Site index curves construction for uneven-aged forest stands. A machine learning simulation approach

Abstract

In research related to forest disturbances, forest structure, ecological diversity, and forest sustainability, the concept of site quality holds significant importance. Site quality can be described as the combination of physical and biological factors that determine a site’s capacity to sustain tree growth. Thus, it becomes crucial to have a comprehensiv e understanding of site quality curves, known as site index curves. This paper endeavors to present a methodology for creating precise and reliable site index curves tailored to uneven- aged stands. To reach this goal, initially standard non-linear regression modeling was applied. Furthermore, in the field of forestry and environmental studies, especially in Greek forests, there are high demands for accurate predictions about forest health, potential and productivity. The rapidly developed field of machine learning can provide solutions to these requirements. To this direction, to directly predict the site index for each tree, the effectiveness of the eXtreme Gradient Boosting (XGBr) ensemble machine learning technique for regression modeling was investigated, aiming to effectively capture the non- linear characteristics of site index curves. In the realm of environmental and forest modeling, the studied simulation approach showed its potential to serve as a crucial foundation for advancing sustainable forest management.