Speakers
Description
Anas Tallou, Vincenzo Giannico, Simone Garofalo, Salvatore Camposeo, Giuseppe Lopriore, Gaetano Alessandro Vivaldi
Abstract
Plant water status is an important factor that needs an accurate temporal and spatial assessment to maintain acceptable yield and quality standards in a changing climate. Integrating remote sensing (RS) technology to estimate stem water potential (SWP, Ψstem) can provide good alternatives to traditional in situ plant water status measurements. Hence, field measurements of Ψstem were taken during two consecutive growing seasons from an irrigated vineyard and olive orchard (Apulia region, Southern Italy) at the time of image acquisition. Multispectral reflectance data from PlanetScope and Sentinel II sensors, corresponding to different reflectance values of the grapevine and olive tree samples’ spectral bands (PBs), were recorded at Ψstem measurements and used to calculate vegetation indices (VIs). Two machine learning (Random Forest and Support Vector Machine) and one Multiple Linear Regression technique (MLR), were used to develop the grapevine and olive SWP prediction model and test their accuracy, independently considering PBs and VIs as predictors. The aim was to develop a robust prediction model of plant water status for sustainable water management. Our results indicate that the use of the RS reflectance values of grapevine and olive trees to develop a Ψstem prediction model using a machine-learning technique can provide a reliable option for plant water status estimation.
