Spectral variables as criteria for selection of soybean genotypes at different vegetative stages

Abstract

Understanding the association between variables obtained by multispectral imaging, such as plant canopy reflectance at different wavelengths and vegetation indices (VIs), and agronomic traits of interest for soybean is crucial for breeding programs. This approach, associated with the evaluation of the best stage for spectral data collection, can identify easier-to-measure variables to be used in the indirect selection of genotypes for more expensive-to-measure traits or those that are expressed later, resulting in a faster, labor-saving, and large-scale selection process. The hypothesis of this study is the possibility of identifying promising variables for the selection of soybean genotypes through the use of vis and that different times of collection of information regarding reflectance interfere in the selection responses. Thus, our objectives were: i) to estimate heritability (h2) for agronomic traits and VIs; ii) to estimate the gains with direct selection (DS) on the VIs and indirect selection (IS) for the agronomic traits selected by the higher heritability; iii) to estimate correlations between spectral variables, VIs, and agronomic traits; iv) to identify the most promising populations for the evaluated variables; and vi) to identify the most adequate phenological stage for spectral data collection. We installed a field experiment in a randomized block design with four repetitions, 28 F3 soybean populations, and four control treatments. Agronomic traits evaluated were: days to maturity (DM), plant height (PH), first pod insertion height (PIH), number of branches (NB), and grain yield (GY). Spectral collections were performed at the V8 (45 days after emergence), R1 (60 days after emergence) and R5 (80 days after emergence) stages. Correlations between agronomic traits, reflectance at 790 nm (near-infrared), 735 nm (RedEdge), 660 nm (red), 550 nm (green), and 450 nm (blue) wavelengths, and VIs were estimated. The heritability were estimated for agronomic traits and VIs. Our findings reveal that it is possible to select soybean genotypes using spectral variables based on heritability and phenotypic aspects. Enhanced vegetation index (EVI) and soil-adjusted vegetation index (SAVI) have high heritability values (>75%) at V8. It is possible to obtain satisfactory selection gains for agronomic traits, especially GY, based on selection on EVI and SAVI at V8 stage. By this approach, the F3 segregating populations that provided the highest selection gains were P7, P8, P9, P14, P27, and P28. Spectral data collection at V8 stage provided higher h2 for VIs and agronomic traits, while R5 is the most suitable stage for estimating associations between VIs and agronomic traits.