Ambient concentrations of metals in surface waters have become an important consideration when establishing water quality criteria and conducting risk assessments. This study sought to estimate amounts of zinc that may be released into freshwater considering ambient concentrations, toxicity thresholds, and bioavailability. Cumulative distribution functions of ambient zinc concentrations were compared statistically for streams and lakes in Europe, North America, and South America to identify differences among mean distribution variables (e.g., slopes, intercepts, and inflection points). Results illustrated that most of the distributions among sites differed significantly. These differences illustrate the variability in ambient zinc concentrations in surface waters because of geographic location, regional geology, and anthropogenic influence. Additionally, water quality data were used to estimate bioavailable zinc concentrations in ambient surface waters (based on predictions using biotic ligand models). The amount of dissolved metal that could be added to surface waters without exceeding toxicity thresholds was calculated by subtracting ambient surface water concentrations from chronic no observable effect concentrations (NOEC; reproduction for Daphnia magna) or 10% effective concentrations (EC10; growth rate for Pseudokirchneriella subcapitata). Because ambient dissolved-zinc concentrations were, on average, below predicted effects thresholds, an average of 57.1 ± 175 μg/L (±SD) of zinc could be added before exceeding the D. magna chronic NOEC or the P. subcapitata chronic EC10. However, numerous sites (17%) were identified as having ambient zinc concentrations in excess of these toxicity thresholds. This article uses existing biotic ligand models for zinc to estimate the potential magnitudes and variabilities of bioavailable zinc concentrations in fresh surface waters from different regions of the world.
Ambient concentrations of metals in surface waters are usually assumed, for regulatory purposes, to be completely bioavailable (i.e., the toxicologically effective metal concentration that is available at the site of toxic action, the biotic ligand) (Di Toro et al. 2001; De Schamphelaere et al. 2005). Metal concentrations stipulated in water quality criteria or guidelines—whether based on total, total recoverable, or dissolved concentrations—have been assumed to be 100% bioavailable, unless adjusted with empirically based translators like water effect ratios (US Environmental Protection Agency [USEPA] 1994; Allen and Hansen 1996; Wood et al. 1997) or other conventions such as the added-risk approach (Crommentuijn et al. 1997; Struijs et al. 1997; Crommentuijn et al. 2000). Criteria based on total, total recoverable, and dissolved metal do not account for site-specific modifying factors (e.g., pH, hardness, dissolved organic matter) and have resulted in water quality criteria/guidelines that are either overprotective or under protective relative to their stated goal of protecting 95% of the species (Allen and Hansen 1996; Wood et al. 1997; Van Genderen et al. 2005). The objective of this study was to estimate the actual degree of bioavailability and aquatic toxicity potential of ambient metal concentrations using the added-risk approach (Crommentuijn et al. 1997; Struijs et al. 1997; Crommentuijn et al. 2000) and biotic ligand models (BLMs; Di Toro et al. 2001).
The added-risk method uses mathematical equations to calculate maximum permissible additions (MPAs) of metals in ambient surface waters, based on assumptions about background concentrations, the fraction of metal that is considered bioavailable, species sensitivities, and the water quality guideline. The MPA is the maximum anthropogenic concentration that can be added to a water body without exceeding the maximum permissible concentration—the concentration estimated to adversely affect 5% of the aquatic species (HC5, the hazardous concentration to 5% of the species). The MPA is analogous to the difference between the water quality guideline (or a toxicity threshold for a sensitive species) and the background or ambient concentration. Crommentuijn et al. (2000) adopted 2.8 μg/L dissolved zinc as the background concentration and considered this concentration to be too low to influence decisions on how much zinc could be released into streams from anthropogenic sources, regardless of bioavailability. This judgment was based on comparing 2.8 μg/L to an estimated toxicological threshold for zinc of approximately 10 μg/L (i.e., HC5; Aldenberg and Slob 1993).
Uncertainties about the bioavailability of zinc in surface waters arguably have been remedied by the development of metal-specific BLMs (e.g., De Schamphelaere and Janssen 2004a; Heijerick et al. 2005; Clifford and McGeer 2009), which provide tools for quantitatively defining the bioavailability of divalent metals like zinc in surface waters, in toxicity tests, and in water quality guidelines (Janssen et al. 2003; Bossuyt 2004; Specht 2005; EU 2006).
