Field-based high-throughput phenotyping enhances phenomic and genomic predictions for grain yield and plant height across years in maize

Abstract

Field-based phenomic prediction employs novel features, like vegetation indices (VIs) from drone images, to predict key agronomic traits in maize, despite challenges in matching biomarker measurement time points across years or environments. This study utilized functional principal component analysis (FPCA) to summarize the variation of temporal VIs, uniquely allowing the integration of this data into phenomic prediction models tested across multiple years (2018-2021) and environments. The models, which included 1 genomic, 2 phenomic, 2 multikernel, and 1 multitrait type, were evaluated in 4 prediction scenarios (CV2, CV1, CV0, and CV00), relevant for plant breeding programs, assessing both tested and untested genotypes in observed and unobserved environments. Two hybrid populations (415 and 220 hybrids) demonstrated the visible atmospherically resistant index's strong temporal correlation with grain yield (up to 0.59) and plant height. The first 2 FPCAs explained 59.3 ± 13.9% and 74.2 ± 9.0% of the temporal variation of temporal data of VIs, respectively, facilitating predictions where flight times varied. Phenomic data, particularly when combined with genomic data, often were comparable to or numerically exceeded the base genomic model in prediction accuracy, particularly for grain yield in untested hybrids, although no significant differences in these models' performance were consistently observed. Overall, this approach underscores the effectiveness of FPCA and combined models in enhancing the prediction of grain yield and plant height across environments and diverse agricultural settings.

Keywords: UAV; field-based high-throughput phenotyping; functional principal component analysis; genomic prediction; grain yield; maize breeding; multikernel prediction; phenomic prediction; plant height.

Fig. 1.

Flight times were illustrated as DAP unit for Pop1 and Pop2 grown in 2018–2019 and 2020–2021. The shading areas highlights the range of flowering times of plots in these trials.

Fig. 2.

Model training scenarios, involving both tested and untested genotypes and environments, along with prediction scenarios. In the upper part of the figure, models were trained using a 4-fold genotype (represented by 4 orange blocks) in environments A and B. Subsequently, predictions were made for both tested and untested genotypes in both environments. The prediction scenarios CV2 and CV1 were then computed. In the lower part of the figure, the models were trained using the same 4-fold genotype, but exclusively in environment A (considered the tested environment). Following this training, predictions were generated for both tested and untested genotypes in environment B (considered the untested environment). The prediction scenarios CV0 and CV00 were then computed. A secondary trait (VARI FPCA1) was consistently present in the training set for M6 (multitrait prediction) across all prediction scenarios. Tested and untested hybrids were the same across all prediction scenarios (CV2, CV1, CV0, and CV00).

Fig. 3.

a) Explained percentage variation accounted for by the variance components in Equation (1), along with the heritabilities for plant height (in cm) and GY (in t/ha). b) Grain yield (t/ha) and plant height (cm) data range for Pop1 and Pop2 across 2018–2021. The x axis separates the growing trial and years for Pop1 and Pop2.

Fig. 4.

Temporal correlation results were illustrated between temporal values of each of 36 VIs and GY (below) and plant height (above) for each trial/year. Whiskers are the standard deviations of temporal correlations.

Fig. 5.

Temporal trajectories of VARI VI for Pop1 grown in 2018 and 2019 and Pop2 grown in drought and optimal trials in 2020 and drought trial in 2021. y axis is the value of the VARI; x axis is the days after planting.

Fig. 6.

Explained percent variation by FPCA 1 and FPCA2 for temporal data of each of 36 VIs in each environment. On the left, FPCA1 and FPCA2 were obtained using temporal data of VIs belonging to all flights. On the right, FPCA1 and FPCA2 were obtained using temporal data of VIs belonging to planting to until flowering times.

Fig. 7.

Prediction ability of the 6 prediction models for GY and plant height for Pop1 and Pop2. Bar plots represent the mean of prediction abilities along with the standard deviations (whiskers). M1 is the genomic prediction model. M2 and M3 are phenomic prediction models that use flight period data from planting to flowering and all flights, respectively. M4 and M5 are multikernel models that utilize M1 + M3 and M1 + M2 predictions, respectively. M6 is the multitrait (phenotype + VARI FPCA1) prediction model.