Marine Pollution Bulletin 89 (2014) 284–295

Contents lists available at ScienceDirect

Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul

A sediment mesocosm experiment to determine if the remediation of a shoreline waste disposal site in Antarctica caused further environmental impacts Jonathan S. Stark ⇑, Glenn J. Johnstone, Martin J. Riddle Australian Antarctic Division, Channel Highway, Kingston, Tasmania 7050, Australia

a r t i c l e

i n f o

Article history: Available online 11 October 2014 Keywords: Metals Infaunal assemblages Monitoring Recruitment Casey Benthic communities

a b s t r a c t A shoreline waste disposal site at Casey Station, Antarctica was removed because it was causing impacts in the adjacent marine environment (Brown Bay). We conducted a field experiment to determine whether the excavation created further impacts. Trays of clean, defaunated sediment were deployed at two locations within Brown Bay and two control locations, two years prior to remediation. Trays were sampled one year before, 1 month before, 1 month after and two years after the excavation. An increase in metals was found at Brown Bay two years after the remediation. However there was little evidence of impacts on sediment assemblages. Communities at each location were different, but differences from before to after the remediation were comparable, indicating there were unlikely to have been further impacts. We demonstrate that abandoned waste disposal sites in hydrologically active places in Antarctica can be removed without creating greater adverse impacts to ecosystems downstream. Crown Copyright Ó 2014 Published by Elsevier Ltd. All rights reserved.

1. Introduction The guiding principles for environmental management in Antarctica come from the Protocol for Environmental Protection to the Antarctic Treaty. On the subjects of waste disposal, waste management and site clean-up, the protocol states that abandoned waste disposal sites should be cleaned up providing removal does not create greater adverse environmental impact than leaving the material in its existing location. Although several waste disposal sites have been removed from Antarctica (e.g. Crumrine, 1992) there have been no published reports of studies to test whether removal was achieved without creating further impacts. The abandoned Thala Valley waste disposal at Casey Station, East Antarctica was identified as a priority for clean-up by the Australian Antarctic Division because the site was hydrologically active and known to be causing adverse environmental impacts in the adjacent marine bay (Cunningham et al., 2005; Stark et al., 2005). Every year the summer melt saw large volumes of water flowing through the waste disposal site, eroding waste material and entraining dissolved and particulate contaminants (Snape et al., 2001). A range of environmental impacts have been reported in the downstream receiving environment, Brown Bay, including ⇑ Corresponding author. Tel.: +61 3 6232 3589; fax: +61 3 6232 3158. E-mail address: [email protected] (J.S. Stark). http://dx.doi.org/10.1016/j.marpolbul.2014.09.045 0025-326X/Crown Copyright Ó 2014 Published by Elsevier Ltd. All rights reserved.

elevated levels of contaminants in sediments, changes in soft-sediment assemblages and impacts on recruitment (Stark et al., 2003a; 2004; 2003b, c). A large scale clean-up and remediation of the site was done in the summer of 2003/04 (Stark et al., 2006b). Such operations are difficult to undertake, are expensive and require new techniques for every component from waste removal to monitoring. We used the opportunity to test and develop techniques to help inform future clean-up operations in Antarctica. A comprehensive monitoring program was designed to look at processes on a range of time scales (Stark et al., 2006b). Over the longest time scales, monitoring will be used to determine whether impacted communities in Brown Bay have recovered and consequently, whether the investment in remediation has delivered the hoped-for environmental improvements. The final sampling for the long-term monitoring has yet to be completed. One of the main concerns in planning the operation was that it could disturb the site to such a degree that a large pulse of contaminants, in particular metals bound in the site matrix, would be released into the adjacent marine environment. To address this, short term monitoring was put into place during the operation to assess if contaminants were released (Stark et al., 2006a). This monitoring was designed to inform operational practices in real-time so that they could be modified and improved if required, however, it would not indicate whether any such release of contaminants caused additional adverse environmental impacts beyond adding to the

285

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

contaminants already in the marine environment, for example adverse impacts on the seabed communities. Here we report a sediment field mesocosm experiment designed to determine whether the excavation operation, and any associated release of contaminants, created any further impacts on the biota in the bay, and hence whether the general obligations under the environmental protocol, of no ‘greater adverse environmental impact than leaving waste material in its existing location’ were satisfied. In designing the study, the main challenge we faced lay in detecting further impacts in already impacted communities, particularly in a situation where we do not have comparable impacted sites that could be used either as controls or replicates. In our previous studies, which identified that Brown Bay was impacted, we addressed this lack of replication of impacted sites by comparing with multiple reference locations (Stark et al., 2003a, b). We predicted that any additional adverse impacts in the already impacted Brown Bay would be subtle and would require an experimental design with a beyond-BACI type approach (Underwood, 1991, 1994). Without an appropriate experimental design and suitable test we would not be able say with confidence whether the clean-up had been completed without causing greater adverse environmental impacts. We designed a manipulative field mesocosm experiment using recruitment to a standard, clean, defaunated sediment. Manipulative experiments offer several advantages over observational or mensurative experimental sampling. They enable the influence of the variable of interest to be examined (e.g. a disturbance such

as pollution) while reducing other sources of variation such as habitat (e.g. grain size in sediments) and patch size. Such experiments are increasingly being used as monitoring tools in ecological and environmental monitoring programs e.g. (Connell, 2001; Cunningham et al., 2003; Glasby, 1998; Powell et al., 2005; Stark, 2008; Stark et al., 2004). In environmental monitoring contexts the hypothesis being tested is often of differences among locations (one or more being impacted). However, natural spatial variation and environmental heterogeneity can make it very difficult to distinguish the effects of anthropogenic disturbance, particularly in soft sediments where assemblages are patchy and the abundance of organisms varies at a range of spatial scales (Barry and Dayton, 1991; Morrisey et al., 1992a,b). Mesocosm recruitment experiments utilising new habitat reduce heterogeneity associated with natural substrata, provide a degree of uniformity, and facilitate replication. They are only affected by ongoing or new disturbances, as opposed to being a result of past disturbances. They offer the means to demonstrate a link between cause and effects where impacts are hypothesised to occur and where there may be evidence of a correlation between patterns of differences and presence of a disturbance such as pollution (Underwood and Peterson, 1988). One situation in which they offer an advantage over mensurative experiments is where an impact (e.g. sediment contamination) is known to exist, but some activity is to take place (such as remediation efforts) that may lead to change in the impact status, e.g. further impacts or recovery. If biological communities at the impacted location are known to be different from controls, and where there is no

(a)

(b)

Fig. 1. (a) Location of Casey Station and (b) deployment locations for the experiment. * = location of Thala Valley remediation operation and former waste disposal site; # = site where sediment was collected for the experiment.

