Performance of optimized hyperspectral reflectance indices and partial least squares regression for estimating the chlorophyll fluorescence and grain yield of wheat grown in simulated saline field conditions.

To overcome the salinity threats to crop production in arid conditions, wheat cultivars should be developed with better performance with regard to key physiological traits. Although different chlorophyll fluorescence (ChlF) parameters, such as maximum quantum PSII photochemical efficiency (Fv/Fm), quantum yield of PSII (ΦPSII), and non-photochemical quenching (NPQ) have been proven to be key physiological traits to improve salt tolerance, their evaluation is time-consuming. In this study, hyperspectral canopy reflectance was used to assess ChlF parameters and grain yield (GY) of two wheat cultivars growing in simulated saline field conditions and exposed to three salinity levels (control, 6.0 dS m-1, and 12.0 dS m-1). Different spectral reflectance indices (SRIs) were formulated as ratios based on contour maps and tested for their relationship with ChlF parameters. The performance of individual SRIs and partial least squares regression (PLSR) models based on ChlF parameters, all examined SRIs, or data fusion of combined ChlF and SRIs to estimate the GY was considered. All examined SRIs failed to assess ΦPSII and NPQ under control condition, but most of them showed a moderate to strong relationship with both parameters under the salinity levels of 6.0 and 12.0 dS m-1. The examined SRIs showed a moderate and strong relationship with Fv/Fm under conditions of 6.0 and 12.0 dS m-1, respectively. Most SRIs correlated better with the three ChlF parameters for the salt-sensitive cultivar Sakha 61 than for the salt-tolerant cultivar Sakha 93. Several SRIs exhibited strong relationships with GY under the salinity levels of 6.0 and 12.0 dS m-1 and for both cultivars. Overall, the PLSR models exhibited additional improvements for estimating and predicting GY in both calibration and validation datasets over that using individual SRIs. The PLSR model based on data fusion was the best model to accurately estimate GY in the validation model even under control conditions. This study, of a type rarely conducted in simulated saline field conditions, indicates that the ChlF parameters could be linked to hyperspectral reflectance data for the rapid and non-destructive assessment of photosynthetic status and prediction of wheat production under salt stress field conditions.