needles in Finland between 1987–2000 are analyzed to build nutrition profiles. The profiles are built from the data by clustering of the Self-Organizing Map. The VS algorithm divides the data into base clusters using region growing and forms a hierarchy from the base clusters. The hierarchy tree is pruned and the final clusters are selected from the pruned tree. We were able to divide the measurements into six groups. In each group the growth of the needles and the amounts of the nutrients were different and thus, different groups represented different kinds of growing conditions. With the help of the domain expert, using the results of the clustering method, it was possible to construct a temporal model that characterizes the development of the forests of Finland.