Baseline metals and excess concentrations in urban sediments--The concentrations of each element at reference (Bodega Bay, Dana Point, Lutak Inlet) and many nonurban sites (e.g., Kamishak, Oliktok, San Luis Obispo, Channel Islands) were plotted against the percent fine sediment and a determination of background concentration was made. Any urban station which displayed an element concentration elevated above this background concentration was considered "contaminated". From the determinations for each element, a list was devised which highlights those stations with concentrations measured above the background level.
We chose percent fine sediment (silt plus clay; <63 µm) for plotting because we believed that naturally and anthropogenically occurring elements would associate mechanistically with some closely allied property of grain size (e.g., TOC or surface area). Both surface area and TOC are physicochemical factors which may control the abundance of some elements. Usually, as the sediment particles become smaller, sediment surface area increases as does organic carbon content because of additional sites for association. Also, iron and manganese, which were positively correlated with decreasing grain size (increasing percentage fines), can control metal concentrations by providing a surface for adsorption via iron and manganese oxides.
Because several parameters are correlated with grain size and this is a property that is easy to measure, we chose it for our independent variable by which we assessed the variation in background element concentration. The correlation of element abundance and grain size has been discussed by Klamer et al. (1990) and reviewed by Horowitz (1991). In our study we analyzed the bulk sediment for total metal concentration instead of separating the sediment into discrete grain size fractions for analysis, as was done by Klamer et al. (1990). Hence, the results may be similar, but excessive dilution by sediment particles greater than 63 µm may skew the results and lead to poorer correlation coefficients than found in other studies. In general, our approach to baseline metal concentrations is rudimentary and readers should consult Luoma (1990) for a more detailed discussion on this subject.
Element regression on aluminum concentration in sediment from the East Coast of the United States was done by Hanson et al. (1993) to determine background or reference concentrations. Because Hanson et al. (1993) found high correlations for most elements with aluminum, we can only conclude that the geochemical makeup or geological histories between the areas (East Coast vs. West Coast) must be different or that our sample size was insufficient to determine such trends. One difference between our study and Hanson et al. (1993) is that they reported many aluminum concentrations below 4%, whereas in our study we had very few values (none of the reference sites) below this level. In our study, aluminum from clean sites moderately correlated with the percent fine sediment particles (r = 0.41) which was less than that found (r = 0.75) by Hanson et al. (1993). Our study does agree with Hanson et al. (1993) in that our strongest respective correlations (%aluminum or %fines) were with copper, iron, lead, and zinc.
The correlations for the Pacific coast reference and nonurban sites are listed in Table 4. Although Port Moller and Nahku Bay are categorized as nonurban, we did not include them in this group because we felt that zinc and lead were elevated at Nahku Bay, and zinc and tin were elevated in Port Moller sediments. The Nisqually Reach reference site was also excluded because several elements exhibited elevated concentrations. Elements which increase in sediments as percent fines increase may do so naturally or because of anthropogenic input. We could not separate natural from anthropogenic accumulation because we did not have a situation where the type of input was known and constant over variable grain size. At this time we cannot separate the natural from anthropogenic inputs but can expect that the reference areas will generally contain lower concentrations of contaminants than the urban sites. For some elements we may assume that contaminated areas with a higher frequency of smaller particles will be expected to accumulate more of the contaminant elements than sites with a sediment containing a lower percent of fine particles. Also, we can not rule out the possibility that even some of the reference sites may contain elevated concentrations due to the global spread (i.e., widespread dissemination from urban areas) of contaminants. Because of this possible global spread, some elements (e.g., cadmium and lead) may be elevated at our nonurban and reference sites which may obscure any relationship between grain size and sediment concentration.