286

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

Table 1 Deployment and sampling dates at each location.

Deployment Time 1 Time 2 Time 3 Time 4

Brown Bay Inner (7–10 m)

Brown Bay Middle (10–15 m)

McGrady Cove (10–15 m)

O’Brien Bay-5 (10–15 m)

20/11/01 7/02/2003 18/11/03 4/02/2004 24/11/2005

28/11/01 12–21/02/2003 18/11/03 4/02/2004 19/11/2005

5/12/01 22/2/03 1/12/2003 6/02/2004 16/11/2005

16/12/01 None sampled 5/12/03 9/01/2004 28/11/2005

pre-impact data, it may be difficult to detect any further change or impacts, particularly if there is a large degree of variation associated with the variables or communities in question. Monitoring of physical and chemical environmental parameters may offer some evidence of potential reduced or increased impacts but provide little ecological information about their biological effects. Manipulative experiments offer a solution, particularly those that manipulate habitat, by providing new habitat (e.g. trays of uncontaminated sediment, settlement panels), or by clearing existing habitat to make it available to ongoing ecological processes of colonisation and recruitment. The use of clean, defaunated sediment in the experiment meant that the communities recolonising the sediment would only be affected by current or ongoing processes and not directly by contaminants already accumulated in surrounding sediments. As colonisation at Casey is largely influenced by the surrounding assemblages (Stark et al., 2004), however, recruiting assemblages are also likely to be different between control and impacted locations. If the remediation process introduced a large pulse of contaminants into the bay, we hypothesised that this would affect the communities established in the trays of clean sediments, in comparison to those at control locations. This effect would be detectable as a significant change in the communities in the trays from before to after the remediation, with no corresponding temporal changes at controls. Thus differences in temporal trends between control and impacted locations are an indication of an impact. Using the principles of BACI (Before, After, Control, Impact) (Underwood, 1991, 1992, 1993, 1994) it is possible to construct tests which specifically examine temporal trends at control and impacted locations and their interactions. Significant interactions of the various terms indicate certain types of impacts from before to after the disturbance (the remediation of the waste dump).

2. Methods

ing slopes runs through the valley and percolates through the site, entraining contaminants before entering the marine environment and depositing them in Brown Bay (Snape et al., 2003, 2001). Sediments in Brown Bay are contaminated by metals and hydrocarbons (Stark et al., 2003b). The cleanup operation ran from November 2003 to February 2004. Waste and contaminated soil was extracted from frozen ground with excavators and bulldozers and placed into leak proof containers. Melt-water was diverted around the site as much as possible, or was contained and treated on the site to prevent contaminants from entering the marine environment. Water that was retained on site was treated in an on-site facility that could clean up to 10,000 L of metal and hydrocarbon contaminated waste water per day (Northcott et al., 2003, 2005). The containers of waste were stored in a holding area until being transported back to Australia where the waste was chemically fixed and buried in Table 2 PERMANOVA results for sediment variables (12 metals and TOM). Source

df

MS

Pseudo-F

P(perm)

P(MC)

C vs I Time Location (C vs I) (C vs I)  Time Location (C vs I)  Time Res Total

1 3 2 3 5 44 58

35.62 2.73 1.82 1.41 1.81 0.59

20.73 1.56 3.07 0.81 3.04

0.001 0.25 0.01 0.56 0.002

0.001 0.22 0.02 0.57 0.003

Estimates of components of variation Source

Estimate

Sq. root

% contribution

C vs I Time Location (C vs I) C vs I  Time Location (C vs I)  Time Res

1.32 0.07 0.09 0.05 0.30 0.59

1.15 0.27 0.30 0.22 0.55 0.77

38 9 10 0 18 25

Location(C vs I)  Time interactions

2.1. Study area Casey Station (66°170 S, 110°320 E) is in the Windmill Islands, East Antarctica (Fig. 1a and b) and is the third station to have been built in this area (Wilkes, Old Casey and the current Casey Station). The Windmill Islands are a region of low lying rocky hills and islands which are partially ice free year round. Low rocky hills and ice cliffs border several large bays around Casey, which contain smaller, inner bays. Sea ice cover in these bays is 1.2–2 m thick and break-out at the sites in this study occurs between December and February in most years. The Old Casey Station waste disposal site was 450 m north-east of the current Casey Station, in Thala Valley, adjacent to Brown Bay (Fig. 1). Waste material was dumped at the seaward end of Thala Valley and directly into the bay between 1969 and 1986 and included ash, gravel, vehicle parts and other metal, batteries, oil drums, glass, plastics, paper, cardboard, wood, rope, clothing, construction materials, asbestos, cement, rubber, insulation batts and drums of unidentified waste chemical and oils (Deprez et al., 1999; Snape et al., 2001). During summer, melt water from the surround-

Time Time Time Time

1 2 3 4

Brown Bay Inner vs. Middle

McGrady vs. O’Brien Bay5

t

P(perm)

P(MC)

t

P(perm)

P(MC)

1.33 0.70 1.32 2.24

0.20 0.71 0.22 0.12

0.20 0.59 0.20 0.05

0.89 1.57 2.31

0.45 0.09 0.02

0.45 0.10 0.04

Bay Inner 0.43 0.26 0.05 0.85 0.05 0.06

0.36 0.26 0.02 0.73 0.09 0.02

Brown 1.14 1.11 1.46 0.73 2.15 2.25

Bay Middle 0.36 0.30 0.29 0.31 0.13 0.15 0.56 0.59 0.07 0.04 0.06 0.02

Time-1 Time-1 Time-1 Time-2 Time-2 Time-3

vs. vs. vs. vs. vs. vs.

Time-2 Time-3 Time-4 Time-3 Time-4 Time-4

Brown 1.06 1.23 2.85 0.51 1.92 2.52

Time-1 Time-1 Time-1 Time-2 Time-2 Time-3

vs. vs. vs. vs. vs. vs.

Time-2 Time-3 Time-4 Time-3 Time-4 Time-4

McGrady 0.95 0.58 1.55 0.03 1.99 0.06 1.33 0.08 1.63 0.07 2.13 0.02

O’Brien Bay-5 0.41 0.11 0.06 0.17 0.09 0.05

Significant test results indicated in bold type.

