Estimation de paramètres structuraux des arbres dans une savane à partir de mesures LiDAR terrestre et d'imagerie à très haute résolution spatiale

Abstract

This thesis takes its place in a context where information on the biophysical state of forest ecosystems at spatial scales only remote sensing can retrieve is in demand more than ever. In order to provide reliable information using validated approaches, the remote sensing research community recognises the need for new and innovative methods, especially in heterogeneous environments like savannas. The recent emergence of terrestrial LiDAR scanners (TLS) and the increase in the computational capability of computers which allow running ray tracing model simulations with a high level of realism hold great potential to improve our understanding of the processes influencing the radiance measured by satellite sensors. This thesis makes use of these two cutting edge technologies for estimating the spatial distribution of tree leaf area, a key element of modeling radiative transfer processes. The first part of the thesis concerns the development of methods for estimating tridimensional leaf area distribution in a savanna environment from TLS measurements. The methods presented address certain issues related to TLs measures affecting the application of classical theories (the probability of light transmission and the contact frequency) to the estimation of leaf area through indirect means. These issues pertain to the cross-section of laser pulses emitted by a TLS and the occlusion effects caused by the interception of laser pulses by material inside the crown. The developed methods also exploit additional information provided by the active nature of the TLS sensor that is not available to passive sensors like hemispherical photography, i.e. the intensity of a pulse return offers the possibility to distinguish between energy interception by wood and foliage. A simplified approach of this method is presented to promote its use by other research groups. This approach consists of a series of parameterisations and represents a significant gain in terms of the required resources to produce the leaf area, estimates. The second part of the thesis explores the combination of the tree representations generated in the first part with a ray tracing model to simulate the interactions of light with tree crowns. This approach is highly innovative and our study showed its potential to improve our understanding of the factors influencing the radiative environment in a savanna. The methods presented offer a solution to map leaf area at the individual tree scale over large areas from very high spatial resolution imagery.