Percentage fine sediment (<63 µm) may be a good measure, or a surrogate, of the variation observed for the concentration of some elements in sediment; however, correlation with grain size determined in more size categories (e.g., 1-2 µm, 2-4 µm, 4-8 µm, etc.) may be more appropriate. In this study, both percent iron and TOC were correlated to percent fines when all Pacific coast sites were considered together (r = 0.57 for iron and r = 0.55 for TOC) (data not shown). At Pacific coast reference sites, iron was still significantly correlated with percent fines (r = 0.67) and TOC less so (r = 0.35) (Table 4). For some of the other data subsets (e.g., West Coast reference sites) the percent fines correlated poorly to iron and TOC, possibly due to low sample sizes. Surfaces such as iron oxides can control metal concentrations by providing a surface for adsorption. Organic carbon can also serve as ligand for some elements; hence, those sediments with higher TOC may contain higher concentrations of certain elements. The mean (sd) percent fine sediment for each site is shown in Figure 44 and the mean (sd) percent TOC for each site is shown in Figure 45, along with the frequency distribution for TOC in Figure 46. One can see that there were many sites with highly variable TOC or percent fines and many of the outliers (high TOC with low percent fines and low TOC with high percent fines) are Alaskan sites.
Each element in reference and nonurban sediments was plotted against percent fine sediment. Two elements (copper and zinc) were strongly correlated and two others (lead and selenium) were moderately correlated with percent fine sediment and therefore varied naturally as a consequence of grain size (Fig. 47) (see Table 4 for correlations). The differences in geochemical makeup may include variable iron and manganese oxides due to differences in sediment grain size and redox state. These oxides can scavenge dissolved metals which may lead to variable sediment concentrations. It is also possible that these nonurban areas experience elevated dissolved concentrations of certain elements which may lead to a slight increase in bulk concentrations and hence a correlation to grain size. The predictive regression equation for each of these elements as a function of percent fine sediment is also shown in Figure 47. The similarity in slope for these equations indicates that a similar process of association of element with sediment may be occurring. Even though the linear correlation for lead was relatively high (r = 0.63), it was not treated like copper, selenium, and zinc because its association with percent fines was curvilinear (Fig. 47c). The baseline concentration of lead was determined, like all the other uncorrelated elements, with a horizontal line above all reference and nonurban concentrations to determine contaminant concentrations.
In most cases, it appeared that elements were randomly distributed over the percentage fine material for all reference sites. In a few cases, it seemed that some reference sites contained excess concentrations of a particular element. For example. chromium and nickel concentrations (Figs. 48a and 48b) at the reference site Bodega Bay were far above those measured at all other reference sites (but not elevated at all Bodega Bay stations) which may indicate a degree of contamination from a local anthropogenic source. Conversely, there may be some natural geological source in this area which explains the elevated concentrations of these two metals because both metals were high at most San Francisco area sites. At this point it is difficult to determine why Bodega Bay contains elevated levels of chromium and nickel without a more thorough investigation of local sources.
Another metal, mercury (Fig. 48c) was elevated in two of the Dana Point samples which may be due to a local source. These mercury concentrations are 4 to 7 times higher than mercury concentrations in most other reference sites. Nisqually Reach (mean (sd): 0.16 (0.3) µg/g) was excluded in baseline determination of mercury because of elevated concentrations. Tin was also elevated at Dana Point (compared to other reference sites) which may be due to a nearby marina and contamination from organotins (Fig. 48d). The data show that 1988 and 1987 levels were higher than those found in samples for 1985 and 1986, indicating an increasing trend. Elevated levels of tin occurred in other urban sites (e.g., San Diego Bay) and may be related to boating activity. Also shown is cadmium for reference sites (Fig. 48e) which is elevated at Dana Point.
Nisqually Reach, Washington has been one of our reference sites, but was excluded in background determination because some elements (e.g., copper, mercury, and zinc) were elevated beyond normal reference-station background concentrations for its low percent of fine sediment. Figures 47a and 47b show the elevated values for copper and zinc at Nisqually Reach. Also because the Bodega Bay reference site contained elevated concentrations of nickel and chromium, it was excluded in determination of background concentrations for only these elements. We left Bodega Bay in the plots for all other elements because these concentrations were below the calculated background concentrations and because the large number of sites added to our discriminatory power for such background determinations. If Bodega Bay was left in for chromium, the background concentration would have been 1100 (vs. 300) µg/g (dry wt.), and the background concentration for nickel would have been 80 (vs. 30) µg/g (dry wt.). Overall it seems that some of our nonurban sites that we have designated as "reference areas" may in fact receive contaminants from local sources or from the global spread (mobilization) of elements. Because both nickel and chromium have been shown to be highly correlated with percent fine sediment (clay plus silt; particles < 63 µm) (Horowitz and Elrick 1987), our lack of correlation could be due to these elevated concentrations at Bodega Bay which would cloud the underlying association. Without Bodega Bay, the correlation of nickel and fine sediment in Pacific coast reference sites improves from r = -0.32 to r = 0.48 and chromium improves from r = -0.59 to r = 0.29.
