Hierarchical partitioning for selection of microbial and chemical indicators of soil quality

Statistical approaches, especially multivariate techniques such as hierarchical partitioning analysis (HP) and redundancy analysis (RDA), can be used to select appropriate variables for soil quality assessment. HP is usually applied to ecological data from plants and animals, but has not been applied to chemical and microbial properties such as those used as indicators of soil quality. Our aim was to show how these methods can be employed to find soil quality indicators, using soil microbiological, chemical and physical data to compare two forest types (native and reforested Brazilian Araucaria forests) in two locations in Southeast Brazil. We used RDA to investigate relationships among variables. Additionally, we quantified the independent effects of predictor variables: location, forest type, two specific seasons and some soil properties and used HP to examine how these environmental variables interact to influence soil microbial and chemical attributes. RDA showed that acid phosphatase and dehydrogenase activity, NO2− oxidizer numbers, basal respiration, metabolic quotient, pH, P and sand content were positive and significantly correlated with the native Araucaria forest, whereas arylsulphatase activity, denitrifier numbers, microbial biomass carbon, microbial quotient, TOC, S and clay levels were positively correlated with the reforested Araucaria. These preliminary results suggest that these variables are the best indicators of soil quality for Araucaria forests. However, HP, used as a complementary tool, showed that only dehydrogenase activity, pH and S variations were due more to forest type than to physical and chemical soil properties, and were resistant to the variation in the two seasons. Overall, our results indicated that dehydrogenase activity, pH and S are potential indicators that can be used to assess or monitor soil health in Araucaria forests. In conclusion, we demonstrated the usefulness of HP to find soil quality indicators. Similarly, other statistical methods, as RDA, could complement HP and increase the reliability of studies that consider simultaneous variables in soil science.