ABSTRACT

The selection of breeding and testing locations can have enormous impact on any crop breeding program. The existence of mega-environments is recognized in many crops. Examples include, northern and a southern corn groups, maturity groups of soybeans, and cotton that has at least four groups with Acala, stripper, and two picker groups with early and late maturity. In developing a breeding program with one of its primary goals to develop cotton for the entire US, Phytogen Seed Company, LLC (Phytogen) started with locations in the San Joaquin Valley, CA and in the Mississippi River delta, MS, both previously selected with a history of producing commercial varieties. Tifton, GA was selected as the main site for developing later maturing cotton varieties to expand the breeding program to cover the entire US.

The primary selection criterion for the outlying locations was to have these sites in the middle of important cotton growing areas delineated by lint production on a county by county basis. Phytogen started a MET to determine if there were duplicate locations within the sites in the Southeast or between them and the sites in the MidSouth, and to determine the best main breeding site for the Southeast. After one year of this test, Phytogen decided to close the Southeast breeding program for economic reasons so this Phytogen test data is only from one year. AMMI analysis in Agrobase21 and the GGEbiplot Pattern Explorer were used to analyze the Phytogen data. Both analyses indicated the existence of possible duplicate locations.

GGEbiplot Pattern Explorer was used for additional analysis of data from county yield histories compiled by the United States Department of Agriculture - National Agricultural Statistics Service (USDA-NASS) and data from the National Cotton Variety Trial (NCVT). GGEbiplot analysis of county yield histories may have value in the selection process of test locations. A pattern of related counties in Georgia was revealed but the underlying factors of the relationship require additional research. Data from the NCVT was collapsed into a genotype / environment two-way dataset and into location / year two-way dataset. The GGEbiplots from these datasets were compared and many of the interrelationships among the locations remain between the two analyses but not all. There is value in a location / year analysis when genotype information is not available.

The GGEbiplot Pattern Explorer also has a biplot that shows the most discriminatory and stable locations compared to an ideal location. The 'best' locations for a main research site are shown for all the datasets that were used.

GGEbiplot analysis was also used to find evidence of a location most similar to Stoneville, MS, an area with an excellent reputation in cotton varietal development. Tifton, GA appeared to be the closest environment in this comparison. This supports Phytogen's original decision to place a station near Tifton as well as Delta & Pine Land Company's decision to locate there.