The stations (each site consists of 3 station measurements for each year analyzed) above the baseline for all years (listed by site) are included in Table 5 which shows the mean (sd) concentration of each element for all overages. If a site had more than one measurement above the baseline, the mean (sd) of all overages were reported, along with the number of overages. The total number of measurements is variable for each station and can be obtained from Figures 3-8. Our assessment of background concentrations of elements in sediment (Table 5) agrees favorably with those devised by Katz and Kaplan (1981) except for chromium, and those elements in our study requiring grain size adjustment (copper, selenium, and zinc). Our background concentrations for some elements also agree closely with values derived from NBSP sites on the East Coast of the United States (Hanson et al. 1993) .
Mean values in Table 5 and Figures 3-8 may not agree in every case because only those stations which were above the background concentration were included in the mean for Table 5, whereas all values were used to generate the means for Figures 3-8. Also, chromium and nickel background concentrations were calculated without Bodega Bay and all background calculations were done without Nisqually Reach as a reference site because of elevated concentrations of many elements. There is no statistical significance to these overage concentrations; the background concentration is an absolute value determined from all reference sites. If the concentration of an element from an urban site was higher than this reference concentration, it was included in Table 5. Overage concentrations that are close to the background concentrations should not be given as much importance as those several times higher than background.
An examination of Table 5 shows some interesting patterns. It can be seen that certain sites, such as Oakland estuary, south San Diego Bay, west Santa Monica Bay, and others, exceeded background levels for several elements. By a weight of evidence argument, it can be concluded that these sites are severely contaminated. In some cases an element's background concentration was exceeded at a site many times, essentially at each station (A, B, and C) for all years. For example, sites such as south and north San Diego Bay, and Hunter's Point show that some elements exceeded background concentrations routinely.
Looking vertically on Table 5, one can see that copper, lead, nickel, and zinc are elevated beyond expected natural concentrations at many sites. Additionally, a number of sites displayed overages with mercury, silver, and tin. Elements such as cadmium, antimony, and selenium do not appear to be elevated at most sites because of single overages and close to background concentrations. Other sites displayed overages for only a few elements but are also considered contaminated, depending on the amount above background. Sites such as Long Beach were slightly over background for some elements (e.g., cadmium) but were very high for others (e.g., lead). These excess concentrations in sediment are very important because sediments can act as a source of elements to the overlying water and enhance dissolved concentrations far above those observed in nonurban areas (Flegal and Sañudo-Wilhelmy 1993). The consequences of this include the potential for increased deleterious effects to water-column organisms, especially those migrating in from outside the estuary that are not adapted to the higher dissolved metal concentrations.
Extractable metals--Shokes and Mankiewicz (1979) and Katz and Kaplan (1981) discuss the hypothesis that most of the acid-extractable metal is that associated with the outside of the sediment particle and hence related to excess additions as a consequence of anthropogenic activity. Therefore, lattice bound elements are part of the particle's natural geochemical structure and are considered as background or baseline. To test this hypothesis, we subjected selected sediments from Cycle VI (1989; not reported in this technical memorandum) to a 1 N HCl acid extraction. Although these sediment samples were not from the years reported in this technical memorandum, the samples were from the same sites and were expected to be essentially the same as those sampled in cycles I - V. We hypothesized that reference areas (e.g., Bodega Bay, Dana Point, etc.) would produce very little acid-extractable metal because most of the total metal would be lattice bound. As a result, sites with a high acid-extractable concentration to total sediment concentration ratio (HCl:Tot) would be expected to contain excess amounts of that particular element on the outside of the particle. This excess accumulation could be due to ligands, such as organic carbon coatings and manganese and iron oxides, complexing metals from the water column where concentrations may be elevated due to natural or anthropogenic processes. Acid at a 1N concentration would release elements complexed to surface ligands but not those associated with the mineral lattices.
