Reduction of Spatial Variation in NDVI Measured at First Bloom in Cotton through Experimental Design and a posteriori Statistical Procedures

High-throughput phenotyping technologies are increasing the number of genotypes breeders are able to evaluate. Thus, accounting for experimental error due to spatial variation, particularly soil heterogeneity, will become increasingly important as more space is required to evaluate breeding materials. Traditionally, experimental error is minimized through experimental design, but research also suggests that a posteriori statistical procedures, such as nearest neighbor and moving means adjustments, in combination with appropriate experimental design improve the prediction accuracy of genotypic performance. The objective of this study was to compare the efficiency of experimental design and a posteriori statistical procedures in reducing experimental error due to spatial variation in normalized difference vegetation index (NDVI). Two hundred and eighty eight recombinant inbred lines were evaluated at Florence, SC in 2014 and 2015. The trial was planted as an incomplete block (alpha-lattice) design with two replications, and NDVI was measured at first bloom. We compare the relative efficiency of the incomplete block field design with a posteriori statistical procedures (treating the experimental design as a randomized complete block), as well as the combination thereof. The genotypic rank changes and relationship between NDVI and yield across the different analyses will also be discussed.