Widlowski, JL, Pinty, B, Lavergne, T, Verstraete, MM, Gobron, N (2006). Horizontal radiation transport in 3-D forest canopies at multiple spatial resolutions: Simulated impact on canopy absorption. REMOTE SENSING OF ENVIRONMENT, 103(4), 379-397.

The divergence of horizontal radiation in vegetation canopies is generally considered to be of negligible consequence to 1) algorithms designed for the physically-based interpretation of space home observations, and 2) field campaigns aiming at the validation of derived surface products, like FAPAR and albedo. However, non-zero horizontal radiation balances are likely to occur if the internal variability of the vegetation target and the typical distances that photons may travel horizontally within such 3-D media extend to spatial scales that are similar to or larger than those of the measuring sensor. Detailed radiative transfer simulations in 3-D coniferous forest environments are presented to show how the magnitude of local net horizontal fluxes (for spatial resolutions ranging from 1 x 1 m(2) to 500 x 500 m(2) forest areas) can reach multiple times the incident solar radiation at the top-of-canopy level. Furthermore, the PDFs of these local net horizontal fluxes (H) are skewed toward negative values (meaning that most local canopy volumes have more radiation exiting than entering via their lateral sides), in particular when the radiative regime is dominated by single-scattering interactions and geometric shading is prominent. In order to maintain the energy balance of the overall forest domain, however, local canopy volumes with rather large positive net horizontal fluxes must also exist, thus underscoring the importance of properly locating local flux measurement equipment. Irrespective of the sign of H, it is shown that the local canopy absorption (A) falls within the A-H (perfect shadowing of the forest floor) and A=H+1 (perfect illumination of the forest floor) domain in the red spectral band. This correlation between A and H implies that the range of local canopy absorption values is far larger than unity which reduces its potential to serve as a proxy in delivering accurate domain-averaged absorption estimates on the basis of spatially incomplete sampling schemes. Instead, it is shown that, for a spatial sampling of 1% of the forest area of interest, local absorption estimates - derived from vertical fluxes only - are sufficient for delivering 2 domain-averaged canopy estimates that lie, on average, within 0.05 of the truth. For forest domains that are smaller than about 30 x 30 in however, horizontal radiation transport will still affect the domain-averaged canopy absorption values and thus a spatially exhaustive sampling of the true local absorption may be more appropriate. (c) 2006 Elsevier Inc. All rights reserved.