In absence of any absolute reference truth RT model performance can be expressed in terms of the deviation between some model m and another model c from among the ensemble of models that performed a given set of spectral (lambda), structural (zeta), illumination (Omega_i) and viewing (Omega_v) conditions. Averaging over an appropriate set of such conditions will allow to compute a (delta_m<->c) model-to-model dispersion statistics [%]:

where

is the total number of all spectral (lambda), structural (zeta), illumination(Omega_i) and viewing (Omega_v) conditions used in the averageing.

• The following Figures depict a series of two-dimensional grids containing information on the various model-to-model BRF differences (blue-red colour scheme in the lower right half of each panel), as well as, the percentage of the total number of BRFs over which the delta_(m<->c) values were derived (black-green colour scheme in upper left half of each panel). More specifically, delta_(m<->c) is shown for those models having submitted the total (top left), single uncollided (bottom left), single collided (bottom right) and multiple collided
• (top right) BRF data for ``structurally homogeneous'' scenarios with finite-sized and turbid medium foliage representations and for spectral conditions in the red and near-infrared (NIR). The blue colour scale increments in steps of 2%, the green colour scale in steps of 10%, and the red also in steps of 10% with the bright red colour indicating values larger than 50%. The maximum number of BRF simulations included in the computation of delta_(m<->c) was 1216 for the ``structurally homogeneous'' canopies.