LEAF CHARACTERISTIC PATTERNS CLUSTERING BASED ON SELF-ORGANIZING MAP

Effects on climate change and global warming have extremely wide space in many ways. One of the affection is directly in the forest. As a result, the forest ecosystems have adapted themselves to response the effect of climate change. Nowadays, the characteristic of forest ecosystems has been found and studied in many research to discover how they can respond and qualify with climate change. Leaf phenology is the main characteristic of the environment reaction of the trees. However, the diversity of each tree species in the dry tropical forest is a time-series data and slightly different between each other. These reasons make each species difficult to separate the pattern of leaf phenology. There are several unsupervised clustering methods that were used to analyze time-series data. Self-Organizing Map (SOM) is one of the unsupervised techniques that was applied in many forest and ecosystem works. The aim of this research is to study the pattern of leaf phenology that covers usual dry season and severe drought in dry dipterocarp forest based on SOM. The performance of SOM algorithm was compared with K-Mean, Hierarchical Clustering and Gaussian Mixture Model (GMM). The result showed that SOM provided a good performance to cluster the leaf characteristic patterns