With the development of BLMs, refinement of governmental regulation of anthropogenic metal loading to surface waters is now a feasible application for certain metals (Janssen et al. 2003; Van Genderen et al. 2008). Following such methods, this study addresses 3 questions concerning the magnitude and toxicity of zinc in a range of ambient surface waters. First, to what extent do ambient zinc concentrations differ regionally? Second, what fractions are bioavailable? Third, how much zinc could be released into these waters without exceeding a chronic toxicity threshold based on water quality and zinc bioavailability (i.e., calculation of sitespecific MPAs)? Assessment of these questions will refine our understanding of the degree to which BLMs can be used in regulatory contexts in a variety of water quality conditions throughout the world.
MATERIALS AND METHODS
Water quality data
Existing surface-water quality data were compiled to estimate ambient zinc speciation and bioavailability in surface waters from several parts of the world. Accordingly, robust data sets (at least 30 samples from a given location) containing information on dissolved-zinc concentrations were collected from electronic sources. Specifically, dissolved-zinc data for North American sites were obtained from the US Geological Survey Website (http://waterdata.usgs.gov/nwis/qw); data for Chilean sites were obtained from Dr Patricio Rodriguez (see also Villavicencio et al. 2005); and data for European sites were obtained from the Forum of the European Geological Surveys Geochemical Database (http://www.gsf.fi/publ/foregsatlas/index.php).
To define zinc speciation and site-specific bioavailability, data sets containing ambient surface water concentrations of the following parameters were used: alkalinity or carbonate, calcium (Ca), chloride (Cl), dissolved zinc, dissolved organic carbon (DOC), magnesium (Mg), pH, potassium (K), sodium (Na), and sulfate (SO4). Data sets in which all these parameters had been measured in the same sample were obtained from freshwater sites in Europe, North America, and South America to evaluate how diverse water qualities affected speciation and bioavailability of dissolved zinc. Freshwater data from sites in Chile and throughout Europe represented multiple locations in which only a single measurement of water quality was reported for each (Table 1). Freshwater data for sites in North America represented multiple water quality measurements for each location (Table 1).
Requests for the toxicological and chemical data sets compiled for this study (nearly 850 ambient zinc observations and BLM simulations for individual surface water samples) should be sent to the corresponding author.
Bioavailability estimation procedures
The BLMs for sensitive aquatic species and endpoints were obtained from published studies: Daphnia magna and Pseudokirchneriella subcapitata were considered representative of the most sensitive aquatic species to zinc in freshwaters (Bodar et al. 2005; EU 2006). Version 2.2.3 of the BLM was downloaded at http://www.hydroqual.com/wr_blm.html (HydroQual). Measured concentrations of pH, DOC, Ca, Mg, Na, K, Cl, SO4, and alkalinity (units of mg/L for DOC and dissolved ions and mg/L as calcium carbonate [CaCO3] for alkalinity) were used as model input parameters. Because water samples used in this study were from surface waters rather than sediment elutriates, the presence of sulfides was not considered significant and a default value of 0.01 μM was used for all simulations (Di Toro et al. 2001; Sukola et al. 2005).
The BLM parameter data and thermodynamic database files for each of the models used here were developed and validated in previous studies. The BLMs for D. magna and P. subcapitata were developed and field-validated using reproductive NOEC and growth EC10 data, respectively, for laboratory and natural waters (De Schamphelaere et al. 2005; parameter files obtained from KD Schamphelaere, Gent University, Belgium). For the D. magna model, DOC concentrations were assumed to represent 61% of measured values, and humic acid was set at 0.01% (KD Schamphelaere, personal communication). Although some research has demonstrated that knowledge of humic acid concentrations may improve predictions (Schwartz et al. 2004; De Schamphelaere and Janssen 2004b), this information was not available for the data sets compiled in this study. Additionally, the improvements are thought to be relatively small. For algae, a simple linear equation is used to calculate the free zinc (Zn2+) concentration associated with the EC10 value (De Schamphelaere et al. 2005):
The BLM was then used to back-calculate a dissolved Zn concentration associated with this free-zinc value and associated water quality. Because the sensitivity of D. magna and P. subcapitata differs depending on water quality (daphnids are more sensitive at low pH, and algae are more sensitive as pH raises above 7), the lowest toxicity value of the 2 simulations was used for further analysis.
