A pattern analysis approach, based on classification and ordination, is presented for the characterization of environmental contributions to differences among lines in mean performance and response across environments. Other approaches to the analysis of line performance and environmental interactions are also considered. A population of soybean lines is used to illustrate the analyses for two characters, seed yield and seed protein percentage. In general, correlation of line performance over environments indicated that only moderate similarity of relative line performance existed over years or locations. These associations did not provide a clear basis for rationalization of test sites. Partitioning of variance components allowed a general strategy for sampling environments to be defined. Marked differences existed among environments for their contribution to environmental interaction, and these generally were consistent among locations for the two years of testing. Classificatory and ordination analyses were applied separately, and the contributions of each of the test environments were determined. These procedures confirmed that large differences in line response existed among environments, and provided additional and complementary information about the contributions of particular test environments to those differences. The effect of abandoning particular test sites on the recognition of differences among lines was examined by pattern analysis. The implications of the information gained through the pattern analysis approach in the development of testing strategies, and in the selection of test environments for specific objectives, are discussed.
Adaptive human aware navigation based on motion pattern analysis