Nickel Phase Partitioning and Toxicity in Field-Deployed Sediments

Jun 7, 2011 - Nickel Producers Environmental Research Association (NiPERA), 2605 Meridian Parkway, Suite 121, Durham, North Carolina 27713, United Sta...
35 downloads 8 Views 2MB Size
ARTICLE pubs.acs.org/est

Nickel Phase Partitioning and Toxicity in Field-Deployed Sediments David M. Costello,†,* G. Allen Burton,† Chad R. Hammerschmidt,‡ Emily C. Rogevich,§ and Christian E. Schlekat§ †

School of Natural Resources & Environment, University of Michigan, 440 Church St., Ann Arbor, Michigan 48109, United States Department of Earth & Environmental Sciences, Wright State University, 3640 Colonel Glenn Hwy., Dayton, Ohio 45435, United States § Nickel Producers Environmental Research Association (NiPERA), 2605 Meridian Parkway, Suite 121, Durham, North Carolina 27713, United States ‡

bS Supporting Information ABSTRACT: The pool of bioavailable metal in sediments can be much smaller than total metal concentration due to complexation and precipitation with ligands. Metal bioavailability and toxicity in sediment is often predicted from models of simultaneous extracted metal and acid volatile sulfide (SEMAVS); however, studies of the applicability of these models for Ni-contaminated sediments have been conducted primarily in laboratory settings. We investigated the utility of the SEM-AVS models under field conditions: Five lotic sediments with a range of sulfide and organic carbon contents were amended with four concentrations of Ni, deployed in streams for eight weeks, and examined for colonizing macroinvertebrates. After four weeks, colonizing macroinvertebrates showed a strong negative response to the Ni-treated sediments and SEM-AVS models of bioavailability differentiated between toxic and nontoxic conditions. By Week 8, relationships deteriorated between colonizing macroinvertebrates and SEM-AVS model predictions. Total Ni in the sediment did not change through time; however, Ni partitioning shifted from being dominated by organic cabon at deployment to associations with Fe and Mn. Combined geochemical and toxicity results suggest that Fe and Mn oxides in surface sediments resulted in Ni being less available to biota. This implies that current SEM-AVS models may overestimate bioavailable Ni in sediments with oxic surface layers and sufficient Fe and Mn.

’ INTRODUCTION Like other divalent metals, Ni is dynamic in sediments with bioavailability altered by flux, speciation, and partitioning. Physicochemical properties of sediments control bioavailability of Ni; complexation of free Ni (Ni2þ) with organic ligands, acid volatile sulfides (AVS), and iron and manganese oxides (FeOX and MnOX) can reduce toxicity.15 Negatively charged functional groups on particulate and dissolved organic matter can complex Ni2þ but decomposition of organic matter in sediment can mobilize Ni2þ, as either a dissolved complex or free ion, to porewater and overlying water.2,6 Sulfur-reducing bacteria in anoxic sediments produce sulfides that precipitate Fe2þ and other divalent metals for which sulfide has a high affinity (i.e., Ni, Cu, Cd, Zn, Pb, Hg). These metal-sulfide precipitates are presumed to be unavailable to benthic organisms.7,8 Mn and Fe oxides, which can be present in oxic sediments at high concentrations, also can scavenge divalent metals and make them unavailable to biota.3,4,9 Most of the recent research on Ni availability in sediments has focused on AVS and organic carbon (OC) as the primary ligands that form insoluble and nontoxic metal complexes. Empirical studies have found that mathematical combinations of simultaneously extracted nickel (SEMNi), AVS, and OC predict Ni toxicity better than total Ni concentrations.4,5,7,8,1013 Two theoretical metal availability criteria, calculated as the molar difference and ratio of SEM metals to AVS (SEMMe-AVS and r 2011 American Chemical Society

SEMMe/AVS, respectively), identify thresholds where if AVS exceeds SEM (i.e., SEMMe-AVS < 0 μmol g1 and SEMMe/ AVS < 1) the metal should occur as an insoluble metal sulfide and would not be toxic.7,8 A third measure accounts for metal binding with OC (i.e., (SEMMe-AVS)/fOC) and identifies nontoxic sediments as those with (SEMMe-AVS)/fOC < 100150 μmol g1.1,5,13 Studies of the applicability of these benchmarks for Ni contaminated sediments have been conducted primarily in laboratory settings;7,10,11 this study tests the utility of SEM-AVS models for Ni under field conditions in a more detailed manner than previously done.12 The objective of this investigation is to test AVS-SEM models of Ni bioavailability for predicting macroinvertebrate toxicity in freshwater lotic sediments. Five sediments with an inferred range of Ni-binding capacities (i.e., AVS and OC content) were amended with Ni, deployed in the field under in situ conditions, and monitored for eight weeks. Field-based manipulations allow for a controlled and replicated experiment while avoiding some artifacts of laboratory studies (e.g., elevated levels of Ni in overlying water, unrealistic flow conditions). This study is a Received: December 29, 2010 Accepted: May 23, 2011 Revised: May 10, 2011 Published: June 07, 2011 5798

dx.doi.org/10.1021/es104373h | Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology

