The European beech (Fagus silvatica L.) is a late-successional shade tolerant species. It is often cultivated and managed with the shelterwood system in Europe, meaning that its natural regeneration takes place in shade. Under low light conditions, seedling size must stay below a threshold where carbon gain and carbon balance remain positive. This is not so easy to determine experimentally. Therefore, a simulation approach using the carbon-based functional-structural model PIAF-1 (Lacointe and Donès, 2007) was used to investigate the relationship between carbon balance and growth of young beech trees under low light availability. Two-year-old beech seedlings were planted in March 2007 in the understorey of natural Scots pine stands in Fontfreyde, Chaîne des puys, France. In summer 2009, nine beech seedlings were sampled under light availability ranging from 7 to 28% of incident light. To acquire their architecture, aboveground parts of each tree were 3D-digitized, including woody parts (growth units) and leaves using PiafDigit (Donès et al., 2006). Then seedlings were harvested in autumn and winter to quantify biomass and starch and sucrose contents of each organ, above and below ground. Roots were separated into 3 classes: taproot, main lateral roots and fine roots. Each part was scanned and the length and diameter of roots were assessed using WinRhizo (Regent Instruments Inc., Canada). These data were used to parameterize and initialize PIAF-1. The digitized aboveground parts were converted from growth units to cylindrical segments bearing leaves in their true place and orientation. Belowground parts were reconstructed based on diameter and length of each root class. Meteorological data (hourly light and temperature) were generated for each tree based on recorded weather data. Transmitted solar irradiance above each seedling was calculated from hemispherical photographs and the global radiation above the stand. Each seedling was visualized on a PIAF-1 3D visualization interface. The quality of the reconstructed seedlings was assessed by comparing the graphical outputs from PIAF-1 with those from VegeSTAR (Adam et al., 2002) using Tree Analyser (Phattaralerphong et al., 2005) to compute synthetic geometrical properties such as total leaf area or density. Potential growth, amount of storage, and reserve capacity of stem, leaf and root are still being analyzed. Results of simulations will be reported and discussed.