Visualizati on and Clustering of Data Derived from Forest Inventory Using Self-Organizing Neural Network (Case Study: District Two Forests of Kacha, Gilan)

Abstract

Forest inventory is essential for all types of management programs, decision-making, and obtaining information about forest lands. Tree density, stand Volume, and diameter at breast height are quantitative forest characteristics that are derived from a significant amount of data through the inventory process. To process and interpret such an extensive set of data, data clustering becomes essential, enabling the identification of diverse data entities. The SOM neural network stands as a valuable tool for data dimensionality reduction and clustering. This tool offers a visualization of a dataset on a two-dimensional plane, acting as a data map. It's particularly effective in discerning relationships among structural variables and pinpointing the role of singular variables in the formation of clusters via the SOM neurons. Within this study, the SOM neural network was harnessed to project and segment data derived from the forest inventory of the District Two Kacha forests. The derived findings highlight that, considering density and stand volume, the study area can be segmented into three distinct clusters: 1(A), 2(B), and 3(C). Notably, samples from Cluster 1(A) exhibit the peak density and stand volume, whereas Cluster 3(C) records the minimum values. Notably, the study found that the SOM neural network could be a valuable tool for analyzing large datasets in forests, particularly in the District Two Kacha.