High-versus low-stress yield test environments for selecting superior soybean (Glycine max L. Merr) Lines

A problem that continues to perplex plant breeders is selecting for yield in stress versus non-stress environments. The purpose of this research was to select superior soybean lines which were best adapted to both low- and high-stress environments by using low- versus high-stress yield test environments.

Forty-eight random F₄ - derived lines were evaluated for yield over a five year period (1982-1986) at two locations [Crossville in East and Milan in West Tennessee] on high fertility, low acidity (low-stress), and low fertility, high acidity (high-stress) nursery environments at each location. Mean yields were 0.50 vs. 2.00 and 0.74 vs. 2.71 Mg ha^-1 for Crossville and Milan, high- vs. low-stress sites, respectively. Five additional locations throughout the state were included in 1985 and 1986 as yield trial testing sites for obtaining information about yield and broader adaptability.

Soil chemical and physical parameters of nursery environments were measured in 1984 and 1985 at each of the four nursery sites. The soil at the Crossville low-stress site was classified as a Typic Hapludult and the high-stress site was classified as an Aquic Fragiudult and Typic Hapludult. The soil at Milan low-stress site was classified as an Aquic Fragiudult over a buried Typic Ochraqualf and the high-stress site was classified as a Typic Fragiudalf and Glossic Fragiudult that intergraded to an Aerie Glossudalf. Soil bulk densities were taken at three soil depths at each site. The density measurements were in an acceptable range (1.3-1.5 m^-3) which indicated that no physical barrier existed which would inhibit root penetration into the soil profile. Soil chemical analyses indicated that high acidity and aluminum concentration were major factors responsible for reduced availability of P, K, Mg, and Ca which caused yield reduction and poor growth in high-stress nursery sites.

Sixteen lines (33% selection intensity) were designated as the "superior" lines because their mean yields were above the grand mean in high-stress environments (0.62 Mg ha^-1) as well as above the grand mean in low-stress tests plus yield trials (2.25 Mg ha^-1). These sixteen lines served as the superior "standard" for the purpose of simulating selection. Eleven of the top 16 soybean lines came from 2 crosses (Essex X (Bay x N73-520 and Essex x Hodgson) and the top 3 superior lines came from the cross Essex x (Bay X N73-520).

Simulated selection, using either nursery type within location(s) for one year of testing, revealed 32 to 54% of the superior lines would have been selected. At least 1 of the top 3 superior lines would have been selected in every case, regardless of nursery type(s), location(s), or year(s) used. Combining high- or low-stress tests over location(s) in a given year or years (2 or 3) did not improve the effectiveness of selection enough to justify the extra time, expense, and labor for the purpose of selecting the highest yielding group ( one third) from a random set of lines in preliminary tests. Comparisons between 33 and 20% selection intensity generally revealed a lower probability in obtaining at least 1 of the top 3 superior lines using 20%, and in some cases the probability was zero. This indicates that a selection Intensity greater than 33% is inadvisable at the preliminary yield testing stage.

Stability analyses revealed that the sixteen superior lines had higher regression coefficients with 10 of those lines having regression coefficients greater that 1.00.

Analyses of plant height and maturity were conducted to determine if selecting the highest yielding lines would have resulted in indirect selection of lines which were shorter and earlier or taller and later or any other combination. No trend of this type would have occurred.