PROSDM: Applicability of PROSPECT model coupled with spectral derivatives and similarity metrics to retrieve leaf biochemical traits from bidirectional reflectance

Authors: WAN, LIANG - Zhejiang University; CENA, HAIYAN - Zhejiang University; ZHANG, JIAFEI - Zhejiang University; XU, YING - Zhejiang University; Huang, Yanbo; LI, XIAORAN - Zhejiang University; ZHAI, LI - Zhejiang University; XU, HAIXIA - Zhejiang University; SUN, DAWEI - Zhejiang University; ZHOU, WEIJUN - Zhejiang University

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/16/2021
Publication Date: 10/27/2021
Citation: Wan, L., Cena, H., Zhang, J., Xu, Y., Huang, Y., Li, X., Zhai, L., Xu, H., Sun, D., Zhou, W. 2021. Applicability of PROSPECT model coupled with spectral derivatives and similarity metrics to retrieve leaf biochemical traits from bidirectional reflectance. Remote Sensing of Environment. https://doi.org/10.1016/j.rse.2021.112761.
DOI: https://doi.org/10.1016/j.rse.2021.112761

Interpretive Summary: Extraction of crop leaf biochemical properties, such as chlorophylls, carotenoids, water content, and dry mass content, from measured spectral data is important for monitoring crop growth. However, there exists inconsistency in bidirectional reflectance spectra led to deviation of spectral measurement. Scientists of Zhejiang University and USDA ARS Crop Production Systems Research Unit at Stoneville, Mississippi have proposed a new plant-leaf model framework to improve model's ability to consistently extract crop leaf biochemical property parameters. The results indicated that this newly proposed model framework provides a new insight for extracting leaf biochemical properties from vegetation remote sensing, which is valuable for fundamental understanding of crop leaf response to optics.

Technical Abstract: PROSPECT model has long been used for the retrievals of leaf biochemical properties from spectral data. However, it does not perform consistently well with bidirectional reflectance factor (BRF) spectra because the anisotropy properties lead to differences between BRF and directional-hemispherical reflectance factor (DHRF) spectra. Recently, some researches have been performed to improve PROSPECT model inversion with leaf BRF spectra based on several new models such as PROCOSINE and PROCWT, but the reliability and the accuracy were not satisfactory. This study proposed a new framework to integrate PROSPECT, Savitzky-Golay filter (SGF) and spectral similarity metric (SSM), called PROSGF-SSM, to improve model inversion. Three different experiments including varied nitrogen (N) fertilizer levels, different growth stages, and cultivars were conducted to collect leaf biochemical constituents and BRF spectra. The results showed that (1) PROSGF-SSM better characterized the difference between measured BRF and simulated DHRF spectra than other inversion models; (2) PROSGF-SSM outperformed PROSPECT, PROCOSINE and PROCWT to retrieve chlorophylls (Cab), carotenoids (Cxc), water content (Cw) and dry mass content (Cm) with root mean square error (RMSE) of 5.06 µg/cm2, 2.01 µg/cm2, 0.0035 g/cm2 and 0.0009 g/cm2, respectively; and (3) compared with PROSPECT, RMSE values of Cab, Cw and Cm retrieved from PROSGF-SSM were reduced by 9.80%, 31.37% and 30.77%, respectively, and a comparable retrieval of Cxc was also obtained. Meanwhile, PROSGF-SSM was also applied to model inversion with DHRF spectra, directional-hemispherical transmittance factor (DHTF) spectra, and both DHRF and DHTF spectra from two public datasets: LOPEX and ANGERS. The results further confirmed that PROSGF-SSM provided a great improvement on model inversion with DHRF and DHTF spectra, which outperformed the results in the previously reported studies. This study indicated that this newly proposed framework provides a new insight for extracting leaf biochemical properties from vegetation remote sensing, which is critical for promoting further application of PROSPECT model inversion in agriculture.