ARTICLE

Table 1. Geochemistry of the Five Sediment Types at Deployment (Day 0) total Ni (mg kg1 dw) AVSa (μmol g1 dw)

OCa (%)

total Fea (g kg1 dw)

Spring

1.0 ( 0.9

0.47 ( 0.03

0.42 ( 0.06

0.15 ( 0.02

2.7

32

157

468

Dow

0.4 ( 0.1

1.1 ( 0.56

0.52 ( 0.11

0.10 ( 0.03

1.8

133

323

1382

sediment type

reference

low

medium

high

St. Joseph

2.5 ( 1.2

1.3 ( 0.48

1.41 ( 0.59

0.29 ( 0.09

2.7

391

649

1170

Raisin

0.8 ( 0.8

0.69 ( 0.17

0.56 ( 0.12

0.23 ( 0.07

4.0

151

328

1143

29.8 ( 36.5

4.9 ( 1.3

21.64 ( 2.26

0.53 ( 0.01

1282

3105

4978

Mill a

total Mna (g kg1 dw)

18

Values are means ((1 SD) of the four Ni-amended sediment treatments.

refinement over previous experiments of sediment Ni bioavailability 7,11,12,1416 due to (1) novel sediment “spiking” methods that produce more realistic Ni distribution between solid and dissolved phases and minimize diffusional loss, (2) a greater number and diversity of test sediments with more combinations of potential binding agents, and (3) test sediment deployment in lotic ecosystems, which is an under-represented ecosystem for testing SEM-AVS theory. Stream chemistry, sediment physicochemistry, and Ni partitioning and flux were measured to model the response of caged organisms and colonizing benthic macroinvertebrate community to sediment Ni. This investigation shows that, although sediment Ni concentrations are static during the experiment, Ni partitioning changes significantly through time. The ability of SEM-AVS bioavailability models to predict changes in the benthic macroinvertebrate community above nontoxic thresholds deteriorated through time. The weakened predictive power of the SEM-AVS model and the increase in importance of Ni associated with Fe and Mn oxides through time suggest that incorporation of Fe and Mn oxides into SEM-AVS models may improve predictions in lotic sediments.

’ EXPERIMENTAL SECTION Sediment Selection. A synoptic survey of AVS and OC in stream sediments in Michigan and Missouri was conducted to select deposits with a range of AVS and OC and, by extension, inferred Ni binding capacity. Five depositional sediments were selected and collected for use (Table 1): Mill Creek deposits had the greatest binding capacity (high AVS and OC), whereas sediment from Spring, Raisin, Dow, and St. Joseph Rivers had lower and contrasting levels of AVS and OC. About 75 L of sediment was collected from each stream with shovels, transferred to plastic-lined buckets sealed with little headspace, stored at 4 °C, and transported to the laboratory for Ni amendment. Ni-Treated Sediments. All sediments were amended with Ni by a “superspike” method that minimizes losses of AVS and large changes of pH.17,18 A small volume of each sediment was amended with NiCl2 (i.e., superspike), equilibrated, and then diluted with untreated sediment to achieve desired concentrations. The pH of initial superspike sediments was adjusted with NaOH and allowed to equilibrate for four weeks with 1 h homogenization once every week. To avoid oxidation of AVS, added solutions were deoxygenated and the headspace of all containers was filled with N2 gas. After the four-week equilibration, superspike sediments were diluted with corresponding untreated sediments to achieve three concentrations of Ni (i.e., low, medium, high). Targeted and measured Ni concentrations (576574 mg/kg dry) were related to predicted binding capacity of each sediment type with greater

binding sediments (e.g., Mill) amended with relatively more Ni than lower binding deposits (e.g., Spring; Table 1). Diluted sediments were equilibrated for an additional four weeks, with 1 h homogenization once every week. Site Selection. Ni-amended sediments were deployed in three streams in Michigan and one in Missouri (Supporting Information (SI) Table S1). Sediments were deployed in multiple streams to account for natural variation of streamwater physico-chemistry (e.g., hardness, turbidity) and regional differences in benthic community species composition. Spring, Raisin, and St. Joseph sediments were redeployed in their respective streams. Mill, Dow, and a second set of Spring sediments were deployed in Little Molasses Creek. Little Molasses Creek was selected because of its relatively low water hardness, which is known to increase the bioavailability of divalent metals.14,19 Water temperature, pH, dissolved O2, specific conductance, and turbidity were measured continuously at each site with sondes. Water hardness and alkalinity were measured at deployment and when sediment baskets were retrieved (three times total). Sediment Deployment. Sediments were deployed in colonization trays or toxicity trays (SI Figure S1)12,13 during stream baseflow conditions in August 2009. Six colonization trays were deployed for each sediment-treatment combination (e.g., Mill Low Ni), with three each removed after four and eight weeks of deployment. Mesh nylon bags were used to reduce loss of sediment during storm events, yet the mesh size (5 mm) was large enough to allow most macroinvertebrates to colonize the trays.13 All trays were placed on the stream bottom (3080 cm depth), flush with the existing sediment, and secured with steel bars. Trays were arranged in streams according to Ni treatment, with reference sediments upstream of Ni treatments that increased in order downstream. In Situ Acute Toxicity. Promptly after sediment deployment, in situ chambers containing Hyalella azteca were placed on toxicity trays for four days to assess acute toxicity (i.e., mortality and feeding rate). In situ chambers12,13 contained 10 H. azteca and ∼30 mg (dry weight) of microbial-conditioned, 1 cm diameter leaf disks (Acer rubrum). Six chambers were filled with streamwater and placed on each sediment-treatment combination with three chambers exposed to the sediment (mesh windows down) and three exposed to the overlying water (mesh windows to the side; SI Figure S1). After four days, the chambers were removed, surviving H. azteca counted, and leaf disks recovered, rinsed, dried, and reweighed. Weight loss of leaf disks was used to estimate H. azteca feeding rates, a measure of sublethal effects. Because limited acute effects were observed at sediment deployment (discussed below), in situ acute toxicity assays were not conducted during later stages of the study. 5799

