Leaf optical properties with explicit description of its biochemical composition: Direct and inverse problems

Abstract

This study presents a methodology to estimate the leaf biochemical compounds specific absorption coefficients and to use them to predict leaf biochemistry. A wide range of leaves was collected including variations in species and leaf status. All the leaves were dried out. The biochemical composition was measured using classical wet chemistry techniques to determine lignin, cellulose, hemicellulose, starch, and protein contents. Concurrently, leaf reflectance and transmittance were measured with a high spectral resolution spectrophotometer in the 800–2500 nm range with approximately 1 nm spectral resolution and sampling interval. In addition, infinite reflectance achieved by stacking leaves was also measured. The PROSPECT leaf optical properties model was first inverted over a selection of wavebands in the 800–2400 nm domain to provide estimates of the scattering characteristics using leaf reflectance, transmittance, and infinite reflectance data. Then, the model was inverted again over all the wavelengths to estimate the global absorption coefficient, using the previously estimated scattering properties. The global absorption coefficient was eventually explained using the measured biochemical composition by fitting the corresponding specific absorption coefficients after substraction of the measured contribution of the residual structural water absorption. Results show that the derived specific absorption coefficients are quite robustly estimated. Further, they are in good agreement with known absorption features of each biochemical compound. The average contribution of each biochemical compound to leaf absorption feature is also evaluated. Sugar, cellulose, and hemicellulose are the main compounds that contribute to absorption. Results demonstrate the possibility of modeling leaf optical properties of dry leaves with explicit description of leaf biochemistry. Estimates of the detailed biochemical composition obtained by model inversion over the 1300–2400 nm spectral domain show poor predictive performances. In particular, the protein content is very poorly retrieved. The retrieval performances of several combinations of the biochemical compounds are investigated. Results show that the total amount of dry matter per unit leaf area is the only variable to be accurately retrieved. Possible improvements of these results are discussed.