Heritability, heterosis, and hybrid/inbred classification ability of maize leaf hyperspectral signals under changing soil nitrogen

Abstract This study investigates the use of leaf hyperspectral data to understand genetic and environmental influences on maize ( Zea mays L.) leaf reflectance and its implications for genetic analysis. The Backcrossed Germplasm Enhancement of Maize (BGEM) panel was grown under two nitrogen regimes, low nitrogen (LN) and high nitrogen (HN), at the University of Nebraska‐Lincoln's Havelock Farm in 2022. Hyperspectral reflectance data were collected using Analytical Spectral Device (ASD) FieldSpec 4 spectroradiometers. Statistical analyses revealed significant genetic and environmental contributions to leaf reflectance, with nitrogen treatments driving substantial variation, particularly in the visible (VIS) spectrum. Wavelengths around 550 and 710 nm showed high sensitivity to nitrogen levels, with reflectance increasing under LN conditions. Leaf reflectance traits demonstrated moderate to high heritability, especially in the VIS and shortwave infrared (SWIR) regions. Six heterotic hotspots were identified along the spectrum, showing relatively high mid‐parent heterosis (MPH). Machine learning models were tested for inbred/hybrid classification based on hyperspectral data, with logistic regression (LOGREG) achieving the highest generalization accuracy (0.60) on independent datasets. Models trained on HN data performed better overall. This research opens avenues for leveraging hyperspectral data in breeding programs and genetic studies. Further work, including genome‐wide association studies (GWAS), is needed to determine the genetic basis of specific wavelengths and their role in heterosis.