We present a new approach to simulate the distribution of natural ligh
t within plant canopies. The canopy is described in 3D, each organ bei
ng represented by a set of polygons. Our model calculates the light in
cident on each polygon. The principle is to distinguish for each polyg
on the contribution of the light coming directly from light sources, t
he light scattered from dose polygons and that scattered from far poly
gons. Close polygons are defined as located inside a sphere surroundin
g the studied polygon and having a diameter D-s. The direct light is c
omputed by projection, The exchanges between close polygons are comput
ed by the radiosity method, whereas the contribution from far polygons
is estimated by a multi-layer model. The main part of computing time
corresponds to the calculations of the geometric coefficients of the r
adiosity system. Then radiative exchanges can be quickly simulated for
various conditions of the angular distribution of incoming light and
various optical properties of soil and phytolelements. Simulations com
pare satisfactorily with those produced by a Monte Carlo ray tracing.
They show that considering explicitly the close neighboring of each po
lygon improves the estimation of organs irradiance, by taking into acc
ount the local variability of fluxes. For a virtual maize canopy, thes
e estimations are satisfying with D-s = 0.5 m; in these conditions, th
e simulation time on a workstation was 25 min for a canopy of 100 plan
ts. (C) 1998 Elsevier Science B.V. All rights reserved.