Assessment of Water Content in Plant Leaves Using Hyperspectral Remote Sensing and Chemometrics, Application: Rosmarinus officinalis

Leaf water content can provide useful information about the leaf’s physiological health. The objective of this study is to demonstrate that water supply and leaf wetness have an impact on the spectral response of plants with an application to the medicinal and aromatic plant species Rosmarinus officinalis. Using hyperspectral remote sensing, our theory is that any meaningful difference in leaf spectral signature, taken under different watering conditions, is a consequence of the amount of water in the plant’s leaves.

Watering was spaced 24 h apart, and then, spectral data were collected 24 h after each watering in the field using the spectroradiometer with a wavelength range from 325 to 1075 nm. Leaf water content has been assessed and predicted by visible-near-infrared spectroscopy and partial least squares regression (PLSR) models.

The results show that watering has an influence on the spectral responses of Rosmarinus officinalis. The reflectance becomes higher after 24 h, and then, it stabilizes 72 h after the first watering. For the external validation, the preprocessing techniques combining standard normal variate and Savitzky-Golay filter showed the best model with high leaf water content prediction performance based on the highest coefficient of regression R² = 0.959 and RPD = 4.97 and lowest root-mean-square error for validation RMSECV = 0.036.

The narrow spectral bands found to be correlated with water are 628, 629, 785, 927, and 928 nm.