Multivariate analysis for soil science

With the growing surplus of data available for modelling in soil science, multivariate analyses offer a path to understand the patterns in large datasets with 100s or 1000s of variables. In this chapter, we explore three classes of multivariate techniques and analysis, and a case study using one technique for each class: dimension reduction with principal component analysis (PCA), supervized classification with linear discriminant analysis (LDA) and unsupervized classification with k-means clustering. While by no means comprehensive, we demonstrate potential uses for multivariate analysis in soil in agricultural and ecological contexts. In addition, we provide some examples of alternatives for each of the classes.