COS 267-4 - Don’t look up(scale): Peanut spectral reflectance enables sub-organ water mapping across species

Abstract: Plant spectroscopy is one of the banners of remote sensing of plant performance and ecosystem function across scales. Plant spectral reflectance contains information of the chemical and structural traits of species, thus being potentially correlated with biological mechanisms and processes in plants. However, the linkage between physiological processes and plant spectra is understudied and must be strengthened before upscaling of physiological processes can be accurately done. We pose that spatial downscaling of physiological processes from whole-leaf to tissue level using spectra may allow us to assess which physiological processes can truly be predicted from spectral reflectance, and which are correlated but not directly linked to plant spectra. Here, we develop partial least squares (PLSR) models using a point spectrometer to predict whole-leaf physiological status (e.g., water potential, relative water content, water fraction, equivalent water thickness, and wilting point) in peanut plants under drought stress. Models were applied to spatially explicit spectral images (450-950 nm) of peanut leaves and of two native Florida species (Frangula caroliniana and Magnolia grandiflora) obtained with drone-compatible spectral cameras. We assessed which spectra-predicted physiological states can map dehydration processes across a leaf regardless of species differences and spectral platform used.

All PLSR models achieved high accuracy (R2 >0.8) during testing and validation and identified key predictive wavelengths for the target physiological variables. However, only spectra-predicted equivalent water thickness (EWT) successfully predicted whole-leaf EWT (R2 >0.98, slope=1.2) when applied across species. Spectra-predicted water potential also correlated with whole-leaf water potential (R2 >0.96, slope = 0.013) and displayed meaningful spatial patterns, but predicted values were significantly shifted upwards relative to measured values. Mapped EWT and water potential showed an inwards pattern of dehydration from leaf tip to base. Our results indicate that spectra-predicted physiological status under drought is tightly linked to water content, and specifically to the amount of water per leaf area unit. Models based on relative water content could not be extrapolated to other species. Predicting shifts in water potential within leaves was possible but values were off, indicating that shifts in water potential are most likely correlated with shifts in EWT. As such, prediction of xylem tension or relative water content per se from spectra may not be possible across species or leaf morphologies. Our study suggests that ecologically relevant upscaling of physiological processes from leaf to landscape through spectra must be constrained to physical or chemical processes that interact with light.