Skip to main content

Yuchen Bai

Thursday, December 12th, 2024

Simulation and machine learning models for bias assessment and reduction in leaf area density estimators in tropical forests

Abstract :

Tropical forests, covering only 7\% of the Earth's land surface, play a disproportionately vital role in biosphere. Better understanding the processes underlying flux seasonality in tropical forests is thus critical to improve our predictive ability on global biogeochemical cycles. Leaf area index (LAI), a key parameter governing water and carbon fluxes, is inadequately characterised, necessitating advancements in monitoring technologies such as aerial and terrestrial laser scanning (LiDAR). Many factors introduce biases in existing LiDAR-based methods for LAI estimation, the primary issue to address is woody component. The first part of thesis focuses on the development of a novel Deep Learning approach named SOUL (Semantic Segmentation On ULs) for leaf/wood semantic segmentation of ULS data in tropical forest environments, relying exclusively on the coordinates of the points, thus ignoring device-specific information such as apparent reflectance, in order to extend its scope to other forests and other sensors. The second part of thesis focuses on analysing the other biases that affect LAI estimation from LiDAR surveys. The DART model is used to simulate ULS data characteristics based on two forest mock-ups: Wytham Woods and Järvselja Birch Stand of RAMI-V. The simulated data provides comprehensive details about the forest structure, LAI, leaf angle, leaf/wood label, etc. By applying SOUL to remove the bias introduced by woody component, and then employing AMAPVox, a ray tracing model, to conduct a quantitative analysis of the remaining biases, a more accurate estimation of forest LAI is obtained. This dissertation contributes to advancing our understanding of tropical forest ecosystem dynamics and offers practical methodologies for accurate and reliable leaf area estimation. 

Date et lieu

Thursday, December 12th at 2pm
Grand Amphitheater of Inria
and  Zoom

Jury members

Sylvie DURRIEU
Chercheur Senior, INRAE (Rapporteure)
Florent LAFARGE
Directeur de recherche, Centre INRIA de l'Université Côte d'Azur (Rapporteur)
Amini MASSIH-REZA
Professeur des universités, Université Grenoble Alpes (Examinateur)
Nicolas BARBIER
C
harge de recherche, Institut de Recherche pour le Developpement (Examinateur)
Jean-Baptiste DURAND
Chercheur Senior, CIRAD  (Directeur de thèse)
Florence FORBES
Directeur de recherche
, INRIA (Co-directeur de thèse)
Grégoire VINCENT
Chercheur Senior
, Institut de Recherche pour le Développement  (Co-directeur de thèse)

Submitted on December 6, 2024

Updated on December 6, 2024