Ecotoxicology and Environmental Safety 114 (2015) 143–149

Contents lists available at ScienceDirect

Ecotoxicology and Environmental Safety

http://d0147-65

n CorrE-m

journal homepage: www.elsevier.com/locate/ecoenv

Evaluation of current copper bioavailability tools for soft freshwaters inSweden

S. Hoppe a,n, J.-P. Gustafsson b,c, H. Borg a, M. Breitholtz a

a Department of Applied Environmental Science (ITM), Stockholm University, Stockholm SE-106 91, Swedenb Division of Land and Water Resources Engineering, KTH (Royal Institute of Technology), Stockholm SE-100 44, Swedenc Department of Soil and Environment, Swedish University of Agricultural Sciences, Box 7014, Uppsala SE-750 07, Sweden

a r t i c l e i n f o

Article history:Received 8 September 2014Received in revised form22 January 2015Accepted 22 January 2015

Keywords:Bioavailability toolsBLMSwedenSoft freshwaterCopper

x.doi.org/10.1016/j.ecoenv.2015.01.02313/& 2015 Elsevier Inc. All rights reserved.

esponding author.ail address: Sabina.hoppe@itm.su.se (S. Hoppe

a b s t r a c t

The Water Framework Directive (WFD) in Europe calls for an improved aquatic ecological status. Bioticligand models (BLM) have been suggested as a possible tool assisting in the regulatory process. The aimof this study was therefore to investigate the applicability of BLM under the WFD to set environmentalquality standards (EQS), in particular regarding copper in Swedish freshwaters of which many are softerthan those used for model calibration. Three different BLMs, one acute and two chronic, were applied towater chemistry data from 926 lakes and 51 rivers (1530 data entries) and evaluated with respect to theircalibration range for input parameters. In addition, the predicted no-effect concentration (PNEC) forcopper was calculated. From the 1530 data entries, 750 ended up outside of the BLM calibration range,when looking at the chemical parameters Ca2þ , alkalinity, pH and DOC, primarily due to low carbonatealkalinity. Furthermore, the calculated Cu PNECs were higher than the suggested Swedish limit for Cu(4 mg L�1) in surface waters for 98% and 99% of the cases concerning lakes and rivers, respectively. Toconclude, our findings show that water chemical characteristics outside of the calibration ranges arequite common in Sweden and that the investigated models differ in how they calculate toxicity con-cerning Cu under these conditions. As a consequence, additional work is required to validate the BLMs byuse of bioassays with representative species of soft waters. Such results will show if these models can beused outside of their calibration ranges and also which of the models that gives the most reliable results.

& 2015 Elsevier Inc. All rights reserved.

1. Introduction

The European Water Framework Directive (WFD) is a step to-wards reaching an improved ecological standard as well as a cor-rective action plan for ecological sustainability to protect aquaticenvironments in Europe (2000/60/EC). The finished Directive (EC,2006) requires all waters in Europe to reach “good status” by theyear 2015. This is to be achieved through massive ecologicalmonitoring and by devising water quality regulations for aquaticpollutants. Concerning the quality regulations, the WFD states that“Community policy on the environment (…) should be based onthe precautionary principle” (Section 11), “Decisions should betaken as close as possible to the locations where water is affectedor used” (Section12) and “Community is to take account to (…)environmental conditions in the various regions of the Commu-nity” (Section13). Based on these paragraphs, every member statehas a substantial influence on how to monitor their waters under

).

the WFD. Annex VIII of the WFD further stipulates “metals andtheir compounds” as key pollutants found in aquatic environmentstoday.

Both in Europe and in the USA, the hardness of water is com-monly considered in environmental risk assessments of metals(Paquin et al., 2002), as major cations protect aquatic organismsagainst free metal ions (Mnþ) by replacing them on biotic bindingsites such as fish gills. However, it has become increasingly evidentthat dissolved organic carbon (DOC) and pH also have a significantimpact on metal bioavailability and toxicity in the aquatic en-vironment (Bianchini and Bowles, 2002; Chapman, 2008; Di Toroet al., 2001; Rozan and Benoit, 1999). Dissolved organic matter(DOM) has a substantial affinity for binding Mnþ , thereby poten-tially reducing the toxicity. Furthermore, the properties and originof DOM have now been showed to influence its binding potential.High-molecular-weight dark aromatic allochthonous DOM canbind more metals than lighter autochthonous DOM (Al-Reasi et al.,2011; Baken et al., 2011; Chappaz and Curtis, 2013; Mueller et al.,2012; Tipping, 1998; Tipping et al., 2002; Tipping et al., 2011),influencing the metal speciation and toxicity. However, the

Fig. 1. Lake and river PNEC values calculated by the Full-BLM, distributed by geo-graphical location ( SMHI-coordinates) in Sweden from south to north compared tothe suggested Swedish freshwater limit value of 4 mg Cu L�1 (Swedish EPA, 2008).

