Improved multivariate analyses to discriminate the behaviour of faba bean varieties

Abstract

Agricultural systems of Southern European regions are often based on short rotations of winter cereals and are thus subject to agro-ecological problems such as decreasing biodiversity, loss of soil fertility and increasing reliance on mineral fertilisers. Introducing new crops such as new varieties of faba bean, and new production methods, e.g. different planting times, may increase the sustainability of farming systems. To advance the use of these methods, both multi-environment field experiments and improved statistical methods to summarise and interpret results are needed. This report summarizes experiments conducted over three years and two locations where we compared phenology, morphology and yield of six faba bean, Vicia faba L., genotypes, sown in November and February. We have analyzed the data using canonical variate and additive main effect multiplicative interaction (AMMI). We demonstrate how such methods may be useful to obtain relevant information about a more successful introduction of faba bean in southern Europe. Our results show that sowing in November is much more suitable to Southern European regions than a February planting. Indeed, for the November planting, beans flower earlier and pods fill before the drought period. Concerning morphology, November sown plants were taller of 0.93 m versus 0.79 m on average; gave a lower insertion of first fertile branch of 0.44 m versus 0.51 m; gave a higher number of lateral branching per plant, of 1.5 versus 0.8; and gave a higher number of pods per plant of 10.2 versus 7.6. On the average yield levels were the highest for November sown varieties, of 3.55 versus 2.66 t ha −1. These findings indicate autumn sown faba bean could be introduced in Southern European regions. Using adequate varietal selection, this crop may improve the agro-environmental sustainability of farming systems. These results also highlight the usefulness of canonical variate and AMMI analysis, as 576 morphological data, e.g. 6 varieties, 2 sowing times, 6 environments and 8 variables, and 72 yield data, e.g. 6 varieties, 2 sowing times and 6 environments, can be summarised in two bi-plots, clearly depicting the effect of sowing dates on crop morphology and yield, across locations and years. Such methods deserve a more widespread use when it is necessary to interpret crop response to environmental and agronomic factors. faba bean / genotypes / Mediterranean environments / canonical variates / AMMI analysis

Durum wheat varieties in N‐deficient environments and organic farming: a comparison of yield, quality and stability performances

“…Readers are referred to the above work for more detail, but the rational behind AMMI is to simplify the complex interaction matrix into a simpler and more meaningful representation, given by the most significant PC scores (Stagnari et al. ). In this case, the first two PCs were selected for systems and years and plotted on an AMMI2 type bi‐plot (Gauch and Zobel ).…”

“…With this model, the 'system 9 year' (where system indicates the combination of genotypes and N levels) interaction matrix is decomposed into the product of two matrices, that is, the system score matrix and the environment score matrix. Readers are referred to the above work for more detail, but the rational behind AMMI is to simplify the complex interaction matrix into a simpler and more meaningful representation, given by the most significant PC scores (Stagnari et al 2007). In this case, the first two PCs were selected for systems and years and plotted on an AMMI2 type bi-plot (Gauch and Zobel 1996).…”

Grasses and legumes in mixture: an energy intercropping system intended for anaerobic digestion

“…Furthermore, the variables related to the chemical composition of silage were jointly considered in a multivariate approach and statistically processed applying a canonical analysis with the two experimental factors (CS and MT) as discriminating sources (Sadocchi, 1981;Stagnari et al, 2007). The first two canonical variables, accounting for the larger part of the data variability, were considered for further data interpretation.…”