Terrestrial biosphere models (TBMs) are invaluable tools for studying plant–atmosphere interactions at multiple spatial and temporal scales, as well as how global change impacts ecosystems. Yet, TBM projections suffer from large uncertainties that limit their usefulness. Forest structure drives a significant part of TBM uncertainty as it regulates key processes such as the transfer of carbon, energy, and water between the land and the atmosphere, but it remains challenging to observe and reliably represent. The poor representation of forest structure in TBMs might actually result in simulations that reproduce observed land fluxes but fail to capture carbon pools, forest composition, and demography. Recent advances in terrestrial laser scanning (TLS) offer new opportunities to capture the three-dimensional structure of the ecosystem and to transfer this information to TBMs in order to increase their accuracy. In this study, we quantified the impacts of prescribing initial conditions (tree size distribution), constraining key model parameters with observations, as well as imposing structural observations of individual trees (namely tree height, leaf area, woody biomass, and crown area) derived from TLS on the state-of-the-art Ecosystem Demography model (ED2.2) of a temperate forest site (Wytham Woods, UK). We assessed the relative contributions of initial conditions, model structure, and parameters to the overall output uncertainty by running ensemble simulations with multiple model configurations. We show that forest demography and ecosystem functions as modelled by ED2.2 are sensitive to the imposed initial state, the model parameters, and the choice of key model processes. In particular, we show that: * Parameter uncertainty drove the overall model uncertainty, with a mean contribution of 63 % to the overall variance of simulated gross primary production. * Model uncertainty in the gross primary production was reduced fourfold when both TLS and trait data were integrated into the model configuration. * Land fluxes and ecosystem composition could be simultaneously and accurately simulated with physically realistic parameters when appropriate constraints were applied to critical parameters and processes. We conclude that integrating TLS data can inform TBMs of the most adequate model structure, constrain critical parameters, and prescribe representative initial conditions. Our study also confirms the need for simultaneous observations of plant traits, structure, and state variables if we seek to improve the robustness of TBMs and reduce their overall uncertainties.