The concentration of the total element in sediment (Tot), the concentration of the element in the HCl-extractable fraction (expressed as total µg extracted normalized to sediment weight), and the ratio of total to extractable for each are shown in Figures 49a-k. For elements such as silver, cadmium, lead, and nickel, the ratio of HCl:Tot was high at most sites, especially those urban sites identified as contaminated (Table 5).
Commencement Bay produced the highest levels of extractable arsenic and the highest HCl:Tot ratio (Fig. 49a), which may be related to a nearby copper smelter which operated in the area for decades and released large amounts of arsenic. Other sites also produced high concentrations of HCl-extractable arsenic, and the lowest levels were found in reference sites (Bodega Bay and Dana Point).
The concentrations of total and HCl-extractable cadmium and the HCl:Tot ratio were generally highest at urban sites and relatively low at the reference sites (Fig. 49b) although the ratio at Bodega Bay was high. This is a strong indication that some of the cadmium in sediments may be due to water-borne or enriched particle input (possibly due to anthropogenic sources) leading to accumulation in sediment. Hence, the level of labile cadmium is high, as evidenced by the acid extractable concentrations. The relatively high ratio at Bodega Bay also indicates some accumulation in the sediments and Figure 48e shows possible enrichment of total cadmium at this reference site. Because some of the Bodega Bay sediments containing a very low percentage of fine particles have total cadmium concentrations less than 0.1 µg/g, other Bodega Bay stations in the range of 0.3 to 0.5 µg/g may be due to excess accumulation.
Extractable chromium at some urban sites like San Diego (both north and south), indicate possible enrichment of the sediment as shown by the HCl:Tot ratios (Fig. 49c). Bodega Bay, even with its very high total chromium concentrations, produced relatively low concentrations of acid-extractable metal indicating that these high concentrations are probably a natural geological component and not contaminant related.
Although acid-extractable copper was very high at some sites, it was surprising that the acid-extractable concentrations for some of the more heavily copper-contaminated sites like San Pedro and San Diego Bays, were relatively low (Fig. 49d). Even when copper concentrations were adjusted for grain size, sites such as south San Diego Bay with very high total copper, produced only a small amount of labile copper. Even though copper may accumulate in sediments, it may be poorly released by acid extraction (Allen et al. 1993) because of strong associations with factors such as iron or manganese oxides, organic carbon, or mineralization. Even if sediment concentrations are much higher than background levels, the low HCl:Tot ratio may indicate low bioavailability; however, these low concentrations may be sufficient to cause toxicity. Sites with total copper concentrations much higher than background levels need to be examined in more detail in order to make conclusions about excess amounts and bioavailability.
Lead was quite labile (and probably available) at the urban sites and less so at the reference sites (Fig. 49e). One exception is the Bodega Bay site with its low total mean (sd) concentration (1.7 (1.2) µg/g) for which a very high HCl:Tot ratio was observed, indicating that a large percentage of the total amount may be due to anthropogenic input. These results are consistent with the high lead concentrations found in white croaker liver from this site (Fig. 31).
Manganese (Fig. 49f) levels were relatively constant in all sites which is not surprising since it is considered to be not a contaminant, but a natural component of the sediment which occurs at high concentration. The slight variation observed may be related to grain size differences found at the various sites.
Nickel also displayed high HCl:Tot ratios at urban sites and less at one reference site, Dana Point (Fig. 49g). Unlike chromium, the HCl:Tot ratio for nickel at Bodega Bay was very high, indicating that the high sediment concentrations of this metal may be due to external complexation on sediment particles. This situation could arise if nickel were high in the water because of some process (natural or human caused) in the river systems, causing this metal to accumulate in nearby coastal sediments.
Selenium, although high at some sites, showed virtually no extractable concentrations (Fig. 49h). The highest HCl:Tot ratios occurred at the sites with relatively low total concentrations which were mainly at urban sites. Even if selenium occurs at excess concentrations in the water column and associates with the exterior of sediment particles, it is possible that the 1N HCl extraction is not the best method to assess the labile portion because of some strong controlling variable that is not affected by the acid. Conversely, selenium may not occur in excess concentrations in the water which would indicate that the variation in sediment concentrations seen in Figure 5 is due to natural geochemical enrichment.
