Remote sensed images for prediction of Stem Water Potential in olive groves

Accurate assessment of plant water status is essential for ensuring optimal crop yield and quality, particularly in arid and semi-arid regions. Stem water potential (Ψ-stem) is a reliable parameter of plant water status in several fruit orchards, including olive trees. However, traditional in situ measurements of Ψ-stem are labor-intensive, time-consuming, and destructive, allowing only limited samples and repetitions. Remote sensing technologies could be a promising alternative to conventional methods for estimating Ψ-stem. In this study, we conducted field measurements of midday Ψ-stem in an irrigated olive orchard located in southern Italy during the 2024 growing season, between June and October. Multispectral remotely sensed reflectance data were also collected simultaneously with the time of Ψ-stem in-situ measurements, acquired by Next-Generation PlanetScope, a new generation of spatial and time high-resolution satellites. Successively, 37 vegetation indices (VI) were computed from the PlanetScope images, and they were tested as proxies for Ψ-stem. A Random Forest (RF) model was used to perform multivariable regression, providing an accurate prediction of olive water status. The correlation coefficient between in-situ measured Ψ-stem and predicted Ψ-stem of the RF model is equal to R2 = 0.91 in training and R2 = 0.73 in test. Moreover, RF has given information about the importance of each VI in the regression estimates, furnishing useful insights about VI more involved in the Ψ-stem estimation.