Abstract/Summary

Vegetation is one of the largest terrestrial sinks of atmospheric carbon dioxide, driven by the balance between photosynthesis and respiration. Understanding the processes behind this net flux is critical, as it influences the global atmospheric carbon dioxide concentration and hence climate change. A key factor determining the carbon flux into the land surface is the absorption of light by vegetation, used to drive photosynthesis. However, climate models commonly represent vegetation canopies as homogenous slabs of randomly positioned leaves. By contrast, real forests generally exhibit large amounts of 3-dimensional heterogeneity. We examine the impact of including measured 3D vegetation canopy structure on modelled gross primary productivity (GPP) by looking at how leaf area is distributed. We introduce a methodology to calculate GPP using output from the explicit Discrete Anisotropic Radiative Transfer (DART) model, following the approach commonly used in land surface schemes. The sensitivity of modelled GPP to canopy structure assumptions in Earth system models is explored, using 3D structural information derived from six forest plots using Terrestrial Lidar Scanning (TLS) data. Here, we use the spatial resolution as a proxy for the canopy structure, with the very coarsest simulations containing no spatial variability in leaf location, with variability introduced as the resolution of the simulations becomes finer. In almost all cases, the simulated GPP is reduced, and with the finest resolution this is up to 25%. This contrasts with recent studies showing the opposite effect. In the few cases where the GPP increased, this was only marginal (< 2.5%). These results suggest that not accounting for the impact of 3-dimensional canopy structure could lead to significant biases in land surface models, particularly in forest’s contribution to the global carbon budget. We suggest that vegetation structure is considered, explicitly or through a correction factor, alongside a comparison to existing clumping approaches.