Trends in mean performance and stability of winter wheat and winter rye yields in a long-term series of variety trials

There is considerable interest in assessing genetic gain from breeding efforts, as well as trends from non-genetic sources such as climate change. Long-term data from official variety trials provide an ideal opportunity to do so. Whereas past work on the subject was mainly focussed on trends in mean performance, little attention has been paid to yield stability. The purpose of the present paper therefore is to provide a framework for assessing trends in stability and to exemplify this using long-term data from German variety trials. Specifically, trends in the mean and variance of winter wheat and winter rye yields associated with genotypic, spatial, and climatic covariates were investigated based on a long-term series of multi-environment trials from 1983/1985 to 2016. Yield stability is assessed based on the variance of yield. For climatic covariates, trends were investigated using growth stage-specific covariates which were obtained by dividing the total growth period into five or ten growth stages within which covariates were aggregated. The analysis was done by linear mixed models and forward selection served to identify trends in mean and variance. Specifically, trends in the mean were selected based on a coefficient of determination, R2, while variance trends associated with a covariate were selected based on the change in variance attributable to the covariate. For spatial covariates, mean yield increased with the field capacity in up to one meter soil depth in case of rye. For both crops, mean yield was negatively affected by a deficit in available water during the development of vegetative parts while water deficit towards the end of the growth period was positively associated with mean yield. Trends were also selected accounting for interactions of climatic covariates and genotypic groups (wheat: quality type; rye: breeding type), revealing that R2 generally increased when interactions were taken into account. The responses to covariates were similar for the different groups, meaning the increase in R2 is mainly attributable to the main effects of the genotypic groups. Variance trends associated with the calendar year were identified for both crops, revealing yield variance slightly increased with time. Based on the selected model, a coefficient of variation was obtained for each year to assess relative yield stability, showing a decrease for wheat and an increase for rye.