Understanding the agricultural potential of a soil is often based only on interpretation by univariate analyses, and this may increase the scale of the problems when selecting appropriate soil management practices. Thus, multivariate analysis is an alternative since it is a set of procedures aimed at grouping individuals and discriminating between these groups. It also serves as an instrument for selection of variables in that those with the highest weight in the construction of the first principal components are likely to better represent the data set under analysis. The aim of this study was to identify soil properties that best explain the spatial variability of production of common bean by means of multivariate analyses. In the 2006/2007 crop year in Selvíria, MS, Brazil, we analyzed common bean yield in relation to some physical and chemical properties of an Oxisol cultivated under high technological management conditions in a minimum tillage system with center pivot irrigation. A geostatistical grid was demarcated for collection of soil and plant data, with 117 sampling points in an area of 2,025 m2 and a homogeneous slope of 0.055 m m-1. Classification into groups was carried out by three methods: the hierarchical grouping method, the non-hierarchical k-means method, and principal component analysis. We may conclude that multivariate analysis combined with precision agriculture is an important tool to assist localized management. Principal component analysis allowed us to identify three groups that explained 86.3 % of the total data variability. These groups consisted of the physical properties of bulk density, total porosity, and gravimetric and volumetric moisture, which showed greater explanatory power for yield variation.
soil management; quality of soil; agricultural sustainability
