The National Benthic Surveillance Project (NBSP) was designed to provide information on levels of chemical contaminants in surface sediments in coastal and estuarine areas, and the levels of these same contaminants in selected marine organisms. Information was also gathered on the prevalence of pollution-associated pathological conditions in representative bottom fish and the relationship of those abnormalities to environmental contaminants. Sediments serve as a repository for a large number of environmental contaminants, including those relatively water insoluble organic compounds and the many metals and metalloids that are released due to human activities (McCain et al. 1988, Flegal and Sañudo-Wilhelmy 1993, Hanson et al. 1993).

Among the goals of the NBSP are assessments of environmental quality of coastal areas in the United States. Such knowledge is essential for effective management of the Nation's highly productive coastal habitats and the resources they support. To date, the NBSP has been highly successful in generating an extensive overview of the present status of environmental quality in coastal waters. However, the variability of many of the measured parameters in urban areas points to the need for continued collection of chemical data to provide a sound statistical basis for the analyses of trends of pollutants in sediment and organisms.

Because of the broad geographical coverage of this study, it has been necessary to sample a limited number of sites in each location (e.g., a major embayment). Hence, the data from these individual sampling sites are not intended to describe the environmental status of entire embayments nor target "hot spots" of contamination. However, these data can be used to identify polluted sites in an embayment which can be investigated more intensively in the future. Included are descriptions of sampling strategies and analytical methods, detailed presentations of the toxic element data, principal findings, and statistical evaluations.

High concentrations of elements in sediment are a potentially serious threat to organisms because of the possibility for uptake through diet or by respiratory exchange from water. Invertebrates, which serve as dietary sources to fish, have been shown to take up elements from sediment (Bryan 1985, Bryan and Langston 1992). High concentrations in invertebrate prey therefore have the potential to be a main route of exposure for fish. Other work has shown that sediments can be a source of dissolved elements to overlying water which facilitates exposure to water-column organisms (Flegal and Sañudo-Wilhelmy 1993). There are very few studies, if any, which examine the relative contribution of water, sediment, and diet to body burden accumulation in fish.

Levels of contaminants in bottom fish generally reflect chemical contamination over a wider geographical area than do levels in sediments or in sedentary organisms due to the mobility of the fish and their consequent ability to act as integrators of certain chemicals from a variety of sources. Bottom fish from a number of chemically contaminated coastal areas have been reported to have pathological conditions, including liver lesions such as neoplasms (McCain et al. 1977; Couch and Harshbarger 1985; Murchelano and Wolke 1985; Myers et al. 1987, 1993) and fin erosion (Wellings et al. 1976, Sherwood and Mearns 1977) which have been shown to be associated with chemical pollution. Thus, the presence of selected chemical contaminants and certain pathological conditions in fish have become useful indicators of environmental degradation. Cause-and-effect relationships between metal contaminants and biological function in various taxa have been inferred from field surveys (for a review see Langston 1990), and further substantiated by long-term laboratory studies (for a review of fish literature see Sorensen 1991).

This technical memorandum summarizes and interprets the results for the first 5 years of the Pacific coast portion of NBSP. It builds upon previous reports and technical memorandums (Varanasi et al. 1988, 1989a), and it complements similar reports on organic chemical contaminants (McCain et al. 1994) and fish pathology (Myers et al. 1993).

The selection of sites for inclusion in the monitoring project was based on the following site characteristics:

• Availability of bottom-feeding fish
• Subtidal, sedimentary-depositional zone
• Site away from point sources or authorized dumpsites
• Site not subject to dredging, scouring, or slumping.

Some sites were specifically located in areas that integrated inputs from multiple sources but were not directly adjacent to any known point sources. The reference and nonurban sites were in areas remote from known local inputs, although they may reflect a wider spread of regional background contamination (i.e., global spread).