Because both the USEPA and the European Union currently use BLM-based approaches for deriving site-specific criteria/standards for zinc in freshwaters, they were compared with the approach presented here. The current USEPA approach calculates an acute zinc criterion for each location (using acute Zn-BLM developed by HydroQual; Santore et al. 2002) and divides it by a constant acute to chronic ratio (5.327; unpublished update submitted by the International Zinc Association to USEPA, July 2006) to estimate a chronic criterion. The European Union has 2 approaches and both were used. First, the most conservative, predicted no-effect concentration (PNEC or HC5) was calculated using the water quality-normalized chronic species sensitivity distribution approach (extrapolated from Bodar et al. 2005). Second, the most conservative BLM-calculated bioavailability factor was applied to the ambient zinc concentration for each site (EU 2006). That is, the ambient zinc concentration was multiplied by a bioavailable fraction that was calculated by dividing the predicted site-specific toxicity and reference water (essentially a water-effect ratio; EU 2006). The chronic criterion estimates from each approach were compared with the ambient zinc concentration within respective samples as a quotient (Eqn. 3) to determine whether an ambient zinc concentration exceeded the site-specific chronic zinc criterion/ standard.
Table Table 1.. Median (dissolved zinc) and average (water quality parameters) values and ranges reported for numerous sitesa
Dissolved Zn μg/L
Alkalinity mg/L as CaCO3
Nr of samples
aA water quality variable with an average but no range (–) is representative of having only 1 measurement within the available data set. This average value was applied to each sample within the respective data set for biotic ligand model simulations. Ca = calcium; CaCO3 = calcium carbonate; DOC = dissolved organic carbon; Mg = magnesium; SU = standard units; Zn = zinc.
Arkansas River (Granite, CO, USA)
Arkansas River (Nathrop, CO, USA)
Columbia River (Kiona, WA, USA)
Mississippi River (Memphis, TN, USA)
Ohio River (Wheeling, OH, USA)
Before estimating bioavailability, differences in data sets were examined. Cumulative distribution functions were constructed to graphically compare dissolved-zinc concentration data sets. Dissolved zinc concentrations within each data set were ranked (ni/nt + 1) from lowest to highest. The nonparametric Kolmogorov–Smirnov test was used to determine differences among dissolved-zinc distributions. To decrease the risk of making a type 1 error (false rejection) among the 105 multiple comparisons, a Bonferroni correction was applied to the alpha value (α = 0.05/105 = 0.00048). In consequence, there was low likelihood of making a type II error of declaring a difference when there was none. The Kolmogorov-Smirnov test is based on magnitudes of differences between data pairs in 2 cumulative distribution functions. When this difference is significantly large, the 2 distributions are considered statistically different.
Additional properties were examined to understand similarities and differences in the distributions. An analysis of covariance was also used to determine pairwise differences (α = 0.05) among cumulative-distribution functions slopes. Tests for homogeneity of regression were performed by estimating the covariate-by-factor interaction for 2 different slopes (SPSS). This comparison is the same as testing for parallel slopes, as described by Neter et al. (1990). To compare 2 regression models, zinc concentrations were considered the independent variable, and the cumulative distribution itself was considered to be the dependent variable. The ambient zinc data were transformed using log10 (dissolved zinc + 1), and the cumulative distribution was transformed using the following logit:
Differences among slopes could represent situations in which anthropogenic or other influences on ambient zinc concentrations cause deviations from the expected normal distribution. That is, although median values among different distributions may appear similar, the tails of these distributions may represent extreme situations (point or nonpoint source contributions), which may be unrepresentative of background alone.
The original goal of determining bioavailability of zinc concentrations was to rely upon multiple measurements made over time at individual locations to represent temporal variation. This was feasible for the North American data sets but not for the Chilean and European data sets, where each datum represented a single grab sample from a cross-section of streams and lakes in each region. Data within each data set were ranked from lowest to highest, relative to measured, ambient, dissolved-zinc concentrations for semilog plotting as cumulative distribution functions. Next, the thermodynamic speciation of zinc in the presence of biotic ligands was calculated based on each site's water quality. In this step, the BLM was not predicting the amount of zinc needed to elicit toxicity to each indicator organism but, rather, calculating the steady-state chemical speciation of zinc, relative to the biotic ligand and the sample's water quality. This analysis estimated ambient concentrations of free-zinc concentrations (Zn2+) and those expected to accumulate on the biotic ligand (De Schamphelaere et al. 2005).
