Fluorescence, temperature and narrow-band indices acquired from a UAV platform for water stress detection using a micro-hyperspectral imager and a thermal camera

The remote detection of water stress in a citrus orchard was investigated using leaf-level measurements of chlorophyll fluorescence and Photochemical Reflectance Index (PRI) data, seasonal time-series of crown temperature and PRI, and high-resolution airborne imagery. The work was conducted in an orchard where a regulated deficit irrigation (RDI) experiment generated a gradient in water stress levels. Stomatal conductance (Gs) and water potential (Ψ) were measured over the season on each treatment block. The airborne data consisted on thermal and hyperspectral imagery acquired at the time of maximum stress differences among treatments, prior to the re-watering phase, using a miniaturized thermal camera and a micro-hyperspectral imager on board an unmanned aerial vehicle (UAV). The hyperspectral imagery was acquired at 40 cm resolution and 260 spectral bands in the 400-885 nm spectral range at 6.4 nm full width at half maximum (FWHM) spectral resolution and 1.85 nm sampling interval, enabling the identification of pure crowns for extracting radiance and reflectance hyperspectral spectra from each tree. The FluorMOD model was used to investigate the retrieval of chlorophyll fluorescence by applying the Fraunhofer Line Depth (FLD) principle using three spectral bands (FLD3), which demonstrated that fluorescence retrieval was feasible with the configuration of the UAV micro-hyperspectral instrument flown over the orchard. Results demonstrated the link between seasonal PRI and crown temperature acquired from instrumented trees and field measurements of stomatal conductance and water potential. The sensitivity of PRI and Tc–Ta time-series to water stress levels demonstrated a time delay of PRI vs Tc–Ta during the recovery phase after re-watering started. At the time of the maximum stress difference among treatment blocks, the airborne imagery acquired from the UAV platform demonstrated that the crown temperature yielded the best coefficient of determination for Gs (r2 = 0.78; p < 0.05) and Ψ (r2 = 0.34; p < 0.001). Among the narrow-band indices calculated, the PRI515 index (reference band = 515 nm) obtained better results than PRI570, with r2 = 0.59 (p < 0.01) for Gs, and r2 = 0.38 (p < 0.001) for Ψ. The BGI1 index calculated from the blue (R400) and green (R550) bands resulted on the highest significance levels (p < 0.001) for both Gs (r2 = 0.62) and Ψ (r2 = 0.49). Out of the structural indices assessed, RDVI, MTVI1 and TVI showed greater sensitivity for Gs (r2 = 0.6; p < 0.01) and Ψ (p < 0.001) than NDVI. Chlorophyll fluorescence calculated from the micro-hyperspectral imagery with the FLD3 method tracked stress levels, obtaining r2 = 0.67 (p < 0.05) with stomatal conductance, and r2 = 0.66 (p < 0.001) with water potential. The work presented in this manuscript demonstrates the feasibility of thermal, narrow-band indices and fluorescence retrievals obtained from a micro-hyperspectral imager and a light-weight thermal camera on board small UAV platforms for stress detection in a heterogeneous tree canopy where very high resolution is required.

► Water stress was investigated using a micro-hyperspectral imager on board an UAV.
► A link between PRI and crown temperature was demonstrated over the entire season.
► Seasonal variation of PRI and Tc-Ta demonstrated a time delay of PRI during recovery.
► Best indicators of stress were crown Tc, F (FLD method) and the BGI1 index.
► It demonstrates the F retrieval using micro-hyperspectral imagers on board UAVs.