Characteristics and Non-parametric Multivariate Data Mining Analysis and Comparison of Extensively Diversified Animal Manure

Abstract

Purpose: This study compared characteristics of different animal manure and examined non-parametric multivariate analysis tools’ suitability for their data mining. This can provide data and methodology support for scientific research and utilization of animal manure raw materials’ characteristics. Methods: Distribution profile testing, statistical calculation, and Spearman correlation analysis—using characteristics of 788 animal manure samples of layer, broiler, pig, dairy, and beef, with fertilizer nutrient compositions, proximate compositions, ultimate compositions, and calorific values—were conducted. Latent associations between different animal manure types’ characteristics were examined through five non-parametric multivariate analyses. Results: All samples’ physicochemical characteristics samples showed different non-normal distributions except potassium. Volatile matter (VM), fixed carbon (FC), ash, carbon, hydrogen, oxygen, and higher/lower heating value (HHV/LHV) were correlated, and nitrogen was positively correlated with phosphorus, potassium, and sulfur. Non-parametric principal component analysis (PCA), non-parametric exploratory factor analysis (EFA), hierarchical cluster analysis (HCA), and non-metric multidimensional scaling (NMDS) obtained similar results: VM, FC, carbon, hydrogen, oxygen, HHV, and LHV had associated attributes (“energy utilization”); phosphorus, potassium, ash, nitrogen, and sulfur had intrinsic associated attributes (“fertilizer utilization”). Conclusions: Animal manure characteristics should be mined and analyzed using non-parametric statistical analysis methods. Non-parametric PCA, non-parametric EFA, HCA, and NMDS are suitable for this purpose. Graphic Abstract: [Figure not available: see fulltext.]