919, DF = 29, p = 0.0001) ( Supplementary data Fig. 3). It can thus be argued that % N is a proxy for organic carbon in St Helena Bay. In order to determine the trace metal concentrations in sediments, sub-samples from each core were dried (60 °C, 24 h) and ground to homogeneity. Approximately 2 g of sediments were then digested using an acid mixture of 4:1 (HCl:HNO3) at 110 °C on a Gerhardt digestion block for 3 h following Morton and Roberts (1999). The supernatant Cabozantinib purchase was then filtered off and diluted to 100 ml with distilled water. A UNICAM SOLAAR M-SERIES Atomic Absorption Spectrometer was used to determine the concentrations of Cu, Zn, Pb, Fe, Cd and Cr in the sediments. The similarity in

the multivariate environment (grain size, and trace metal concentrations) at the different pipeline and non-pipeline sites in the two locations was calculated using Euclidean distance, following log10(x + 1) and normalisation of the data. This matrix was visualised by ordination using non metric multidimensional scaling (nMMDS) in PRIMER v6. In order to determine whether there were a priori differences between pipeline and non-pipeline sites in the environment at each location, and between locations (factors), the multivariate data were analysed using the PERMANOVA Target Selective Inhibitor Library order routine in PRIMER v6. PERMANOVA

tests the simultaneous response of variables to one or more factors in an analysis of variance (ANOVA) experimental design on the basis of a resemblance measure, using permutation methods ( Anderson et al. 2008). The routine partitions the total sum of squares according to the specified experimental design, including appropriate treatment of factors that are fixed or random, crossed or nested, and all interaction terms. Here the different sample sites are nested within either Casein kinase 1 pipeline or non-pipeline factors (both considered random), which in turn are nested by location (fixed). A distance-based pseudo-F statistic is

computed (analogous to the F statistic for multi-factorial ANOVA models) and p-values are subsequently obtained by permutation. In order to determine the relationship between the measured environmental variables, non-parametric Spearman Rank Order correlations were performed in STATISTICA v. 11 and significance values were adjusted using Bonferroni correction (Townend, 2002). The similarity between samples in terms of their foraminifera was calculated using the Bray-Curtis Index (Clarke and Gorley, 2006), following root-root transformation of the abundance data. Living and dead assemblages were treated separately and all analyses were computed using PRIMER v6 software. The similarity matrices were subsequently visualised using nMMDS plots. Living foraminifera are presumed to respond to the environment in which they are found, whilst dead individuals provide an indication of post-mortem and taphonomic processes such as advection (Murray, 1991).