Abstract:

Forests are the largest contributors to the global land CO2 sink, sequestering 29% of global emissions annually. However, climate change and anthropogenic pressures have reduced CO2 uptake in land ecosystems by 17% over the past decade due to higher temperature extremes and land cover change, with future trends uncertain. In response, tree planting and forest carbon offset programs, have gained traction, but these programs face methodological challenges including poor monitoring, carbon budgeting issues, and doubts about effectiveness. A key obstacle in accurately quantifying carbon sequestration in afforestation and reforestation programs is the lack of transparency and accuracy in the methods currently employed, which are error-prone, lack uncertainty reporting, and have theoretical shortcomings. Moreover, there is limited scientific validation of their performance. 
This project aims to address such limitation via modelling the CO2 metabolic pathways of single trees. By integrating ecosystem flux estimates from micrometeorological methods, laser-derived green biomass structure, and plant physiology, I will model each tree’s contribution to the net forest ecosystem CO2 uptake. This approach will allow to pinpoint individual trees’ contributions to carbon sequestration while keeping aligned with existing micrometeorological theory and infrastructure. 
The goal of this project is to enhance our quantifiable understanding of tree physiology by transitioning from composite observations of CO2 exchange at the forest-scale level to testable quantitative analysis at the single tree-level. The core methodological contribution of this project is to bring together existing measurements as well as theory from various fields that are rarely integrated, in one same statistical modelling framework. Particularly, the project proposes the use of Generalized Additive Models embedded within a Bayesian hierarchical framework to evaluate different carbon metabolic pathways and assess single sources of uncertainty. Via Dirichlet regressions, the sum of the contributions of single trees to the net forest uptake of CO2 will be made compatible with existing ecosystem level estimates and measurements. 
The project presents a statistical modeling framework as a necessary milestone to meet two domains that are to date considerably apart: (a) the highly specific field-based ecosystem datasets and (b) the general application of AI models scalable to biogeographic applications.