0.92 1.56 1.43

0.50 0.07 0.11

0.45 0.13 0.13

287

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

2

Brown Bay Inner Brown Bay Middle McGrady O'Brien Bay-5 2

PC2

4

0

2 3 4 12 13 12 3 3 3 4 3411223 3 1 43 3312221 1 333 42 2 1 22 4 34 4 22 344 4

3 3 13 4 23 3 1 3 2 1 3222 13 3 2 4 21 311 12 32 3 2 42 13 4 1 431 4 4 44 4 44

324 4 2 4 2

Pb

1

4

Sn Cu Fe Ba OM

2

4

4

Mg 4

4

Sr

-2

-2

0

2

4

6

PC1 Fig. 2. PCA ordination of concentrations of 12 metals and TOM, with vector plot overlayed on ordination, indicating single variable correlations with PCA axes.

a managed landfill. All of the waste and contaminated soil was excavated and approximately half was transported back to Australia for remediation, with the remainder stockpiled at Casey on bentonite geofabric to prevent leakage of water and contaminated particles, which was returned to Australia in 2011. Further information on the clean up operation and history of the site can be found in Stark et al. (2006b).

we hypothesised that there would be significant time  location interactions, with the impacted location Brown Bay having a different time  location response to the control locations. In particular there would be significant differences from before to after the remediation in communities in the experimental trays at Brown Bay that were not seen at controls or were of a different magnitude. 2.3. Sediment analysis

2.2. Experimental design In order to standardise the sediment used in the experiment (i.e. grain size, organic content, no contaminants and absence of macrofauna) it was collected only from one uncontaminated control site. Marine sediment was collected by divers using 15-L polyethylene buckets from 16 to 19 m depth from a site in O’Brien Bay (Fig. 1). Sediment was sieved through a 500-mm mesh to remove infauna and left to settle overnight in plastic bins (500  300  400 mm). Defaunated sediment was transferred into plastic trays (34  23  12 cm) lined with a 300 mm mesh. Trays of defaunated sediment were deployed at 4 locations: two adjacent to the waste disposal site in Brown Bay, Brown Bay Inner (approx 50 m from the shore, 7–10 m deep) and Brown Bay Middle (approx 200 m from shore, 10–15 m deep); and two control locations: McGrady Cove (10–15 m deep) in Newcomb Bay, and O’Brien Bay-5 (10–15 m deep) (Fig. 1). Trays were placed on the seabed where they settled into the sediment to a depth of 5–7 cm. At each location 20 trays of sediment were deployed in an area of approximately 25  25 m in 4 groups of 5 trays, each group separated by approximately 20 m. At each sampling time 4 trays of sediment were collected from each location, 1 tray randomly selected from the unsampled trays in each group (see Table 1 for deployment and retrieval dates). Replicate cores were pushed into the tray of sediment by divers on the sea bed and capped and the tray was then returned to the surface and the cores removed and capped. Four macrofauna cores (10 cm diameter) were collected from each tray and two cores (5 cm diameter) were collected for sediment chemical analyses. There were 4 sampling times (Table 1): Time 1 was collected approx 1 year before the cleanup, after a deployment period of approx 14 months, Time 2 was collected at the beginning of the cleanup (24 months deployment), Time 3 was collected in the same season at the end of the cleanup (26.5 months deployment), and Time 4 was collected approx 2 years after the cleanup (4 years deployment, Table 1). The hypothesis being tested concerned the effects of the remediation operation. If the operation created additional impacts

Sediments were analysed for metals and total organic matter (TOM). Sediment cores were frozen after sampling and prior to analysis the top 2 cm of the core was sectioned. This was extracted for 4 h with a 1:10 w/v 1 M HCl. A 3 g sub-sample of homogenised wet sediment was mixed with 30 mL of 1 M HCl (prepared from Merck Suprapur 30% HCl: 167 mL of concentrated acid was added to 1.6 L of Milli-Q water, made to 2 L with Milli-Q) in a polypropylene centrifuge tube (30 mL, Sarstedt AG & Co) and tumbled for 4 h at room temperature. The mixture was then centrifuged for 15 min at 12,000 rpm and the supernatant filtered through a 0.45 um membrane cartridge filter (Minisart, Sartorius AG) into a clean tube. The filtered supernatant was then stored at 4 °C in acid washed nalgene containers before analysis by ICP-MS at the Central Science Laboratories (CSL), University of Tasmania. Upon receivership the samples were diluted (1/10) and indium (100 ppb) was added to each as the internal standard. Instrument settings and method parameters are those described by Townsend et al. (2007). TOM was determined by the Loss on Ignition method, using the ash free weight after incineration. A 2 g homogenised wet subsample was weighed into a pre-combusted crucible and dried at 105 °C. The dried sample was reweighed before being placed in a muffle furnace for 4 h at 550 °C. 2.4. Statistical analysis A 3 factor design was used in multivariate tests using the program PERMANOVA (PRIMER 6.1.13 & PERMANOVA + 1.0.3 Plymouth Marine Laboratories). The first factor was Control vs. Impact (C vs. I) which was fixed. The second factor was Location which was nested in C vs. I, with 2 locations within each level of C vs. I (Control: McGrady and O’Brien-5; Impact: Brown Bay Inner and Middle). The third factor was Time which was orthogonal and fixed, with four times: 2 sampling times before the remediation and 2 after the operation (Table 1). As it was not known whether there would be immediate effects of the cleanup, times were not nested in a before/after factor, so tests could be made among all

288

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

Table 3 Results from 3 factor PERMANOVAs for metals (log(x + 1) transformed). Source

DF

C vs. I Time Location (C vs. I) C vs. I  Time Time  Loc (C vs. I) Residual Total

1 3 2 3 5 44 58

C vs. I Time Location (C vs. I) C vs. I  Time Time  Loc (C vs. I) Residual Total

1 3 2 3 5 44 58

C vs. I Time Location (C vs. I) C vs. I  Time Time  Loc (C vs. I) Residual Total

1 3 2 3 5 44 58

MS

Pseudo-F

P

MS

Pseudo-F

P

Lead 24.71 0.64 0.19 0.58 0.86 0.19

128.67 0.78 1.03 0.70 4.60

0.01 0.55 0.36 0.59 0.002

Copper 2.39 0.10 0.21 0.16 0.17 0.04

12.16 0.64 5.17 0.99 4.05

0.09 0.64 0.01 0.50 0.01

Iron 0.52 0.04 0.21 0.05 0.10 0.05

2.59 0.38 4.58 0.48 2.16

0.24 0.77 0.02 0.71 0.07

Antimony 0.00 0.00 0.00 0.00 0.00 0.00

49.04 0.99 1.80 1.80 3.24

0.02 0.47 0.16 0.26 0.02

Tin 5.70 0.21 0.39 0.21 0.18 0.05

15.93 1.19 7.07 1.19 3.36

0.05 0.41 0.003 0.39 0.02

Zinc 0.75 0.02 0.04 0.02 0.04 0.02

18.90 0.52 2.57 0.54 2.77

0.04 0.70 0.10 0.69 0.03

Significant Time  Loc (C vs. I) interactions Time Pb – Time (BB Inner) Cu – Time (BB Inner) Sb – Time (BB Inner) Sn – Time (BB Inner)

