Multivariate analysis to model yield variability for defined management zones in a banana agroecosystem

The delineation of management zones is based on the spatial behavior of a few soil variables selected and evaluated previously, and usually not correlated in situ with yield. Since the soil-plant system is multivariate, the analysis of its complexity requires statistical tools of equal size. These tools are convenient in providing an intuitive interpretation of the relationship between variables and the ordering of sampling sites. This study aims at the identification of management zones in a banana agroecosystem, starting from the overall analysis of soil variables with crop performance components, using multivariate statistical tools. Three clusters of sites were identified based on soil variables, dry mean weight diameter, pH, and (Ca+Mg)/K ratio, all correlated with crop yield. The groupings allowed to delineate management zones whose production has a uniform spatial behavior, significantly different between zones (P < 0.01)