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology Sediment Colonization and Geochemistry. Colonization trays were recovered from the streams and sampled after four and eight weeks of deployment.12 Results for trays deployed in Spring River are limited to four weeks only, as a result of flooding during the sixth week that buried the trays under natural sediment. Three colonization trays from each sediment-treatment combination were sampled destructively at each sampling time. One third of each tray was subdivided into surface (02 cm) and deep (26 cm) sediment and combined for geochemical analysis while the remaining two-thirds of each tray were placed into a separate 1 L bottle containing 90% ethanol for assessment of the colonizing benthic macroinvertebrate community. Preserved macroinvertebrates were removed by sieving (45 μm), identified to the family level,20 and enumerated. Benthic communities were described by six indices: taxa richness (family level), abundance, Shannon diversity, Chironomidae abundance, Gammaridae abundance, and Ephemeroptera, Plecoptera, and Trichoptera (EPT) abundance. Surface and deep sediment reserved for geochemical analysis was homogenized in the field gently and quickly (i.e., to avoid oxidation), aliquoted into cups, and frozen for analysis of AVS, SEMMe (Ni, Mn, and Fe), % solids, wet density, OC, and CaCO3. Remaining surface and deep sediment was combined and refrigerated for analysis of porewater Ni, total metals (Ni, Fe, Mn), and dissolved organic carbon (DOC). Standard gravimetric techniques were used to measure water, organic (loss-on-ignition at 550 °C for >1 h), and carbonate contents (1000 °C heating of remaining material for >1 h) of sediments.21 Loss-on-ignition was converted to OC assuming the Redfield ratio for C in organic matter (0.36).22 AVS and SEMMe were measured with methods and apparati described by Allen and colleagues,23 which involves acidification (1N HCl) of sediment followed by trapping of H2S in NaOH. Sulfide was determined colorimetrically23 and SEMMe was measured by either flame atomic absorption spectrometry (FAAS)24 or inductively coupled plasma mass spectrometry (ICP-MS)25 after 0.45 μm filtration. Fe and Mn oxides (FeOX þ MnOX) were estimated as the molar quantities of Fe and Mn in excess of AVS ([SEMMn þ SEMFe]  AVS) in SEM extracts. For porewater Ni and DOC, sediments were centrifuged, supernatant filtered through 0.45 μm polycarbonate membranes, and filtrate analyzed for Ni by ICP-MS and DOC with an infrared combustion analyzer.26 Total recoverable metals in sediment were determined by FAAS after microwave digestion of sediment in concentrated nitric acid.27 All results from QA/QC analysis were within acceptable ranges (SI Table S2). Statistical Analysis. Statistical analyses were conducted using R 2.8.1.28 In situ survival and feeding of H. azteca were analyzed with separate one-way ANOVAs for each sediment type and exposure compartment (overlying water vs sediment). To meet the assumptions of ANOVA, all survival data were arcsine square root transformed and feeding rates were square root transformed when variances were unequal. Changes in sediment geochemistry through time were assessed using ANOVA with sediment type and Ni treatment as blocking factors. The design of the experiment (multiple sediment types with few treatment levels) allowed for the benthic data to be analyzed with regression techniques. Due to complexities of field experiments (i.e., limited replication and highly variable controls), it is difficult and impractical to perform ANOVAs for particular sediment types. However, by grouping multiple sediments in a single regression analysis, the power of the statistical test is greatly improved and it

