Understanding the integrated effect of external factors (e.g., human activities) and internal factors (e.g., hydrodynamics, sediment properties) on metal(loid) distributions is necessary for relevant decision-makers to develop river basin management strategies. In attempts to understand the possible causes of the distribution of nine target metal(loid)s in riverbed sediments across Xijiang River basin in China, we grouped and portrayed the multiple metal(loid) distributions by calculating metal(loid)'s normalised-and-weighted average concentrations, and then canonical correlation analysis combined with a series of statistical operations, collectively called optimized CCA analysis, was applied to quantify the strength of relationship between multiple metal(loid) distribution and integrated effect of internal-external factors. Results showed that the target metal(loid)s can be divided into three groups according to their distribution patterns: Group A (including Zn, As, Cd, Sb and Pb), Group B (including Cr, Ni and Cu) and Group C (including Tl). Among them, metal(loid)s in Group A was significantly enriched in comparison with the reference values of Chinese sediments, and the wide-ranging accumulation of Cd and Sb in the whole study area needs paying great attention to. For those metals in natural states (e.g., metals in Group B), the affinity of sediment (e.g., Fe and Mn oxides) is responsible for their distributions. By contrast, when metal(loid)s (e.g., metal(loid)s in Group A and Group C) had obvious anthropogenic sources, the interferences of anthropogenic inputs (e.g., non-ferrous metal enterprises' waste-discharging activities) and the specific sedimentary characteristics (e.g. karst topography and low-energy hydrodynamic depositional conditions) in study area can weaken the correlation between the binding affinity of sediment and the contents of metal(loid)s. The optimized CCA analysis can be an alternative and advantageous statistical operation for determining the main types of causes of multiple metal(loid) distribution in the case of observations with relatively low case-per-variable ratios.
Read full abstract