Forest ecosystems are an essential component of the Earth?s surface and are influenced by disturbance events (e.g. harvests, fires) that create available growing space for trees and affect the composition of forest ecosystems.
Disturbance events influence the value of Leaf Area Index (LAI), Canopy Cover (Cv), and Chlorophyll content (Cab). These parameters can thus be used for disturbance monitoring. In practice, they can be effectively retrieved for large areas using remote sensing data. Retrieval methods that rely on the inversion of radiative transfer models are most appropriate since they enable to retrieve estimates for several parameters at once. The inversion, however, is an ill-posed problem: the number of unknowns is greater than the number of observations and there is no analytical solution. Numerical algorithms have been used to retrieve solutions but the accuracy of the estimates is limited. The objective of this study will be to improve the accuracy of the estimates of LAI, Cv, and Cab by reducing the ill-posedness thanks to three techniques: using a coupled model, using prior information and increasing the number of observations. Reflectance models for soil, leaf, canopy, and atmosphere will be coupled to simulate images at Top Of Atmosphere (TOA) level, from where satellites collect data. First, a sensitivity analysis of the coupled model will be carried out. Second, a new method to integrate prior information, such as GIS data, in the retrieval process will be proposed. Third, a method will be developed to increase the robustness of the data assimilation by concurrently using multi-sensor data. Finally, the knowledge and techniques acquired in the previous steps will be integrated in a single system and used to retrieve LAI, Cv, and Cab in the context of an operational application.