Using multivariate analysis to design management zones

Abstract

Soil chemical and physical attributes are important in any agricultural cropping system, but in precision agriculture they are more relevant due to the possibility of application using different management practices along a produc­tion field. However, the correlation between these attributes has been little explored in the delineation of man­agement zones. This work aims to maximize the use of joint spatial variability for soil attributes. Its secondary objectives were 1) reduction of spatial variability dimensionality among all attributes and 2) assessment of agree­ment between univariate and multivariate management zones. The management zones resulting from the inter­polation of attribute values, as well as from the scores of each of the three main components, were delineated using the Fuzzy c-means algorithm. The fuzzy performance and modified partition entropy indexes were used to determine the optimal number of management zones. The Kappa index was used to evaluate the agreement of management zones obtained from attributes with those obtained from principal components. By using principal component analysis, it was possible to reduce the dimensionality of the number of variables that contribute to the joint spatial variability existing in the study area. There was no complete agreement between the uni- and multi­variate management zones outlined, which is why further studies on the subject are needed.