Redundant analysis predictingmetal distribution using landscape variables
produced a triplot similar to that of the RDA of metal distribution
based on sediment physicochemical variables (p = 0.009). RDA axis 1
and axis 2 both support a significant effect on the variability of sediment
metal (Fig. 2B) and accounted 72.52% of variance in landscape variables.
Imperative area (p = 0.034), paddy field (p = 0.002), forest land
(p = 0.001) and mudflat (p = 0.004) are significant factors that influence
sediment metal concentrations. According to the length of arrows
in RDA triplot, the importance of significant variables in turn is forest
land, paddy field, imperative area and muddy flat. Along RDA axis 1,
Cd, Fe, Ni and Cr stand out with positive axis 1 and axis 2 scores significantly
correlated with forest land. While, along RDA axis 2, As, Pb and
Zn showed positive axis 1 and negative axis 2 indicative of close relationships
with high percentages of paddy fields. Mn has a positive
score significantly correlated with imperative areas and mudflat.