S. Hoppe et al. / Ecotoxicology and Environmental Safety 114 (2015) 143–149144

protective effect of metal–DOM binding will be reduced if the pHvalue decreases, introducing competition on active DOM bindingsites between Mnþ and Hþ ions (Di Toro et al., 2001). In waterswith low pH, DOM–metal complexes will be weaker, releasingbioavailable Mnþ ions, affecting the metal toxicity to the organism.However, the addition of Hþ connected with a lower pH can alsoinhibit Mnþ accumulation at the biotic ligand, thereby preventingthe expected increased toxicity. In the future, a more adequateapproach for determining metal toxicity in freshwaters is needed.Biotic Ligand Models (BLMs) have been suggested as potentialassisting tools under the WFD. Most BLMs use the model WHAM(Tipping, 1994, 1998) to calculate metal binding to DOM. TheWHAM model is a discrete site/electrostatic model in which dif-ferent sites are recognized (e.g., monodentate, bidentate, tri-dentate), which have different affinities for binding cations. Thesemodels combine water chemical properties, taking DOC, pH andhardness into account, with biological organism responses basedon the metal able to accumulate on the organisms' receptor site,thereby generating reality-based site specific toxicity predictions(Paquin et al., 2002). BLMs have been used in voluntary metal riskassessments for evaluation and control of risks (Vijver et al., 2008),and guidance documents for Registration, Evaluation, Authoriza-tion and Restriction of Chemical (REACH substances) recommendBLM as a tool in setting generic exposure scenarios for metals(ECHA, 2008). However, for European regulation of emission limitsconcerning metals in wastewater from e.g. mining, the BLM hasnot yet been accepted as a regulatory tool. The current generationof BLMs has been calibrated largely using freshwaters with rela-tively high hardness levels compared to a large part of the onesfound in Scandinavia. This as a substantial number of Swedishfreshwaters are soft to ultra-soft, with low levels of Ca2þ com-bined with high levels of DOC, Al and Fe, causing numerous lakesto fall outside the BLM calibration ranges. Furthermore, there arestudies indicating that the BLM-predicted toxicity for both Cu andZn can be underestimated in waters with high levels of DOC, Aland Fe combined with low pH (De Schamphelaere et al., 2003a,2003b).

In Europe, a few countries (France, The Netherlands and UK)have used BLMs to derive environmental quality standards (EQS).In France, the effect of considering bioavailability in EQS has beenassessed by means of a “User-friendly BLM” with varying resultsdue to a large variability in the water chemical properties. In theUK, a modified version of the same User-friendly BLM, taking DOMbinding into account, has been proposed as a tool for regulatingagencies and in the Netherlands assessment of bioavailability isincorporated into policy documents (David et al., 2012).

With the overarching aim to investigate the applicability ofBLM for regulatory purposes in Sweden, particularly regardingcopper, the objectives of the present study were:

1)

To investigate to what extent the water chemistry of Swedishlakes and rivers falls within the calibration range of the Cu-BLMs available today.

2)

To compare results from three different Cu-BLMs in terms ofthe influence of DOC, Ca2þ and pH.

2. Materials and methods

2.1. Studied freshwaters

In the present study, Swedish water chemistry data from 926lakes and 51 rivers sampled in 2009 were used, resulting in 1530data entries. Lake water chemistry data originated from theSwedish national lake survey program and river chemistry from

the Swedish river monitoring program (http://webstar.vatten.slu.se/db.html), covering the main part of the country from north tosouth (Fig. 1).

Each of the rivers were sampled 10–12 times annually, result-ing in 604 different data entries, and the lakes were sampled onceduring the autumn turnover (September–December). Waterchemistry variables were analysed by the accredited laboratory ofDepartment of Aquatic Sciences and Assessment, Swedish Uni-versity of Agricultural Sciences (SLU). All analyses were performedfollowing the quality assurance routines specified in the accred-itation. As BLM requires alkalinity to be expressed as CaCO3, thiswas calculated by the authors from the mEq of alkalinity measuredin the waters. The samples for metal determination within theSwedish national monitoring program are normally not filtered.Several studies show that filtering through the often re-commended pore size 0.45 mm, allowing most of the colloidalmetal forms to pass, will have very little influence on the Cu-concentrations (Borg 1986; Lithner et al., 2003; Ingri et al., 2004;Köhler, 2010; Österlund et al., 2012). Al and Fe concentrations maybe more influenced by filtration but since these parameters are notused as input parameters in BLM this will not influence the cal-culated results in the current study.

