Hyperspectral response characteristics and correlation analysis of grape leaf tissue structure

Objectives The spectral response of leaf tissue structure in different grape cultivars was investigated. The study aimed to identify the main factors responsible for spectral differences in grape leaves and advance spectral diagnosis accuracy in leaf nutrition.

Methods The experiment was conducted in Langfang vineyard, Hebei Province. The leaves of four grape cultivars, Summer Black, Italia, Ruby Seedless and Autumn Black, were sampled regularly during the study period. The spectral data and N content were determined synchronously using FieldSpec FR2500 spectrometer and chemical method, respectively. The leaf tissue structures were observed and measured using scanning electron microscopy (SU8010) cryopreservation technology. The correlation between the spectrum and tissue structure of the leaves was calculated.

Results The spectral difference between the leaf′s front and back was caused by the distribution of stomata in the reverse side of the grape leaves. In the visible band, the leaf back′s spectral reflectance was higher than that of the leaf front. However, it is the opposite in the near-infrared band, and the leaf front's spectral reflectance was higher than that of the leaf back. The cultivars exhibited differences in spectral reflectance when leaf N content was similar due to variation in the number and distribution of stomata on the leaf surface, the thickness of palisade tissue cells, and the thickness of sponge tissues. The correlation between the red edge parameters λ_red of the front (i.e., the wavelength at which the first differential value of spectral reflectance reaches the maximum range in 660–770 nm) and the leaf thickness of the different varieties was significantly different. In addition, there was a strong correlation between the spectral red edge parameters and other leaf structural parameters, among which the thickness of palisade tissue and spongy tissue was a non-negligible factor in the spectral response of the grape varieties.

Conclusions The differences in cell morphology on the front and back surfaces of the leaves, internal structure of leaves and their correlation with spectral characteristics are determined. We find that the red edge parameter λ_red can show the leaf thickness of different grape varieties. If the variety factor is considered, some parameters with better correlation can also be selected for each variety. These results provided a basis for establishing and optimising the leaf nutrient diagnosis spectral model in grape. For improving the accuracy of the leaf nitrogen diagnostic model, the influence of leaf structure should be considered when using the spectral technique to diagnose leaf nutrition.