Land surface models (LSMs) are an important tool for understanding land-surface-atmosphere dynamics and interactions and climate-carbon feedbacks. Perhaps the partitioning of solar radiation between various compartments of the surface constitutes the most important step to further quantify the role of vegetation in redistributing energy and drive related biogeophysical processes, as photosynthesis, evapotranspiration, changes in leaf and soil temperature and snowmelt. The advent of the Two-Stream (TS) solution brought more accuracy to LSMs as the radiation treatment was divided into two directions (upward and downward) and properties (direct and diffuse). It is known that diffuse solar radiation can enhance photosynthetic capacity, as it penetrates more efficiently into the vegetation canopy. Firstly, a statistical evaluation of the diffuse light effect on energy and carbon fluxes over Amazon is conducted. However, even the broadly used TS scheme is not realistic enough to account for effects on radiation propagation related to canopy architecture. The vegetation structure varies considerably for different ecosystems and it has an important impact on the amount of absorbed radiation and surface albedo, which directly affects the surface energy and carbon balances. 3D radiation transfer (RT) models are accurate tools to obtain energy propagation and they have been widely used by the satellite community in order to derive land surface characteristics, but because of their complexity and computational costs, they are unsuitable for large areas and long period runs. Secondly, to bring the reality of 3D RT models into an efficient operational physically-based LSM, a parameterization accounting for vegetation structure is introduced and tested.