DART-based model for three-dimensional scaling of energy balance from individual leaves to plant canopies

Abstract. Spatial heterogeneity of vegetation canopies significantly influences the ecosystem energy exchange and gross primary production (GPP) by affecting processes, such as, the photosynthetic light absorption, energy distribution, and carbon fixation. Conventional one-dimensional (1D) models cannot account for horizontal structural diversity of complex canopies as, for instance, sparse row crops or crops in early growth stages. Therefore, the primary aim of this study was to enhance the accuracy of GPP estimations by incorporating three-dimensional (3D) plant structural complexity. To achieve this goal, a first version of the 3D energy balance DART-EB model was developed by combining the Discrete Anisotropic Radiative Transfer (DART), effectively capturing full 3D canopy structure, with modified energy balance (EB) equations originating from the 1D model SCOPE. Unlike the 1D approach, DART-EB iteratively resolves EB for individual elements of a 3D soil–vegetation scene (i.e., soil surface, plant leaves, and stems) separately. High computational efficiency of DART’s LuxCorRender Monte Carlo bi-directional photon-path tracing algorithm allows to compute emitted and absorbed radiation per side of a leaf-representing triangular facet. To verify DART-EB performance, we first cross-compared its results with outputs of the widely used SCOPE model for a virtual homogeneous canopy. A strong consistency between the two models was confirmed by their high agreement in simulated net photosynthesis (d-index = 1; RRMSE = 3 %). Once verified, DART-EB was tasked to perform a comparison between geometrically-explicit vertically-homogeneous and spatially-heterogeneous vineyard canopies. Omitting the spatial heterogeneity of a vine canopy structure led to 44 % RRMSE in simulated net photosynthesis. Further comparison of DART-EB outputs against eddy-covariance measurements of an alfalfa crop in Spain revealed a significantly lower error (RRMSE = 8 %) for its genuine than for a homogeneous canopy representation (RRMSE = 20 %). The spatially-explicit digital 3D representations of alfalfa field were reconstructed from their in-situ RGB photographs. The error reduction was caused by more accurate modelling of a total absorbed photosynthetically active radiation by chlorophylls, enabled by a realistic 3D spatial distribution of the alfalfa foliage. The presented first version of DART-EB continues to be further developed, aiming to increase the computational accuracy, efficiency, and applicability to structurally highly complex canopies (e.g., forests) and large landscapes.