Additional supporting data are found in appendices available in the online edition of IEAM Volume (2), Number (3). DOI: 10.1897/2005–035.1.
Sediment trend analysis (STA) is a technique that determines the net patterns of sediment movement and their dynamic behavior or stability. The data required are the complete particle size distributions obtained from bottom grab samples collected in a regular grid over the area of interest. Appendix 1 provides the particular details of how STA is undertaken. Because many contaminants are known to associate with the natural particles contained in sedimentary deposits, STA can provide additional weight-of-evidence in ecological risk assessment, remedial investigation, remediation itself, and litigation issues. The STA was applied to 242 sediment samples collected from the Hylebos Waterway, Tacoma, Washington, USA, in support of remedial action planning, contaminant source identification, and ultimately allocation of legal liability for contamination. The Waterway itself comprises a narrow shipping channel extending 3 miles from Commencement Bay (Puget Sound) where it ends in a dredged turning basin (Upper Turning Basin). A 2nd dredged turning basin (Lower Turning Basin) is located about three-quarters of the distance down its length. Both sides of the channel are home to an extensive industrial complex associated with significant contaminant releases into the water. The area was declared a Superfund Site in the early 1980s. The results of the STA showed a consistent pattern of sediment transport directed from the mouth of the Waterway to the turning basin at its head. Divided into 5 separate transport environments (TEs), the sediments within the Waterway progress from transport in Dynamic Equilibrium near the mouth, to Total Deposition (type 1) in the vicinity of the Lower Turning Basin, followed by Total Deposition (type 2) in the Upper Turning Basin. Assuming that contaminants associate preferentially with the finer, rather than the coarser, components of the grain size distributions, a probable behavior of contaminants that can be contained in the sediments is proposed for each TE. Maps showing the spatial distributions of existing contaminant data appear to conform very well to the patterns that might be expected from the STA results. This evidence was primarily used to demonstrate that potentially responsible parties (PRPs) located at the head of the Waterway could not be responsible for contaminated sediments toward its mouth. The findings, for example, effectively dismissed the assumption by the Natural Resource Damage Trustee agencies that contaminated sediments from a particular source would be as likely to migrate down the Waterway as up the Waterway. As a result, major documented sources of contamination near the mouth should be expected to bear a larger share of the total cleanup compared with sources farther toward the head. Furthermore, the STA provided explanations for apparent anomalies such as how hot spots of polychlorinated biphenyls (PCBs) could be located near a property where PCBs had never been released into the environment. If sediment gradient pattern analysis alone were used to allocate liability among PRPs, those located near such hot spots would receive a disproportionate share of liability.
Additional supporting data are found in appendices available in the online edition of IEAM Volume (2), Number (3). DOI: 10.1897/2005–035.1.
Contaminated sediments do not always remain in place in the environment. The mobility of sediments and the particle-associated contaminants contained within them are a complicating factor in ecological risk assessment, remedial investigation, remediation itself, or ultimately in litigation. Rational decision making must take into account the probable stability of contaminated sediments, their sources, transport, and ultimate fate (e.g., US Environmental Protection Agency [USEPA] 2002; Apitz et al. 2005).
One technique that might help provide such information is sediment trend analysis (STA), an empirical method that examines relative changes in the complete grain size distributions of aquatic sediments to determine their net transport pathways together with their dynamic behavior (i.e., accretion, erosion, dynamic equilibrium). Because many contaminants adsorb onto the particles that make up natural sediment, this information can aid in assessing the relationship between contaminant loadings and their sources, as well as provide an understanding of the fate and behavior of contaminants contained in the sediments.
To support remedial action planning, source identification, and ultimately allocation of legal liability for contamination, STA was performed on 242 surficial sediment samples collected in July 2001 from the Hylebos Waterway in Tacoma, Washington, USA (Figure 1). The purposes of the study were principally to identify individual sediment transport environments on the basis of sediment characteristics and their dynamic behavior and to assess the probable extent of contamination from specific sources.