S. Hoppe et al. / Ecotoxicology and Environmental Safety 114 (2015) 143–149 145

2.2. Models

Three different BLMs were used in this study; an acute Bioticligand model (2007) (http://www.hydroqual.com/wr_blm.html)referred to in the following as HQ-BLM, a chronic BLM using alarger dataset, used in the voluntary risk assessments (ECHA,2007), V.0.0.0.17 referred to as the Full BLM and a chronic User-friendly BLM (Bio-met, 2011; Peters et al., 2011). The HQ-BLM usesan acute to chronic ratio (3.2) to produce recommended values forchronic toxicity, so-called CCC (criteria continuous concentration)values, which are commonly used by the US EPA (EPA, 2011). TheFull BLM includes literature-based ecotoxicological data for a largevariety of species, whereas the User-friendly BLM is a down-scaledsimplified version of the Full BLM, requiring fewer chemical vari-ables (pH, Ca2þ and DOC). Both the Full- and the HQ-BLM modelsuse the WHAM model by Tipping (1994, 1998) for calculation ofthe free Cu-ion concentration. The BLM calculations of this studywere made without explicitly considering whether the waterchemistry was within the calibration range of the models. Thethree models estimate toxicity based on different principles. TheHQ-BLM uses CCC values as an endpoint, which is extrapolatedfrom acute values. The Full BLM calculates chronic endpoints fromthe metal concentration where 95% of the species included in itsdatabase is protected (HC5 curves). These HC5 curves are based onspecies sensitivity distributions (SSDs), containing the species in-cluded in the model's database (Wheeler et al., 2002). The User-friendly BLM produces EQS recommendations, which is equivalentto PNEC values. Comparing the water chemistry to the BLM cali-bration range, the models lowest limit was used (Table 1). Forexample for alkalinity, the lower limit of 10 mg CaCO3 was applied.

2.3. Data treatment

SPSS v.18 was used to investigate to what extent the data setwas suitable for the models calibration range. Calculated PNECresults from the three BLMs were compared using One-way AN-OVA (SPSS v.18) and post-hoc; Tukey or Dunnett's tests. To assessparameter sensitivity, 10 data points each for DOC, Ca2þ and pHwere separately inserted to the models together with default va-lues (DOC: 5 mg C L�1, Ca2þ: 20 mg L�1 and pH: 7) and the re-sulting PNECs were plotted. Only the parameter of interest wasvaried while the others were kept constant to capture the impactof change for that specific parameter. Parameters were varied by astepwise inclination of 2 mg DOC L�1 (between 0.5 and 36),0.5 units of pH (2–11.5) or 2 mg Ca2þ (0.5–300). Effect values(LC50: HQ-BLM and NOEC: Full BLM) for Daphnia pulex were cal-culated for 20 of the total 926 lakes and plotted against DOC. These20 lakes were analysed for metals 8 times per year. This was doneto compare the effect values between the models and as DOC isone of the most important parameters when calculating toxicity, itwas used as grouping variable.

Table 1Mean values with standard deviations for lake and river water chemistry as well as the

Lakes (n¼926)

Mean St.dev. Min/max

Ca2þ (mg L�1) 6.5 79.9 0.06/107DOC (mg C L�1) 12 77.5 0.54/69pH 6.6 70.7 4.0/8.3Al (mg L�1) 151 7166 6/2100Fe (mg L�1) 679 7729 10/5500Alkalinity (mg CaCO3 L�1) 19 730 0.27/412Hardness (mg CaCO3 L�1) 14 734 0.28/412

3. Results

3.1. BLM calibration limits

Comparing chemical data from the freshwaters included in thisstudy to the most conservative BLM calibration window showedthat out 750 out of 1530 freshwaters, fell outside the intendedcalibration range of the BLM software (Table 2; Fig. 2) for theparameters Ca2þ , pH, DOC and alkalinity.

The chemical parameters that were most frequently outside ofthe BLM calibration range were alkalinity and Ca2þ (69%) and pH(46%). Iron (Fe) and aluminum (Al) were not included as inputparameters in the models, but there is a recommendation by thedevelopers of the software not to use the models for concentra-tions that exceed 300 mg L�1 Al and Fe, respectively (S1). In Ta-ble 2, the results have been broken down into three differentgeographical regions of Sweden, the north, central (“mid”) andsouthern parts. The region with the largest percentage of lakesdeviating from the BLM calibration range depended on the para-meter under study. In the central and southern parts of thecountry, the percentage of lakes outside of the pH calibrationrange exceeded 40%. For the alkalinity, which together with pHand TOC, are the most important parameters for calculating toxi-city with BLMs, the geographical area with the largest divergencewas northern Sweden, where 69% of the lakes fell under the ca-libration range. For the rivers, the discrepancy was less pro-nounced and independent of geographical location. ComparingBLM-calculated PNEC values to the current Swedish freshwaterEQS of 4 mg Cu L�1 (Swedish EPA, 2008) resulted in 99% of therivers and 98% of the lakes being above the limit.

