Predicting grain yield using canopy hyperspectral reflectance in wheat breeding data

Abstract

BackgroundModern agriculture uses hyperspectral cameras to obtain hundreds of reflectance data measured at discrete narrow bands to cover the whole visible light spectrum and part of the infrared and ultraviolet light spectra, depending on the camera. This information is used to construct vegetation indices (VI) (e.g., green normalized difference vegetation index or GNDVI, simple ratio or SRa, etc.) which are used for the prediction of primary traits (e.g., biomass). However, these indices only use some bands and are cultivar-specific; therefore they lose considerable information and are not robust for all cultivars.ResultsThis study proposes models that use all available bands as predictors to increase prediction accuracy; we compared these approaches with eight conventional vegetation indexes (VIs) constructed using only some bands. The data set we used comes from CIMMYT’s global wheat program and comprises 1170 genotypes evaluated for grain yield (ton/ha) in five environments (Drought, Irrigated, EarlyHeat, Melgas and Reduced Irrigated); the reflectance data were measured in 250 discrete narrow bands ranging between 392 and 851 nm. The proposed models for the simultaneous analysis of all the bands were ordinal least square (OLS), Bayes B, principal components with Bayes B, functional B-spline, functional Fourier and functional partial least square. The results of these models were compared with the OLS performed using as predictors each of the eight VIs individually and combined.ConclusionsWe found that using all bands simultaneously increased prediction accuracy more than using VI alone. The Splines and Fourier models had the best prediction accuracy for each of the nine time-points under study. Combining image data collected at different time-points led to a small increase in prediction accuracy relative to models that use data from a single time-point. Also, using bands with heritabilities larger than 0.5 only in Drought as predictor variables showed improvements in prediction accuracy.

Increasing Predictive Ability by Modeling Interactions between Environments, Genotype and Canopy Coverage Image Data for Soybeans

“…Jarquin et al [10] utilized the reaction norm model for genomic prediction where the genetic and environmental values were replaced by the regression on the markers, and in the interaction between the markers and the environmental covariates, respectively. Dealing with high-throughput phenotypic information, several authors [11][12][13] have shown improvements in predictive ability with the inclusion of these sources of information in the models for wheat and maize. Montesinos-Lopez et al [14] showed that accounting for the band (hyper-spectral image data)-by-environment interaction also improved yield predictability in wheat when compared with those models that did not include this component in the models.…”

Aerial High-Throughput Phenotyping Enabling Indirect Selection for Grain Yield at the Early-generation Seed-limited Stages in Breeding Programs

“…While VIs alone may not show a sufficiently strong and consistent genetic correlation with a restricted selection index for GY and DTHD to provide significant increases in GS predictive ability, it is likely that breeders will have access to more extensive suites of phenotypic traits in addition to VIs as HTP systems become increasingly automated, customizable, and scalable. For example, hyperspectral reflectance phenotypes, which record spectral reflectance at a large range of wavelengths, were shown to increase prediction accuracy for GY when compared to VIs (Montesinos-López et al, 2017). Beyond spectral reflectance-related traits, deep learning algorithms have been developed to determine DTHD from proximal imagery of wheat canopies (Wang et al, 2019), UAV imagery has been used to estimate lodging in wheat as a function of changes in plant height throughout the growing season (Singh et al, 2019), and convolutional neural networks have been trained to identify foliar diseases in maize from aerial imagery (Wu et al, 2019).…”

Green-gradient based canopy segmentation: A multipurpose image mining model with potential use in crop phenotyping and canopy studies

“…in the local to international breeding programs (e.g. see Lopes et al, 2014;Mondal et al, 2016;Crespo-Herrera et al, 2017;Montesinos-López et al, 2017;and Lopes et al, 2018). Furthermore, when the number of total GSMcurve attributes were limited to the GSM-coverage series (more than 80% reduction in coefficients of exponential red or blue curves).…”