Président du jury Gilbert, Marius
Promoteur Bogaert, Jan
;De Cannière, Charles Publication Non publié, 2014-10-24
| Résumé : | Human pressure on forest resources increased significantly during the past decades through land use and land use change, especially in the tropics where forest clearing is a major source of CO2 release in the atmosphere. Consequently, forests are the focus of international environmental policies and discussions aiming to reduce emissions from deforestation and forest degradation (i.e., REDD+). The capacity of participating countries to regularly provide accurate forests C stocks measurements at a national scale thus represents an important challenge to address. In dense forests, generally only the above ground biomass (AGB) is measured as it accounts for more than 50% of total C stocks. However, important gaps remain at each scale of measurement, i.e. from felled tree to regional mapping, with the resulting errors propagation through these different scales being probably the most concerning issue. In the present work, we propose to address these issues by using a multi-scale approach in order to improve our global understanding of AGB variations in dense tropical forests of Central Africa. In particular, we studied (i) forest AGB prediction from remote-sensing textural analysis, (ii) the potential role of largest trees as predictor of the entire forest-stand AGB and (iii) intra- and inter-individual radial variation of wood specific gravity (WSG, i.e. oven-dry mass divided by its green volume) and its potential consequences on the estimation of the AGB of the tree. First, we analyzed the potential use of textural analysis to predict AGB distribution based on very high spatial resolution satellite scenes. In particular, we used the Fast Fourier Transform Ordination (FOTO) method to predict AGB from heterogeneous forest stands of the Democratic Republic of the Congo (DRC). Here, based on 26 ground plots of 1-ha gathered from the field, plus a successful combination of Geoeye and Quickbird contrasted scenes, we were able to predict and to map AGB with a robust model (R² = 0.85; RMSE = 15%) based on textural gradients. Secondly, the research of AGB indicators was focused on the dissection of the role played by largest trees. Here we found largest trees not only hold large share of forest carbon stock but they contain the print of most of forest-stand structure and diversity. Using a large dataset from western Cameroon to eastern DRC, we developed a non-linear model to predict forest carbon stock from the measurement of only a few large trees. We found the AGB of the 5 % largest stems allow to predict the AGB of the entire forest-stand yielding an R² of 0.87 at a regional scale. Focusing on largest trees species composition, we also showed only 5 % of species account for 50 % of total AGB. In the end, we investigated inter- and intra-individual WSG variations. Despite recognized inter- and intra-specific variations along the radial axis, their ecological determinants and their consequences on trees aboveground biomass assessments remain understudied in tropical regions. To our knowledge, it has never been investigated in Africa. Using a 3-D X-Ray scanner, we studied the radial WSG variation of 14 canopy species of DRC tropical forests. Wood specific gravity variance along the radial profile was dominated by differences between species intercepts (~76%), followed by the differences between their slope (~11%) and between individual cores intercept (~10%). Residual variance was minimal (~3%). Interestingly, no differences were found in the comparison of mean WSG observed on the entire core and the mean WSG at 1-cm under the bark (intercept ~0; coefficient = 1.03). In addition, local values of WSG are strongly correlated with mean value in the global data base at species level. I deeply believe these results favor the development of promising tools to map and to estimate accurately the AGB of tropical forest-stands. The information provided by largest trees on the entire forest-stand is particularly interesting both for developing new sampling strategies for carbon stocks monitoring and to characterize tropical forest-stand structure. In particular, our results should provide the opportunity to decrease current sampling cost while decreasing its main related uncertainties, and might also favor an increase of the current sampling coverage. |