Samples were collected from 51 Pacific coast sites, with complete sets of samples collected from 37 of these sites (Table 1). The 14 remaining sites were sampled for exploratory purposes only. The locations of these sampling sites, the types of samples collected, the frequency of sampling, and the site abbreviations are presented in Table 1. In this technical memorandum, we consider the Pacific coast to include all sites, including Alaska, and the West Coast to include only those sites in Washington, Oregon, and California. (Three sites were actually from the Arctic but are considered with the Pacific coast sites for simplicity.) Figures 1 and 2 illustrate the location of the sampling sites. Of the 37 complete sites, 27 of these sites were in or near urban embayments, and the remaining 10 sites were in nonurban embayments, four of which served as reference (comparison) sites. The reference sites were selected for minimal contamination while having the same fish species as obtained at the urban sites. Examination of these reference area sediments and fish thus assisted in interpreting the significance of elevated concentrations that were observed in the respective sediments and fish from the urban sites.

Following is a brief description of each sampling site.

Alaska

The Oliktok Point site is in shallow water west of Oliktok Point which is the eastern edge of the Colville River delta and near one of the North Slope sealift barge staging areas in the Beaufort Sea.

The Endicott Field site is located west of an earthen causeway serving petroleum production facilities at Endeavor Island in the Beaufort Sea north of the Prudhoe Bay oil field.

The Chukchi Sea site is offshore of the location near Kotzebue where a major ore (lead and zinc) transshipment facility will be constructed for the world-class Red Dog Mine.

The Port Moller site is in western Bristol Bay just off the entrance to the community of Port Moller which has been recommended for development as a support base for southern Bering Sea petroleum exploration, development, and production.

The Dutch Harbor site is near a major support base for an international bottom-fish fleet which includes major container loading piers and small shipyards built on the site of a large World War II naval facility.

The Kamishak Bay site is located south of the active Augustine Island volcano and is an important commercial fishing and integration site for Cook Inlet, a major marine petroleum production area.

The Port Valdez site is along the northern side of a narrow fjord at the head of Prince William Sound which serves as the marine terminal for the Trans-Alaska Pipeline System and its associated ballast water treatment facility. The new town of Valdez on the north side has developed a small boat facility and receives commercial freight traffic.

The Lutak Inlet site (near Skagway) is a reference site for the Alaskan sites. It is located between the ports of Haines and Skagway in an arm off Lynn Canal.

The Nahku Bay site is in an embayment about 2 km north of the port of Skagway near the head of Taiya Inlet.

The Skagway site is in the harbor area at the port of Skagway adjacent to the ore loading docks which recently have seen a shift from mineral and commercial commerce to large cruiseship operations.

The Boca de Quadra site is in a deep fjord once selected as a potential location for submarine mine tailings discharge from the proposed world-class Quartz Hill molybdenum mine.

Washington

The Elliott Bay site is in a major commercial shipping harbor of the port of Seattle north of Harbor Island between the West and East Waterways that constitute the mouth of the Duwamish River.

The Commencement Bay site is in the major commercial shipping harbor of the Port of Tacoma located between the mouths of the heavily industrialized Hylebos and Blair Waterways.

The Nisqually Reach site is the reference site located off the mouth of the forest- and rural-bordered Nisqually River in southern Puget Sound.

The Columbia River Estuary site is north of the town of Astoria, Oregon between Desdemona Sands and the Washington shore representing a dynamic but depositional environment of the lower Columbia River.

Oregon

The Youngs Bay exploratory site is a small tidal estuary immediately west of Astoria impacted by earlier fish processing, lumbering, and coal gas processing facilities.

The Coos Bay site represents three stations along the estuarine portions of the Coos River from the shipping port of Coos Bay, with its lumber related industries and light manufacturing, to its northern reach off the town of North Bend.

California

The Humboldt Bay exploratory site covers the estuary from South Bay north to Arcata Bay, including the Mad River Slough and the Arcata and Eureka Channels.

The Bodega Bay reference site is in the bight between Bodega Rock and the Doran Beach shoreline. This rural area is experiencing an accelerated population growth; however, no known pollution point sources exist, except for residential inputs.

The Farallon Islands exploratory site is approximately 30 miles west of San Francisco Bay and just southwest of Southeast Farallon Island.

The Islais Creek exploratory site is a heavily industrialized, partially-dredged ship channel in southern San Francisco Bay which receives a combined sewer overflow discharge (Chapman et al. 1986).

