Determining vertical structure of forests in Poland using a semi-automated approach based on ALS data

Abstract

The vertical structure of the forest is one of the key characteristics of forest management, influencing biodiversity, resource competition, and various ecological processes. Despite its importance, determining the vertical structure of stands over large areas is still a challenge. This study presents the first country-wide assessment of the vertical forest structure in Poland using airborne laser scanning (ALS) data. The research introduces a semi-automatic method for classifying forest stands into vertical structure classes without the need for extensive fieldwork. The main strength and innovation of our approach is demonstrating that a model can be built and validated almost without field-acquired data. By processing point-cloud metrics derived from ALS data and employing machine learning techniques, particularly the random forest algorithm, the method generated a high-resolution vertical structure map across the country. The five-class model developed had a overall accuracy of 0.78. The results show that Polish forests are predominantly characterized by a single-story vertical structure, influenced by the dominance of Scots pine, with more complex structures in mountainous and biodiversity-rich areas. The methodology significantly reduces costs and time associated with traditional forest surveys, offering a scalable tool for forest monitoring, biodiversity assessment, and sustainable management, particularly under changing environmental conditions and habitat loss.