Spatial Forest Stand Dataset for Graph Convolutional Network Applications

This dataset supports sustainable forest management research by integrating <b>Geographic Information Systems (GIS)</b> and forest register data with <b>Graph Convolutional Networks (GCN)</b>. It includes spatial and attribute data from the <b>Nishikawa forest area</b>, located in Saitama Prefecture, Japan, covering terrain features, road networks, forest stand boundaries, and timber characteristics.Two graph-based models were constructed to represent spatial relationships among forest stands: the <b>Selective Connection Model</b>, emphasizing road proximity and terrain constraints, and the <b>Full Connection Model</b>, capturing adjacency relationships between stands to promote balanced forest management. Node features include forest age, timber volume, and terrain attributes, while edge weights reflect accessibility, slope, and distance.This dataset enables grouped forest stand-level analysis, supporting the development of efficient and sustainable logging plans that incorporate spatial and terrain considerations. It is ideal for applications in forestry optimization, machine learning, and spatial analysis.