Exploring Drivers of Structural Variability in Norway Spruce Stands Using Airborne LiDAR
Year: 2026
Type: article
Abstract: Airborne LiDAR provides a powerful framework for objectively describing forest structure across multiple spatial scales. From LiDAR points to a unit of area, to descriptors of individual trees, to cohorts of trees, to the stand, or at a landscape level, its analytical value can create a hierarchical understanding of drivers affecting the structural variability of trees and their spatial organisation on multiple levels. In this study, we analysed the structure of relatively pure Norway spruce (Picea abies (L.) Karst.) stands in the Ciucaș Mountains (Brașov County, Romania) using airborne LiDAR data. The point cloud was processed in R using an open-source workflow comprising ground classification, terrain modelling, height normalisation, canopy height modelling, treetop detection, and crown delineation. Structural metrics were derived hierarchically, from individual trees to summary statistics for tree cohorts within 500 m2 grid cells, and subsequently to stand-level statistics across grid cells within spruce stands. The geospatial dataset consisted of 58,708 grid cells grouped in 147 stands; for inferential analyses, a filtered dataset of 42,015 cells was retained, covering a total area of 2075.6 hectares. We applied a four-part analytical framework: (i) hierarchical mixed-effects models at the cell level, (ii) quantification of within-stand structural heterogeneity, (iii) stand-level structural typology based on principal component analysis and k-means clustering, and (iv) integration of disturbance history using combined Potapov (pre-2001) and Hansen (2001–2022) forest loss layers. The analysis revealed several patterns consistent with ecological expectations: mean canopy height increased with stand age, slope, and topographic position, whereas it decreased significantly with elevation. Canopy cover increased with elevation, slope, and topographic position, whereas the number of LiDAR-detected trees per cell increased with elevation but decreased with age and slope. The null model intraclass correlation coefficients ranged from 0.252 to 0.408, indicating appreciable between-stand variation alongside a larger within-stand residual component. Structural heterogeneity increased strongly with slope and, to a lesser extent, with northness. The stand typology identified five structural types: low-stature dense uniform, mid-height heterogeneous, tall closed uniform, tall open simplified, and very tall closed complex. Mapped tree cover loss occupied a small fraction of the retained forest sample and was associated with lower mean canopy height (approximately 1 m), lower canopy cover (approximately 4 percentage points), smaller mean crown area (approximately 2 m2), and a slightly lower number of LiDAR-detected trees per cell. These associations do not identify the causes of disturbance or its overall historical importance. In the Ciucaș Mountains, LiDAR-derived metrics provide a spatially explicit baseline for characterising spruce stand structure and its associations with stand age, topography, and mapped disturbance history.
Source: Remote Sensing
DOI: 10.3390/rs18223862
Language: English
Subfield: Forestry
Domain: Life Sciences
Open Access status: gold
Estimated APC paid: $3,258
OpenAlex ID: W7225747210