The theory to predict the relative changes that will occur in particle size distributions of sediments through erosion, transport, and deposition was 1st presented in McLaren and Bowles (1985). On the basis of their theory, several methods to carry out STA have been developed. The McLaren and Bowles approach is 1 dimensional, whereby the changes in grain size distributions along individual sample sequences are tested for validity with the Z score statistic to determine the preferred transport direction. A practical assessment of this approach is discussed in Hughes (2005). Gao and Collins (1991, 1992) and Gao (1996) proposed a 2-dimensional vector approach to determine trends, some elements of which were revised by Chang et al. (2001). A different vector approach altogether was produced by Le Roux (1994) and Le Roux et al. (2002). A summary of the various techniques is provided in Rios et al. (2003).
Regardless of the technique used to derive sediment trends, the original theory of how grain size distributions should change with transport has remained undisputed since the 1985 paper by McLaren and Bowles. Briefly, the theory demonstrated that, when 2 sediment samples (d1 and d2) are taken sequentially in a known transport direction (e.g., from a riverbed, where d1 is the up-current sample and d2 is the down-current sample), the sediment distribution of d2 can become finer (case B) or coarser (case C) than d1; if it becomes finer, the skewness of the distribution must become more negative. Conversely, if d2 is coarser than d1, the skewness must become more positive. The sorting will become better (i.e., the value for variance will become less) for both cases B and C. If either of these 2 trends is observed, sediment transport from d1 to d2 can be inferred. If the trend is different from the 2 acceptable trends (e.g., if d2 is finer, better sorted, and more positively skewed than d1), the trend is unacceptable and it cannot be supposed that transport between the 2 samples has taken place.
In the above example in which the transport direction is unequivocally known, d2(s) can be related to d1(s) by a function X(s), where s is the grain size. The distribution of X(s) can be determined by
where X(s) provides the statistical relationship between the 2 deposits and its distribution defines the relative probability of each particular grain size being eroded, transported, and deposited from d1 to d2. It is the shape of the X(s) distribution relative to the shapes of the d1(s) and d2(s) distributions that determines the dynamic behavior (stability) of the sediments. The 5 defined categories for dynamic behavior are 1) Net Erosion, 2) Net Accretion, 3) Dynamic Equilibrium, 4) Total Deposition (type 1), and 5) Total Deposition (type 2). Appendix 1 (see Figure A6) contains a more thorough analysis of the theory and the procedures used in the derivation of the transport pathways.
Sediment grab samples were collected from the Hylebos Waterway during the period 12–15 July 2001 with a Van Veen-type grab sampler. This device samples the top 10 to 15 cm of sediment, from which a mixed sample of 200 to 300 g encompassing the full depth of the sampler was obtained. As discussed in Appendix 1, the mixed sample is assumed to represent an average of all the sediment derived from an unknown number of sources or directions. All samples were collected from a 12-foot (3.7-m), hard-bottom inflatable speedboat equipped with a depth sounder, a small electric winch, and a grab sampler. Positions were obtained with a differential Global Positioning System receiver with 2-m accuracy in differential mode (Trimble DS212L). In most instances, samples were obtained at predetermined locations; however, where shoreline structures (e.g., docks and marinas) and vessels interfered with navigation, samples were collected as close as practicable to the planned position. Each sample was stored in a plastic zip-lock bag and transported to the GeoSea laboratory in Brentwood Bay, British Columbia, Canada, for grain size analysis.
Samples were collected on a regular hexagonal grid with a spacing of 110 m in the outer portion of the Waterway (from Commencement Bay to the start of the Lower Turning Basin) and 55 m to include the Lower Turning Basin landward to the Upper Turning Basin (Figure 1). A total of 251 sample sites were visited, of which 9 were found to be hard ground (usually occurring when bark mulch covered the bottom) and no sample could be taken. A site was designated “hard ground” after 3 separate drops of the grab sampler failed to retrieve sediment.
All samples were analyzed for their complete grain size distribution (range 1,800–1 μm) with a Malvern MasterSizer 2000 laser particle sizer. The laser-derived distributions were combined with sieve data for grain sizes of more than 1,500 μm by a merging algorithm. The distributions were entered into a computer equipped with appropriate software to establish sediment trends and transport functions.