Location 4 > 3 = 2 = 1* 4 > 3 = 2 = 1* 4 > 1* 4 > 3 = 2 = 1*

Pb – Location (Time 4) Cu – Location (Time 4) Sn – Location (Time 4)

BB inner > BB Middle* BB inner > BB Middle* BB inner > BB Middle*

Significant test results indicated in bold type.

pairwise comparisons of time. Replicate sub samples were taken within each tray and averaged to produce values for each tray. Univariate analysis of environmental variables and of biological variables was also done using the PERMANOVA program, with a Euclidean distance similarity matrix for single variables, which produces results equivalent to those of traditional univariate ANOVA (Anderson, 2001a,b). Distance based redundancy analysis methods were used to determine which environmental variables (metals and TOM) were important in explaining variation in community patterns at each site. The routines DISTLM and dbRDA were used in the program PRIMER 6.1.13 & PERMANOVA + 1.0.3 (Plymouth Marine Laboratories). Distance based linear models were tested using the AICc selection criteria as it is recommended for situations where the number of samples is small relative to the number of predictor variables (Anderson et al., 2008) as was the case here. As AICc is a highly conservative criteria it usually produced models with only one or two variables as the most parsimonious, however these did generally not explain much of the total variation (ca. 50%) and which, when a dbRDA ordination was done of the fitted model, bore a closer resemblance to an MDS of community data. These models were used to determine which variables were important in explaining community patterns. 3. Results 3.1. Sediments Multivariate analysis of concentrations of 12 metals and total organic matter (TOM), using PERMANOVA, showed a strong

difference between control and impacted groups and between locations within control and impact groups (Table 2, Fig. 2). There was also a strong interaction of differences among locations within the control and impact groups (Location(CvsI)  Time), and some differences among trays within each location and time (Fig. 2). Estimates of the components of variation show that differences between the control and impact groups was the largest source of variation, followed by the residual variance (differences among trays, particularly at the impacted locations – Table 2). Brown Bay Inner and Middle were not significantly different from each other at Times 1, 2 and 3 but were at Time 4. There were also large differences between trays at Time 4 at Brown Bay Inner, which reduced the differences between the two impacted sites. While there was also a significant difference between the two control sites at Time 4, the magnitude was smaller (less distance between groups in PCA, Fig. 2) and was in a different plane of the PCA ordination, indicating different metals were responsible. There was also some overlap of multiple times between the two control sites, while at Brown Bay Inner the Time 4 samples stand well apart (Fig. 2). There were also significant differences within locations between times, in particular for a difference between Time-4 and other times at Brown Bay Inner and Middle (Table 2). Again these differences were partially obscured by large differences between trays. A PCA ordination of all 13 environmental variables (PC1 and PC 2 explain 83% of variation) shows a distinct separation between sediments at control and impacted sites along the PC1 axis, which, as shown in the vector plot, is correlated with concentrations of lead, tin and copper and to a lesser extent iron (Fig. 2). The PC1 axis also clearly delineates samples from Brown Bay Inner at Time 4 and reflects an increase in these elements at this site after the cleanup. The PC2 axis was correlated with concentrations of magnesium, strontium and barium and TOM, and separates samples within control locations, particularly in regard to time, but it can

289

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

Cu

Pb

15

60 50 40 30 20 10 0

10 5 0 1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

1 2 3 4

1 2 3 4

Sn

1 2 3 4

2 3 4

1 2 3 4

2 3 4

1 2 3 4

2 3 4

1 2 3 4

2 3 4

Sb

6 5 4 3 2 1 0

0.08 0.06 0.04 0.02 0 1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

1 2 3 4

-1

1 2 3 4

Fe

Zn

2000

30 25 20 15 10 5 0

1500 1000 500 0 1 2 3 4

1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

1 2 3 4

Mg

Ba

5000

25

4000

20

3000

15

2000

10

1000

5

0

0 1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

1 2 3 4

1 2 3 4

TOM

Sr

5

60 50 40 30 20 10 0

4 3 2 1 0 1 2 3 4

BBi

1 2 3 4

BBm

1 2 3 4

McG

2 3 4

1 2 3 4

1 2 3 4

1 2 3 4

OB5

BBi

BBm

McG

2 3 4

OB5

Fig. 3. Average (+SE) metal concentrations in sediments in trays at each location at each sampling time (times 1, 2, 3, 4), n = 4.

also be seen there is a general pattern of an increase in these variables across locations in Newcomb Bay (Brown Bay and McGrady) at Time 4 (Fig. 2). Univariate analysis of metals in sediments in trays showed that the impacted locations (Brown Bay) accumulated significantly greater concentrations of lead (Pb), tin (Sn), antimony (Sb) and zinc (Zn) in comparison to controls (Table 3, Fig. 3). There was also a significant interaction between Time and Location (C vs. I), with Brown Bay Inner accumulating significantly greater concentrations of Pb, Cu, and Sn than Brown Bay Middle by Time-4 (Fig. 3). At Brown Bay Inner concentrations of Pb, Cu, Sb and Sn were significantly greater at Time-4 than at Times-1, -2 and -3 (Fig. 3, Table 3). However, for a number of other metals (Ba, Mg, Sr) and TOM there was a significant increase for the three locations in Newcomb Bay (Brown Bay Inner and Middle and McGrady), which were not observed at the location in O’Brien Bay. O’Brien Bay-5 generally showed a decrease in the concentrations of these variables, except magnesium which showed a slight increase at time-4 (Fig. 3). 3.2. Macrofaunal communities As expected macrofaunal communities recruiting to trays were different at each location and this was consistent over the duration

of the experiment. An MDS ordination of communities in each tray (cores averaged) at each time shows the clear differences between locations (Fig. 4a). If the remediation operation had an impact on communities in Brown Bay we would expect to see this as a change in communities from before to after the cleanup that was different from changes at control sites. This could be evidenced as differences at Brown Bay from before to after in any of: (i) a difference not seen at control sites (a significant interaction between Time and Location (C vs. I)); (ii) a larger difference than seen at controls (effect size); (iii) a difference in the community development/succession (seen as a different trajectory in multivariate ordinations). The multivariate method PERMANOVA, with a 3 factor design as for the metals, was used to test whether there was a significant interaction between Time and Location (nested in Control vs. Impact) that related to the timing of the remediation (Time-1 and -2 were before the operation, Time-3 and -4 after). There was a significant Time  Loc(C vs I) interaction (Table 4) which showed that: (i) within the control and impacted groups, locations were significantly different at each time (e.g. controls Time 2, OB5 different to McG); and (ii) there were significant differences between most times at each location, the exceptions being no difference between Time-2 and -3 at Brown Bay Middle and at McGrady Cove between Time-2 and -3 and Time-3 and -4.