ARTICLE

is possible to find unified relationships that are applicable to a wide variety of field conditions. The partitioning of Ni between porewater and solid phases was examined by calculating a partitioning coefficient (Kd; L kg1).29 Greater Kd values indicate relatively more Ni absorbed to the solid phase and lower values signify proportionately more Ni in the porewater. Log Kd was compared among sampling dates with one-way ANOVA, followed by a Tukey’s HSD post hoc test. Multiple linear regression was used to predict log Kd from measured physicochemical parameters in surface and deep sediments (i.e., total Fe, total Mn, SEMMn, SEMFe, FeOX þ MnOX, OC, AVS, CaCO3, and density). All explanatory variables were log transformed. Stepwise selection was used to determine parameter inclusion, and Akaike’s Information Criterion with sample size correction (AICc)30 was used for selection of the “best” model(s) for Ni Kd at each sampling date (more detail in the Supporting Information). Due to potential multicollinearity between our predictor variables, any potential variable with a variance inflation factor (VIF) >2 was removed from the list of potential parameters.31 Two-way ANOVA followed by Tukey’s HSD was used to compare FeOX þ MnOX concentrations among sediment types and through time. The colonizing macroinvertebrate community indices also were analyzed with stepwise multiple linear regression with AICc. Potential explanatory variables for our benthic community indices included all the measures of Ni bioavailability in surface and deep sediments (total Ni, porewater Ni, SEMNi, SEMNi/ AVS, SEMNi-AVS, (SEMNi-AVS)/fOC, and Kd), and additional sediment physicochemical parameters (AVS, OC, total Fe, total Mn, SEMFe, SEMMn, FeOX þ MnOX, CaCO3, and density). When statistically significant, a blocking factor for stream was added to the model to account for differences in the resident benthic community available to colonize the baskets. All predictor variables were log transformed except for SEMNi-AVS and (SEMNi-AVS)/fOC, which were log þ1 transformed after negative values, which are predicted to be nontoxic,8,13 were converted to 0. Stepwise and AICc procedures for benthic indices were completed as noted above. A doseresponse curve for Gammaridae, which was the most sensitive benthic index, was produced with binomial logistic regression.

’ RESULTS AND DISCUSSION Ni-Treated Sediments. At deployment, sediment geochemistry and Ni content differed substantially among sediment types and Ni treatments (Table 1). Each sediment type had at least two treatments at Day 0 with Ni concentrations that exceeded nontoxic benchmarks based on SEM-AVS models (SEMNi-AVS > 0 μmol g1, SEMNi/AVS >1, and (SEMNi-AVS)/fOC > 120 μmol g1). Through time, total recoverable Ni concentrations did not change significantly (F2,61 = 0.09, p = 0.9, SI Tables S2 and S3). In Situ Acute Toxicity. Although amphipods are very sensitive to Ni,16 no significant mortality was observed for H. azteca exposed to either sediment or overlying water (SI Table S6). Reduced survival, though only marginally significant, was observed on Spring sediments deployed in Spring River and Dow sediments. H. azteca on Spring sediments in Spring River showed declines in survival at medium and high sediment Ni treatments (40 and 47%, respectively) relative to reference and low sediments (93 and 97%, respectively). On Dow sediments, the marginally significant difference between treatments was due 5800

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology

ARTICLE

Figure 1. Relationship between Ni partitioning coefficient (Kd) and significant sediment physicochemical variables as predicted by stepwise multiple linear regression during three sampling times. For Day 0 and Week 8 plots (A and C, respectively), the x-axes are predicted Kd values from the best-fit models. The solid line in A and C is a 1:1 line and predicted values that fall on this line are the same as observed. At Week 4 (B), total Fe, which was the only sediment physicochemical variable that predicted Kd, is on the x-axis and the solid line represents the best-fit line from least-squares regression.

to Ni-treated sediments having greater H. azteca survival than reference sediments. Reduced H. azteca survival on Spring sediments, which had one of the lowest inferred binding capacities for Ni, was not unexpected, yet the Spring sediments placed in Little Molasses Creek exhibited no acute toxicity. Statistically insignificant toxicity and inconsistency between streams suggests that these sediments were not acutely toxic. There was no observed change in H. azteca feeding rates for organisms exposed to overlying water. However, feeding rates for H. azteca exposed to low and high treatments in Raisin River (150 and 1140 mg Ni kg1 dw, respectively) were 3 higher than organisms on reference sediment. There was no measured change in the feeding rate of H. azteca exposed to the surface of other sediments. Collectively, the absence of reduced feeding rates and the lack of significant mortality indicate that these Ni-amended sediments are not acutely toxic to caged benthic organisms, although the potential for toxicity is predicted by SEM-AVS theory. However, these sediments do have the potential to negatively affect benthic organisms, and this was observed in the colonizing benthic macroinvertebrate community (discussed below). Ni Partitioning. Ni partitioning changed through time with proportionately more Ni in the porewater on Day 0 and relatively more Ni in the solid phase at Weeks 4 and 8 (F2,67 = 4.5, p = 0.01). Relatively lower Kd at deployment resulted from amended Ni being in excess of solid-phase binding sites and this Ni was either mobilized to the porewater or complexed by the sediments as the deposits aged. Physicochemical variables controlling Ni Kd changed through time as the sediments experienced diagenetic alterations. Added Ni appeared to be bound almost exclusively to sediment organic matter at deployment. On Day 0, Ni Kd was predicted by OC, SEMMn, and CaCO3, with OC being the primary driver of Ni partitioning (Figure 1A, SI Table S7). AICc indicated that a second model including just OC and SEMMn had a similar predictive power (Figure 1A, SI Table S7). By Week 4, Ni partitioning likely shifted from being bound to carbon to associations with Fe. Ni Kd was related to total Fe (Figure 1B) and AICc indicated that models with additional parameters for AVS (deep; 26 cm) and OC (surface; 02 cm) offered only slight improvement in predictive power (SI Table S7).

