Informative bands used by efficient hyperspectral indices to predict leaf biochemical contents are determined by their relative absorptions

Hyperspectral indices have proven to be useful for quantifying various plant parameters including a number of biochemical components. However, it remains a great challenge to unveil the physical and physiological mechanisms of wavelengths used by these statistically/empirically identified indices, and this unveiling is a critical step towards developing generally applicable indices. In this study, we have introduced a dummy variable to the well-accepted leaf scale radiative transfer model (PROSPECT-4) and have conducted a series of virtual experiments by tweaking different absorption features (intensity, peak wavelength, and half-width) of the dummy variable for investigating the underlying mechanisms of those wavelengths used by efficient indices. Results clearly indicated that the informative bands used in efficient indices, instead of the target parameter's absorption peaks, are generally concentrated among wavelengths that the targeted parameter has relatively high specific absorption coefficients compared with other biochemical components. This finding has also been validated by replacing the dummy variable with carotenoids based on PROSPECT-5B simulated dataset. Further analysis reveals that although the concentrations of chlorophyll are higher than other pigments in most leaves, the commonly recommended wavelength of 550 nm may be disturbed by other pigments and may only be applicable when anthocyanins are minor. Results obtained in this study have largely explained why specific wavelengths are used in efficient indices, and thus this should lay a basis for understanding the underlying mechanisms and help to develop robust indices for estimating vegetation biochemical parameters.