3.2. Difference between the BLMs

The two chronic BLMs produced the highest PNEC values for50% of the river observations, the average difference betweenthem being 8–10 mg Cu L�1. However, when including the HQ-BLMthe difference was much greater, at the most 163 mg Cu L�1 be-tween calculated PNECs. The parameter sensitivity assessment forpH and DOC showed a similar positive PNEC correlation for pH andDOC between the models, although with different slopes, the HQ-BLM displaying a lower slope than the other two models. However,for Ca2þ the models gave completely different slopes (Fig. 3). Anegative correlation was observed in the Full BLM, a weak positiveresponse was observed in the HQ-BLM, whereas in the User-friendly model a scattered response was observed.

All the models showed a similar response to the variation of pHin the range 3–8 (Fig. 3). At higher levels, the PNEC values calcu-lated by HQ-BLM were positively correlated with pH, which wasnot the case for the other two models. When the acute HQ-BLMwas compared to the chronic models there was a statisticallysignificantly difference between the PNECs (po0.05). However,there was no difference between the two chronic models (p¼0.4-

BLM calibration interval related to when comparing Swedish water chemistry.

Rivers (n¼51) BLM calibration interval

Mean St.dev. Min/Max Narrowest window

15 725 1.3/144 3.1–1609.8 74.8 1.3/36 0.05–307.0 70.4 5.8/8.2 6.0–8.5

145 7129 4.3/2000 0.1–300697 7725 23/9000 0.1–30032 757 1.3/322 10–36041 765 4.3/386 –

Table 2The percentage of water chemical data from lakes and rivers falling outside theBLM calibration range in Sweden altogether and in three regions, divided by the xcoordinates; North: 4x: 7,000,000; Mid: x: 6,700,000–7,000,000; South: x:o6700,000. However since Al and Fe concentrations were not filtered the filteredconcentrations could decrease the percentage of freshwaters outside of the BLMcalibration range.

Ca2þ DOC pH Alkalinity

n % n % n % n %

Lakes (n: 927)Sweden (total) 394 43 155 17 326 35 407 44North (n: 321) 221 69 20 6 102 32 223 69Mid (n: 90) 50 56 15 17 41 46 57 63South (n: 518) 124 24 123 24 186 36 129 25

Rivers (n: 605)Sweden (total) 50 8 11 2 23 4 196 32North (n: 207) 53 26 0 0 20 10 131 63Mid (n: 69) 0 0 0 0 0 0 9 13South (n: 331) 0 0 23 7 3 1 60 18

S. Hoppe et al. / Ecotoxicology and Environmental Safety 114 (2015) 143–149146

0.9). The PNECs of the User-friendly BLM were higher (mean dif-ference 8.13 mg Cu L�1) in 57% of the cases compared to the FullBLM. For the lakes where the Full BLM gave higher PNECs themean difference was 4.8 mg Cu L�1. However, comparing all threemodels, the difference in PNECs could be as high as a 107 mg Cu L�1 between the models, resulting in a statistically significantdifference (ANOVA: po0.05). Applying a one-way ANOVA oncalculated PNECs for the lakes and rivers, there was a statisticaldifference (po0.001) comparing the results from the HQ-BLM tothe two chronic models. The calculated toxicity for D. pulexshowed that the models predict toxicity differently when con-sidering DOC as an influencing factor (Fig. 4).

4. Discussion

The present study demonstrates that a majority of the Swedishfreshwaters investigated falls outside of these BLMs calibrationranges. The most crucial of the water chemical parameters in-vestigated are Ca2þ , alkalinity and pH. Furthermore, almost all(98% and 99% of lakes and rivers, respectively) of the BLM

Fig. 2. Distribution of the water chemical parameters Ca2þ

generated PNEC values (which do not include any assessmentfactor) were higher than 4 mg Cu L�1, which is the current Swedishguideline for Cu in freshwater (Swedish EPA, 2008).

There are significant differences in terms of how the three in-vestigated BLMs calculate PNEC values for Swedish lakes and riv-ers, leading to a question which one is the most appropriate to usein risk assessments? This was further observed concerning theimpact from water chemical variables (pH, Ca2þ and DOC) on thepredicted toxicity, which also differed significantly (po0.05) be-tween the HQ and chronic software.

This difference between the three models in how they use thechemical parameters ability to influence bioavailability is probablywhy they provide different PNEC values while using the samewater chemistry input data (Fig. 3). Since the HQ BLM is an acutetoxicity model that uses a conversion factor of 3.2 to transformacute to chronic toxicity (ECHA, 2007) and the chronic toxicitymodels are programmed using chronic toxicity data, discrepanciesare to be expected, and in a general sense the chronic BLMs shouldbe considered more reliable. However, the difference between themodels was relatively large, and statistically significant, making italmost impossible to deduce which, of any, software that is the‘best’, without the use of bioassays to verify the model results,especially outside the validation range of the software.