By Week 8, Ni partitioning was associated principally with Fe and Mn oxides. Ni Kd was predicted by FeOX þ MnOX and CaCO3 and AICc indicated a second model including just FeOX þ MnOX had similar predictive power (Figure 1C, SI Table S7). Although Ni phase association appeared to change through time, the slope of the Kd relationships did not differ among sediment types as each followed the same diagenetic process. Physicochemical variables strongly predicted Ni Kd on Day 0 but these relationships weakened (i.e., r2 declined) through time (SI Table S7). Although Kd analysis suggests that AVS was of minimal importance for predicting Ni phase partitioning, AVS was a significant factor for predicting bioavailability between deployment and Week 4. Benthic Colonization. The colonizing benthic macroinvertebrate community exhibited a strong response to the Ni-amendments four weeks after deployment: five out of six of the benthic indices were influenced negatively by at least one measure of Ni in sediment (Table 2, SI Table S5). Invertebrate abundance was best predicted by SEMNi, whereas taxa richness was best predicted by AVS corrected Ni (i.e., SEMNi/AVS) and diversity by AVS and OC corrected Ni (i.e., (SEMNi-AVS)/fOC) (Table 2, SI Table S8). The relationship between richness and deep sediment SEMNi/AVS is linear (Figure 2) and there is no sharp decline in richness as SEMNi/AVS exceeds the proposed threshold above which toxicity may occur (SEMNi/AVS > 1).5,7,8,13 Shannon diversity is reduced when (SEMNi-AVS)/fOC exceeds the proposed threshold of 100150 μmol g1;13 however, Shannon diversity was reduced at a (SEMNi-AVS)/fOC below the threshold on Dow sediment (31 μmol g1; Figure 3), which was characterized as a low affinity for Ni. Although SEM-AVS theory indicates only the potential for toxicity about the nontoxic threshold, all sediment treatments above the (SEMNi-AVS)/ fOC threshold exhibited reduced diversity (Figure 3). Individual colonizing taxa exhibited diverse responses to Ni at Week 4. As expected, Gammarus were very sensitive to Ni;16,20 EPT taxa also exhibited sensitivity to Ni, whereas Chironomids did not respond to Ni at the concentrations used in this study. Declines in EPT and Gammarus were best explained by SEMNi in deep sediment and Gammaridae colonization exhibited a strong doseresponse relationship (p < 0.001) (Figure 4). 5801

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology

ARTICLE

Table 2. Model Selectiona Summary of Sediment Physicochemical Variables That Best Predicted Indices of the Colonizing Benthic Macroinvertebrate Community time week 4

week 8

1° variableb

benthic index

2° variable c

richness

SEMNi/AVS (deep) []

SEMMn (surface) [þ]

abundance diversity

SEMNi (deep) [] (SEMNi-AVS)/fOC (surface) []

SEMNi/AVS (surface) [þ] AVS (deep) [þ]

EPT

SEMNi (deep) []

AVS (deep) []

chironomids

FeOxþMnOx (surface) []

gammarus

SEMNi (deep) []

SEMMn (surface) []

richness

total Fe []

density [þ]

abundance

SEMNi (surface) []

total Fe []

diversity EPT

density [þ] TOC (surface) []

SEMMn (deep) [þ] AVS (deep) [þ]

chironomids

SEMMn (deep) []

density []

gammarus

none

3° variable Kd []

AVS (surface) [þ]

SEMNi (surface) []

a

Parameter selection was completed by stepwise multiple linear regression followed by model comparison with Akaike’s Information Criterion. b The ranking of variables (1° to 3°) indicates the order in which they were added to the model during stepwise selection. c The sign in brackets indicates whether the coefficient was positive or negative.

Figure 2. Relationship between log SEMNi/AVS of deep sediment and taxa richness of the colonizing macroinvertebrate community at Week 4. The solid line represents the best-fit line determined from least-squares regression. SEM-AVS theory predicts that for sediments with a SEMNi/ AVS ratio below 1 (dotted vertical line) there should be no free metal and thus no toxicity. Sediment types are represented by different symbols: Dow in Little Molasses (LM) (9), Spring in LM (b), Mill in LM (2), Raisin (), St. Joseph ()), and Spring in Spring (O). Ni treatments are represented by different colors: reference (blue), low (green), medium (orange), and high (red).

The AICc model selection suggests that Ni in deeper sediment and in the SEM fraction were most likely to negatively influence the macroinvertebrate community. For the five benthic indices that responded to Ni, four indices indicated an inverse relationship with Ni in deep sediment with only Shannon diversity responding to Ni in surface sediments (Table 2). Although laboratory studies have found that Ni in surface sediments caused toxicity,10 it was observed in this study that measures of benthic colonization responded to deeper sediments. Ni in deeper sediments may have been a source to surficial sediments. Although total recoverable Ni was closely correlated with SEMNi (Pearson R = 0.80, p < 0.001), SEMNi always produced better models than

Figure 3. Relationship between (SEMNi-AVS)/fOC of surface sediment and Shannon diversity of the colonizing macroinvertebrate community at Week 4. (SEMNi-AVS)/fOC values were log þ1 transformed after negative values were converted to 0. SEM-AVS theory predicts that for sediments with (SEMNi-AVS)/fOC below 100150 μmol g1 (shaded box) should be nontoxic. Sediment types are represented by different symbols: Dow in Little Molasses (LM) (9), Spring in LM (b), Mill in LM (2), Raisin (), St. Joseph ()), and Spring in Spring (O). Ni treatments are represented by different colors: reference (blue), low (green), medium (orange), and high (red).

