2.2.1 Collecting hyperspectral images
The field data collected in 2020 for validating GECROS were also
utilized to validate the methodology proposed in this study. Thus, to
obtain the remote sensing predictions, the hyperspectral imaging data
were collected in 2020 at the same date before each destructive sampling
(Table 3). Details about canopy hyperspectral reflectance measurements
were previously given (Wang et al., 2023). In brief, a DJI M600 PRO
hexacopter (DJI, Shenzhen, China), equipped with a Cubert S185
hyperspectral snapshot camera (Cubert GmbH, Ulm, Baden-Württemberg,
Germany), was flown over the experimental field between 10 a.m. and 2
p.m. We captured 125 spectral bands in the range of 450-950 nm with a
sampling interval of 4 nm. Compared with using reflectance only, the
feature set combining reflectance, vegetation indices and texture
information worked better when predicting leaf traits (Wang et al.,
2023). Thus, the complete dataset used for remote sensing prediction in
this study was consistent with the combined features from hyperspectral
image and targeted three leaf traits, includingW leaf, N leaf and LAI.