Inverse Retrieval of Chlorophyll From Reflected Spectra for Assimilating Branches of Drought-Tolerant Tamarix ramosissima

Abstract

Leaf chlorophyll content is a critical indicator for the better understanding of the featured carbon cycle and ecosystem functions in arid regions. The potential applicability of the inverse retrieval of chlorophyll via the radiative transfer model PROSPECT has been intensively examined in this study for a dominant species in desert ecosystems, Tamarix ramosissima, in central Asia, which has distinct structural features of xeromorphism and its young twig functioned as “assimilating organisms.” This study revealed that the performance of the inverse retrieval based on the original version of PROSPECT-4 was poor with very low accuracy. As a comparison, its calibrated version was capable of accurate retrieval of chlorophyll content, which recorded an R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (coefficient of determination) of 0.47 with a root mean square error (RMSE) of 28.79 mg/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , over the entire measured chlorophyll range, from 57.37 to 202.27 mg/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> . The model calibration is not considered to be overfitting since the tenfold cross-validation shows a close RMSE value of 26.19 mg/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> . The effects of the percentage of training dataset for calibration and the number of calibration repeats have also been intensively investigated in this study. Despite some inherent defects of applying broadleaforiented PROSPECT on assimilating branches, it remains a feasible selection to inversely retrieve the chlorophyll content from the reflectance and thus provides a base for investigating the chlorophyll content of desert plants using hyperspectral remote sensing.