total Ni. Small concentration differences between total Ni and the SEM fraction appear to be crucial to influencing Ni bioavailability to benthic organisms. Moreover, although SEM-AVS theory suggests that exposure to porewater metals is the cause of toxicity,1,8 SEMNi always outperformed direct measures of porewater Ni in predicting the benthic response. These results suggest that measurements of Ni in a dilute acid extract of sediments, even without knowledge of other binding agents (e.g., AVS and MnOX), could greatly improve estimates of bioavailable Ni. 5802

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology

Figure 4. Relationship between log SEMNi of deep sediment and Gammarus abundance (as a percent of reference sediments) at Week 4. The solid line represents the best-fit line from binomial logistic regression and the dotted line is the 95% confidence interval of the line. Sediment types are represented by different symbols: Dow in Little Molasses (LM) (9), Spring in LM (b), Mill in LM (2), Raisin (), St. Joseph ()), and Spring in Spring (O). Ni treatments are represented by different colors: reference (blue), low (green), medium (orange), and high (red).

Although measures of Ni were the primary drivers of reduced benthic indices, multiple regression identified other sediment physicochemical variables that modified the benthic community response. Models for Shannon diversity and EPT abundance were improved by the inclusion of AVS and models for taxa richness and Gammarus abundance were improved by including SEMMn. Together with modifications to Ni bioavailability (e.g., SEMNi/AVS), at least one measure of AVS was included in four of the five benthic indices that had responded to Ni at Week 4 (Table 2). In comparison, Mn was selected for two models and OC for just one model. Collectively, these models suggest that for colonizing benthic macroinvertebrates AVS is the most likely ligand reducing Ni bioavailability followed by Mn oxides then organic carbon. The colonizing benthic macroinvertebrate community response to Ni changed significantly by the second sampling period (Week 8). Median total Ni and deep SEMNi declined minimally (10% and 3%, respectively) from Week 4 to 8, yet for the most part, the benthic macroinvertebrates were not adversely affected by Ni. Of the six benthic indices measured, only two were inversely related to Ni (Table 2). Total macroinvertebrate abundance was best predicted by surface SEMNi and models of Chironomidae abundance included surface SEMNi as the tertiary variable in a three parameter model (Table 2, SI Table S9). Taxa richness, Shannon diversity, and EPT abundance were all related to sediment physicochemical variables other than Ni, and Gammaridae abundance was not related to any of the measured sediment variables (Table 2, SI Table S9). The temporal trend in this study is in contrast to that observed by Nguyen and colleagues,12 who found Ni toxicity occurred nine months after sediment deployment. In this study, Ni was distributed more realistically between porewater and solid phases and the lotic sediments had greater Ni-binding capacity (more representative of natural sediments), which may explain the temporal differences between the two studies. The minimal response by the benthic community at Week 8 in the presence of elevated total Ni concentrations suggests that the

ARTICLE

Ni was bound to ligands that made it unavailable to biota. The current SEM-AVS bioavailability models account for the binding action of AVS and OC, but all of the SEM-AVS measures of bioavailability predicted that many of the sediments exceeded the nontoxic thresholds. Additionally, porewater measurements from Week 8 indicate that there is mobile Ni at depth (e.g., mean 120 μg L1 in high treatments). At the week 4 sampling period, sediments that exceeded the nontoxic threshold exhibited toxicity (e.g., reduced Shannon diversity), whereas, during the week 8 sampling, sediments continued to exceed nontoxic benchmarks but now showed no adverse benthic response (Figure 3, Table 2, SI Tables S3S5). Fe and Mn oxides in surface sediments increased significantly through time (F2,51 = 3.7, p = 0.03) with more Fe and Mn oxides in surface sediments at Week 8 compared to Day 0. Due to increases in Mn and Fe oxides and their greater influence on the partitioning of Ni at Week 8, we suggest that Mn and Fe oxide surfaces may be binding Ni and reducing bioavailability.3,9 As lotic sediments diagenetically age, suspend and deposit, or are buried, oxic surfaces are created or destroyed. Sediments that develop an oxic layer may contain more Fe and Mn oxides that can scavange metals and make the sediments less toxic. Because we observed minimal toxicity in our sediments at Week 8, even at the highest Ni concentrations, the Fe and Mn oxide pool must be sufficiently large to bind most of the available Ni. SEM-AVS models of bioavailability do not account for protective effects of Fe and Mn oxides and overestimate the bioavailable fraction of Ni. This suggests that for lotic sediments with sufficient Fe and Mn, a Mn and Fe oxide corrected measure of Ni bioavailability may offer improved predictive power of metal toxicity for concentrations above nontoxic thresholds. The temporal aspect of relationships between Ni and benthic indices in this study raises questions about the application of laboratory-derived bioavailability models for sediment risk assessment. Sediment risk assessment goals typically assess metal exposure risk over the long-term yet our study indicates that currently used laboratory methods may not replicate crucial Ni dynamics. We found that the distribution and effects of Ni in situ are temporally dynamic, with an overall trend toward decreasing toxicity and bioavailability over time. Laboratory bioassays, which assume a constant relationship between sediment Ni phases and toxicity through time (i.e., steady-state approach), do not represent lotic field conditions and may overestimate bioavailability and effects of Ni. Specifically, we suggest further examination of the temporal and spatial dynamics of oxidation of contaminated lotic sediments as it related to toxicity. Most likely, sediment risk assessment on laboratory toxicity data represents a worst case scenario, and uncertainty analyses accompanying Ni sediment risk assessments need to consider the dynamic nature of Ni in natural sediments. The careful and thorough methods used in this study allowed for elucidations of trends between sediment geochemistry and benthic macroinvertebrates that are representative of natural in situ conditions and can be used to create protective bioavailability models. In this study, Ni-amended sediments followed a predictable pattern of diagenesis after deployment in the streams with Ni bound initially to OC then switching to associations with AVS and Fe and Mn oxides. Similar to laboratory results,7,10,11 SEM-AVS models predicted invertebrate responses at Week 4 but these models were not sufficient by Week 8. Temporal inconsistency in the effectiveness of SEM-AVS models when predicting toxicity of sediments above nontoxic thresholds 5803

