Contents - Index


Avoiding/Exploiting GE through selection for explanatory traits

 

Previous topic: Mega-environment classification based on a trait-effect biplot

Comparison between the GGE biplot and the trait-effect by environment biplot indicates that four traits, namely, kernel weight, lodging, days to heading, and days to maturity, adequately explained the GGE pattern of yield. This implied that yield may be effectively manipulated by selecting for these four traits. 

The only way to avoid negative GE and to exploit positive GE is to divide the target environment into meaningful mega-environments and to select specifically adapted varieties for each.

Selection strategy for the Eastern mega-environment

The following biplot contains environments only from the eastern Canadian provinces ON, QC, PEI, and MB:

Interpretation: A trait has positive associations on yield in an environments if the trait and the environment have an acute angle in the biplot. Thus, kernel weight had positive associations with yield in all environments in the Eastern mega-environment, implying that selecting for greater kernel weight in eastern Canada should improve barley yield, at least for this Harrington * Tr306 population. On the contrary, lodging had negative associations with yield in all eastern environments, so did days to heading. Therefore, selecting for greater kernel weight, better lodging resistance, and early heading should improve barley yield in eastern Canada, at least in this breeding population. Days to maturity, however, did not have consistent associations with yield in eastern Canada. This trait cannot be used in breeding for higher yield in eastern Canada. 

Selection strategy for the Western mega-environment

The following biplot includes only the Western locations (Alberta, Saskatchewan, Washington, Montana, and Alaska).

Interpretation: First, the environments took both positive and negative values for both axes, implying that the trait-yield relations varied dramatically among environments within this mega-environment. Consequently, no trait had positive associations with yield in all environments. Secondly, environments from the same locations (Alberta and Saskatchewan) were scattered across the biplot, indicating that the environments cannot be further divided into repeatable sub-groups. Therefore, the Western mega-environment is a single mega-environment with large unpredictable trait-yield relations. As a result, it does not seem feasible to improve yield by indirect selection for any of the traits; selection for yield per se in multiple environments is the only way to improve yield in this mega-environment.