Above-ground net primary productivity (NPP) of individual arabica coffee plants (60 coffee resprouts, of various ages (0 to 5 years, after pruning), located below or far from the shade trees) was assessed during two years in the field from repeated biomass estimations (via branch scale allometry) and litter harvest (30 litter traps) (Coffee Flux Observatory, http://www5.montpellier.inra.fr/ecosols/Recherche/Les-projets/CoffeeFlux). Suprisingly, NPP was not influenced by the distance to the shade trees. MAESTRA, a 3D light interception model was applied to map shade tree transmittance and to calculate yearly light budgets (absorbed photosynthetically active radiation, aPAR) of the coffee plants (Charbonnier et al., 2013). Light use efficiency (LUE) of coffee plants was calculated dividing their annual NPP by plant aPAR. MAESTRA showed that aPAR decreased severely for coffee plants located under shade tree crowns (down to 70%). However, we obtained a 2-fold increase in LUE for coffee plants located under shade trees, and a spatial gradient of LUE according to the distance to the shade tree. The analysis revealed that the increase in LUE totally compensated the expected reduction of NPP due to the reduction in aPAR. We will discuss the possible causes of such an increase in LUE and the genericity of this finding. We will also emphasize on the role played by the 3D light interception model in the demonstration of this crucial property for AFS.