Genetic dissection of dynamic leaf area index variation in maize using UAV-based phenotyping and time-series genome-wide association studies

Abstract

Abstract The leaf area index (LAI) is a key determinant of canopy architecture and yield potential in maize, primarily through its influence on photosynthetic efficiency. Although unmanned aerial vehicle (UAV) technology has greatly advanced field-based phenotyping, its potential for deciphering the genetic mechanisms underlying dynamic and complex trait development remains underexplored. In this study, multispectral UAV images were collected from a diverse maize panel across eight developmental stages in four environments over two consecutive years. Using multi-temporal data, a random forest model accurately predicted LAI (R² = 0.82–0.83), significantly outperforming models based on single time-point data. By integrating high-throughput phenotypic predictions with time-series genome-wide association studies (GWAS), 36 dynamic SNPs associated with LAI variation were identified. Principal component analysis (PCA) of temporal LAI data revealed two principal components that together explained 84.2–86.5% of the total phenotypic variance. GWAS based on these components identified an additional 51 SNPs, seven of which overlapped between the two analytical approaches. Among the 72 candidate genes identified, Zm00001d048615 exhibited significant variation in both phenotype and expression among different inbred lines. The heterologous overexpression of Zm00001d048615 in Arabidopsis induced leaf curling and a significant reduction in leaf size, indicating its potential role in regulating leaf development. Collectively, these findings establish a robust framework that integrates UAV-based phenomics with temporal GWAS to identify key genes regulating complex dynamic traits. This approach provides valuable insights and genetic targets for improving maize canopy architecture and yield potential through molecular breeding.

Plant phenotyping relevance

UAVマルチスペクトル画像と時系列データからLAIを推定するモデルを開発・評価し、高スループット表現型解析に中核的に用いているため。

Code and data availability

The supplied preprint blocks describe UAV multispectral image collection, random forest LAI prediction, and time-series GWAS, but contain no data availability statement, no public deposit of phenotype datasets, UAV images, trained models, or author analysis code, and no authors' public URLs for such assets. The only in

No evidence-backed public reproduction asset is currently recorded.