Physiological factors associated with genotype by environment interaction in wheat

Wheat cultivars often show highly significant genotype by environment interaction (G x E) for yield, even when comparing different years within a relatively stable location. This study attempts to explain some of the physiological bases of G x E in two experiments: (i) historic yield potential trials (HYPTs) of bread wheat (Triticum aestivum L.), durum (T durum Desf.) and triticale (X Triticosecale Wittmack) cultivars grown under agronomically optimal conditions; (ii) an elite spring wheat yield trial (ESWYT) of 30 bread wheat genotypes cultivated at 27 international locations. For the HYPT, the main objectives were to determine the environmental variables during different phenological stages associated with: (i) G x E among the three crop species, (ii) G x E within each species, and (iii) underlying physiological causes of G x E. For ESWYT, meteorological data were not available and so mean site values of certain crop parameters were used as proxy environmental data to determine whether conditions either pre- or post-anthesis were more influential in determining G x E. Partial least-squares analysis and factorial regression models were used to identify the environmental factors best explaining G x E independent of the main effects. Of the three crops, durums were shown to be the most sensitive to conditions pre-anthesis, requiring higher radiation and cooler average temperatures in order to set high grain number. Triticale, despite having the highest average yield and biomass, performed relatively poor when conditions from spike growth stage onwards were sunny and warm. Bread wheat appeared to be the most robust of the three species. Considering yield, biomass, and yield components, it was apparent that the spike primordia growth stage was generally the most sensitive to environmental factors causing G x E. Results for the ESWYT suggested that conditions post-anthesis were more influential on G x E than conditions pre-anthesis. Implications for how such analysis may assist with both conventional and molecular approaches to breeding are discussed. (C) 2002 Elsevier Science B.V. All rights reserved.