Colloquium

Understanding Crown Shyness at a 3D level in Tropical Forests

Organised by Laboratory of Geo-information Science and Remote Sensing
Date

Tue 21 May 2024 10:30 to 11:00

Venue Gaia, building number 101
Droevendaalsesteeg 3
101
6708 PB Wageningen
+31 (0) 317 - 48 17 00
Room 1

By Guillermo González Fradejas

Abstract
Crown shyness describes the phenomenon by which tree crowns avoid growing into one another, forming remarkable puzzle-like complementary patterns with gaps between each other. Previous studies point to abrasive contact of the outer parts of the crowns as one of the main drivers behind this phenomenon. Slender trees are more prone to abrasive contact, as they oscillate more under windy conditions. Therefore, the slenderness (the ratio between height and DBH) of trees is expected be positively correlated with crown shyness.

Previous studies have look at this phenomenon at a 2D level and through indirect measures. However, there is a gap in our understanding of it at a 3D level. LiDAR technology can help us overcome that gap, by looking at 3D representations of trees in the form of point clouds. This thesis is a follow-up study based upon the work of van der Zee et al., 2021 in which the relationship between surface complementarity and tree slenderness of fourteen pairs of trees was computed for a tropical forest plot in Guyana. TLS LiDAR was used to obtain individual point clouds for each tree.

In this thesis, crown shyness was assessed in a at a group level for another plot of the same study area through two variables: Mean Group Surface Complementarity (SCg) and Mean Group Intercrown Distance (IDg). SCg values range from -1 to 1, with 1 indicating perfect shape complementarity and negative values indicating crown overlap. On the other hand, IDg measures the average size of the gaps between crowns within a group. The aim of this thesis is to scale-up the previous approach from tree pairs to groups of trees. The goal is to assess the dependency of crown shyness on four tree parameters: slenderness, height, stem diameter, and crown area by computing linear models between SCg and IDg, and those tree parameters.

The dataset utilized in this study comprised 884 individually segmented point cloud objects. s. Pre-processing steps included filtering point clouds representing objects shorter than 10 meters and quality assessment to filter out defective point clouds. Following this process, 31 groups were made by clustering 44 trees. Groups were made based on proximity. Each group was formed by at least three trees: one central tree and two or more surrounding ones. Slenderness, height, stem diameter, and crown area were computed for each of the 44 trees. Average values of the same variables were computed for each group.

SCg and IDg were computed for each group in a one-to-way manner. This means that only the interactions between the surrounding trees and the central one were considered. The process involved separating the crown and the stem of the trees and modelling the interacting parts of each crown as α-shapes. After that, Sixteen linear models were generated to evaluate the connection between two crown shyness measures, SCg and IDg, and four tree parameters (slenderness, height, stem diameter, and crown area). The models considered two sets of independent variables: one based on the mean values of tree parameters within a group, and the other focused solely on the central tree of each group.

Mean Group Surface Complementarity (SCg) did not show significant correlations with neither mean group slenderness (Adj-R2 = 0.0824; p-value = 0.08) or slenderness of the central tree (Adj-R2 = 0.0692; p-value = 0.09). However, it was significantly and positively related to the other group variables: mean group height, stem diameter, and crown area, being the relationship with mean group heigh the strongest one (Adj-R2 = 0.454; p-value < 0.0001). There were also significant and positive relationships between SCg and the central tree parameters. Similarly, Mean Group Intercrown Distance (IDg) did not exhibit significant correlations with neither mean slenderness (Adj-R2 = 0.0875 ; p-value = 0.06) or slenderness of the central tree (Adj-R2 = -0.0198; p-value = 0.51) but showed significant and positive correlations with mean group height, stem diameter and crown area. However, there were no significant relationships found between IDg and central tree parameters.

These results contrast with the findings of van der Zee et al., 2021 for pairs of trees and suggest that drivers of crown shyness vary across scales, i.e.: relations observed at pairwise level do not determine patterns at a higher scale. At the group level studied here, size was found to be a much more important factor than slenderness, as crown shyness was related to groups formed by trees on average taller, thicker, and with wider crowns. Complementary crown growth allows trees to maximize light use compared to overlapping crowns. Previous research indicates a positive link between crown shyness and biomass, suggesting that complementarity enhances photosynthesis and biomass production. This implies complementarity may drive tree size, altering the causal direction explored in this study. Finally, a major limitation of this thesis was the quality of the segmentation of the point cloud into individual trees. Future research should focus on improving this area and in developing more automated methods for quality assessment to facilitate scaling up to bigger groups.