Vegetation canopy functioning can be studied combining both Soil-Vegetation-Atmosphere Transfer (SVAT) models and remote sensing data. These models describe energy and mass transfers in the soil-plant-atmosphere continuum. Remote sensing provides useful information for driving such models. The main objective of this work was to determine the contribution of multispectral remote sensing data in the functioning of a complex SVAT model. The chosen model was SiSPAT (Braud, 1995), considering coupled heat and moisture flows in the soil. It was coupled with two canopy radiative transfer models in order to simulate at field scale main surface processes and remote sensed data (bi-directional reflectance and directional brightness temperature). In the visible and the near infrared, the 2M-SAIL model (Weiss et al., 2001) was chosen for its ability to account for the development of yellow and green vegetation layers throughout the crop cycle. In the thermal infrared, the directional model proposed by François (2001) was used. In the microwaves domain (passive or active), the contribution of remote sensing data was only studied through the surface soil water content. This new developed model was called SiSPAT-RS (Simple Soil Plant Transfer and Remote Sensing) and was applied on two wheat field dataset, acquired during the ReSeDA experiment in 1997 in the South France. First, a sensitivity analysis was performed over 60 parameters and initial state variables, using a stochastic Monte Carlo sampling and a multicriteria methodology based on a Pareto ranking. Results allowed to detect the most influent parameters on the simulation of several state variables, and to reduce significantly their associated uncertainty intervals. The model calibration was performed considering different assumptions, related to the experimental knowledge of soil properties and surface variables available. This step allowed to (1) validate the model on the other wheat field and (2) propose and apply an assimilation method, based on the knowledge of thermal infrared brightness temperature and the surface soil water content. In this last context, it was possible to estimate the main surface processes with a good accuracy and to quantify the model errors associated to the parameter uncertainties.