The Hylebos is the northernmost of several man-made waterways that make up the industrial port facilities of the Tacoma Tideflats. Extending from Commencement Bay, it runs southeast for about 5 km and is generally about 200 m across. A small creek (Hylebos Creek) enters the Waterway at its extreme southeast end. The waterways themselves are constructed in old channels of the former Puyallup River delta, which now flows into Commencement Bay through the Puyallup Waterway. Although considerably changed by 20th century industrialization, the river still carries an active sediment load that is presently forming a delta at its mouth in Commencement Bay. Its suspended load is quite clearly visible in the Bay and, at the time of sampling, was seen to extend far into the Hylebos Waterway. Drogue studies in the Waterway have shown a net inflow of saline water below 6 m and a net outflow at 2 to 6 m. The surface also displays a net inflow, although wind from the southeast can reverse this (Loehr et al. 1981).
With the exception of some tidal flats on the north side of the outer portion, the banks of the Hylebos Waterway are lined with an extensive industrial complex (Figure 2). Because of past contaminant releases into the area, the Tacoma Tideflats was declared a Superfund Site in the early 1980s.
The sediments in the study area range from sandy gravel to mud, with the largest proportion (57%) being sandy mud (Figure 3). The latter sediment type is common throughout the Waterway. Significant patches of mud are found in both the Lower and Upper Turning Basins.
In searching for patterns of sediment transport (the full rationale and technique are described in Appendix 1), it was found that the best and most consistent pathways could be generated with the use of all samples as a single complete data set. Attempts to isolate specific sediment types (e.g., mud or muddy sand) did not yield satisfactory results. It is likely, therefore, that despite the range and mixture of particle sizes present in the Waterway, all samples represent a single facies from which the transport relationships among them can be determined.
Following the calculation of numerous sample sequences, a mutually supportive and coherent pattern of transport that could account for all the sediment grain size distributions required a total of 35 lines or sample sequences (Figures 4 and 5). The trend statistics for each trend line are provided in Appendix 2. These include the Z score statistic, R2, and the derived interpretation of the dynamic behavior as determined by the X(s) function. For ease of discussion, the pathways are grouped into areas defined as transport environments (TEs; Figure 5). A TE is an area within which transport lines are associated both geographically and by their dominant dynamic behavior. Generally, transport lines cannot be continued from one TE into another (i.e., the trend from one TE into the other will become statistically unacceptable, or ambiguous), so a region in which transport lines naturally end (and begin) defines the intervening boundary.
The following provides a brief description of each TE (Figure 5) together with examples of d1, d2, and X distributions.
This TE is confined to Commencement Bay. Evidence that the samples making up this environment could be related to the sediments found immediately inside the Hylebos Waterway was lacking. This environment is composed of relatively few samples and should be accepted with caution. The trends suggest that this part of Commencement Bay is under the influence of a clockwise gyre, with the sediments originating from shoreline glacial deposits on the north side of the Bay. The nearshore trends show Net Erosion (Figure 6), suggesting that foreshore lowering with consequent coastal erosion is likely occurring. In the offshore deeper waters, sediments in the gyre generally fine to muddy sediments, and Total Deposition (type 1) appears to be taking place (Figure 7).
This environment originates at the entrance to the Waterway and extends to slightly more than one-half its length. Most of the trends are in Dynamic Equilibrium (Figure 8), a finding that is relatively rare in muddy sediments, which because of their cohesive properties, tend to form deposits of Total Deposition. The finding of Dynamic Equilibrium suggests that processes capable of resuspending the muddy sediments enable transport to continue up the Waterway. A probable process that would result in resuspension is propeller wash.
Originating in a slightly narrower part of the Waterway, these lines continue with transport up the channel and terminate in the mud associated with the Lower Turning Basin. Nearly all the lines are defined by Total Deposition (type 1) dynamic behavior (Figure 9). Unlike the outer Waterway (TE2), these sediments are evidently less susceptible to resuspension, probably because 1) the sediment has become finer and is therefore more cohesive than the sediments in the outer Waterway, 2) shipping activity might decrease with increasing distance from the mouth of the Waterway, and 3) currents are known to decrease with distance up the Waterway (Norton and Barnard 1992).
