Speaker
Description
Given its unique penetration capabilities and sensitivity to woody elements, radar remote sensing techniques at P and L bands are certainly the most promising tools for wide scale retrieval of forest parameters. Nonetheless, the radar signal sensitivity is prone to multifactorial
effects, and the retrieval methods (mostly correlative or network based) are often limited by an insufficient number of training samples_
For a better understanding of these multifactorial effects, sensitivity analysis can be performed trough the use of Electromagnetic (En) models (e.g MIPERS-4D hereinafter), providing likelihood estimates of the retrieved parameters. In this study, a specific focus is given on forest height and biomass retrieval from P and L bands.
Regarding forest height retrieval, several assumptions are investigated, such as the uncorrelation between volume and ground scattering (being a key hypothesis behind the so-called RVoG and more generally two-components models), and the 3dB threshold assumption coming with the exploitation of Tomo SAR based reflectivity profiles.
More challenging, we also see that the retrieval of forest biomass depends on a wider range of unknown parameters, such as the intrinsic relation between the canopy fresh matter and the total dry biomass (forest AGB).
These results will be also discussed and cross compared with respect to experimental data and estimates derived from recent airborne and tower-based experiment. Interstingly, we will also see that this cross analysis with the latter contribute to the spatialisation of temporal results observed at local scale, providing semi-empirical relations adapted to regional retrieval.
