Combining biophysical parameters, spectral indices and multivariate hyperspectral models for estimating yield and water productivity of spring wheat across different agronomic practices

Abstract

Manipulating plant densities under different irrigation rates can have a significant impact on grain yield and water use efficiency by exerting positive or negative effects on ET. Whereas traditional spectral reflectance indices (SRIs) have been used to assess biophysical parameters and yield, the potential of multivariate models has little been investigated to estimate these parameters under multiple agronomic practices. Therefore, both simple indices and multivariate models (partial least square regression (PLSR) and support vector machines (SVR)) obtained from hyperspectral reflectance data were compared for their applicability for assessing the biophysical parameters in a field experiment involving different combinations of three irrigation rates (1.00, 0.75, and 0.50 ET) and five plant densities (D 1 : 150, D 2 : 250, D 3 : 350, D 4 : 450, and D 5 : 550 seeds m -2 ) in order to improve productivity and water use efficiency of wheat. Results show that the highest values for green leaf area, aboveground biomass, and grain yield were obtained from the combination of D 3 or D 4 with 1.00 ET, while the combination of 0.75 ET and D 3 was the best treatment for achieving the highest values for water use efficiency. Wheat yield response factor (ky) was acceptable when the 0.75 ET was combined with D 2 , D 3 , or D 4 or when the 0.50 ET was combined with D 2 or D 3 , as the ky values of these combinations were less than or around one. The production function indicated that about 75% grain yield variation could be attributed to the variation in seasonal ET. Results also show that the performance of the SRIs fluctuated when regressions were analyzed for each irrigation rate or plant density specifically, or when the data of all irrigation rates or plant densities were combined. Most of the SRIs failed to assess biophysical parameters under specific irrigation rates and some specific plant densities, but performance improved substantially for combined data of irrigation rates and some specific plant densities. PLSR and SVR produced more accurate estimations of biophysical parameters than SRIs under specific irrigation rates and plant densities. In conclusion, hyperspectral data are useful for predicting and monitoring yield and water productivity of spring wheat across multiple agronomic practices.

Comparative Performance of Spectral Reflectance Indices and Multivariate Modeling for Assessing Agronomic Parameters in Advanced Spring Wheat Lines Under Two Contrasting Irrigation Regimes

“…The red-edge region was often used as an indirect stress indicator, especially when the plants suffer stress. This region carries important information about biomass quantity and leaf area index, and therefore, could be used to distinguish plant health and yield (Smith et al, 2004; Gitelson et al, 2011; El-Hendawy et al, 2019b). The five wavelengths extracted within this region were associated with DW and GY under LM and FL+LM ( Table 2 ).…”

“…Several published SRIs have been used to successfully estimate different parameters such as aboveground biomass and water content, leaf area index, gas exchange and transpiration rates, stomatal conductance, ion and pigment contents, carbon isotope discrimination, yield components, and grain yield in several field crops under either normal or abiotic stress conditions (Erdle et al, 2013; Li et al, 2014; Lobos et al, 2014; El-Hendawy et al, 2015; Bayat et al, 2016; Becker and Schmidhalter, 2017; Garriga et al, 2017; Kawamura et al, 2018; El-Hendawy et al, 2019a; El-Hendawy et al, 2019b). For example, in diverse studies, several SRIs, which are related to plant biomass, plant water status, and plant photosynthetic efficiency, such as the green normalized difference vegetation index (GNDVI), normalized difference vegetation indices (NDVIs), SRIs related to normalized water indices (NWI-1, NWI-2, NWI-3, and NWI-4), and normalized difference moisture index (NDMI: 2200; 1100) showed significant correlation with final grain yield and explained more than 70% of yield variability under contrasting water irrigation regimes (Shanahan et al, 2001; Aparicio et al, 2002; Prasad et al, 2007; Lobos et al, 2014; Elazab et al, 2015; El-Hendawy et al, 2017a).…”

Digital image processing technology under backpropagation neural network and K-Means Clustering algorithm on nitrogen utilization rate of Chinese cabbages

“…The color feature is a common feature in image analysis. Compared with other features, it is very robust to image rotation changes and translation [ 21 ]. The morphological feature is an important indicator to describe the image.…”

Potential of Hyperspectral and Thermal Proximal Sensing for Estimating Growth Performance and Yield of Soybean Exposed to Different Drip Irrigation Regimes Under Arid Conditions

“…Due to the close relationship between specific wavelengths in these three parts of the electromagnetic spectrum (VIS, NIR, and SWIR) and the different biophysical and biochemical characteristics of the canopy, these specific wavelengths have been exploited to calculate specific spectral reflectance indices (SRIs) using simple mathematical operations (normalized or ratio formulas). These SIRs have been employed in estimating and monitoring several plant phenotypic traits such as aboveground dry and fresh biomass, grain yield and its components, pigment contents, and most plant measurements that reflect the status of plant water [ 6 , 37 , 43 , 45 , 46 , 47 , 48 ].…”