Object-based approach for mapping complex forest structure phases using LiDAR data

Managed forests are important components of the landscape comprising different stand structural phases (stand initiation, stem exclusion and understorey re-initiation) and delivering important ecosystem functions such as timber production and biodiversity. This paper focuses on the development of a classification method for determining structural phases for conifers and broadleaves using LiDAR data and object-based image analysis (OBIA) approach over a large study area (110 km(2)). Firstly, using OBIA, homogenous stands were segmented with minimum area of 100 m(2). Tree tops were detected from a canopy height model and gap area between trees determined. Secondly, stand parameters such as tree density and tree height statistics (mean, standard deviation and percentiles) were calculated. The final classification was based on the analysis of stands with known structural phases where the best classifiers were 60th and 80th tree height percentile, tree density and area of gaps between trees. In the study area more than 13,000 stands were allocated and 9,616 of them classified into the three phases, the area proportions being: stem exclusion 68%, understorey re-initiation 28% and stand initiation 4%. The range of stand sizes varied from 100 to 80,476 m(2) across all phases. Our approach shows that it is feasible to classify forest stands into structural phases on a large scale that would have value in forest management planning.