Uso da abordagem estatística procrusteana em Ecologia de Solo: caso de estudo envolvendo sistema de integração lavoura-pecuária-floresta no Cerrado

Abstract

This thesis is part of a multiple scientific effort seeking to support the replacement of degraded brazilian pastures by systems which integrate different land use types such as crop, pasture, and forest plantation (collectively known as iCLF systems). Here, the focus was also to discuss the potentialities of an unusual statistical multivariate approach called “Procrustes Analysis” in the plant and soil ecology framework. The current thesis has three chapters through which details of the Procrustes analysis are presented on both technically e intuitively manner. The first chapter describes roadmaps showing how the procrustean residual vector (so-called PAM: Procrustean association metric), representing the multivariate correlation between two or more data tables, can be used as an univariate variable in more user-traditional statistical approaches such as ecological ordination, regression analysis and ANOVA followed by mean comparisons. The second chapter discussed a case study and had as the general objective to use PAMs, depicting the relationships between distance matrices from individual soil microbial structure (PLFA: Phospholipids Fatty Acid) and distance matrices form soil properties variables (chemical and physic), as response variables in an ANOVA framework with land use type as categorical predictor (degraded pasture, improved pasture, native fragment and iCLF system). The hypothesis in this case was that the fungi:bacteria ratio given by PLFA analysis, a good index of changes in microbial structure as response to land use alteration and associated to more conservative soils in terms of carbon mineralization, is favored by the man – introduced vegetal heterogeneity which characterizes the integration crop – livestock – forest. The last chapter was entirely dedicated to answer some technical questions which arose after the publication of the first chapters. Basically the two most common questions were: i) Does the increasing number of columns/variables within a data table affect Procrustes outcomes? ii) Can the procrustean residual vector, the PAM, translate differences between treatments in terms of multivariate correlation as it is used in mean comparisons? Specifically for these questions, Procrustes was useful in supporting iCLF systems as potential alternative to degraded pasture by raising insights that the man – introduced vegetal heterogeneity in such integrated agroecosystem, favor shifts in microbial structure toward fungal dominance.