Semester of Graduation

Summer 2026

Degree

Master of Agriculture (MAgr)

Department

SPESS

Document Type

Thesis

Abstract

Sheath blight, caused by Rhizoctonia solani AG1-IA, ranks among the most damaging diseases of southern U.S. rice. Current knowledge indicates that resistance to this disease is quantitative, polygenic, and strongly influenced by genotype-by-environment (G × E) interaction, as well as confounding factors such as canopy architecture. These conditions make phenotyping the primary bottleneck in resistance research. This thesis evaluated whether different approaches to measuring resistance altered conclusions regarding its biology, heritability, and genetics.

A panel of 241 long-grain genotypes was evaluated across five environments and two seasons. Phenotypes derived from visual severity, epidemiological models, and UAV multispectral imagery were compared for heritability, genotype selection, and genome-wide association study (GWAS). Epidemiological analysis demonstrated that a single late-season visual assessment (D70) captured the signal with the highest heritability (H² = 0.79) and provided the greatest genetic gain per unit of effort, while multi-time-point and model-derived metrics added complexity without improving selection. G × E remained the primary limitation. UAV-based phenotyping confirmed that genotype ranking depended on the timing of the flight. A raw vegetation index near D70 ranked genotypes as effectively as machine learning models, yet no UAV-derived phenotype reconstructed the cumulative disease progress (AUDPC). The least valid UAV-based proxy was also the trait with the highest heritability in the study (H² = 0.90), a result that decouples precision from validity. GWAS revealed that the phenotype definition and the timing of assessment altered both the number and the genomic location of the detected associations. Visual severity at D60 yielded no associations, despite high heritability, whereas epidemiological and UAV-based phenotypes converged on a single common region (chromosome 1, near 24 to 27 Mb). However, each approach captured signals that the others did not detect. Collectively, this work reframes phenotyping from a measurement problem into a conceptual one. A well-timed late-season assessment is the most efficient phenotype, and any proxy must be validated against the target trait rather than trusted on heritability alone.

Date

7-17-2026

Committee Chair

Famoso, Adam

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Available for download on Friday, July 15, 2033

Share

COinS