The Hunters Point site is located along the edge of the tidal flats south of the site of the former San Francisco Naval Shipyard and east of Candlestick Park where it receives the shoreside discharges from the communities and industries along the west side of San Francisco Bay. Stations are located on both the tidal flats (less than 1.5 m depth) and in the undredged channel (greater than 6 m).

The Redwood City site is at the southern extremity of San Francisco Bay in the undredged channel off the mouth of Redwood Creek; this location integrates the surface discharges from the communities around San Jose and Silicon Valley.

The Oakland Estuary site, sampled in Cycles IV and V, is the restricted partially-dredged Inner Harbor channel between Oakland and Alameda near Coast Guard Island which represented the contaminant impacts from historical heavy and light industry, wharfs, docks, and marinas, and urban development.

The Oakland site, sampled in the first cycle, is immediately west of the Alameda Naval Air Station south of the Oakland Inner Harbor entrance.

The Southampton Shoal site is just east of the shipping channel to upper San Francisco and San Pablo Bays located about equidistant between the refinery town of Richmond and Angel Island. This location is influenced on the surface by outflowing fresh water from the Sacramento and San Joaquin River systems and at depth by upwelling of oceanic water entering through the Golden Gate.

The Castro Creek site is a shallow tidal estuary with known industrial and refinery contamination northwest of Point San Pablo on the southern entrance to San Pablo Bay.

The San Pablo Bay site is in eastern San Pablo Bay south of Mare Island Naval Shipyard where Carquinez Strait discharges the Sacramento and San Joaquin Rivers and is locally impacted by agriculture, urbanization, shipping, and petroleum refining.

The Moss Landing exploratory site is at the head of the submarine Monterey Canyon off Moss Landing where the primary potential impacts are a power plant, a marina, and local agricultural runoff.

The Monterey Bay site is northeast of Monterey in open water with only modest localized residential and agricultural impacts now that the once-thriving sardine canneries have closed.

The Estero Bay exploratory site on the central California coast is offshore southwest of the entrance to Morro Bay beyond the 10 m depth contour. English (Pleuronectes vetulus) and petrale sole (Eopsetta jordani) were found here but no white croaker (Genyonemus lineatus). A limited sampling effort more inshore during Cycle I was unsuccessful.

The San Luis Obispo exploratory site is in open water at 50 m depths due south of Point San Luis. White croaker were found here but no English sole were caught. This and the previous site were investigated as potential central California reference sites.

The Channel Islands site, an exploratory site between Santa Cruz and Santa Rosa Islands south of Santa Barbara, was investigated as a possible reference site but was found to have insufficient target fish species.

The West Santa Monica site is in deep water (>50 m) north of the Hyperion sewage discharge.

The East Santa Monica site is off Manhattan Beach between the El Segundo power plants and the Redondo Beach marina.

The San Pedro Canyon site sampled during Cycle I is in open water south of the San Pedro breakwater.

The San Pedro Outer Harbor site is located in the undredged outer harbor between the San Pedro breakwater and the main section of Los Angeles Harbor and is impacted by heavy industry, shipbuilding and repair, fishing and recreational boating, shipping, petroleum production and refining, and urban runoff.

The Long Beach site is in the main undredged Long Beach Harbor approach (one of the largest ports on the West Coast) with similar types of contaminant inputs as for the San Pedro Outer Harbor site. The Long Beach Naval Shipyard and Naval Station is located between this and the previous site.

The Cerritos Channel site is located in the east basin turning area, which is just off the main channel in San Pedro Bay.

The Seal Beach site sampled in Cycle I is nearshore between the breakwater to Alamitos Bay and the Island Chaffe petroleum production platform.

The Dana Point reference site is in open water immediately off the Dana Point Marina breakwater. No local pollution point sources (besides occasional recreational boating discharges) are known for this area; however, the southward flowing longshore current may carry materials from the Los Angeles area.

The Dana Point Inside site is an exploratory site inside the breakwater at Dana Point and close to a large marina.

The Oceanside exploratory open-water site is 30 miles north of San Diego and was sampled in Cycle IV. This sampling was an unsuccessful attempt to obtain black croaker (Cheilotrema saturnum) for reference comparison to the San Diego sites.

The Outside Mission Bay site (Cycle V) offshore of Pacific Beach in the San Diego area was a more successful site for obtaining black croaker.