To calculate how much zinc measured at each site was bioavailable, BLM-predicted toxicities were estimated based on each site's water quality and the BLM. This value was then compared with each site's dissolved-zinc concentration using the following index:
The MPA was estimated using the BLM as the difference between the ambient dissolved concentration and estimated toxicity of dissolved zinc to each indicator organism:
Negative values resulting from this calculation suggest that the ambient zinc concentration was in excess of the predicted concentration that would result in sublethal effects on freshwater and saltwater organisms.
Another method for calculating an MPA has been developed by Crommentuijn, Struijs, and their coworkers (Struijs et al. 1997; Crommentuijn et al. 2000), using the following log-based equation:
where α (1.98) and β (0.2) are estimates of species sensitivity distributions (EU 2006) for freshwater organisms exposed to specific toxicants; the PAFmax is an additional effect to aquatic biota that could be realized without exceeding the sensitivity of the 5th percentile species; and ϕ is the fraction of the preindustrial, background zinc concentration (Cb) that is bioavailable.
RESULTS AND DISCUSSION
Ambient water quality
The 15 sites varied greatly in water quality. Mean concentrations of DOC, hardness, and pH changed by latitude in Europe, with the Chilean and US sites falling within the range found in Europe (Table 1). For example, average DOC values were less than 5 mg/L for 12 locations, between 5 and 10 mg/L for 2 locations, and greater than 10 mg/L for 1 location (Finland). Total hardness increased from means of 13 to 24 mg/L CaCO3 in the Nordic countries to greater than 250 mg/L CaCO3 in Italy and Spain. Following the same pattern, mean pH levels were less than 7 in the Nordic countries and between 7 and 8 in the remainder.
Ambient dissolved-zinc concentrations
The magnitudes and distributions of ambient dissolved-zinc concentrations from around the world appear very different, likely reflecting differences in climate, geology, geography, and land use (Figure 1 and Table 1). Zinc concentrations were highest (median ∼60 μg/L) at the 2 US sites on the Arkansas River that were closest to historic and current lead-zinc mining. Moreover, median concentrations at the remaining US sites (16–20 μg/L) were appreciably more than those in Chile (3 μg/L) and Europe (1.7–4.5 μg/L) (Figure 1 and Table 1). Some bias may be included for North American sites because these data sets include older data, where analytical detection limits were higher. Nevertheless, their inclusion in this analysis fulfills the objectives of identifying potential locations where further assessment may be needed. Medians between sites varied nearly 40-fold (2.4-95 μg/L) and within sites by 8-fold to 760-fold (Figure 1).
Despite visual similarities among distributions, 63% of the 105 pairwise comparisons among distributions were statistically different (p < 0.00048) using the Bonferroni-corrected Kolmogorov-Smirnov test (data not shown). This suggested that only a few sites, those with overlapping distributions, had similar ambient zinc concentrations. This conclusion contests the assumption that a single ambient-zinc concentration can be assigned to different sites or water bodies, even within the same region (Struijs et al. 1997; Crommentuijn et al. 2000). Probable explanations for this would be differences in land use (e.g., forested, agricultural, urban, mining), geology (e.g., bed rock, sedimentary rock, metamorphic and igneous rocks, fluvial and lacustrine deposits), and climate (McLaughlin and Smolders 2001).
Based on distribution slopes (logit-transformation; Eqn. 2), 87% (13 of 15) of the sites were considered statistically similar (Figure 1B). Visually, these sites all have relatively consistent trends with small tails at either end of the distribution (Figure 1A). However, for those distribution slopes where the slopes were different—Italy and the United Kingdom—the distributions appear to have very little variation in the tails relative to the overall distribution (UK) or to relatively moderate slope (Italy). This finding suggests that most of the sites within Europe or North America have relatively consistent ranges in ambient zinc concentrations.