Most sites studied displayed very high HCl:Tot ratios for silver except at Dana Point and Coos Bay, two sites with low total silver (Fig. 49i). Bodega Bay also contained low total silver, but a high extractable to total concentration ratio (HCl:Tot) which may be related to the low TOC and percent fines associated with this site. Even though the total silver concentration at Bodega Bay may be low, this total concentration was relatively high considering its sandy nature. Though the background concentration for silver is listed at 0.5 µg/g (Table 5), it appears that some sites with concentrations below this level have very high HCl:Tot ratios, which may be due to anthropogenic or natural accumulation.
Tin in sediment, although very high at some sites, especially those which may contain high levels of butyltins, did not produce high extractable concentrations (Fig. 49j). Because butyltins are unaffected by 1N HCl and have no affinity for the acidified matrix, we would not expect organotins to be mobilized into the extractable phase. However, an urban site (e.g., Hunters Point) which may contain organotins, could exhibit elevated concentrations of inorganic tin (which is mobilized by 1N HCl) as a consequence of bacterial degradation of butyltins (tributyltin, dibutyltin, and monobutyltin) (TBT > DBT > MBT > inorganic tin).
Zinc was high at many of the urban sites and the highest concentrations produced the highest levels of extractable metal (Fig. 49k). Of the four highest total concentrations, only Oakland had a total zinc concentration at the expected value, given its percentage of fine-grain sediment particles (Table 5, Fig. 47b). The total zinc concentration at the other sites, north and south San Diego Bays and San Pedro Outer Harbor, exceeded their expected concentrations by 25 to 60 µg/g . In comparing the HCl:Tot ratio for Oakland and the other high total-concentration sites, no correlation to HCl:Tot ratios could be found. The only discernible pattern was with Dana Point and Bodega Bay which contained the lowest total concentrations and produced the lowest HCl:Tot ratios. One possibility is that zinc, like copper, is too strongly bound to one or more controlling ligands to be released by 1N HCl.
In summary, it appears that acid-extraction of elements from sediments may be a useful way to determine which locations contain excess amounts of metals and metalloids of concern. However, while this technique may work for many elements, some elements, while associated with sediment because of excess concentrations, may not be released by this technique. In several cases (cadmium, lead, nickel, and silver) the high ratios of extractable metals to total metals were from urban sites that are suspected of being contaminated by excess metals; however, some metals, such as copper and zinc, may not be amenable to such extraction due to undefined complexing properties of the sediment.
Correlations of elements in sediment--A correlation matrix of log10 concentrations was used to form a dendrogram in cluster analysis in order to search for groups of elements that cooccur. We examined several subsets of the data, including Pacific coast urban sites, West Coast urban sites, Pacific coast nonurban sites (including reference sites), West Coast nonurban sites, all West Coast sites, and all Alaskan sites. Based on the metric of 1 - r (r is the product-moment (Pearson) correlation coefficient), we grouped elements with similar distributions over sites. From the dendrogram we chose 40 as the value for forming groups because it corresponded to a correlation coefficient of around 60 and because there were natural groupings below this value and few in the 40 to 50 range. When all Pacific coast urban sites (excluding reference and nonurban sites) were clustered (n = 250), one group emerged which included copper, iron, lead, manganese, and zinc (Table 6; Fig. 50a). The high correlation of copper and lead has been noted before by Meador (1990) in sediments from Puget Sound, Washington and Rice et al. (in press) in sediments from the Hudson-Raritan estuary, New York. The cause of such association of elements in sediment is due to either a natural geochemical association or common pollutant sources. It is tempting to speculate that copper, lead, and zinc might generally occur in urban areas due to common inputs (i.e., urban areas have high inputs of these metals and nonurban areas have low inputs). Conversely, manganese and iron, which are known to act as ligands which control the sediment concentration of some metals, cooccur in such high concentrations that their association is probably due to a natural geological process which varies with grain size.
