Improving the PROSPECT Model to Consider Anisotropic Scattering of Leaf Internal Materials and Its Use for Retrieving Leaf Biomass in Fresh Leaves

• Published in: IEEE transactions on geoscience and remote sensing Vol. 56; no. 6; pp. 3119 - 3136
• Main Authors: Qiu, Feng, Chen, Jing M., Ju, Weimin, Wang, Jun, Zhang, Qian, Fang, Meihong
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.06.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Absorption; Anisotropic scattering; Anisotropy; Biological system modeling; Biomedical optical imaging; Computer simulation; hyperspectral remote sensing; leaf internal scattering; leaf mass per area (LMA); leaf optical properties; Leaves; Light scattering; Near infrared radiation; Optical properties; Optical sensors; PROSPECT; Reflectance; Retrieval; Simulation; Spectra; Uncertainty; Water; Water absorption; Wavelength; Wavelengths
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2018.2791930

The PROSPECT model has been widely used to estimate leaf biochemical constituents, but retrieval of leaf mass per area (LMA) in fresh leaves has proved to be difficult due to the predominant water absorption in the infrared spectral region. At wavelengths where water absorption is low, both LMA absorption and light scattering are relatively high. Therefore, the uncertainty in scattering simulation at these wavelengths will lead to a relatively large error in LMA estimation. In this paper, we introduce a wavelength-independent factor to represent the first-order effect of anisotropic scattering in the elementary layer in the modified model PROSPECT-g, aiming at appropriately simulating leaf optical properties in spectral regions with high scattering and thus reducing the uncertainty in LMA estimation. In order to avoid introducing a new variable to be retrieved in model inversion, this factor is an intermediate variable derived from measured near infrared region spectral data and other existing model parameters. Results show that about 30%-40% of the tested samples are well simulated using PROSPECT-5, while for the rest of the samples simulation is greatly improved with PROSPECT-g. Leaf reflectance and transmittance reconstructions using PROSPECT-g are improved, especially at wavelengths with high scattering such as 750-1400 and 1500-1850 nm. LMA retrieval is significantly improved, with the average root-mean-square error decreasing from 38.7 (PROSPECT-5) to 16.6 g/m 2 (PROSPECT-g) for 628 leaves after considering anisotropic scattering in the elementary layer. Improvements are particularly noticeable for leaves with extremely high LMA contents.