AN ANALYTICAL AND COMPUTATIONALLY EFFICIENT REFLECTANCE MODEL FOR LEAF CANOPIES

An explicit analytical model for calculating vegetation canopy reflect ance is developed in this paper, based on radiative transfer theory th rough separating the roles of incident direct and diffuse radiation, s ingly and multiply scattered radiation by foliage and soil. Using prin ciples of scattering from a point source, the contributions of diffuse sky radiation to both first-order and multiple scattering reflectance are accurately specified. In addition, this analytical model incorpor ates the effects of nonrandom spatial dispersion of foliage, noncircul ar shape and nonhorizontal orientation of leaves on the canopy hotspot and then on reflectance distributions. Although the reflectance by mu ltiple scattering is simply estimated with the two-stream approximatio n (Nilson, 1991), the angular distributions of canopy reflectance prod uced by the model agree well with those measured above soybean and whe at canopies. Some original results from model sensitivity analysis are sketched as follows: (1) leaf shape and orientation have a considerab le influence on the canopy hotspot except for planophile canopies or i n near nadir viewing directions. The contributions of soil reflectance and multiple scattering increase with the modal inclination angle of leaves; (2) the foliage spatial dispersion pattern changes the magnitu de and angular distribution of the canopy reflectance strongly in the RED band; (3) when the ratio of diffuse sky radiation to the total inc ident radiation is less than 20%, its contribution to canopy reflectan ce is less than 6% and 8% on average for RED and NIR bands, respective ly. This provides a means to approximately calculate the two component s signifying the role of diffuse sky radiation and then to conditional ly simplify this reflectance model, and therefore improve its practica l applicability; (4) the sensitive regions of different structural par ameters are usually different and generally change with waveband and l eaf orientation. This implies that only by using the reflectance data in its sensitive region, can a specified parameter be accurately estim ated by model inversion. Finally, further improvements needed for anal ytical approaches are briefly discussed.