An examination of the cluster dendrogram for West Coast reference and nonurban sites (Fig. 50b, Table 7) and only Alaskan sites (Fig. 50c, Table 8) shows an interesting pattern. In the West Coast nonurban sites, copper and zinc were weakly correlated (r = 0.50) compared to all Pacific coast urban sites (r = 0.87). Lead was also less strongly correlated with copper and zinc in the reference sites when compared to urban sites but was weakly correlated with cadmium. For the Alaskan sites, the copper and zinc correlation was weak and the strong correlations consisted of a group containing aluminum, antimony, iron, lead, manganese, and zinc. For all Pacific coast urban sites, zinc and lead correlated at r = 0.67 (Table 6), which was much stronger than in West Coast reference and nonurban sites (r = 0.26) but essentially the same at Alaskan sites (0.74). Also, chromium and nickel, were highly associated in all urban sites and West Coast reference and nonurban sites, but less so at Alaskan sites (Table 8). For the West Coast reference sites, there was a weak correlation of elements to fine sediment which may be due to the limited range in grain size for these sites as compared to that observed for all reference (including Alaska) and nonurban sites (Table 4). Conversely, the correlation of elements and grain size may occur only under conditions where excess elements occur in solution and can associate with sediment. If we assume that the native concentration of an element in sediment is a function of the whole particle, then the a priori assumption that surface area (as approximated by grain size) would control element concentration may not be appropriate because concentration would be a function of particle weight.
By plotting iron and manganese, two nonpollutant associated metals, separately for all West Coast (non-Alaskan) sites (r = 0.77) and all Alaskan sites (r = 0.96), it can be seen that both display a similar pattern of high correlation (Figs. 51a and 51b) leading us to believe that the natural geochemical makeup of the Alaskan sites is basically the same or very similar to that found at West Coast sites. Robertson and Able (1990) also found a good correlation (r = 0.60 to 0.72) between iron and manganese in their samples from the Gulf of Alaska. They also found a high correlation between aluminum and iron at these sites as we did in our Alaskan sites; however, we did not see a high correlation for our West Coast sites.
Figures 52a and 52b show the strong correlation between copper and zinc in all West Coast urban sites (r = 0.92) (no table shown) and weakly correlated in West Coast reference and nonurban sites (r = 0.50) and Alaskan sites (r = 0.40) (Fig. 52c). (If Nisqually Reach is included with West Coast nonurban sites, the correlation of copper and zinc increases to r = 0.69.) Notice also that the association between copper and zinc was very strong at high (contaminant) concentrations (Fig. 52a) which were in excess of those seen in Figure 52b (> 1.3 log ppm). Due to the natural geochemical similarity between Alaskan and West Coast sediments and the described pattern for copper and zinc, we conclude that these two elements have a common source in West Coast urban sites that is only weakly present in West Coast nonurban and Alaskan sites.
Lead and copper were also highly associated for all Pacific coast (r = 0.71, Table 6) and West Coast urban sites (r = 0.75) , especially at high concentrations (Fig. 53a) but only weakly correlated at Alaskan (Fig. 53b) and West Coast reference sites (Fig. 53c). We conclude from this that copper and lead are associated because of common urban sources which are only weakly present in nonurban West Coast and Alaskan sites. The basic pattern of association is present in Alaskan sites, but is weak because of a few sites (e.g., Skagway and Nakhu Bay) which are elevated with lead only. Although not high by site standards, the Alaskan sites, Skagway and Nakhu Bay, appear to be contaminated with lead, possibly from mining operations which are known to occur in the area. Lead and zinc are also highly correlated (r = 0.74) in all Alaskan sites which seems to be driven by the sites Skagway and Nakhu Bay and may be due to a common source, such as mining (Figure 54) .
It is interesting to note that the most highly correlated elements (copper, zinc, selenium and lead) are also correlated with grain size, which occurs singly for each element (Fig. 47 and Table 4). Examination of urban sites indicates a correlation of percent fine sediment with either copper, lead, or zinc leading us to the conclusion that when elevated above background concentrations, these metals cooccur due to their accumulation in fine sediment of urban sites. This is in addition to the association of these metals to each other and sediment fines in reference and nonurban sites which may be due to either natural geochemical associations or accumulations from distant anthropogenic sources. It is possible that there is a global spread of certain elements which can cause elevations in concentrations at remote sites (e.g., our reference and nonurban sites) which would lead to similar, but weaker, correlations observed for urban sites. It is possible that some property associated with small grain size (e.g., surface area, manganese or iron coatings, or TOC) allows these elements to accumulate when they occur together from a common source (e.g., an urban area). One can readily see that urban areas with high concentrations of copper, zinc, and lead, could lead to high correlation coefficients for these elements.