We conduct a sensitivity/uncertainty analysis of the land-surface scheme JULES focusing on biophysical parameters and model processes that govern light propagation and canopy photosynthesis. We find that current simulations of productivity/energy-exchange are limited as much by their approximate representation of complex physical processes as they are by the accuracy to which their biophysical parameters can be specified. This inference is made for three forest biomes: sparse, boreal needleleaf; dense, tropical broadleaf; and temperate broadleaf of intermediate density. Within the present study, the most influential biophysical parameters are light-limited quantum efficiency (α), the Rubisco-limited rate of photosynthesis at the top of the canopy (V cmax ) and the near-infrared transmittance of vegetation (T NIR ). Assuming respective, current uncertainties of 0.02, 20 5mol/m2/s and 0.16 in these aforementioned parameters, predictions of Gross Photosynthetic Product (GPP), latent heat (LE), sensible heat (H) and upwelling thermal radiation (R) vary by ~20%, ~10%, ~35% and ~0.3%, respectively. Current representations of canopy photosynthesis and stomatal conductance yield comparable uncertainties in simulated GPP and LE and a ~60% uncertainty in H. For canopy photosynthesis, explicit treatment of sunfleck penetration and leaf orientation are important elements in the calculation. Of the meteorological variables that drive the land-surface scheme, the downwelling fluxes of radiation in the shortwave and longwave vie in importance with the most influential biophysical parameters. The results from our study are partly biome-dependent. Thus ground albedo and leaf area index (LAI) assume greater importance in sparsely vegetated systems.