Robust multivariate analysis of compositional data of treated wastewaters
- 54 Downloads
A dataset of water samples collected behind a biological wastewater treatment plant (BWWTP) during a year was processed as compositional data by a log-ratio transformation and then analysed by a robust principal component analysis (RPCA) and the robust Mahalanobis distances (RMDs). For this purpose, covariance matrices were computed using a minimum covariance determinant (MCD) algorithm. Raw and transformed 11 physico-chemical parameters were reduced to 4 robust principal components (RPCs). Correlations between centre log-ratio (clr)-transformed parameters and RPCs were found to be more realistic than those between the parameters and RPCs of raw data. The first and second RPCs represented nitrogen and phosphorus compounds, respectively. Their temporal changes were explained by some processes occurring during biological wastewater treatment. A nitrification process was also demonstrated by the temporal changes of the raw and clr transformed concentrations of ammonium. The robust and classical Mahalanobis distances were computed from the raw and isometric log-ratio (ilr)-transformed data to show the overall temporal changes of treated wastewater composition and to detect outlaying samples.
KeywordsCompositional data Log-ratio transformation Treated wastewaters Multivariate analysis
The author thanks Dr. Zdeněk Matěj (Lund University, Sweden) for his help with the MATLAB subroutines. This work was financially supported by the project “Institute of Environmental Technology—Excellent Research” (CZ.02.1.01/0.0/0.0/16_019/0000853) provided by the Ministry of Education, Youth and Sports of the Czech Republic.
- Aitchison J (1982) The statistical analysis of compositional data. J Roy Stat Soc Ser B (Methodol) 44:139–177Google Scholar
- Drew LJ, Grunsky EC, Schuenemeyer JH (2008) Investigation of the structure of geological process through multivariate statistical analysis—the creation of a coal. In: Bonham-Carter G, Cheng Q (eds) Progress in geomathematics. Springer, Berlin, pp 53–77. https://doi.org/10.1007/978-3-540-69496-0_5 CrossRefGoogle Scholar
- Praus P (2005a) SVD-based principal component analysis of geochemical data. Cent Eur J Chem 3:731–741Google Scholar
- Praus P (2005b) Water quality assessment using SVD-based principal component analysis of hydrological data. Water SA 31:417–422Google Scholar