Commencing in the narrows up-channel from the Lower Turning Basin, landward transport continues into the Upper Turning Basin. Nearly all the pathways produced X distributions indicative of Total Deposition (type 2; Figure 10). Such behavior suggests that much of the coarser grain sizes in the channel have already been deposited, undoubtedly in the Lower Turning Basin, and the remaining sizes left to continue up-channel are extremely fine (i.e., all remaining sizes have an equal probability of being deposited).
Consisting of a single line of samples taken in Hylebos Creek, this environment indicates coarse sediment in Dynamic Equilibrium flowing toward the Upper Turning Basin. The influence of the creek does not appear to be particularly significant with respect to the marine sedimentation that characterizes the Waterway.
|Sediment dynamic behavior (stability)||The shape of X(s) relative to d1(s) and d2(s) (see Figure A6)||Contaminant dynamic behavior|
|Net Erosion||The mode of X(s) is coarser than the d1(s) and d2(s) modes. More grains are eroded than deposited, and sediment coarsens along the transport path.||Contaminant levels decrease rapidly down the transport path and are dispersed to areas of deposition.|
|Net Accretion||The mode of X(s) is finer than the modes of d1(s) and d2(s). More grains are deposited than eroded, and accretion occurs down the transport path.||Contaminant levels increase down the transport pathway.|
|Dynamic Equilibrium||The modes of all 3 distributions are the same. The probability of finding a particular grain in the deposit is equal to the probability of its transport and redeposition (i.e., replacement is grain-by-grain along the transport path). The bed is neither accreting nor eroding and is, therefore, in Dynamic Equilibrium.||Contaminated sediments will move down the transport pathway while remaining as a coherently defined hotspot.|
|Total Deposition (type 1)||The X(s) distribution increases monotonically over the complete d1(s) and d2(s) distributions. Sediment fines in the direction of transport; however, the bed is no longer mobile. Once deposited, no further transport occurs.||Contaminated sediments form localized hotspots that undergo no further transport.|
|Total Deposition (type 2)||X(s) is horizontal. This type of X distribution is found only in extremely fine sediments when the mean grain size is very fine silt or clay, Such sediments are usually found far from their source (compared with Total Deposition (type 1) sediments. Deposition is no longer related strictly to size sorting. All sizes have an equal probability of being deposited down the transport path.||Contaminated particles have an equal probability of being deposited anywhere in this type of environment. Hot spots do not form; rather, contaminant concentrations will be relatively equal throughout the depositional area.|
Despite considerable efforts to find reversals or a more complex pattern of sediment transport in the Waterway, the derived patterns of net sediment transport for the Hylebos show movement entirely in an up-channel direction (from the mouth toward the head of the Waterway). Furthermore, the dynamic behavior of the transport regime changes in a regular way from Dynamic Equilibrium in TE2 (outer Waterway), followed by Total Deposition (type 1) in the sediments associated with the Lower Turning Basin (middle Waterway TE3), and ending with extremely fine sediments in Total Deposition (type 2) in TE4 (inner Waterway). This progressive change in dynamic behavior suggests that sediments become increasingly finer (and increasingly cohesive) toward the upper end of the Waterway together with a decrease in current velocities. The latter is reported in Norton and Barnard (1992), in which velocities at the mouth were up to 10 cm/s compared with those at the head, where they were generally less than 2 cm/s. Floyd and Snider Inc. (1998) described the presence of underlying dense water masses that move regularly into the Hylebos Waterway as a result of upwelling from Commencement Bay. A similar stratification was observed by Loehr et al. (1981), in which flood tidal currents prevailed near the bottom and surface, with a midlayer favoring the ebb. Both current velocities and the stratification can be disturbed by vessel activities inducing mixing and possible scour (Floyd and Snider Inc. 1998).
It is quite likely that the observed changes in dynamic behavior toward the head is also influenced by a decreasing number of vessel passages in the channel (i.e., given that docks line most of the channel, more vessels can be expected to pass back and forth in TE2 (outer Waterway) creating conditions favorable for resuspension and Dynamic Equilibrium than in TEs 3 and 4, in which Total Deposition prevails). Such resuspension events in TE3 might provide an opportunity for further transport from TE3 into TE4, after which no further area is left to which sediment can be transported regardless of vessel traffic.