The ambient zinc concentrations representing 1 SD from the distribution medians, ranged from 3.7 to 140 μg/L (+1 SD) and 0.62 to 30 μg/L (-1 SD). The overall range in ambient zinc concentrations was 0.090 to 620 μg/L (Table 1). Although differences in ranges and slopes among individual distributions were evident, the intercontinental consistency was apparent, considering that the locations represent both temporal variability (North America) and spatial and regional variability (Chile and Europe) (Figure 1). These findings agree with similar analyses performed for ambient copper concentrations (Van Genderen et al. 2008) and suggest that any effect associated with anthropogenic contributions was small, relative to each site's overall distribution.
Zinc speciation and bioavailability
Water quality parameters used in the BLM simulations are summarized in Table 1. They were used to develop quantitative estimates for 3 parameters: distributions of ambient dissolved-zinc concentrations in natural surface waters, the bioavailable fraction of zinc, and the BLM predicted MPA.
In all surface waters, median dissolved-zinc concentrations were 1 order or magnitude higher than estimated free-zinc concentrations and 5 orders of magnitude higher than concentrations of zinc bound to the biotic ligand. Median concentrations for all data sets were approximately 3.6 μg/L dissolved Zn, 0.76 μg/L free Zn, and 2.7 × 10−5 μg/L biotic ligand-bound Zn.
Dissolved organic carbon had a preeminent influence on the bioavailability and toxicity potential of zinc in the data sets evaluated (Figure 2). Despite zinc bioavailability being low in the majority of surface waters, sites whose ambient zinc concentrations exceeded toxicity thresholds had less than 5 mg/L of DOC present (Figure 2A). Although pH can also significantly influence zinc bioavailability (De Schamphelaere 2005), its influence was not as evident for these sites. Bioavailability indices ranged from 0.0017 to 51 (Figure 2A). On average, this suggests that ambient dissolved-zinc concentrations represented 86% (average bioavailability index of 0.86 for entire data set) of the dissolved-zinc concentration required to meet or exceed the predicted effects thresholds for sensitive aquatic biota. Similarly, the dissolved-zinc toxicity threshold varied systematically with DOC through the range of 0.070 to 72 mg/L as C (Figure 2B). The chronic NOEC (D. magna) or EC10 (P. subcapitata) values, which ranged from 2.7 to 3200 μg/L, varied with DOC concentration (Figure 2B). This finding supports the need to consider DOC concentration when determining site-specific standards for zinc, as emphasized recently by several researchers (Santore et al. 2002; De Schamphelaere et al. 2005; Van Genderen et al. 2008).
Moreover, 17.4% (145 of 834) of the surface water samples exceeded the toxicological thresholds for the indicator species and endpoints (i.e., indices exceeding 1.0; Figure 2A). Notably, 122 (84.1%) of the 145 exceedances occurred in the North American locations. Of the 23 sites in Europe where there were exceedances, 13 (56.5%) had dissolved-zinc concentrations below the overall average for the entire data set (15.7 μg/L). Using both the USEPA and European Union BLM approaches, approximately 12% to 27% of the ambient zinc concentrations appeared to exceed the regulatory standards (data not shown). In particular, chronic criterion estimated using the USEPA, European Union full species sensitivity distribution-normalization, and European Union bioavailability factor approaches identified 12.2%, 17.5%, and 26.5% of the surface-water samples as exceeding the toxicological thresholds for the indicator species and endpoints, respectively. This suggests that the BLM-based approach used in this study and 3 independent regulatory approaches might produce similar results. Comparisons among BLM-based MPA approach and established regulatory methods were also consistent when this approach was evaluated for copper (Van Genderen et al. 2008). However, none of the regulatory approaches provides a mechanism to estimate metal magnitudes that could be added to a water body and still protect aquatic life.
Potential regulatory applications
Because most (82.6%) of the bioavailability indices was less than 1.0 for the surface waters evaluated, it is assumed that these sites could assimilate additional zinc loadings without exceeding toxicity thresholds. The BLM-based MPAs ranged from a negative value (-608) to 3150 μg/L and averaged approximately 57 μg/L, meaning that these concentrations theoretically can be added to the water bodies without risk of adverse effects to the species studied (Figure 3). Negative MPA values indicate the dissolved-zinc concentrations exceeded the toxicity threshold based on the site's water quality. In fact, by relating both the bioavailability index and the MPA for each site, the 2 metrics can identify situations in which toxicity may result from elevated ambient-zinc concentrations (Figure 3).