290

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

(a)

Stress: 0.1 2 2 4

3

4 4

4

3 22

4 4 3 44 3 2 2 44 3 3 3 2 32322 3 2

4

3

3

4

11

4 2 4

2 3

1

3 3 24 3

1

2

2

1 1 1

1

1 4

1

Brown Bay Inner Brown Bay Middle McGrady Cove O’Brien Bay-5 1 = Time 1 2 = Time 2 3 = Time 3 4 = Time 4

1

PCO2 (20.9% of total variation)

(b)

2 4

20

3 4

0

3 42 2 1 3 1

4 13 2

-20

-40 -60

-40

-20

0

20

40

PCO1 (50.2% of total variation) Fig. 4. (a) MDS of soft sediment communities in trays (based on square root transformed abundances and Bray-Curtis similarities). Replicate samples in trays averaged, and (b) PCO ordination of group centroids for each time at each location.

Table 4 PERMANOVA table of results for macrofaunal assemblages. Source

df

MS

Pseudo-F

P(perm)

P(MC)

C vs. I Time Location (C vs. I) C vs. I  Time Loc (C vs. I)  Time Residual Total

1 3 2 3 5 44 58

28,357 2161.9 10,592 1085 668.15 221.7

2.9225 3.342 47.777 1.6772 3.0138

0.001 0.003 0.001 0.101 0.001

0.115 0.001 0.001 0.083 0.001

Significant test results indicated in bold type.

To examine differences in effect size, the centroids for each group of replicates (time  location groups) were calculated from the Bray-Curtis similarity matrix (of square root transformed abundances) and the distance among centroids was then calculated and used to generate a similarity matrix which was then used to do a PCO ordination (Fig. 4b). Distances among centroids can be used as an estimate of relative effect size and visualized in the PCO ordination (Anderson et al., 2008). Differences among locations are clearly distinct on the first PCO axis which explains the majority of the variation (50.2%), with the impacted locations on the far right. Differences among sampling times are discernible on PCO axis 2, explaining ca. 21% of variation, with a general progression from time 1 to time 4 from the bottom to the top of the PCO plot for each group. Importantly the distance between times (e.g. between time 2 and 3 and times 3 and 4 with each location) are quite similar, indicating a similar effect size of time at each location. Thus there does not appear to be an effect of the remediation at Brown Bay in comparison to controls. The direction of change in the PCO ordination, from time 1 through to time 4 is different at each location (Fig. 4b) and also indicates no general difference attributable to the remediation. Using this approach there was no indication of a

difference in temporal trends between control and impacted locations (they were generally all different at each time) and thus at the assemblage level no evidence of an impact due to the remediation. The mean number of taxa per core did not show any relationship to time, including from before to after the remediation, but was significantly greater at the control locations (Fig. 5, Table 5). The mean total number of individuals per core showed contrasting patterns at each location, with little change over time at Brown Bay Inner, a decrease over time at Brown Bay Middle and an increase over time at the control locations (Fig. 5). Crustaceans dominated the assemblages at all locations, ranging from 54% to 97%. Gammarid amphipods comprised the greatest proportion of crustaceans (Fig. 5). Polychaetes were generally low in abundance and showed no pattern through time with the exception of a large peak in the abundance of Capitella sp. at Brown Bay Middle at time 1 which subsequently declined (Fig. 5). The patterns of recruitment of over 20 individual taxa were tested, none of which appeared to be related to the cleanup operation with the possible exception of one species, the isopod Austrosignum c.f. grande, which was significantly less abundant at Brown Bay after the remediation than before, with no such change at controls, although abundances were much lower overall at controls (Fig. 5, Table 5). When multiple tests are performed in this way some significant differences are expected by chance (the Type I error rate = a, in this case 0.05 or 1 in 20 significant by chance) (Quinn and Keough, 2002) and do not provide strong evidence of an impact. Other recruitment patterns included: taxa more abundant at either impacted (e.g. the amphipod Orchomenlla franklini, Fig. 5) or control locations (the cumacean Eudorella gracilor Fig. 5) and taxa that had peaks of abundance at some locations or were consistently more abundant at one location than others, but without any control vs impact relationship.

291

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

Total taxa

25 20 15 10 5 0

1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

Total polychaetes

500 400 300

Mean per core

200 100 0

1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

Austrosignum grande

200

Total individuals

1200 1000 800 600 400 200 0

1200 1000 800 600 400 200 0

1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

Total crustaceans

1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

Total gammarids

500 400

150

300

100

200

50

100

0

1 2 3 4

1 2 3 4

1 2 3 4

2 3 4

0

1 2 3 4

Eudorella gracila

1 2 3 4

1 2 3 4

2 3 4

Orchomenella franklini 400

300 250 200 150 100 50 0

300 200 100 1 2 3 4

BBi

1 2 3 4

BBm

1 2 3 4

McG

2 3 4

OB5

0

1 2 3 4

BBi

1 2 3 4

BBm

1 2 3 4

McG

2 3 4

OB5

Fig. 5. Mean abundance per core (+SE) of selected taxa in trays at each location at each sampling time.

Table 5 Results from 4 factor ANOVAs for biota. Source

DF

C vs. I Time Location (C vs. I) C vs. I  Time Time  Loc (C vs. I) Residual Total

Total taxa 1 154.24 3 27.69 2 13.29 3 8.72 5 1.69 44 2.81 58

MS

F

P

P(MC)

12.45 15.91 4.72 5.01 0.60

0.17 0.01 0.02 0.05 0.70

0.07 0.004 0.02 0.04 0.70

C vs. I Time Location (C vs. I) C vs. I  Time Time  Loc (C vs. I) Residual Total

Austrosignum grande 1 290.11 10.60 3 13.19 3.94 2 29.64 9.01 3 23.86 7.12 5 3.35 1.02 44 3.29 58

0.15 0.09 0.00 0.02 0.42

0.05 0.08 0.002 0.03 0.41

Significant test results indicated in bold type.

