Abstract : In optical remote sensing, radiative transfer (R-T) models aim to simulate radiometric measurement of spatial sensors that spot “Landscape-Atmosphere” system. The modeling of vegetation canopies measurements is generally limited to landscape representation and R-T modeling. This thesis allowed transforming a 3-D R-T model DART (patent : PCT/FR 02/01181) to multispectral model simulating optical remote sensing images (thermal infrared included) of any natural and urban landscape with relief and atmosphere for any spatial/airborne sensor. This new model was validated by models intercomparison (RAMI-3 experience, Joint Research Centre, Italy) and with in situ and airborne measurements (Czech Republic). In fine, an inversion method was developed. It allowed using DART for evaluating the modeling impact on LAI estimation of coniferous forest using hyperspectral VHR images.