Another method for calculating the MPA, termed the added-risk approach by Crommentuijn, Struijs, and their coworkers (Eqn. 5; Struijs et al. 1997; Crommentuijn et al. 2000), applies a single estimate of the background zinc concentration (e.g., 2.8 μg/L) to all water bodies being studied without computing the actual fraction of bioavailable zinc. Because the MPA equation is log-based and applies an assumed background concentration to all sites, the model always returns positive values. Unless the maximum permissible concentration (MPA plus background concentration) for a given water body is less than the calculated MPA, the addedrisk approach allows additional zinc loadings to a system, however small the amount. For example, an Italian sample with a modest ambient zinc concentration of 4.7 μg/L was flagged by the bioavailability index (1.2) and BLM-based MPA (-0.79 μg/L) as exceeding the toxicity threshold; however, the added-risk approach suggested that an additional 20.2 μg/L zinc could be added to the site, even when the ambient zinc level was assumed to be 100% bioavailable. This outcome underscores the importance of considering sitespecific water quality on zinc bioavailability regardless of the assumptions concerning the magnitudes of dissolved-zinc concentrations in surface waters.
A comparison of the BLM-based and the added-risk MPA approaches is shown in Figure 4. The BLM estimates toxicity based on site-specific water quality and ambient dissolved-zinc concentrations (Figure 4A). In contrast, the added-risk approach is based on the concentration estimated to affect 5% of the species and assumptions about the fraction of background zinc that is bioavailable: none, half, or completely available (Figure 4B). Although a variety of bioavailable fractions can be assumed for a range of circumstances, sitespecific water quality influences are not considered in the added-risk approach. More important, conditions suggesting toxicity, such as the 145 sites exceeding toxicity thresholds (Figure 3), may be overlooked if the assumption about the bioavailable fraction is too low. However, in situations where the ambient zinc concentration is equal to, or greater than, the maximum permissible concentration (i.e., the water quality standard), an exceedance would be realized and the added-risk approach would not be used. An example of this is illustrated in Figure 4B, where the added-risk-MPA curves inflect at ambient zinc concentrations approximating the median effects level for the chronic ecotoxicity species sensitivity distribution (i.e., > 100-200 μg/L depending on the assumed bioavailable fraction). Regardless of magnitude, the BLM can provide decision makers with a tool for accounting for ambient zinc concentrations in risk assessments.
In the past, regulatory decision makers have had to rely on inexact measures to protect aquatic life from anthropogenic pollutants. Originally, these measures were total concentrations of metals and then, as our understanding of bioavailability became more refined, dissolved metals. Now, for metals like dissolved copper and zinc, we have a measure—bioavailable metal defined using the BLM—that not only provides the most accurate in situ measure available but also can be used to determine how much metal in ambient waters is bioavailable and how much can be allowed to enter water bodies and still protect aquatic life. This article has attempted to illustrate how this new measure can be applied using data from a variety of freshwater locations in Europe, North America, and South America.
Existing approaches to account for the influence of ambient aqueous concentrations of metals for risk assessment have attempted to estimate preindustrial background concentrations without regard to site-specific water quality actual ambient metal concentration at a site. Because significant differences in ambient zinc concentrations were observed for diverse water bodies from North America, South America, and Europe, the BLM was used to determine that an average site-specific maximum permissible addition of 57 μg/L could be realized for nearly 700 sites, owing to limited zinc bioavailability in more than 80% of the water bodies. In addition, approximately 150 sites were identified that had ambient zinc concentrations that exceeded chronic effects thresholds. The BLM projections could be used to target or improve local risk reduction efforts. The risks posed by ambient zinc concentrations to sensitive indicator species in surface waters can now be readily computed site-specifically using the BLM. This practice will facilitate accounting for the bioavailabilities of both background and anthropogenically derived zinc in regulating point and nonpoint sources of metal.
Funding for this study was provided by Rio Tinto and the International Lead Zinc Research Organization.