4.1. Applicability of BLM for Swedish conditions

As the HQ- and Full-BLMs require a larger number of inputparameters compared to the User-friendly BLM, they may be moredifficult to apply as a regulatory tool in cases where water che-mical variables are limited. In such cases, the User-friendly modelcould prove an easier choice. However, the User-friendly and theHQ-BLMs assume that Ca2þ protects organisms against free andlabile metal fractions even when the Ca2þ concentrations are highenough to cause toxicity (Douglas and Horne, 1997). Opposite towhat would be expected, a negative slope was found betweenPNEC and Ca2þ , already below 50 mg Ca2þ L�1, using the Full-BLM. This pattern is indicative of the predicted Cu-toxicity in-creasing with increased water hardness, contradictory to what isgenerally experienced at levels below 50 mg Ca2þ L�1.

The calibration range of the tested BLM models only covered aminority of the lakes used in this study. A plausible reason for thisis that these available BLMs have been calibrated to include mostlycentral and southern European freshwater conditions whereas

, alkalinity and pH in the lakes included in this study.

Fig. 3. Variations in the PNEC values between the three Cu-BLMs investigated in this study, based on variations of the parameters Ca2þ , DOC and pH.

S. Hoppe et al. / Ecotoxicology and Environmental Safety 114 (2015) 143–149 147

there were not many waters included that represented the pre-vailing Scandinavian freshwater conditions (i.e. low levels of Ca2þ

and high levels of TOC, Al and Fe). This is particularly important forthe User-friendly BLM, as this model will provide the same EQS forall sites outside its calibration range, thereby providing uncertainPNEC values which cannot be relied upon. The HQ-BLM is vali-dated for waters below 300 mg Al and Fe L�1 but since theseparameters are not input variables, the model will not considerany variations below these concentrations. However, also lowerconcentrations of Al and Fe can compete for DOM binding sites,influencing Cu bioavailability (Al-Reasi et al., 2011; Baken et al.,2011; Chappaz and Curtis, 2013; EPA, 2011; Mueller et al., 2012;Tipping, 1998; Tipping et al., 2002; Tipping and Lofts, 2011) andthereby the toxicity, which leads to an uncertainty when trying touse the models in Al rich waters (De Laender et al., 2005). Highconcentrations of Al and Fe are often connected to a high per-centage of dark allochthonius carbon. The metal-binding proper-ties of the dissolved organic matter (DOM) can differ, where a dark

high molecular allochtonous DOM often are able to complex me-tals to a larger extent than a light autochthonous DOM.

A few studies have applied BLM to Swedish freshwaters (DeLaender et al., 2005; ECHA, 2007) but these studies focused onfreshwaters with water characteristics inside the model calibrationwindow. Hence, for freshwaters outside of the BLM calibrationrange it may be necessary to apply relatively large assessmentfactors for the conversion of PNECs to EQSs, at least until the BLMshave been validated using ecotoxicological studies. Such studiesshould preferably use soft water species representing severaltrophic levels. This may extend the calibration range of the currentBLMs, and the studies may also show which software is the mostappropriate to use.

4.2. Swedish Cu criteria

As the BLM-calculated PNEC values significantly exceeded theSwedish suggested criteria for freshwater of 4 mg Cu L�1, in 99% of

Fig. 4. Cu LC50 and NOEC for D. pulex plotted against DOC, calculated by HQ BLMand Full BLM, respectively. Numbers indicate water hardness (mg CaCO3 L�1). Dataoriginating from lakes in the Swedish national monitoring program.