The Outside San Diego site is in open water west of Silver Strand Beach (Coronado) and was to represent an integration area immediately outside the San Diego Bay entrance; however, adequate numbers of target fish species were not available.

The North San Diego site is an urban site south of Lindbergh Field off the east end of Harbor Island in the northeast corner of north San Diego Bay.

The South San Diego site is outside and northeast of the main dredged channel, immediately adjacent to the piers largely devoted to shipbuilding and repair, between the Coronado Bridge and the 28th Street Pier (approximate northern boundary of the San Diego Naval Station).

The National City site is along the west edge of the channel in south San Diego Bay opposite the Seventh Street Channel. Stations included both tidal flats (<4 m) and channel (>10 m).

The Shelter Island site is north of the dredged channel adjacent to Shelter Island in northwestern San Diego Bay.

The West Harbor Island site is immediately adjacent to the western end of Shelter Island in north San Diego Bay in proximity to numerous recreational boating marinas.

The sites along the coasts of Washington, Oregon, and California were sampled using the NOAA ship McArthur (S-330), the research vessel (R/V) Harold W. Streeter or the R/V Sea Otter. The NOAA ship Miller Freeman (R-223) and the U.S. Fish and Wildlife Service vessel Curlew were used in Alaska. To maximize comparability of data over space and time, standardized collection gear and sampling methods were used. All vessels were equipped with navigational equipment to determine the location (latitude/longitude) of all the sampling stations. Sampling was conducted during a May-September field season each year. We collected sediment and fish in each of the 5 years covered by this technical memorandum, but report fish tissue concentrations only for the first four years (1984 through 1987). Sediment concentrations were determined for each year; however, fish tissues were not analyzed for cycle 5 (1988).

At each sampling site, surface sediments (top 2-3 cm) were collected with either a modified Van Veen grab sampler (0.1 m²), Smith-McIntyre grab (0.1 m²), or a box corer (0.3 m²). Three grabs were taken at each of three stations and three cores (15 to 19 cm x 3 cm diameter plastic tubes) were taken from each grab (total of 27 core tubes per site). Cores were put on ice and frozen at -20°C within four hours for future analyses.

Bottom-dwelling fish were collected with an otter trawl (7.5 m opening, 10.8 m total length, 3.8 cm-mesh in the body of the net, and 0.64 cm-mesh in the liner at the cod end, none of which was chemically treated) using a trawl time of 5-15 minutes. Fish of a minimum size (generally > 15 cm in length) were randomly selected from each haul and assigned a unique identification number. These fish were kept alive (to prevent autolysis of tissues) until necropsies were performed. The necropsy procedure for each fish involved a) weighing, b) measuring, c) removing otoliths for age determination, and d) excision of the liver and stomach contents for elemental analysis. Approximately 30 fish per target species (Table 2a) were necropsied at each site. The liver tissue was placed in a separate, acid-rinsed plastic vial and frozen at -20°C for analyses at a later time. The remainder of the fish tissues were reserved for organic chemistry, histopathology, and bioindicator analyses. The total contents of stomachs from at least 10 fish per target species at several sites (Table 1) were removed and composited in an acid-rinsed glass jar and frozen for chemical analyses.

Analysis of Elements in Sediment and Tissue

The particle size characteristics and total organic carbon (TOC) content of sediment samples were determined by personnel of the Southeast Fisheries Science Center, National Marine Fisheries Service (NMFS).

For elemental analysis in sediment, one core tube from each grab was thawed at room temperature and the excess water drained from the bottom (ca. 30 min.). The top 2 cm of the sediment core was extruded with the outermost surface layer of sediment in contact with the plastic core tube being discarded and the wet sediment from each grab oven dried at 85°C. Equal weights of the three dried samples from replicate cores were composited into a single sample for each individual station. Hence, we analyzed three station composites (three grabs per station) for a total of three separate analyses per site. The sediment was subjected to a total acid digestion with 6 mL of hydrofluoric acid (HF) and 2 mL of aqua regia (hydrochloric acid (HCl): nitric acid (HNO₃) 3:1) (HF, HCl and HNO₃, ultrapure, Seastar Chemicals, Sidney, British Columbia). The sediment/acid mixture was sealed in a Teflon vial and heated in a stainless steel Parr bomb at 120°C for 36-48 hours. The digestate was quantitatively transferred to a 50 mL volumetric flask and brought to volume using saturated boric acid (H₃BO₃ (Puratronic), Aesar/Johnson Matthey, Ward Hill, Massachusetts).

