Potential of the existing and novel spectral reflectance indices for estimating the leaf water status and grain yield of spring wheat exposed to different irrigation rates

Abstract

Hyperspectral sensing technique can provide an exact and expeditious manner for effective management of deficit irrigation through assessment of changes in plant water status in a large scale and real-time. This hypothesis was tested in this study using the hyperspectral signatures of the canopy in the visible- (VIS), near- (NIR), and shortwave-infrared (SWIR) to estimate and predict the leaf water status, measured here in terms of leaf water potential (LWP), relative water content (RWC), and equivalent water thickness (EWT), and grain yield (GY) of wheat cultivars exposed to 1.00, 0.75, and 0.50 of the estimated evapotranspiration (ETc). Results showed that the three parameters of leaf water status exhibited strong correlations with GY under 0.75 (except LWP) and 0.50 ETc treatments. This indicates that these parameters can be used as early indicators for management of deficit irrigation. Based on the relationships between these four phenotypic parameters and original canopy spectral reflectance within 350–2500 nm, the sensitive spectral wavelengths that exhibited strong correlations with all parameters existed mainly within the NIR and SWIR regions, with peak-wavelengths around 351, 518, and 687 nm in the VIS, 762, 974, 1100, and 1240 nm in the NIR, and 1392, 1515, 1930, and 2273 nm in the SWIR regions. These peak-wavelengths were used to build new two- and three-band normalized spectral reflectance indices (NDSIs). The NDSIs that combine NIR and VIS, NIR and NIR, SWIR and NIR, and SWIR and SWIR wavelengths were more effective for tracking changes in leaf water status and GY than those that combine only VIS wavelengths. The high fit between the observed and predicted values for phenotypic parameters based on twelve newly developed and published NDSIs indicates that the most recently developed NDSI models were more precise and accurate, and thus could be used for monitoring the changes in leaf water status and wheat production caused by deficit irrigation. The performance of partial least square regression (PLSR) based on either eleven wavelengths or different NDSIs as a predictive approach was the same and sometimes better than the individual NDSIs for assessment of phenotypic parameters. The results of spectral reflectance data and PLSR tools can serve as rapid and non-destructive alternative approaches for monitoring the water status and wheat production, and can be used to develop certain spectral indices for management of deficit irrigation in arid regions.