We developed a double-digitising method combining a hand-held electromagnetic digitizer and a non-contact 3D laser scanner. The former was used to record the positions of all leaves in a tree and the orientation angles of their lamina. The latter served to obtain the morphology of the leaves sampled in the tree. As the scanner outputs a cloud of points, software was developed to reconstruct non-planar (NP) leaves composed of triangles, and to compute numerical shape parameters: midrib curvature, torsion and transversal curvature of the lamina. The combination of both methods allowed construction of 3D virtual trees with NP leaves. The method was applied to young beech trees (Fagus sylvatica L.) from different sunlight environments (from 1 to 100% incident light) in a forest in central France. Leaf morphology responded to light availability, with a more bent shape in well-lit leaves. Light interception at the leaf scale by NP leaves decreased from 4 to 10% for shaded and sunlit leaves compared with planar leaves. At the tree scale, light interception by trees made of NP leaves decreased by 1 to 3% for 100% to 1% light, respectively.