Recent work has shown that fish length can have a strong influence on the concentration of some elements in tissues (Evans et al. 1993); therefore, we sampled the livers of the three largest female fish to give a total of three analyses per site (Table 2b). The fish lengths from most sites were relatively close to the mean with little variation; however, high variation occurred in starry flounder because occasionally we would catch a very large individual.

For analysis of elements in fish tissue, livers from three individual female fish were completely digested with HNO₃ (7 mL acid for ~ 0.25 g of tissue, dry weight). The liver/acid mixture was sealed in a Teflon vial, placed on top of a drying oven set at 110°C and left overnight. After this initial digestion, the sealed Teflon vials were placed inside another container and were subjected to a three-step microwave digestion (1 min. at 50% power, short pause, 4 min. at 50% power, 15-30 min. pause, and finally 4 min. at 50% power). When the digestate had cooled, a total of 3 mL of hydrogen peroxide (H₂O₂) (ultrapure, Fluka Chemical, Ronkondoma, New York) was added in increments of 1 mL over three 1-hour time intervals. The final digestate was brought to 25 mL volume using Milli-Q water (Millipore Corporation, Bedford, Massachusetts).

Stomach contents samples for each site were obtained from a composite of 10 fish. Analysis of elements in stomach contents was performed with a protocol which combined the sediment and tissue digestion procedures. First, 0.25 g dry weight of stomach contents were digested according to the liver protocol. Then 6 mL of HF and 1 mL of HCl were added to each vial. The sealed vials were then placed in the stainless steel Parr bombs and heated at 120°C for 48 hours. After cooling the digestate was brought to 50 mL volume using saturated H₃BO₃.

The analyses of all three sample types were conducted via the techniques of atomic absorption spectroscopy (AAS): flame, heated graphite atomization, hydride generation, and cold vapor for mercury (see Table 3). Initially a Perkin-Elmer Model 5000 atomic absorption spectrophotometer, equipped with deuterium background correction, was used. For later cycles, some elements were analyzed with a Perkin-Elmer Model 5100 atomic absorption spectrophotometer equipped with Zeeman background correction. Instrument calibration curves were made with a minimum of five concentrations using single-element standards prepared by diluting 10 mg/mL of certified standard solutions (National Institute of Standards and Technology (NIST)). Computerized least squares linear or quadratic fit was determined for all samples falling within the end points of the calibration curves. All elemental concentrations in sediment and tissue have been reported based on a dry weight basis. Additional details on sample preparation and AAS analysis can be found in Robisch and Clark (1993).

Quantification was achieved by using solution blanks and certified reference materials (CRM) provided by the National Research Council of Canada (NRCC), Ottawa, Ontario, Canada; the U.S. Environmental Protection Agency (EPA), Cincinnati, Ohio; and the National Institute of Standards and Technology, Gaithersburg, Maryland. These materials are called standard reference materials (SRMs) by NIST. The CRMs included both marine sediments and various fish and invertebrate tissues and were processed in the laboratory alongside our field samples. These reference materials were used to correct our sample concentrations for loss or enhancement caused by sample preparation and digestion or instrument interferences. These corrections were accomplished by a response factor (RF) which was our measured concentration of the CRM divided by the published certified value. These RF were applied to all measured values.

For tissue, we analyzed the CRMs DORM-1, DOLT-1, and TORT-1 (NRCC), 1566 and Bovine liver (NIST), and "Trace Metals in Fish" (EPA). For sediments we analyzed the CRMs MESS-1, BCSS-1, and PACS-1 (NRCC), in addition to 1645 and 1646 (NIST).