3.3. Relationships between biological and environmental data At each site several variables were clearly important in explaining variation between times in biological communities as determined by multivariate redundancy analysis. At Brown Bay lead and tin each explained up to 24% of the total variation in communities recruiting to trays (marginal tests, Table 6), were highly correlated (r = 0.96) and were not included together in any models

due to this collinearity, but used separately to represent each other. A model consisting of tin, strontium, TOM, iron and zinc explained over 60% of the total variation at Brown Bay Inner. A dbRDA plot of the fitted model can be seen in Fig. 6b and the time groupings match that of the MDS plot of the biological community in Fig. 6a, particularly at time-4 which is correlated with tin (and thus lead) and strontium. The vectors show increasing concentrations of tin (and thus lead) and strontium at time 4 and this model aligns reasonably well with the MDS of community structure. The best model for Brown Bay Middle contained 5 variables including TOM, cadmium, nickel, copper and vanadium and explained ca. 70% of variation with the fitted model matching the MDS of community structure well (Fig. 6c and d). An increase in magnesium and TOM separates time-4 from other times. The most influential single variables were magnesium and TOM (Table 6). The control sites are different from both the Brown Bay sites. At McGrady the most parsimonious model that best matched the MDS of community structure (Fig. 6e and f) was of 3 variables: copper, magnesium and cadmium, which explained 56% of total variation. The best single variables in marginal tests were magnesium, TOM and iron (Table 6). The fitted model shows that magnesium and copper increase at time 4 (Fig. 6f). At O’Brien Bay-5 the best model contained 3 variables: magnesium (which increased at time 4), barium and TOM (which decreased from time 2) and explained 68% of total variation and matched the MDS of community structure well. The best single variables were barium, magnesium and cadmium (Table 6).

292

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

Table 6 Results from marginal tests of single variables at each location from DSTLM analysis, indicating significance of fit to community data and proportion of variance explained. Variable

Prop.

Variable

Brown Bay Inner Sn 0.002 Pb 0.004 Ba 0.01 Sb 0.01 Sr 0.01 Cu 0.02 Mg 0.02 TOM 0.27 V 0.40 Zn 0.60 Ni 0.61 Cd 0.73 Fe 0.78

P

0.24 0.24 0.22 0.20 0.19 0.18 0.17 0.08 0.07 0.05 0.05 0.04 0.04

Brown Bay Middle Mg 0.001 Sr 0.003 Fe 0.01 Zn 0.01 Cd 0.02 TOM 0.04 Ni 0.08 Sb 0.17 Cu 0.25 Sn 0.41 Ba 0.51 V 0.70 Pb 0.82

0.28 0.26 0.20 0.20 0.19 0.16 0.12 0.10 0.09 0.07 0.06 0.04 0.03

McGrady Cove Mg TOM Fe Pb Ba Sr V Sb Cu Cd Zn Ni Sn

0.30 0.26 0.21 0.20 0.18 0.17 0.15 0.15 0.13 0.12 0.10 0.07 0.02

O’Brien Bay-5 Ba Cd Cu Fe Mg Ni Pb Sb Sr TOM V Zn Sn

0.27 0.30 0.26 0.08 0.20 0.15 0.37 0.30 0.04 0.28 0.19 0.24 0.00

0.000 0.001 0.01 0.01 0.03 0.02 0.05 0.05 0.08 0.11 0.16 0.39 0.90

P

0.03 0.01 0.02 0.43 0.07 0.15 0.00 0.01 0.83 0.02 0.08 0.03 1.00

Prop.

4. Discussion We conclude that the remediation operation did not create greater adverse environmental impacts and at worst had a minor further impact on the already impacted assemblages in Brown Bay. Some contaminated material was released during the operation and post remediation as evidenced by increases in some metals at the Brown Bay sites, in particular at Brown Bay Inner, where concentrations of lead and tin increased. There is no evidence to suggest that there was a greater impact on the already impacted assemblages in Brown Bay post remediation. Differences from before to after the operation are comparable to those at the control sites, in terms of both size and trajectory of community development. The abatement methods in place during the operation (Stark et al., 2006b) were successful in limiting the contaminated material discharged to levels insufficient to have a detectable impact on sediment assemblages. Prior to the remediation operation (Times 1 and 2), significantly greater quantities of most metals had accumulated in trays in Brown Bay, evidence that the melt water runoff entering into the bay was contaminated by material in the waste disposal site matrix. However, there was very clearly a large flux of contaminants into Brown Bay some time after the remediation operation. This increase in sediment metal concentrations had two notable characteristics: it was very spatially variable, with large difference among trays, and it was greater at Brown Bay Inner than at Brown Bay Middle. There are two plausible explanations for an increase in the sediment metal concentrations in the Brown Bay trays. The metals may have come from the contaminated terrestrial site via meltwater runoff in summer, or it may have been resuspended from the existing contaminated sediments in Brown Bay, via currents, bioturbation, or physical disturbance from ice scour. We believe that the most likely source was the waste disposal site, largely because concentrations were much higher at Brown Bay Inner than Middle. Fresh melt water with entrained contaminated

particulates and dissolved metals, would have a greater sediment deposition rate closer to the terrestrial source than further away. Upon entering the marine environment particulates would begin to settle out, particularly the larger clumps of particulates which would presumably also be more contaminated, as well as small particles of metal. Deposition would also be patchy and likely to vary with tides and currents, contributing to spatial variation in deposition patterns and thus in metal flux to the seabed. This fits well with the observed patterns of metal accumulation in Brown Bay (Stark et al., 2005). Ice scour would be the main form of resuspension as currents and tides are very low and there are few large animals in Brown Bay to contribute to bioturbation. Furthermore, physical disturbance by ice scour is less likely in the Inner part of Brown Bay as it is protected from iceberg ingress by fast ice for most of the year, and as it is shallower, large bergs are likely to ground further out before reaching the inner part of the bay. This model of contaminated sediment deposition into trays from the terrestrial source also fits with another experiment done at the time of the remediation operation which demonstrated there was a pulse of contamination released during the remediation operation (Stark et al., 2006a). This was detected as elevated body burdens of several metals in amphipods held in water column mesocosms. The fast-ice itself is unlikely to be a source of contaminants as it only begins to form in late March, well after the melt streams into the bay have stopped flowing, thus it would not contain any contaminated melt water. Despite the variable increase in metals in Brown Bay, there is very little evidence of a biological impact. As the existing communities in Brown Bay are known to be different from those at control locations (Stark et al., 2003a,b), assemblages recruiting to defaunated sediment are also likely to be different, as is demonstrated in this experiment at Time-1 and -2. If the remediation operation did create further impacts, this would be apparent in the sediment trays at Brown Bay as a temporal change coincident with the remediation operation or after it. This was not apparent in multivariate analysis of assemblages recruiting to trays. There was some evidence of possible impact to a few taxa, whereby their abundances decreased at Brown Bay after the remediation operation in comparison to controls, but there is no clear pattern for the majority of taxa and it is more likely these changes are chance effects expected when analysing multiple species individually or are related to succession dynamics. For example the isopod Austrosignum cf. grande was an early colonist at Brown Bay, recruiting in large numbers (and is found in sediments in the bay in high abundances), which subsequently declined over time. It is not possible to determine whether this decline was due to the remediation as they did not recruit in high numbers at control sites. This decline may have been due to some other process such as competition with later colonists. This recolonisation experiment highlights an important feature of this type of experiment: that there are likely to be very strong differences in colonisation and community development at different locations. This is likely to be due to strong local control of recruitment by locally reproducing, self sustaining populations (e.g. Osman and Whitlatch, 1998). Only a time series of sufficient length can determine whether the temporal trajectories of control and impact sites are significantly different from before to after a disturbance. This requires establishing an experiment well before a disturbance to allow time for communities to establish, and allowing sufficient sampling time post disturbance to determine if there are significant impacts. Such constraints means that the application of such experiments is only potentially useful if planning allows sufficient deployment time pre-disturbance and that resources are available to enable repeat sampling. Technologies to clean up Antarctic contaminated sites are still in development and are the subject of active discussion in the

