Improving the retrieval of solar-induced chlorophyll fluorescence at canopy level by modeling the relative peak height of the apparent reflectance

Abstract

Recently, a great number of algorithms have been used in connection with the retrieval of solar-induced chlorophyll fluorescence (SIF) with the aim of detecting the physiological status of vegetation. In these SIF retrieval algorithms, methods based on Fraunhofer line discrimination (FLD) are widely used. However, all these FLD-based methods require that assumptions are made about the shape of the real reflectance around the absorption bands. These assumptions are a significant source of error in the retrieval of SIF and mean that satisfactory results at the O2-B band cannot be obtained if the spectral resolution (SR) is relatively coarse. In order to enable more accurate SIF retrieval at the O2-B band at canopy level, we have developed an alternative solution that uses a statistical empirical model to retrieve the SIF inside the absorption band, bypassing the assumptions about the shape of the real reflectance made in the FLD-based methods. In the proposed approach, called the reflectance height method (RHM), a model of the relative peak height of the apparent reflectance using two variables including the absorption depth of the irradiance and the gradient of the smoothed reflectance around the absorption bands is constructed using simulated training samples covering the most commonly occurring vegetation conditions. Validation of the RHM using both simulations and in situ observations showed that the RHM can effectively improve the SIF retrieval accuracy at the O2-B band at canopy level compared to traditional FLD-based methods, especially for cases using sensors with relatively coarse SR (SR≥1 nm) and high signal-to-noise ratio (SNR) (SNR≥1000).