Extractable Elements

A select number of sediments from the 1989 (Cycle VI) sampling effort for NBSP were subjected to acid leaching in order to assess the difference between total and labile concentrations. (These total concentrations were not part of other statistical analyses in this technical memorandum.) We attempted to test the hypothesis that the concentration of an element in the weak-acid extractable fraction, as a percentage of the total concentration found in the bulk sediment, may indicate excess amounts due to urban inputs when compared to background concentrations. Katz and Kaplan (1981) propose that leachable metals may be an indication of anthropogenic input because the excess metal would be associated with the exterior of the sediment particle and would not be a part of the lattice structure which could be released only through harsh acid digestion. This is exclusive of metal particles from human activities which would not yield high concentrations of leachable metals because of their refractory nature. The amount associated with the exterior of sediment particles may be available for bioaccumulation either through uptake from water (of that portion in equilibrium with surrounding water (overlying or interstitial water)) or available to organisms from ingested sediment.

Sediment samples from all three stations (A, B, and C) at seven sites (plus two reference sites; Bodega Bay and Dana Point) were analyzed. We took about 5 g of wet sediment, added 50 mL of HCl to make the final solution of 1 N, agitated the slurry for 3 hours, and allowed it to stand for 24 hours. The pH was measured after 24 hours of standing and was always 0.3 to 0.9. Solutions were analyzed by AAS, corrected for matrix effects, and related to the sediment concentrations in micrograms per gram dry weight (µg/g dry wt).

A variety of statistical methods were used to analyze the data from the chemical analyses in order to present the large amount of data in a graphical format. This was done to facilitate interpretation and allow us a way to evaluate possible interrelationships among the concentrations of the elements in sediments and fish.

Screening of Chemical Data for Accuracy

In accordance with Little and Smith (1987), chemical data were initially examined using graphical methods for multivariate outlier detection. Outlying samples were located and identified using three-dimensional scatterplots (see Huber 1987), and the accuracy of their constituent values was confirmed by reference to original records. Where analytical errors were identified, original data were recalculated. Our data contained only a few values (2-3% of the total) below their respective detection limits; hence, we did not use any censored-data algorithms in estimation of statistical parameters. Those concentrations below their detection limit were treated as equal to the detection limit value.

Calculation of comparison intervals in graphs--In order to compare the concentrations of elemental contaminants in sediment and fish tissues at sites along the entire Pacific coast, comparison interval graphs were constructed. For comparison of elements over sites and over species, we used GT2 plots and floating-bar plots which are modified GT2 plots (Landahl 1994). Below is a general discussion on GT2 plots followed by specific information for each type of plot.

While it is relatively easy to calculate 95% confidence intervals for mean concentrations based on the number of samples (n) and the variability for each mean, it is not as easy to infer whether two means from a group of means are the same or different by examining plots of confidence intervals. Instead, a statistical technique, similar to the confidence interval but more closely related to an analysis of variance, was employed for graphical comparisons. Therefore, we generated GT2 plots which involved calculating and plotting a "comparison interval" (Gabriel 1978, Sokal and Rohlf 1981) for each mean, based on:

• the number of samples for that mean,
• the variability about that mean, and
• the number of means being compared.

When sample sizes for the means of interest are unequal, it is desirable to use the GT2 method for calculating the comparison intervals. Also, the span of a comparison interval depends on the within-group variability in the entire data set and on the number of samples in each category. For a particular data set, categories with the same number of samples have identical comparison interval spans regardless of their individual means and standard deviations.

When data for only two stations at a site are available, the comparison interval will of necessity be quite large for the same reason that a confidence interval would be. That is, the value of the t-statistic is large, because of the uncertainty of variability about the mean when n = 2. When n = 1, the comparison interval is indeterminate. Thus, the method was used only when sample data were available for at least two stations at a site.

Floating-bar plots--Some of the data have been presented as a modified boxplot showing the arithmetic mean (thick horizontal bar) and one standard deviation (top and bottom of box). The horizontal bar on the vertical line ("whisker" ) represents the geometric mean of the data and the length of the whisker is the 95% comparison interval calculated by the GT2 method using ln (x+1) transformed data values (Landahl 1994). The position of the horizontal bar on the whisker gives some indication of the distribution of the data around the geometric mean. The number of analyses are listed (usually above) for each site and the bottom of the horizontal gray line across most figures is the upper comparison interval (UCI) for the contiguous West Coast reference site (either Bodega Bay, Dana Point, or Nisqually Reach) having the highest concentration for the element plotted. On these plots, two sites are significantly different when their comparison intervals (whiskers) do not overlap. The numbers for these plots were back transformed to arithmetic values for plotting and hence have asymmetrical comparison whiskers.

