Advancing Vineyard Irrigation in the San Joaquin Valley, California, with Hyperspectral-Based Plant Water Status Mapping
Eve Laroche-Pinel, Kaylah Rachel Vasquez, and
Luca Brillante*
*Department of Viticulture and Enology, California State
University Fresno, 2360 E Barstow Ave, Fresno, CA, 93740
(lucabrillante@csufresno.edu)
Remote sensing has emerged as a valuable tool to enhance the precision of water supply budgeting using both spectral and spatial data. A study conducted in a Vitis vinifera L. cv. Cabernet Sauvignon vineyard in the San Joaquin Valley, California used a variable rate automated irrigation system across 12 distinct water regimes replicated in four randomized sets, resulting in 48 experimental zones. The primary objective of this experimental setup was to introduce variability in grapevine water status, facilitating the generation of a comprehensive data set for modeling purposes. Over the course of the growing season, spectral data within these zones were collected using a near infrared (NIR) – short wavelength infrared (SWIR) hyperspectral camera (ranging from 900 to 1700 nm) mounted on an unmanned aircraft vehicle. Given the significant water absorption characteristics within this spectral range, the sensor was deployed to evaluate grapevine stem water potential, Ψstem, a standard metric for assessing plant water status, from pure grapevine pixels in the hyperspectral images. Concurrently, Ψstem values were measured in the field from bunch closure to harvest and subsequently modeled using machine-learning techniques, leveraging the remotely sensed NIR-SWIR data as predictors in both regression and classification frameworks (with classes representing varying levels of physiological water stress). Hyperspectral images underwent conversion to bottom-of-atmosphere reflectance using standard ground panels and the quick atmospheric correction method, with ensuing results subjected to comparative analysis. The most effective models used ground panel-derived data and predicted Ψstem values with an R2 of 0.54 and an RMSE of 0.11 MPa, as estimated through cross-validation. Additionally, the optimal classification approach achieved an accuracy rate of 74%. This endeavor aims to pioneer novel methods for precise monitoring and management of irrigation in vineyards, while concurrently furnishing valuable insights into the physiological responses of plants to deficit irrigation practices.
Funding Support: American Vineyard Foundation, CSU-Agriculture Research Institute
