Canopy microclimate is a critical component of most biophysical processes in plants. Understanding the role of microclimate across scales in canopies with complex, heterogeneous architectures is challenging, as it is difficult to represent the relevant range of scales. In this study, a model was developed and validated to accurately predict the three-dimensional distribution of microclimate-related quantities (e.g., net radiation, surface temperature, evapotranspiration, flux partitioning) in complex canopy geometries. The modeling strategy was to aggregate individual leaves into isothermal sub-volumes, as resolving all leaves over whole-canopy scales is unfeasible. The model takes leaf-level models for convection, evapotranspiration, and radiative absorption/emission and integrates them over a discrete volume using the leaf angle probability distribution function. The model is built on a framework designed to utilize graphics processing units (GPUs) in order to offset the expense associated with model complexity, which can allow for the simulation of canopy-scale problems at sub-tree resolution. Additional cost-saving strategies are also suggested. Three-dimensional canopy energy transfer models have traditionally been difficult to validate. Two validation experiments were designed to simultaneously measure virtually all model outputs, often using multiple methods at many spatial locations. The model was able to reproduce point, 3D distributed, and bulk measurements at high accuracy, with average model errors within expected measurement errors. living on or within plant tissues. These processes can have high sensitivity to 4 leaf temperature, as temperature differences of only a degree can have critical 5 implications on physiological function (Jones, 2004). 6 Leaf temperature can vary dramatically over short spatial distances due to 7 step changes in radiation interception. In dense, closed canopies, radiative fluxes 8 are relatively homogeneous in horizontal directions, which results in average 9 temperature distributions that are primarily one-dimensional (Monteith and 10 Unsworth, 2008). Many canopy energy models have been successfully developed 11 which treat the canopy as a single 'big leaf ' (e.g., Sinclair et al., 1976; Sellers 12 et al., 1996), or several layers of big leaves (e.g., Leuning et al., 1995; Dai et al., 13 2004). However, in heterogeneous landscapes (e.g., urban canopies, savannas), 14 temperature can vary in the horizontal direction by 10 • C or more within a 15 single tree crown (Tyree and Wilmot, 1990). This necessitates a fully three-16 dimensional (3D) description of leaf temperature in these types of canopies.
17Modeling 3D canopy microclimate presents several challenges, the foremost 18 of which is accurately predicting net radiation fluxes, as radiation drives leaf 19 temperature. As a result, leaf temperature is very sensitive to errors in pre-20 dicted net radiation. During the day, radiative fluxes are usually very large, 21 and dominate the leaf energy balance....
