Assessment of Forest Biodiversity in Poland Using Unsupervised Learning and Satellite Data

Forest biodiversity is an essential factor that shapes the forest’s structure and quality and affects the entire environment. Biodiversity assessment is performed based on species composition. The Shannon-Wienner index (SWI), which is calculated using species composition, is the most commonly used. Species composition can be assessed based on field observations or the classification of satellite imagery. Classification is a supervised process. It requires a learning set and is subject to error. This paper presents an approach to assessing forest biodiversity using cluster analysis (unsupervised learning) of data from satellite images. The results obtained were compared with an index calculated from field observations for 5278 forest districts. The results show that the relationship is linear with Pearson’s correlation coefficient 0.861. The overall RMSE error is 0.136 with BIAS equals 0.098. The presented novel approach indicates the possibility of directly applying satellite imaging to assess the biodiversity of forest areas.