Value of Locations for Representing Mega‐Environments and for Discriminating Yield of Watermelon in the U.S.

Abstract

In a crop breeding program, multiple‐location trials can be used to define target regions and mega‐environments that, in turn, will help the breeder develop stable cultivars. In addition, locations can be chosen that are efficient for distinguishing among cultivars (genotypes) and that are good representatives of the target regions. The objectives of this study were to study mega‐environments and identify test locations that were both discriminating and representative of target regions. Watermelon [ Citrullus lanatus (Thunb.) Matsum. & Nakai] fruit yield and yield components were evaluated in 3 yr and eight locations using replicated, multiple‐harvest trials. Data were analyzed using genotype main effect and genotype × environment interaction (GGE) biplot model as well as other methods for stability analysis. Marketable yield and percentage early fruit had a nonrepeatable crossover pattern and thus, formed a single and complex mega‐environment. Two key locations, (Kinston, NC, and Charleston, SC) were efficient representatives of two mega‐environments for fruit count. Locations at Woodland, CA, and College Station, TX, can be used interchangeably for identifying genotypes with high percentage cull fruit. There was only one mega‐environment for fruit size. Identification of mega‐environments for watermelon in the southern United States has implications for future breeding and genotype evaluation in the United States including the use of specialized genotypes for high performance in specific locations.