UAV-derived plant height estimates enhance indirect selection for grain yield in maize under contrasting nitrogen availability

This study aimed to evaluate the potential of UAV-based plant height estimation to identify high-yielding maize genotypes under contrasting nitrogen availability. The experiment was conducted using a RCBD in a split-plot arrangement, with two nitrogen levels and 47 genotypes. Plant height was estimated from digital surface models at four phenological stages and compared with field-measured plant height at the VT stage. The estimates were evaluated based on bias, and Pearson correlation analysis was performed between estimated and field-measured plant height. Joint selection was performed using a 20% selection intensity based on estimated plant height and grain yield. At the VT stage, R² values were 0.44 and 0.64, with bias values of −0.35 and −0.78 under high- and low-nitrogen conditions, respectively. The q95 metric showed correlations ranging from 0.48 to 0.71 and from 0.57 to 0.84 with field-measured plant height, and from 0.11 to 0.36 and from 0.30 to 0.65 with grain yield, under high- and low-nitrogen conditions, respectively. Joint selection (20% selection intensity) identified genotypes consistently selected by both criteria, particularly under low nitrogen availability. RGB imagery and the q95 metric demonstrated potential for high-throughput phenotyping and as an indirect selection criterion for high-yielding maize genotypes, particularly under nitrogen-limited conditions.