LAUSR

Abstract Stomatal conductance is a critical regulating factor in plant water relations and responses to abiotic stress. Abscisic acid (ABA) is one of the plant hormones that regulates stomatal conductance and leaf transpiration. The presence of endogenous or exogenous ABA induces stomatal closure, which reduces leaf transpiration rates and increases tolerance to abiotic stress. In this study, visible near-infrared (Vis-NIR) spectroscopy, as well as proximal multispectral and thermal imaging were used to evaluate changes in stomatal conductance through exogenous ABA applications to apple trees. ABA was applied twice at 500 mg kg−1 in 2016, with five control and five ABA-treated trees in a three-year-old apple orchard. Proximal Vis-NIR spectral reflectance (350–2500 nm) data, and multispectral and thermal infrared images were acquired from control and treated trees after 1–3 days of exogenous ABA application to the trees. Ground reference stomatal conductance was also measured to compare the data with proximal sensing data. Partial least square regression (PLSR), linear support vector machines (SVM), and quadratic SVM algorithms were applied to classify the control and ABA-treated leaves, before and after feature selection using rank features technique and stepwise regression analysis. The average classification accuracy ranged between 80 and 85% at 3 days after treatment with the entire Vis-NIR spectra, while the accuracies ranged between 74 and 80% with five selected spectral bands. The ABA treatment effects could not be observed with crop water stress index extracted from thermal images, although the leaf temperature in ABA-treated trees were higher than the untreated control trees. Green normalized difference vegetation index extracted from multispectral images also did not show any differences between control and ABA-treated trees. Overall, results suggest that the hyperspectral Vis-NIR sensing was able to acquire spectral changes pertinent to the dynamic processes such as stomatal conductance, independent from non-responsive traditional vegetation indices that lacked responsive spectral bands.