Testing Locations in Regional High Quality Tests for Cotton Seed Quality Traits

Abstract

Significant genotype (G) × environment (E) effects for cotton (Gossypium hirsutum L.) seed quality traits have been identified in previous studies. Significant G × E interactions necessitate multiple-location tests to evaluate seed quality traits, which add cost to the tests. Reduction of testing locations could trim costs if the analysis of G × E interactions and the efficiency in breeding are not dramatically affected. The objectives of this study were: 1) to determine an appropriate reduction of testing locations without significant loss in power for detecting G × E effects; 2) to determine an appropriate reduction of testing locations without significant loss in accuracy for estimating strain means; and 3) to identify a possible mega-environment for evaluation of seed traits using GGE biplot. Historical data of Regional High Quality (RHQ) tests from 2005 through 2013 were used to address the objectives for three seed quality traits including oil content, N content, and free-gossypol. Significant G × location (L) interactions were detected in most cases. However, with averages of the three traits, less G × L interactions were detected with 7.3% and 9.1% non-significance, when two and three locations were omitted, respectively. Reduction of locations up to three, increased standard error to 25% of those with zero locations omitted. There was no clear mega-environment identified for seed traits. However, the locations of Lubbock, TX, Stoneville, MS, Florence, SC, and Portageville, MO were identified as being more representative than others for evaluation of the N content.