Optical specifications for a proximal sensing approach to monitor the vine water status in a distributed and autonomous fashion

In agriculture, increasing attention is being paid to collect data in a non-destructive way using optical systems which can be field distributed in a completely interconnected network. To improve the irrigation scheduling management, the control of the plant's water status is crucial. This work focused on the definition of optical specifications (wavelength-selection in vis/NIR region) for the development of cost-effective sensors, giving an initial bulk of information to design optical devices to be used in a network of distributed field sensors. The analyses were performed on vines of cv. Pinot Blanc. Optical data were collected on leaves before the analysis of water potential and moisture content. Pearson-correlation analysis between predawn water potential (ѰPD) and moisture content was performed (r = 0.47 and p-value<0.05) highlighting a non-highly correlation between the two parameters. The optical data (350–2500 nm) were used to build a PLS-model with vis/NIR and ѰPD (RMSEP = 0.056 MPa, R2 = 0.7). The study identified the most significant wavelengths related to the water potential at the leaf level to design a chemometric model that was compared to the model based on the whole spectra. Therefore, related VIP-scores were used to calibrate another PLS-model after the selection of most relevant optical bands (530 ± 20 nm, 700 ± 20 nm, and 1400 ± 20 nm). Good predictive performance was obtained with an RMSEP = 0.056 and an R2 = 0.60. These results paved the ground for further development of integrated optical sensors capable to monitor vine water status in the field in a distributed and autonomous fashion.