S. Hoppe et al. / Ecotoxicology and Environmental Safety 114 (2015) 143–149148

the rivers and 98% of the lakes (Fig. 1), it is apparent that there is alarge deviation between this suggestion to EQS for Cu in Swedenand the BLM-predictions presented above. Since many of thefreshwaters fell outside of the BLM calibration range it is difficultto speculate about the reasons for this discrepancy. The criteria forCu presented by the Swedish EPA (2008), was developed usingthree different approaches. The first approach used an SSD curveobtained from the VRA (European Copper Institute, 2006), basedon 22 different species. Here, the NOEC varied between 8.5 and122 mg Cu L�1. The obtained HC5 from the SSD curve was 8.2 mg L�1. The Swedish EPA used an assessment factor of 2, because ofremaining uncertainties of the BLM-applicability in soft waters,resulting in an EQS of 4.1 mg L�1. The second approach uses a laterversion of the model from 2007 (European Copper Institute, 2007).Applying 7 typical scenarios for European waters, the lowest HC5was 7.8 mg L�1 (UK river). By using an assessment factor of 2, asabove, the EQS became 3.9 mg L�1. In the third calculation, a FullBLM version from 2007 was used to develop a worst case PNEC forSweden. Data from the national and regional monitoring programswere used (data from 9 years), and only sites within the modelcalibration window were included, yielding a lowest HC5 of 17.8 mg L�1 (the median 74.0 and the maximum 183 mg L�1). In this casean assessment factor of 4 was used since among other things, TOCdata was used instead of DOC, which resulted in a value of 4.45 mg L�1. Consequently the three different approaches resulted in avery similar Cu concentration, around 4 mg L�1, which was sug-gested as an EQS for Cu (Swedish EPA, 2008). In order to elucidatethe reasons for the deviating results, bioassays with representativespecies in these kinds of waters need to be conducted and com-pared with BLM predicted toxicity.

4.3. Conformity between the models

The Full BLM and the acute HQ-BLM gave a similar response tochanges in pH, up to approximately 8, as well as to DOC, indicatingthat the two models produce similar results under these condi-tions. As the Full BLM has partly been based on the HQ-BLM andsimilarities in toxicity calculations may therefore be expected.These since the both models mainly use the same chemicalparameters to calculate toxicity and as the HQ-BLM has a functionfor calculating chronic values. However, as shown in this study,

they still deviate in terms of their predictions of PNECs. The HQ-BLM almost consistently calculate significantly lower PNECs thanthe Full BLM (po0.05), especially during different DOC con-centrations (Fig. 3). Surprisingly, a comparison between the twomodels concerning LC50 values (HQ-BLM) and NOEC values (FullBLM) for D. pulex showed that the NOEC values from the Full BLMin fact were higher than the LC50 values calculated by the HQ-BLMin all the cases, deviating by a factor of more than 10 (Fig. 4). Sincethe NOEC level is supposed to represent the level of Cu in thewater where no effect will be observable in the Daphnia popula-tion and the LC50 value represents the lethal concentration for 50%of the population, the opposite relation between the two predic-tions would be expected. However, as the NOEC endpoint is hardto assess and is not considered a good way of measuring toxicresponse (Jager, 2012) an ECx endpoint would be easier toevaluate.

The purpose of the User-friendly BLM is to serve as a simplifiedtool compared to the Full BLM as it requires fewer chemicalparameters to calculate toxicity. As these two models are theore-tically exchangeable, their results were expected to agree reason-ably well, with some loss of accuracy due to fewer chemicalparameters and the lack of a speciation model in the User-friendlyBLM. However, this was not the case when the two models wereapplied to the Swedish lake and river data used in this study. TheUser-friendly model computed higher PNECs for a majority of thecases and the dependence on water chemical properties differed,especially regarding Ca2þ even within the calibrated range of themodel. The Ca2þ curve from the User-friendly BLM indicates that itcalculates the protective abilities from Ca2þ in intervals, giving thesame PNEC values for a span from 50 up till 150 mg Ca2þ L�1. Themodel then increases the recommended PNEC from 16 to 23 mg Cu L�1, thereby increasing its expected protective abilities andresulting in an irregular curve. The influence of Ca2þ on PNECcalculations by the two models was very different and as such theassumption that these two models are exchangeable may bequestionable.

4.4. Major conclusions

A data analysis of 1530 data entries from Swedish lakes andrivers showed that a large number of Swedish freshwaters falloutside of the calibration range of three current Cu-BLMs. Fur-thermore, it was shown that a vast majority of BLM calculatedPNECs were above Swedish EPA limits concerning Cu in surfacewaters. The calculated toxicity on D. pulex and the calculatedPNECs of the three investigated models gave different results,sometimes diverging with more than 160 mg Cu L�1 from eachother, indicating difficulties when deciding which model to use. Asa large portion of the freshwaters used in this study were found tobe outside the BLM calibration ranges, further validation studiesare needed before the current BLMs can be used without extraassessment factors. Furthermore, additional ecotoxicological stu-dies are suggested as a means to extend the calibration ranges ofthe BLMs or the development of models applicable in soft waterconditions.

Acknowledgements

The authors would like to thank The Swedish Research Councilfor Environment, Agricultural Sciences and Spatial Planning(FORMAS, Contract 2006-638) for funding this work. Furthermorewe would like to thank the three anonymous reviewers for theirvaluable comments.

S. Hoppe et al. / Ecotoxicology and Environmental Safety 114 (2015) 143–149 149

Appendix A. Supplementary material

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.ecoenv.2015.01.023.

References

Al-Reasi, H.A., Wood, C.M., Smith, D.S., 2011. Physicochemical and spectroscopicproperties of natural organic matter (NOM) from various sources and im-plications for ameliorative effects on metal toxicity to aquatic biota. Aquat.Toxicol. 103, 179–190.

