|Title||Assessing biomass and architecture of tropical trees with terrestrial laser scanning|
|Author(s)||Lau Sarmiento, Alvaro Ivan|
|Source||Wageningen University. Promotor(en): M. Herold, co-promotor(en): H. Bartholomeus; K. Calders. - Wageningen : Wageningen University - ISBN 9789463434720 - 157|
Laboratory of Geo-information Science and Remote Sensing
|Publication type||Dissertation, internally prepared|
Over the last two decades, terrestrial light detection and ranging (LiDAR), also known as terrestrial laser scanning (TLS) has become a valuable tool in assessing the woody structure of trees, in a method that is accurate, non-destructive, and replicable. This technique provides the ability to scan an area, and utilizes specialized software to create highly detailed 3D point cloud representations of its surroundings. Although the original usage of LiDAR was for precision survey applications, researchers have begun to apply LiDAR to forest research. Tree metrics can be extracted from TLS tree point clouds, and in combination with structure modelling, can be used to extract tree volume, aboveground biomass (AGB), growth, species, and to understand ecological questions such as tree mechanics, branching architecture, and surface area. TLS can provide a robust and rapid assessment of tree characteristics. These characteristics will improve current global efforts to measure forest carbon emissions, understand their uncertainties, and provide new insight into tropical forest ecology. Thus, the main objective of this PhD is to explore the use of 3D models from terrestrial laser scanning point clouds to estimate biomass and architecture of tropical trees. TLS-derived biomass and TLS-derived architecture can potentially be used to generate significant quality data for a better understanding of ecological challenges in tropical forests.
In this thesis, a dataset of forest inventory with TLS point clouds and destructive tree harvesting were created from three tropical regions: Indonesia, Guyana, and Peru. A total of 1858 trees were traditionally inventoried, 135 trees were TLS scanned, and 55 trees were destructively harvested. In this thesis, procedures to estimate tree metrics such as tree height (H), diameter at breast height (D), crown diameter (CD), and the length and diameter of individual branches were developed using 3D point clouds and 3D modelling. From these tree metrics, I infer AGB, develop allometric models, and estimate metabolic plant scaling of individual tropical trees. All these metrics are validated against a traditional forest inventory data and destructively harvested trees.
Chapter 2 presents a procedure to estimate tree volume and quantify AGB for large tropical trees based on estimates of tree volume and basic wood density. The accurate estimation of AGB of large tropical trees (diameter > 70 cm) is particularly relevant due to their major influence on tropical forest AGB variation. Nevertheless, current allometric models have large uncertainties for large tree AGB, partly due to the relative lack of large trees in the empirical datasets used to create them. The key result of this chapter is that TLS and 3D modelling are able to provide individual large tree volume and AGB estimates that are less likely to be biased by tree size or structural irregularities, and are more accurate than allometric models.
Chapter 3 focuses on the development of accurate local allometric models to estimate tree AGB in Guyana based solely on TLS-based tree metrics (H, CD, and D) and validated against destructive measurements. Current tropical forest AGB estimates typically rely on pantropical allometric models that are developed with relatively few large trees. This leads to large uncertainties with increasing tree size and often results in an underestimation of AGB for large trees. I showed in Chapter 2 that AGB of individual large trees can be estimated regardless of their size and architecture. This chapter evaluates the performance of my local allometric models against existing pantropical models and evidenced that inclusion of TLS-based metrics to build allometric models provides as good as, or even better, AGB estimates than current pantropical models.
Chapter 4 provides an insight into the architecture and branching structure of tropical trees. In Chapter 2, I demonstrated the potential of TLS to characterize woody tree structure as a function of tree volume, but little is known regarding their detailed architecture. Previous studies have quantitatively described tree architectural traits, but they are limited to the intensity of quantifying tree structure in-situ with enough detail. Here, I analysed the length and diameter of individual branches, and compared them to reference measurements. I demonstrated that basic tree architecture parameters could be reconstructed from large branches (> 40 cm diameter) with sufficient accuracy. I also discuss the limitations found when modelling small branches and how future studies could use my results as a basis for understanding tree architecture.
Chapter 5 describes an alternative approach to estimating metabolic scaling exponents using the branching architecture derived from TLS point clouds. This approach does not rely on destructive sampling and can help to increase data collection. A theory on metabolic scaling, the West, Brown & Enquist (WBE) theory, suggests that metabolic rate and other biological functions have their origins in an optimal branching system network (among other assumptions). This chapter demonstrates that architecture-based metabolic scaling can be estimated for big branches of tropical trees with some limitations and provides an alternative method that can be implemented for large-scale assessments and provides better understanding of metabolic scaling.
The results from this thesis provide a scientific contribution to the current development of new methods using terrestrial LiDAR and 3D modelling in tropical forests. The results can potentially be used to generate significant quality data for a better understanding of ecological challenges in tropical forests. I encourage further testing of my work using more samples including other types of forests to reduce inherent uncertainties.