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology presents a challenge for the management of contaminated ecosystems. The protective effects of OC expected by the SEM-AVS models were rarely observed, but rather there was evidence that association with Fe and Mn oxides reduced the bioavailability of Ni. This implies that future models of metal bioavailability should explicitly include pools of FeOX and MnOX as potential ligands able to reduce toxicity.

’ ASSOCIATED CONTENT

bS

Supporting Information. Available information includes additional information about AICc and model selection process, images of the acute toxicity and benthic colonization chambers, table of surface water physical and chemical parameters measured during the study, results from QA/QC analysis, tables of sediment geochemistry and benthic data, statistics from in situ acute toxicity, and statistical tables for the suite of models used in the three model selection procedures (AICc). This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: 734-763-3602; e-mail: [email protected].

’ ACKNOWLEDGMENT We thank Anthony Honick, Kevin Custer, Michael Eggleston, Miling Li, Michelle Sawyers, Stephanie Tubbs, Tom Aepelbacher, Rachel Bredernitz, Avani Naik, Melissa Tabatchnick, Katlin Bowman, Robbie Weller, and Matt Konkler for field and lab assistance. Chris Ingersoll, Bill Brumbaugh, and John Besser from the USGS in Columbia, MO assisted with the sediment spiking and field deployments. Four anonymous reviewers provided useful comments on a previous version of this manuscript. Funding was provided by NiPERA. ’ REFERENCES (1) Di Toro, D. M.; McGrath, J. A.; Hansen, D. J.; Berry, W. J.; Paquin, P. R.; Mathew, R.; Wu, K. B.; Santore, R. C. Predicting sediment metal toxicity using a sediment biotic ligand model: Methodology and initial application. Environ. Toxicol. Chem. 2005, 24 (10), 2410–2427. (2) Doig, L. E.; Liber, K. Nickel speciation in the presence of different sources and fractions of dissolved organic matter. Ecotox. Environ. Safe. 2007, 66, 169–177. (3) Takematsu, N. Sorption of transition metals on manganese and iron oxides, and silicate minerals. J. Oceanogr. Soc. Jpn. 1979, 35, 36–42. (4) Burton, G. A.; Green, A.; Baudo, R.; Forbes, V.; Nguyen, L. T. H.; Janssen, C. R.; Kukkonen, J.; Leppanen, M.; Maltby, L.; Soares, A.; Kapo, K.; Smith, P.; Dunning, J. Characterizing sediment acid volatile sulfide concentrations in European streams. Environ. Toxicol. Chem. 2007, 26 (1), 1–12. (5) Procedures for the Derivation of Equilibrium Partitioning Sediment Benchmarks (ESBS) for the Protection of Benthic Organisms: Metal Mixtures (Cadmium, Copper, Lead, Nickel, Silver and Zinc); U.S. Environmental Protection Agency: Washington, DC, 2005. (6) Shaw, T. J.; Gieskes, J. M.; Jahnke, R. A. Early diagenesis in differing depositional-environments - the response of transition-metals in pore water. Geochim. Cosmochim. Acta 1990, 54 (5), 1233–1246. (7) Di Toro, D. M.; Mahony, J. D.; Hansen, D. J.; Scott, K. J.; Carlson, A. R.; Ankley, G. T. Acid volatile sulfide predicts the acute toxicity of cadmium and nickel in sediments. Environ. Sci. Technol. 1992, 26 (1), 96–101.

