Repetibilidade da produtividade de linhagens de soja em ambientes do estado do Mato Grosso

Abstract

The recommendation of new cultivars, better than those existent in the state of Mato Grosso, Brazil, is a considerable challenge. In this state, more than nine million hectares of soybean are cultivated, under a wide range of climate, soil and management conditions. Thus, it is expected that the interactions between cultivar and environment be expressive. For plant breeders, it is not only important to quantify this interaction, but to propose measures for its mitigation. Within this context, this work was conducted with the objectives of: quantifying the genotype x environment interactions and estimate the repeatability of the performance of lines in the environments, in pairs, as selection strategy, aiming at improving the selection process. We used data obtained from a seed company. We evaluated nine lines, conducting the experiments using the 2010/2011 harvest at 3 locations, 2011/2012 harvest at 12 locations and 2012/2013 harvest at 11 location. A significant difference was detected between the lines in 76.9% of the experiments. No line presented superior performance at all locations. The line x location interaction was superior to the location x harvest interaction. The estimated for line performance correlation in the environments, in pairs, was negative or null, demonstrating that line performance repeatability is very low. Under conditions such as those of the experiment, the decision of the plant breeders in recommending a new cultivar is very fragile. Security can only be increased if evaluating a higher number of lines in the highest number of environments possible.