Can wetland plant functional groups be spectrally discriminated?

24 January 2019

Plant functional traits (PFTs) underpin ecosystem processes and therefore ecosystem service provision. If PFTs are possible to detect and discriminate spectrally, then it may be possible to use remote sensing applications to map ecosystem processes or services within and across landscapes. As a first step towards this application, we explored whether functional groups of 22 dominant South African wetland species were spectrally separable based on their PFTs. We measured 23 biochemical and morphological PFTs in combination with spectra from 350 to 2349 nm using a handheld radiometer. First, we evaluated the possibility of accurately predicting morphological and biochemical PFTs from reflectance spectra using three approaches: spectrum averaging, redundancy analysis (RDA), and partial least squares regression (PLSR). Second, we established whether functional groups and species were spectrally distinguishable. We found seven PFTs to be important in at least two of the three approaches: four morphological and three biochemicals. Morphological traits that were important were leaf area (PLSR: r2=0.40, regression: r2=0.41), specific leaf area (r2=0.67), leaf mass (r2=0.43, r2=0.38), and leaf length/width ratio (r2=0.62). Biochemical traits that play a role in the structural composition of vegetation, like lignin content (r2=0.98, r2=0.54), concentration (r2=0.45) and cellulose content (r2=0.57, r2=0.49), were found to be important by at least two of the analyses. Three other traits were important in at least one of the analyses: total biomass (r2=0.56), leaf C/N ratio (r2=0.99), and cellulose concentration (r2=0.76). Redundancy analysis suggests that there is a large percentage (52%) of the spectrum not explained by the PFTs measured in this study. However, spectral discrimination of functional groups, and even species, appears promising, mostly in the ultraviolet A part of the spectrum. This has interesting applications for mapping PFTs using remote sensing techniques, and therefore for estimating related ecosystem processes and services.