Hyperspectral retrieval of leaf physiological traits and their links to ecosystem productivity in grassland monocultures

• Published in: Ecological indicators Vol. 122; p. 107267
• Main Authors: Zhao, Yujin, Sun, Yihan, Lu, Xiaoming, Zhao, Xuezhen, Yang, Long, Sun, Zhongyu, Bai, Yongfei
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.03.2021; Elsevier
• Subjects: Aboveground biomass; canopy; carbon; carotenoids; chlorophyll; Ecosystem function; ecosystems; Functional trait; grasslands; Hyperspectral image; leaf area index; leaves; nitrogen; Partial least squares regression (PLSR); photosynthesis; prediction; specific leaf weight; spectrometers; Unmanned aerial vehicle (UAV); unmanned aerial vehicles; Partial least squares regression (PLSR); Unmanned aerial vehicle (UAV); Hyperspectral image; Functional trait; Ecosystem function; Aboveground biomass
• DOI: 10.1016/j.ecolind.2020.107267
• ISSN: 1470-160X

•Leaf traits should be hyperspectrally estimated on an area basis in grasslands.•LAI-based upscaling in spectral estimates of canopy traits outperforms biomass basis.•Plot-level leaf trait values and intraspecific variations link traits to productivity. Plant functional traits are closely associated with key ecological processes and ecosystem functions. Recent studies have demonstrated that plant functional traits, especially physiological traits, can be successfully derived from hyperspectral images. Plant physiological traits are frequently quantified either as area-based content [μg cm−2] or mass-based concentration [mg g−1 or %]. However, it remains unclear whether the two metrics of traits can be quantified using remote sensing approaches. We quantified area- and mass-based foliar physiological traits to compare the prediction accuracy of the two metrics based on leaf spectra using partial least squares regression (PLSR) at a grassland monoculture experiment. These two metrics were then scaled up to canopy traits, respectively, based on leaf area index (LAI) and biomass to test their performance at the canopy level. The canopy physiological traits with high prediction accuracy (R2 ≥ 0.60) were selected for mapping using the unmanned aerial vehicle (UAV)-based UHD185 spectrometer. Biomass and LAI were also estimated and mapped using the PLSR method. The mapped leaf traits (canopy traits divided by the corresponding LAI), were used to explore the relationships between the interspecific and intraspecific variations in leaf physiological traits and ecosystem productivity (i.e., aboveground biomass). The results showed that the retrieval of leaf physiological traits using leaf spectra and canopy spectra or remote sensing was better performed on an area basis rather than a mass basis, especially for the physiological traits related to photosynthesis. Model selection results also indicted that remotely sensed physiological traits (chlorophyll a, chlorophyll b, carotenoid, carbon, nitrogen, and leaf mass per area (LMA)) and their intraspecific variations (coefficient variation (CV) for a single trait and functional richness (FRic) for multiple traits) were significant predictors of community aboveground biomass across grassland monocultures. Our study highlights the potential of hyperspectral images for trait mapping and estimating ecosystem productivity at large scales. Our findings also provide a vital insight for disentangling the links of functional traits and intra- and interspecific trait variations to key ecological processes and functions.