Baken, S., Degryse, F., Verheyen, L., Merckx, R., Smolders, E., 2011. Metal com-plexation properties of freshwater dissolved organic matter are explained by itsaromaticity and by anthropogenic ligands. Environ. Sci. Technol. 45,2584–2590.

Bianchini, A., Bowles, K.C., 2002. Metal sulfides in oxygenated aquatic systems:implications for the biotic ligand model. Comp. Biochem. Physiol. 133, 51–64.

Bio-met, 2011. Bio-met Bioavailability Tool; User Guide (Version 1.4). ⟨http://bio-met.net/evidence-base/⟩.

Hydroqual, 2007. Biotic Ligand Model, Version 2.2.3. ⟨http://www.hydroqual.com/wr_blm.html⟩.

Borg, H., 1986. Metallers förekomstformer i sjövatten i Rönnskärsområdet. SwedishEPA Report, p. 3124.

Chapman, P.M., 2008. Environmental risks of inorganic metals and metalloids: Acontinuing, evolving scientific odyssey. Hum. Ecol. Risk Assessment. 14, 5–40.

Chappaz, A., Curtis, J., 2013. Integrating empirically dissolved organic matter qualityfor WHAM VI using the DOM optical properties: a case study of Cu�Al�DOMinteractions. Environ. Sci. Technol. 47, 2001–2007.

David, M., Batty, J., Neibeek, G., Van Assche, F., Schlekat, C., Perceval, O., RodriguezRomero, J., Delbeke, K., 2012. Workshop on Metal Bioavaliability Under theWater Framework Directive: Policy. Science and Implementation of RegulatoryTools, Brussel.

De Laender, F., De Schamphelaere, K.A.C., Verdonck, F.A.M., Heijerick, D.G., VanSprang, P.A., Vanrolleghem, P.A., Janssen, C.R., 2005. Simulation of spatial andtemporal variability of chronic copper toxicity to Daphnia magna and Pseudo-kirchneriella subcapitata in Swedish and British surface waters. Hum. Ecol. RiskAssess. 11, 1177–1191.

De Schamphelaere, K.A.C., Heijerick, D.G., Janssen, C.R., 2003a. Refinement and fieldvalidation of a biotic ligand model predicting acute copper toxicity to Daphniamagna. Comp. Biochem. Physiol. C—Toxicol. Pharmacol. 134, 243–258.

De Schamphelaere, K.A.C., Vasconcelos, F.M., Heijerick, D.G., Tack, F.M.G., Delbeke,K., Allen, H.E., Janssen, C.R., 2003b. Development and field validation of apredictive copper toxicity model for the green alga Pseudokirchneriella sub-capitata. Environ. Toxicol. Chem. 22, 2454–2465.

Douglas, W.S., Horne, M., 1997. The interactive effects of essential ions and salinityon the survival of Mysidopsis bahia in 96-h acute tests of effluents discharge tomarine and estuarine receiving waters. Environ. Toxicol. Chem. 16, 1996–2001.

Di Toro, D.M., Allen, H.E., Bergman, H.L., Meyer, J.S., Paquin, P.R., Santore, R.C., 2001.Biotic ligand model of the acute toxicity of metals; 1 technical basis. Environ.Toxicol. Chem. 20, 2383–2396.

EC, 2006. Proposal for a Directive of the European Directive Parliament and of theCouncil on Environmental Quality Standards in the Field of Water Policy andAmending Directive 2000/60/EC. European Commission, Brussels. ⟨http://ec.europa.eu/environment/water/water-dangersub/pdf/com_2006_397_en.pdf⟩.

ECHA, 2007. Voluntary Risk Assassment of Copper, Copper II Sulphate Pentahy-drate, Copper(I)oxide, Copper(II)oxide, Dicopper Chloride Trihydroxide; Eur-opean Union Risk Assessment Report. European Copper Institute. ⟨http://echa.europa.eu/copper-voluntary-risk-assessment-reports⟩.

ECHA, 2008. Guidance on Information Requirements and Chemical Safety Assess-ment. Environmental Risk Assessment for Metals and Metal Compounds (Ap-pendix R.7.13-2). European Chemicals Agency, Brussels. ⟨http://echa.europa.eu/

documents/10162/13632/information_requirements_r7_13_2_en.pdf⟩.European Copper Institute, 2006. Van Sprang, P., Vangheluwe, A., Van Hyfte, A.,

Heijerick, D., Vandenbroele, M. and Verdonch, F. European Union Risk Assess-ment Report on Copper, Copper II Sulphate Pentahydrate, Copper(I)oxide,Copper(II)oxide, Dicopper Chloride Trihydroxide, Voluntary Risk Assessment,Environmental Effects, Chapter 3.2 (Part 1), Draft, February 2006.

