Simple models for light competition within agroforestry discontinuous tree stands: are leaf clumpiness and light interception by woody parts relevant factors?

Predicting the temporal and spatial variability of radiation intensity under wide-spaced tree stands is required for many applied issues in savannah-like ecosystems, orchards, agroforestry and urban forestry systems. Numerous authors have advocated the use of simple light interception models that approximate the crown shape with ellipsoids. They have suggested taking into account leaf clumping to improve the efficiency of these simple models, but this was never assessed. We tested this hypothesis together with the impact of including predictions of light interception by woody parts (trunks, branches). We calibrated and evaluated the model using cross-validation across eight walnut trees with field measurements of radiation intensity and spatial heterogeneity using hemispherical photographs. Leafless trees were efficiently modelled using Wood Area Density (WAD, m 2 m −3 ) for branches and an opaque cone for the trunk. We introduced a clumping parameter (μ) but this proved inefficient, clumping being highly variable amongst trees. This results from the limitations of representing the crown as an ellipsoid, a procedure too coarse to be improved by using a clumping parameter. The model proved efficient to predict the light pattern around an average tree, but was not fit for simulating the variability of individual trees. We finally discuss practical recommendations for modelling light competition in integrated agroforestry models simply.