It is interesting that sediments in TE1 (Commencement Bay) could not be related by transport to sediments in the Waterway. In other words, there must be a significant source present in the Waterway that is not as important in the Commencement Bay environment. Two possible explanations are probably both operating. The 1st is that suspended sediment associated with the Puyallup River plume, although negligible compared with the source of sediments that are provided by a lowering foreshore along the north side of Commencement Bay, becomes a significant sediment input once inside the Waterway. Second, industrial activity and the presence of various outfalls are likely contributing a miscellany of sediment types unique to the Waterway (note that many of the isolated patches of coarse and mixed sediments are found along its banks; Figure 3).
The relationship between contaminant levels contained in sediments with the texture and stability of the sediments themselves is now known to be highly complex and is the subject of considerable research (e.g., Apitz et al. 2004). Site-specific conditions can result, for example, in a uniform distribution of contaminants throughout the particle size range of their associated sediments. In other instances, the distribution of contaminants could show bimodality, with modes associated with both fine and coarse sediment fractions. It has, however, long been recognized that many contaminants tend to associate preferentially with the finer sediment fractions as opposed to the coarser sizes and that pollutants tend to follow the same transport pathways of sedimentary material, tending to be transported to depositional sinks regardless of the exact source of contamination (Young et al. 1985).
In the context of STA, and on the basis of the assumption that contaminants will preferentially follow net sediment transport pathways, McLaren and Little (1987) predicted the accumulation and dispersal of hydrocarbons and heavy metals throughout a small estuary in southwest Wales. It was found that the relationship between the predicted concentrations in different portions of the estuary with actual measured concentrations produced a highly significant correlation (Spearman's rank correlation coefficient of 0.98, where 1.0 indicates complete agreement between the expected order of contaminant concentrations with the observed order of concentrations). Since this finding, the empirically derived relationships between contaminant levels contained in sediments and the results of STA have both improved and been supported in a number of studies (e.g., McLaren et al. 1993; Pascoe et al. 2002). These relationships, based on all the assumptions made in carrying out STA (see Appendix 1), are summarized in Table 1.
On the basis of the correlation between sediment stability and contaminant behavior (Table 1), it is instructive to consider the probable behavior of contaminated particles in the Waterway in the absence of local contaminant sources. Assuming that a source of contaminated particles enters the Waterway at its mouth, the 1st environment encountered (TE2, outer Waterway) is predominantly in Dynamic Equilibrium. Contaminated particles deposited in this environment will tend to have an equal probability of continuing up-channel transport as on a conveyor belt. Hot spots might develop at random but will tend to move toward the Lower Turning Basin given sufficient time.
The conveyor belt form of transport ends at the start of TE3 (middle Waterway), an environment dominated by Total Deposition (type 1). Here, contaminated particles will come out of transport to form 1 or more hotspots that are unlikely to be easily dispersed. Thus, the Lower Turning Basin is expected to be an important contaminant sink.
Finally, TE4 (inner Waterway) is dominated by Total Deposition (type 2), an environment in which the remaining particles in transport are sufficiently fine that they escaped deposition in the Lower Turning Basin and now have an equal probability of deposition anywhere in the Upper Turning Basin. Specific hotspots would be unlikely; however, a ubiquitous contaminant level throughout the environment would more probably be observed. It is emphasized that throughout all 3 of the Hylebos TEs, it would be very unlikely for a contaminated particle contained in the sediments to have the opportunity to move in the reverse direction toward the mouth.
In reality, the high level of industry surrounding the Waterway has resulted in numerous contaminant sources. Like sediment, the greater the amount of contaminant entering the environment, the greater the probability of its deposition in the sediment regardless of the dynamic behavior. For example, a significant contaminant source in TE2 (outer Waterway), in which the sediments are predominantly in Dynamic Equilibrium, could well form local hotspots by simply overwhelming the sedimentary environment. Although the hotspot might be dispersed in the up-channel direction, without an effective source control program, the original hotspot will be continually replenished.
To relate the findings of the STA with known contaminant levels, a contaminant database was made available to GeoSea (Striplin and Associates, Olympia, WA, USA). Because of the large size of the database and the extensive number of organic and heavy metals available, it was necessary to make various practical decisions. Some data collected during particular surveys were not described well enough in the provided documentation to be used with confidence. The database was ultimately edited to include only surface samples.
