Spatio-temporal modeling of high-throughput multi-spectral aerial images improves agronomic trait genomic prediction in hybrid maize

Authors: MORALES, NICHOLAS - Cornell University; ANCHE, MAHLET - Cornell University; KACZMAR, NICHOLAS - Cornell University; Lepak, Nicholas; NI, PENGZUN - Cornell University; ROMAY, MARIA CINTA - Cornell University; SANTANTONIO, NICHOLAS - Cornell University; Buckler, Edward - Ed; GORE, MICHAEL - Cornell University; MUELLER, LUKAS - Boyce Thompson Institute; ROBBINS, KELLY - Cornell University

Submitted to: bioRxiv
Publication Type: Pre-print Publication
Publication Acceptance Date: 10/21/2022
Publication Date: 10/21/2022
Citation: Morales, N., Anche, M.T., Kaczmar, N., Lepak, N.K., Ni, P., Romay, M., Santantonio, N., Buckler IV, E.S., Gore, M.A., Mueller, L.A., Robbins, K.R. 2022. Spatio-temporal modeling of high-throughput multi-spectral aerial images improves agronomic trait genomic prediction in hybrid maize. bioRxiv. https://doi.org/10.1101/2022.10.18.512728.
DOI: https://doi.org/10.1101/2022.10.18.512728

Interpretive Summary: This study aimed to improve the understanding of how different factors influence crop growth in agricultural field experiments. The researchers used advanced techniques, including high-resolution aerial imagery captured by drones, to measure various aspects of the crops. They focused on a parameter called the normalized difference vegetation index (NDVI), which indicates plant health and growth. To analyze the data, the researchers developed a two-stage approach. In the first stage, they separated the effects of environmental factors from the genetic influences on crop growth over the entire growing season. This allowed them to understand how the environment impacted the crops separately from the inherent genetic traits. In the second stage, they combined the information about environmental effects with genetic data to create more accurate predictions of crop performance. They tested their approach using both computer-generated simulation data and real-world maize field experiments conducted in different years.

Technical Abstract: Design randomizations and spatial corrections have increased understanding of genotypic, spatial, and residual effects in field experiments, but precisely measuring spatial heterogeneity in the field remains a challenge. To this end, our study evaluated approaches to improve spatial modeling using high-throughput phenotypes (HTP) via unoccupied aerial vehicle (UAV) imagery. The normalized difference vegetation index (NDVI) was measured by a multi-spectral MicaSense camera and ImageBreed. Contrasting to baseline agronomic trait spatial correction and a baseline multi-trait model, a two-stage approach that quantified NDVI local environmental effects (NLEE) was proposed. Firstly, NLEE were separated from additive genetic effects over the growing season using two-dimensional spline (2DSpl), separable autoregressive (AR1) models, or random regression models (RR). Secondly, the NLEE were leveraged within agronomic trait genomic best linear unbiased prediction (GBLUP) either modeling an empirical covariance for random effects, or by modeling fixed effects as an average of NLEE across time or split among three growth phases. Modeling approaches were tested using simulation data and Genomes-to-Fields (G2F) hybrid maize (Zea mays L.) field experiments in 2015, 2017, 2019, and 2020 for grain yield, grain moisture, and ear height. The two-stage approach improved heritability, model fit, and genotypic effect estimation compared to all baseline models. Electrical conductance and elevation from a 2019 soil survey significantly improved model fit, while 2DSpl NLEE were most correlated to the soil parameters and grain yield 2DSpl effects. Simulation of field effects demonstrated improved specificity for RR models. In summary, NLEE increased experimental accuracy and understanding of field spatio-temporal heterogeneity.