293

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

Brown Bay Inner 1

Stress = 0.07 4 4

2 22 2

1 1

4

3 3 4 1

3 3

(b) dbRDA2 (25.4% of fitted, 15.4% of total variation)

(a)

20 4

OM

10 1

1 4

Sr 4 1 2 2 3 3 2 2 Zn Fe 3 3

1

0

-10

Sn 4

-20 -20

-10

0

10

20

30

dbRDA1 (61.2% of fitted, 37.2% of total variation)

Brown Bay Middle 4

(c)

Stress = 0.13

1 1

1

4 2

3

2

(d) dbRDA2 (28.5% of fitted, 20% of total variation)

4 4

3

1 3 3 2 2

20

4 4

1

10

OM

Ni Cd 1 1

0

4

12 2

4

3 3

Cu V

-10

2

3

3

2

-20 -30 -20 -10 0 10 20 30 dbRDA1 (58.7% of fitted, 41.2% of total variation)

McGrady Cove

3 2 2 2 3 3 2

1

3

1

4 4

1

4

(f)

dbRDA2 (21% of fitted, 11.8% of total variation)

Stress = 0.09 4

(e)

20

0

4

32 3 3 3 2 12 2 4 4 1 Cu

1

4

-20 Mg Cd

-40 -40

-20

0

20

40

dbRDA1 (62.5% of fitted, 35% of total variation)

O’Brien Bay-5

(g) 2

2 2

4 4

2

4 4

3 3 3 3

(h) 20

dbRDA2 (18.6% of fitted, 12.6% of total variation)

Stress = 0.06

Ba 22 2

10

2 4 Mg

0

4

4

3 3 OM 3

-10

4

3

-20 -20

-10

0

10

20

30

dbRDA1 (75.7% of fitted, 51.4% of total variation) Fig. 6. (a), (c), (g), (e) – MDS plots of communities recruiting to trays at each location; (b), (d), (f), (g) – dbRDA plots of selected models with fewest variables that best matched MDS plots at each location.

Committee for Environmental Protection which reports to Antarctic Treaty Consultative Meetings, including work towards a Clean-up Manual to share best practise (CEP, 2013). Part of the

process of developing best practise is to identify where effort and money are best invested to deliver environmental improvements. Ultimately it is the clean-up and remediation of sites that delivers

294

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295

environmental improvements – not the monitoring. Our reason for directing significant effort to the monitoring in this case is that we hope it will inform future clean-up operations. We do not expect, or believe it is desirable, that the level of monitoring reported here be required routinely for similar operations. Our original purpose was to develop and apply an experimental test with sufficient statistical power (low risk of a Type II or false no impact error) to determine whether a clean-up operation, that involved significant precautions to prevent further off-site transport of contaminants, could be done without resulting in greater adverse environmental impacts. Our short-term, real-time monitoring (Stark et al., 2006a) had previously demonstrated that, despite the precautions, a pulse of contaminants was released to the marine environment during the clean-up and that they were taken up by amphipods held in the water column. The sediment mesocosm experiment reported here provides further evidence that contaminants were released during clean-up and additionally, that they reached the sediments in levels sufficient to cause detectable increases in the mesocosm sediments. However, neither is evidence that the clean-up operation and the associated release of contaminants had created greater adverse impacts than leaving the material in its original location. Our analysis of the biota in the sediment mesocosms, designed to provide a sensitive test of ecosystem impacts, indicated that the clean-up operation caused no impacts in addition to those already documented in Brown Bay. There is a subtle but important difference between additional or further impacts i.e. additional but similar in severity to preexisting impacts, and greater adverse impacts referred to in the general obligations of the environmental protocol, which effectively establishes the principle that clean-up should proceed unless it would make matters worse. As our experiment was designed to detect any additional impacts attributable to the clean-up, it set and satisfied a more stringent test than that established by the environmental protocol and is therefore more precautionary than required. We recognise that the measures we took to prevent contaminant release may exceed those available for routine clean-up operations. Their partial failure does provide the opportunity to assess whether the level of effort we directed towards preventing further mobilisation of contaminants was necessary in a place with a history of past contamination and, by extension, whether we would recommend this as essential for clean-up under similar circumstances in future. We encourage doing whatever is practical to contain contaminants on site during clean-up operations. However, we conclude that the requirement to prevent greater adverse environmental impacts can be satisfied even if some contaminant release cannot be prevented during clean-up and should not deter cleanup initiatives. In the longer term, we expect the communities in Brown Bay to recover now that the source of contaminants from Thala Valley waste disposal site has been removed. Further monitoring surveys are planned and we hope to be able to report that not only did the clean-up result in no greater adverse impacts but that it has delivered significant environmental improvements to the ecosystems downstream. We hope this study reduces concerns about the risk of creating greater adverse impacts that might deter future clean-up initiatives in Antarctica. Acknowledgements The authors are very grateful to the many people that assisted in this experiment including P. Goldsworthy, A. Tabor, A. Cawthorn, S. Stark, A. Palmer, R. Connell, and C. Sampson. We also thank I. Snape and K. Stark for their continuous support. This research was funded by the Australian Antarctic Division (AAS projects 2201 and 4180).

