Kaylah Vasquez, Guadalupe Partida, Eve Laroche-Pinel, and Luca Brillante*
*Department of Viticulture and Enology, California State University Fresno, 2360 E
Barstow Ave, Fresno, CA, 93740 (lucabrillante@csufresno.edu)

Remote sensing is a developing component of sustainable water management, particularly by identifying spectral and spatial information. In 2022, we began a study on Cabernet Sauvignon located in the San Joaquin Valley. We installed an automated irrigation system to execute variable irrigation across 48 watering zones, encompassing 12 irrigation regimes with four randomized replicates each. Irrigation schedules were a fraction of the grower control, which corresponded to the maximum water allocation received in the area. We collected spectral information across all zones throughout the growing season using an unmanned aircraft vehicle (UAV)-based hyperspectral imaging in the near infrared (NIR) – short-wave infrared (SWIR) from 900 to 1700 nm. Within this range of wavelengths are high-water absorption bands that were used in machine-learning regression models to predict plant water status. Plant water status measurements (stem water potential and leaf gas exchange) were taken as contemporarily as possible to the UAV flights every two weeks from June to harvest (five flights and ~1000 individual readings) to ground-truth spectral information. The collected measurements were averaged by experimental zone. Data analysis consisted of extraction of the pure canopy signal from the images using segmentation methods (accuracy >99%) and then averaging each experimental zone. This information was then used to train and predict water status measurement values using a random forest modeling approach. Recursive feature elimination was performed to reduce the number of predictors by 65%, to a total of 80 wavelengths. We obtained a 0.53 R2 and root means square error (RMSE) of 0.12 MPa in a 10-fold cross-validation routine. This project is a step toward developing new methods to precisely monitor and manage irrigation in vineyards.

Funding Support: American Vineyard Foundation; California State University – Agricultural Research Institute