European Copper Institute, 2007. Van Sprang, P., Vangheluwe, A., Van Hyfte, A.,Heijerick, D., Vandenbroele, M. and Verdonch, F. European Union Risk Assess-ment Report on Copper, Copper II Sulphate Pentahydrate, Copper(I)oxide,Copper(II)oxide, Dicopper Chloride Trihydroxide, Voluntary Risk Assessment,Environmental Effects, Chapter 3.2 (Part 1), Draft, July 2007.

EPA, 2011. ⟨http://water.epa.gov/learn/training/standardsacademy/upload/module_aquaticlife.pdf⟩.

Ingri, J., Nordling, S., Larsson, J., Rönnegård, J., Nilsson, N., Rodushkin, I., Dahlqvistc,R., Andersson, P., Gustafsson, Ö, 2004. Size distribution of colloidal trace metalsand organic carbon during a coastal bloom in the Baltic Sea. Mar. Chem. 91,117–130.

Jager, T., 2012. Bad habits die hard: the NOEC's persistence reflects poorly onecotoxicology. Environ. Toxicol. Chem. 31, 228–229.

Köhler, S.J., 2010. Comparing Filtered and Unfiltered Metal Concentrations in SomeSwedish Surface Waters. Rapport 2010:04. Institutionen för vatten och miljö,SLU.

Lithner, G., Holm, K., Ekström, C., 2003. Metaller och organiska miljögifter i vat-tenlevande organismer och deras miljö i Stockholm 2001. ITM Rapport 108,ISBN 91-631-3758-5.

Mueller, K.K., Lofts, S., Fortin, C., Campbell, P.G.C., 2012. Trace metal speciationpredictions in natural aquatic systems: incorporation of dissolved organicmatter (DOM) spectroscopic quality. Environ. Chem. 9, 356–368.

Paquin, P.R., Gorsuch, J.W., Apte, S., Batley, G.E., Bowles, K.C., Campbell, P.G.C., Delos,C.G., Di Toro, D.M., Dwyer, R.L., Galvez, F., Gensemer, R.W., Goss, G.G., Hog-strand, C., Janssen, C.R., McGeer, J.C., Naddy, R.B., Playle, R.C., Santore, R.C.,Schneider, U., Stubblefield, W.A., Wood, C.M., Wu, K.B., 2002. The biotic ligandmodel: a historical overview. Comp. Biochem. Physiol. C—Toxicol. Pharmacol.133, 3–35.

Peters, A., Merrington, G., de Schamphelaere, K.A.C., Delbeke, K., 2011. Regulatoryconsideration of bioavailability for metals: simplification of input parametersfor the chronic copper biotic ligand model. Integr. Environ. Assess. Manag. 7,437–444.

Rozan, T.F., Benoit, G., 1999. Intercomparison of DPASV and ISE for the measure-ment of Cu complexation characteristics of NOM in freshwater. Environ. Sci.Technol. 33, 1766–1770.

SPSS. SPSS v.18.0, Chicago, IL, USA.Swedish EPA, 2008. Naturvårdsverket. Förslag till gränsvärden för särskilda för-

orenande ämnen; Rapport 5799.Tipping, E., 1994. WHAM-A chemical-equilibrium model and computer code for

waters, sediments, and soils incorporating a discrete site electrostatic model ofion-binding by humic substances. Comput. Geosci. 20, 973–1023.

Tipping, E., 1998. Humic ion binding model VI: an improved description of theinteractions of protons and metal ions with humic substances. Aquat. Geochem.4, 3–48.

Tipping, E., Rey-Castro, C., Bryan, S.E., Hamilton-Taylor, J., 2002. Al(III) and Fe(III)binding by humic substances in freshwaters, and implications for trace metalspeciation. Geochim. Cosmochim. Acta 66, 3211–3224.

Tipping, E., Lofts, S., Sonke, J.E., 2011. Humic ion-binding model VII: a revisedparameterisation of cation-binding by humic substances. Environ. Chem. 8,225–235.

Vijver, M.G., De Koning, A., Peijnenburg, W.J.G.M., 2008. Uncertainty of watertypespecific hazardous copper concentrations derived with biotic ligandmodels. Environ. Toxicol. Chem. 27, 2311–2319.

Wheeler, J.R., Grist, E.P.M., Leung, K.M.Y., Morritt, D., Crane, M., 2002. Speciessensitivity distributions: data and model choice. Mar. Poll. Bull. 45, 192–202.

Österlund, H., Gelting, J., Nordblad, J., Baxter, D.C., Ingri, J., 2012. Copper and nickelin ultrafiltered brackish water: labile or non-labile? Mar. Chem. 132–133,34–43.