References Anderson, M.J., 2001a. A new method of non-parametric multivariate analysis of variance. Aust. Ecol. 26, 32–46. Anderson, M.J., 2001b. Permutation tests for univariate or multivariate analysis of variance and regression. Can. J. Fisheries Aquatic Sci. 58, 626–639. Anderson, M.J., Gorley, R.N., Clarke, K.R., 2008. PERMANOVA+ for PRIMER: Guide to software and statistical methods. PRIMER-E, Plymouth, UK. Barry, J.P., Dayton, P.K., 1991. Physical heterogeneity and the organization of marine communities. In: Kolasa, J., Pickett, S.T.A. (Eds.), Ecological Heterogeneity. Springer-Verlag, New York, pp. 270–320. CEP, 2013. Antarctic Clean-Up Manual, ATCM XXXVI /CEP XVI. Committee for Environmental Protection, Brussels. Connell, S.D., 2001. Urban structures as marine habitats: an experimental comparison of the composition and abundance of subtidal epibiota among pilings, pontoons and rocky reefs. Mar. Environ. Res. 52, 115–125. Crumrine, K.Z., 1992. Surface remediation at McMurdo Station, Antarctica. In: Melander, O., Fontana, L.R. (Eds.), 5th Symposium on Antarctic Logistics and Operations. Buenos Aires, Dirección Nacional del Antártico. Argentina San Carlos de Bariloche, Argentina, pp. 41–60. Cunningham, L., Snape, I., Stark, J.S., Riddle, M.J., 2005. Benthic diatom community response to environmental variables and metal concentrations in a contaminated bay adjacent to Casey Station, Antarctica. Mar. Pollut. Bullet. 50, 264–275. Cunningham, L., Stark, J.S., Snape, I., McMinn, A., Riddle, M.J., 2003. Effects of metal and petroleum hydrocarbon contamination on benthic diatom communities near Casey Station, Antarctica: an experimental approach. J. Phycol. 39, 490– 503. Deprez, P.P., Arens, M., Locher, H., 1999. Identification and preliminary assessment of contaminated sites at Casey Station, Wilkes Land, Antarctica. Polar Record 35, 299–316. Glasby, T.M., 1998. Estimating spatial variability in developing assemblages of epibiota on subtidal substrata. J. Mar. Freshwater Res. 49, 429–437. Morrisey, D.J., Howitt, L., Underwood, A.J., Stark, J.S., 1992a. Spatial variation in soft sediment benthos. Mar. Ecol. Prog. Ser. 81, 197–204. Morrisey, D.J., Underwood, A.J., Howitt, L., Stark, J.S., 1992b. Temporal variation in soft sediment benthos. J. Exp. Mar. Biol. Ecol. 164, 233–245. Northcott, K.A., Snape, I., Connor, M.A., Stevens, G.W., 2003. Water treatment design for site remediation at Casey Station, Antarctica: site characterisation and particle separation. Cold Reg. Sci. Technol. 37, 169–185. Northcott, K.A., Snape, I., Scales, P.J., Stevens, G.W., 2005. Contaminated water treatment in cold regions: an example of coagulation and dewatering modelling in Antarctica. Cold Reg. Sci. Technol. 41, 61–72. Osman, R.W., Whitlatch, R.B., 1998. Local control of recruitment in an epifaunal community and the consequences to colonization processes. Hydrobiologia 375–376, 113–123. Powell, S.M., Snape, I., Bowman, J.P., Thompson, B.A.W., Stark, J.S., McCammon, S.A., Riddle, M.J., 2005. A comparison of the short term effects of diesel fuel and lubricant oils on Antarctic benthic microbial communities. J. Exp. Mar. Biol. Ecol. 322, 53–65. Quinn, G.P., Keough, M.J., 2002. Experimental Design and Data Analysis for Biologists. Cambridge University Press, Cambridge. Snape, I., Riddle, M.J., Filler, D.M., Williams, P.J., 2003. Contaminants in freezing ground and associated ecosystems: key issues at the beginning of the new millennium. Polar Record 39, 291–300. Snape, I., Riddle, M.J., Stark, J.S., Cole, C.M., King, C.K., Duquesne, S., Gore, D.B., 2001. Management and remediation of contaminated sites at Casey Station, Antarctica. Polar Record 37, 199–214. Stark, J.S., 2008. Patterns of higher taxon colonisation and development in sessile marine benthic assemblages at Casey Station, Antarctica, and their use in environmental monitoring. Mar. Ecol. Prog. Ser. 365, 77–89. Stark, J.S., Johnstone, G.J., Palmer, A.S., Snape, I., Larner, B.L., Riddle, M.J., 2006a. Monitoring the remediation of a near shore waste disposal site in Antarctica using the amphipod Paramoera walkeri and diffusive gradients in thin films (DGTs). Mar. Pollut. Bullet. 52, 1595–1610. Stark, J.S., Riddle, M.J., Simpson, R.D., 2003a. Human impacts in soft-sediment assemblages at Casey Station, East Antarctica: spatial variation, taxonomic resolution and data transformation. Aust. Ecol. 28, 287–304. Stark, J.S., Riddle, M.J., Smith, S.D.A., 2004. Influence of an Antarctic waste dump on recruitment to near-shore marine soft-sediment assemblages. Mar. Ecol. Prog. Ser. 276, 53–70. Stark, J.S., Riddle, M.J., Snape, I., Scouller, R.C., 2003b. Human impacts in Antarctic marine soft-sediment assemblages: correlations between multivariate biological patterns and environmental variables. Estuarine, Coastal Shelf Sci. 56, 717–734. Stark, J.S., Snape, I., Riddle, M.J., 2003c. The effects of hydrocarbon and heavy metal contamination of marine sediments on recruitment of Antarctic soft-sediment assemblages: a field experimental investigation. J. Exp. Mar. Biol. Ecol. 283, 21– 50. Stark, J.S., Snape, I., Riddle, M.J., 2006b. Abandoned waste disposal sites in Antarctica: monitoring remediation outcomes and limitations at Casey Station. Ecol. Manage. Restorat. 7, 21–31. Stark, J.S., Snape, I., Riddle, M.J., Stark, S.C., 2005. Constraints on spatial variability in soft-sediment communities affected by contamination from an Antarctic waste disposal site. Mar. Pollut. Bullet. 50, 276–290.

J.S. Stark et al. / Marine Pollution Bulletin 89 (2014) 284–295 Townsend, A.T., Palmer, A.S., Stark, S.C., Samson, C., Scouller, R.C., Snape, I., 2007. Trace metal characterisation of marine sediment reference materials MESS-3 and PACS-2 in dilute HCl extracts. Mar. Pollut. Bullet. 54, 23–29. Underwood, A.J., 1991. Beyond BACI: experimental designs for detecting human environmental impacts on temporal variations in natural populations. Australian J. Mar. Freshwater Res. 42, 569–587. Underwood, A.J., 1992. Beyond BACI: the detection of environmental impacts on populations in the real but variable world. J. Exp. Mar. Biol. Ecol. 161, 145–178.

295

Underwood, A.J., 1993. The mechanics of spatially replicated sampling programmes to detect environmental impacts in a variable world. Australian J. Ecol. 18, 96– 116. Underwood, A.J., 1994. On beyond BACI: sampling designs that might reliably detect environmental disturbances. Ecol. Appl. 4 (1), 3–15. Underwood, A.J., Peterson, C.H., 1988. Towards an ecological framework for investigating pollution. Mar. Ecol. Prog. Ser. 46, 227–234.