GT2 plots--GT2 plots were constructed for the figures comparing tissue, sediment, and stomach contents. The top of each bar is the mean concentration for an element over all years and sites. The vertical whisker is the comparison interval. Nonoverlapping bars indicate significantly different means at alpha<0.05. For these GT2 plots we used log₁₀ values for calculation of the mean and comparison intervals and plotted them as log values. These plots were constructed with log values because we were mainly interested in comparisons, whereas in the floating-bar plots we were interested in showing the different concentrations for each site in addition to making comparisons. The symmetrical comparison interval whiskers about the mean also allowed easier judgment of variability among compartments. The left ordinate shows the plotted log values in parts per million (µg/g) and the right ordinate on each plot shows the arithmetic equivalent .

Determining background sediment concentrations--To determine if an element's abundance was correlated with percent fine sediment we plotted each element from all reference sites (except Nisqually Reach) plus all nonurban sites and a few exploratory sites that were judged acceptable (11 sites total) against the percent fines (percent of clay plus silt; i.e., < 63 µm particle diameter). This is similar to Hanson et al. (1993) who used aluminum as a comparison variable; however, we tested every element separately with percent aluminum and found only one (antimony) was significantly correlated.

Many elements varied significantly with percent fines including iron and manganese. For those elements (copper, selenium, and zinc) that displayed a strong correlation to percent fines, a linear regression equation was developed with log concentrations in order to express the relationship and to characterize the amount of that element that would be expected given variable levels of percent fine sediment. For these elements a line was drawn above and parallel to the line given by the regression equation in order to include all reference sites. This line, extracted from the regression equation, was greater than two times the upper 95% confidence interval for the regression. For those elements which showed no correlation to percent fines, a horizontal line was drawn which was above all reference stations. If the concentration of an element from an urban station was found to be above this line, we considered that station to be contaminated, that is, above the background concentration that was expected for that element in a relatively pristine, nonurban area. If the concentration of an element was above this baseline concentration for a station in any given year, it was included in a table which lists sites by elements and their background concentration. If a site contained more than one station measurement above the baseline, its mean and standard deviation (sd) was reported, along with the number of measured values (3 stations analyzed/site/year). The total number of measurements is variable for each site and can be obtained from Figures 3-8.

Methods for correlations--Product-moment (Pearson) correlation coefficients between elements in the different matrices (sediment and tissue) over sites were determined in order to discover which elements, if any, were highly correlated. For example, correlation coefficients were determined for all pair combinations (n=153) of the 16 elements measured in all sediment samples plus percent fines and TOC. These coefficients are displayed in tables and the significant correlations are highlighted bold. This information was used to assess sources of elements and geochemical associations. Significant correlations were determined by setting alpha = 0.05 and then correcting for the number of comparisons to allow for comparisonwise error (alpha + number of comparisons) (Milliken and Johnson 1984). For the example above of n=153, the alpha level for significance became 0.00033.

Cluster analysis was performed by the following method. First a product-moment correlation (r) matrix was calculated for all log₁₀ transformed elements of interest (usually 16 elements) and for selected cases (e.g., Alaska only or reference sites only). The few missing values were replaced by mean substitution, which had little or no effect on the correlations. This matrix was then used to calculate a similarity matrix based on 1- r, which was used as the distance metric in the clustering calculations. From this, a hierarchical tree was constructed using single linkage amalgamation. The clusters are groups of elements joined by nodes and the closer the node is to vertical axis (elements) the more closely related the elements are in that cluster. The horizontal axis is scaled in percentage of the maximum to minimum distance that the clusters are joined (link/max)*100. Groups of elements were made from the dendrogram based on similarity by selecting those which fell below the 40% value which generally corresponds to a similarity (r) of approximately 0.60.

Due to the large variability in concentration found at the stations within a site and the incomplete sampling scheme (few sampling years for a given site), trend analysis was not attempted. We felt that the incomplete database would not produce reliable patterns for determination of temporal trends. When additional data are collected, we may have enough data points to test hypotheses regarding such trends.