The effect of interplot competition on yield assessment in triticale trials

Correction models for intergenotypic competition in winter wheat

“…4 Only significant covariates (at the 5% level) are mentionned. The efficiency of height covariates in adjustment for competition in wheat had already been observed [10,14]. Our results showed that the covariate using height at heading time (only measured in 1993) led to a slightly, but significantly (P = 0.03) better correction than the covariate for height at maturity (r = 0.63 instead of 0.58).…”

“…m -2 . Competition has sometimes been found in 6-rowplot trials within heterogeneous samples of triticale [14] or wheat genotypes [16]. Yet we assumed here that competition effects in the 6RP trial are negligible as regard to those in the SRP trial, as it was also reported by Brabant et al [5] and Goldringer et al [10].…”

“…Hence, a reference trial seems essential to test the relevance of any new model or any neighbouring covariate based on a new trait, as well as when studying a new species. Nevertheless, there are numerous studies in wheat showing the efficiency of neighbouring covariate based on height in correcting for competition [10,14,16]. Therefore, we believe that neighbouring covariate based on height could be used routinely, without reference trial, when it leads to a significant reduction of the residual mean squares.…”

“…The two earliness traits were not efficient. The search for other traits more efficient as a covariate should include the measure of plant height at different stages during the development as done by Kempton et al [14] in triticale trials. Other traits such as tiller number, leaf number, length and height of the flag leaf have been described as possibly involved in competition phenomena [11,17].…”

“…Different models have been developed for correcting competition effects [4]. When competition can be associated with any plant character, one (or more) covariate equal to the difference for the character between the value of the plot and the mean of the adjacent plots can be introduced in the model [14]. If the yields of adjacent plots are negatively correlated, an autoregressive model may be successfully fitted [12].…”