ARTICLE

(8) Di Toro, D. M.; Mahony, J. D.; Hansen, D. J.; Scott, K. J.; Hicks, M. B.; Mayr, S. M.; Redmond, M. S. Toxicity of cadmium in sediments: the role of acid volatile sulfide. Environ. Toxicol. Chem. 1990, 9 (12), 1487–1502. (9) Tebo, B. M.; Bargar, J. R.; Clement, B. G.; Dick, G. J.; Murray, K. J.; Parker, D.; Verity, R.; Webb, S. M. Biogenic manganese oxides: properties and mechanisms of formation. Annu. Rev. Earth Planet. Sci. 2004, 32, 287–328. (10) Vandegehuchte, M. B.; Roman, Y. E.; Nguyen, L. T. H.; Janssen, C. R.; De Schamphelaere, K. A. C. Toxicological availability of nickel to the benthic oligochaete Lumbriculus variegatus. Environ. Int. 2007, 33, 736–742. (11) Lee, J.-S.; Lee, J.-H. Influence of acid volatile sulfides and simultaneously extracted metals on the bioavailability and toxicity of a mixture of sediment-associated Cd, Ni, and Zn to polychaetes Neanthes arenaceodentata. Sci. Total Environ. 2005, 338, 229–241. (12) Nguyen, L. T. H.; Burton, G. A.; Schlekat, C. E.; Janssen, C. R., Field measurement of nickel sediment toxicity: role of acid volatile sulfide. Environ. Toxicol. Chem. 2010, 30 (1), 162172. (13) Burton, G. A.; Nguyen, L. T. H.; Janssen, C.; Baudo, R.; McWilliam, R. A.; Bossuyt, B.; Beltrami, M.; Green, A. Field validation of sediment zinc toxicity. Environ. Toxicol. Chem. 2005, 24 (3), 541–553. (14) Nowierski, M.; Dixon, D. G.; Borgmann, U. Effects of water chemistry on the bioavailability of metals in sediment to Hyalella azteca: implications for sediment quality guidelines. Arch. Environ. Contam. Toxicol. 2005, 49, 322–332. (15) Borgmann, U.; Neron, R.; Norwood, W. P. Quantification of bioavailable nickel in sediments and toxic thresholds to Hyalella azteca. Environ. Pollut. 2001, 11, 189–198. (16) Milani, D.; Reynoldson, T. B.; Borgmann, U.; Kolasa, J. The relative sensitivity of four benthic invertebrates to metals in spikedsediment exposures and application to contaminated field sediment. Environ. Toxicol. Chem. 2003, 22 (4), 845–854. (17) Hutchins, C.; Teasdale, P.; Lee, S.; Simpson, S. The effect of sediment type and pH-adjustment on the porewater chemistry of copperand zinc-spiked sediments. Soil Sediment Contam. 2009, 18 (1), 55–73. (18) Simpson, S. L.; Angel, B. M.; Jolley, D. F. Metal equilibration in laboratory-contaminated (spiked) sediments used for the development of whole-sediment toxicity tests. Chemosphere 2004, 54 (5), 597–609. (19) Pagenkopf, G. K. Gill surface interaction model for trace-metal toxicity to fishes: role of complexation, pH, and water hardness. Environ. Sci. Technol. 1983, 17 (6), 342–347. (20) Clements, W. H.; Carlisle, D. M.; Lazorchak, J. M.; Johnson, P. C. Heavy metals structure benthic communities in Colorado mountain streams. Ecol. Appl. 2000, 10 (2), 626–638. (21) Heori, O.; Lotter, A. F.; Lemcke, G. Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results. J. Paleolimnol. 2001, 25, 101–110. (22) Redfield, A. C., On the proportions of organic derivitives in sea water and their relation to the composition of plankton. In James Johnstone Memorial Volume; Daniel, R. J., Ed.; Liverpool University Press: Liverpool, U.K., 1934; pp 176192. (23) Allen, H. E.; Fu, G.; Boothman, W.; DiToro, D. M.; Mahony, J. D. Determination of Acid Volatile Sulfide and Selected Simultaneously Extractable Metals in Sediment; U.S. Environmental Protection Agency: Washington, DC, 1991; p 18. (24) U.S. Environmental Protection Agency. Method 7000B: Flame Atomic Absorption Spectrophotometry; U.S. Environmental Protection Agency: Washington DC, 2007. (25) U.S. Environmental Protection Agency Method 6020A: Inductively Coupled Plasma-Mass Spectrometry; U.S. Environmental Protection Agency: Washington, DC, 2007. (26) Sharp, J. H.; Benner, R.; Bennett, L.; Carlson, C. A.; Fitzwater, S. E.; Peltzer, E. T.; Tupas, L. M. Analyses of dissolved organic-carbon in seawater—The JGOFS EqPac methods comparison. Mar. Chem. 1995, 48 (2), 91–108. 5804

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805

Environmental Science & Technology

ARTICLE

(27) U.S. Environmental Protection Agency Method 3015A: Microwave Assisted Acid Digestion of Aqueous Samples and Extracts; U.S. Environmental Protection Agency: Washington, DC, 2007. (28) R Development Core Team R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2008. (29) Manahan, S. E.Environmental Chemistry, 8th ed.; CRC Press: Boca Raton, FL, 2005; p 783. (30) Burnham, K. P.; Anderson, D. R. Multimodel inference understanding AIC and BIC in model selection. Sociol. Method. Res. 2004, 33 (2), 261–304. (31) Graham, M. H. Confronting multicollinearity in ecological multiple regression. Ecology 2003, 84 (11), 2809–2815.

5805

dx.doi.org/10.1021/es104373h |Environ. Sci. Technol. 2011, 45, 5798–5805