Numerous maps of various contaminants were constructed with the aid of Surfer®, a contouring and 3-dimensional surface mapping software package made by Golden Software (Golden, CO, USA). It was found that separate maps of the organic compounds generally produced similar patterns, as did separate maps of the trace metal data. For this paper, a map of total pesticides and polychlorinated biphenyls (PCBs) has been chosen to provide an appropriate illustration of the findings (Figure 11).
The relationship between the defined TEs (Figure 5) and contaminant levels (Figure 11) appears to be consistent with the expected findings as described in Table 1. In TE2 (outer Waterway), several isolated hotspots exist (shown as A, B, C, and D in Figure 11). Not all the hotspots are necessarily related to an immediate shoreline source (e.g., hotspots B and C), and these might well be interpreted as random locations in an environment of Dynamic Equilibrium, in which the contaminants are contained within a sediment conveyor belt moving toward the head of the Waterway. Hotspots E, F, G, and H are nearly all out in midchannel (i.e., no immediate shoreline source) and appear relatively isolated, as expected in an environment of Total Deposition (type 1). Finally, TE4 (inner Waterway) shows a more or less equal spread of values throughout, which is also expected in an environment of Total Deposition (type 2).
Sediment trend analysis assumes that the probability of sediment movement is based on particle size and that the relationship between transport pathways and contaminant levels is dependent on contaminants preferentially associating with the finer rather than coarser size fractions of the available sediment. In situations in which such assumptions might not be valid, any correlation between the pathways and contaminants would be unlikely. When such a correlation does exist, however, the findings, together with other available evidence, could help to resolve liability issues. For the Hylebos Waterway, the results of the STA were applied as part of an analysis of the potential legal liability of parties who might have been sources of the contamination found in the sediments. The potential legal liability included liability to the USEPA for the cleanup of the contaminated sediments, liability to Natural Resource Trustee agencies for natural resource damages (NRDs) caused by the contaminated sediments, and liability allocation among the potentially responsible parties (PRPs) for these costs.
The most striking finding of the STA from a liability standpoint was the net sediment transport from the mouth of the Waterway toward the head. This evidence was used in liability allocation negotiations to demonstrate that PRPs located at the head of the Waterway might not be responsible for contaminated sediments at the mouth of the Waterway. Similarly, PRPs located sediment downstream (i.e., toward the head) of the highly contaminated Lower Turning Basin are unlikely to be responsible for the contaminated sediments there.
The information on the defined TEs, particularly the areas of dynamic equilibrium, was used in allocation negotiations to demonstrate that the sources of several of the isolated hotspots up the Waterway were actually from PRPs located farther down near the mouth of the Waterway, and not from the immediately adjacent properties. From an allocation perspective, the major documented sources of contamination at the mouth of the Waterway should be expected to bear a large share of the total cleanup and NRD liability for the entire Hylebos, because of the dynamic migration of sediment contamination from these sources toward the Head.
The STA information was used to supplement analyses of the patterns of sediment concentration gradients that had previously been used to identify likely source properties. In many cases, the STA analysis helped explain apparent anomalies in the sediment concentration gradients. For instance, the STA explained how hot spots of PCBs could be located near a property in which PCBs had never been released to the environment. If sediment gradient pattern analysis alone were used to allocate liability among PRPs, the PRPs located near these hot spots would receive a disproportionate share of liability.
The STA results were also used during negotiations to rebut an assumption by the NRD Trustee agencies that contaminated sediments from a particular source would be as likely to migrate down as up the Waterway. The STA showed that this sediment distribution assumption was invalid. Such a change in assumptions had a dramatic effect on allocation calculations.
Other information useful to liability analysis included the conclusion that sediments outside of the Waterway in Commencement Bay are separate from the Waterway. This information shows that PRPs responsible for contaminated sediments within the Waterway are unlikely to be sources of contamination found in sediments outside of the Waterway, even in the immediate vicinity of the mouth of the Hylebos.
The authors thank S. Heise of the Technische Universität Hamburg-Harburg, Germany, and 1 anonymous reviewer for particularly constructive comments on the manuscript. This project was funded by one of the private industial stakeholders of the Hylebos Waterway.