Predicting grapevine stem water potential (Ψ stem ) using UV ‐visible‐ NIR spectroscopy across contrasting vineyard environments

BACKGROUND Climate change is intensifying water scarcity in viticulture, increasing the need for rapid and reliable tools to monitor vine water status. Spectroscopic techniques have emerged as promising approaches for rapid assessment of plant water status, although their robustness across different environments remains a key challenge. This study aimed to predict grapevine stem water potential (Ψ stem ) using UV‐visible (UV‐Vis) and near‐infrared (NIR) spectroscopy (190–2600 nm), and to evaluate the robustness and transferability of predictive models across grapevine varieties, vineyards in Galicia (Spain), growing seasons and contrasting climatic conditions. Leaf reflectance spectra (UV‐Vis‐NIR) and Ψ stem measurements were obtained from 193 dried leaf samples collected from five vineyards representing five grapevine cultivars over three growing seasons under contrasting environmental conditions and analysed in the laboratory. Partial least squares (PLS) regression models were subsequently developed using different preprocessing techniques. RESULTS The best‐performing global model, based on spectroscopic preprocessing over the full spectral range, achieved R 2 val = 0.85 and root‐mean‐square error of prediction = 0.17 MPa, demonstrating strong predictive capacity across varieties and sites. However, performance decreased in specific vineyards (Leiro; R 2 val = 0.37), highlighting the influence of local environmental and management conditions. Region‐specific models showed variable performance, with Quiroga providing the highest accuracy ( R 2 cval = 0.84; square error of cross‐validation = 0.13 MPa). Exploratory analyses revealed consistent relationships between Ψ stem , climatic variables and spectral responses, and model outputs were successfully translated into agronomically meaningful water status classes. CONCLUSIONS UV‐Vis‐NIR spectroscopy combined with PLS provides a rapid and reliable approach for estimating Ψ stem and supports precision irrigation. Although global models show strong overall performance, site‐specific calibration may improve accuracy under heterogeneous environmental conditions.