The Fate of Copper Added to Surface Water: Field, Laboratory, and Modeling Studies

Abstract The fate and effects of copper in the environment are governed by a complex set of environmental processes that include binding to inorganic and organic ligands in water, soil, and sediments. In natural waters, these interactions can limit copper bioavailability and result in copper transport from the water column to the sediment. In the present study, data on the fate of copper added to lakes, microcosms, and mesocosms were compiled and analyzed to determine copper removal rates from the water column. Studies on copper behavior in sediment were also reviewed to assess the potential for remobilization. A previously developed, screening‐level fate and transport model (tableau input coupled kinetic equilibrium transport–unit world model [TICKET–UWM]) was parameterized and applied to quantify copper removal rates and remobilization in a standardized lake setting. Field and modeling results were reconciled within a framework that links copper removal rates to lake depths and solids fluxes. The results of these analyses provide converging evidence that, on a large scale, copper is removed relatively quickly from natural waters. For the majority of studies examined, more than 70% of the added copper was removed from the water column within 16 d of dosing. This information may be useful in the context of environmental hazard and risk assessment of copper. Environ Toxicol Chem 2019;38:1386‒1399. © 2019 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals, Inc. on behalf of SETAC.

No copper remobilization from Lake Greifen; partial copper release from Lake Sempach likely enhanced by artificial oxygenation. Copper released redeposited with precipitated iron/manganese oxides. van den Berg et al. (1999) River Meuse in the Netherlands Degree of trace metal (including copper) remobilization less than 2% amount deposited. Hunt and Smith (1983) Seawater microcosms using sediments from Narragansett Bay, Rhode Island, USA Observed a release of copper and other metal from contaminated sediments but cautioned that extrapolation to natural systems would require consideration of additional factors. Zwolsman et al. (1997) Scheldt estuary in the Netherlands and Belgium.
Dissolved copper and other metals were mobilized in estuary through the deoxidation of sulfides in suspended matter upon encountering the higher oxygen concentrations associated with the lower estuary. Shaw et al. (1990) Sites of the coast of California, USA Metal transport to and from the sediment is complex; sample location showing Cr, Co, Ni, and Cu release had concentrations lower than 2 µg/L Klinkhammer et al. (1982) two deep ocean sites in the central equatorial Pacific Copper flux from sediment to overlying water was 5-10 greater than flux to sediment; however, copper concentrations near the sediment interface were low (0.03 -0.45 µg/L) 27 28 29 30 31 32 33 SI-7 All concentration in units of μg/L 118 A summary of TICKET-UWM parameters used for the EUSES Model Lake calculations is shown in Table  119 SI-3. The chemical composition associated with the three water chemistries in Table SI-3 is generally 120 consistent with summaries prepared from EU monitoring databases (ARCHE, 2013). 121 Data from EUSES model lake (EC, 2004;European Chemicals Agency, 2010). Surface area and volume values quoted refer to a regional scale assessment. They do not influence the removal simulations.
b Calculated using the settling velocity, suspended solids concentration, sediment bulk solid concentration, and the burial (net sedimentation) rate shown in the table using a steady-state solids balance (Chapra, 1997). c EUSES pore water side mass transfer coefficient. Mass transfer resistance is assumed to be all on the sediment side (Di Toro et al., 1981). d 10 th percentile AVS concentration for a Flemish dataset (Vangheluwe et al., 2005) e Water chemistries for the three pH values were taken from Annex 10 of the Globally Harmonized System of Classification and Labelling of Chemicals (United Nations, 2017).

122
The resuspension rate (Table SI-3) was calculated from a solid mass balance in the active sediment layer 123 assuming steady-state conditions (no accumulation/depletion of sediment solids). The following equation 124 where vs is the settling velocity (2.5 m/d), m is the water column suspended solids concentration (15 mg/L), 127 ma is the sediment solids concentration (0.5 kg/L), and vb is the burial rate (0.3 cm/yr). To be consistent 128 with laboratory/mesocosm/field removal tests, the diluting effect of flow into the system was minimized by 129 increasing the hydraulic residence time of the model lake system from the EUSES value of 40 days (0.11 130 year) to 300 years. 131 It should be noted that the EUSES Model Lake is prescribed to have a sediment compartment that is 10% 132 aerobic. At present, TICKET-UWM simulations utilize a single sediment layer and cannot directly 133 reproduce this condition. The redox state (oxic or anoxic) of the sediment layer in the TICKET-UWM is 134 specified by the user. This selection determines the sorption/precipitation reactions included in the 135 simulation. For the EUSES Model Lake calculations, an anoxic sediment layer was used with the AVS 136 SI-9 concentration set to the 10 th percentile value from EU monitoring data (Vangheluwe et al., 2005)    These ranges were calculated using data prior to the time at which 70% removal was achieved.   These ranges were calculated using data prior to the time at which 70% removal was achieved.

Detailed Sediment Simulations 178
The water column in the TICKET-UWM is represented as being oxic with a negligible sulfide 179 concentration. In accordance with the equilibrium mass action law, any copper sulfide solid resuspended 180 from the sediment layer to the water column immediately dissolves, releasing copper to re-equilibrate 181 between the settling particles the dissolved phase in the water column. Since the immediate redistribution 182 supplies dissolved copper in the water column, the model calculations are closer to a worst-case scenario 183 for dissolved copper removal. 184 For the detailed sediment simulations, the EUSES Model Lake water chemistry (Table SI-3) was used for 185 the surface water. Model simulations used bulk and pore water sediment chemistry from several field 186 studies (Table SI-8). The base case sediment pH was 7.56. Copper log KD values in sediment were specified 187 using the Calculated KD approach (described above). Simulations were performed for an oxic sediment and 188 an anoxic sediment to assess the two redox end members occurring in natural sediment. For the oxic 189 sediment simulations, sulfide production and copper sulfide precipitation were not included. Copper was 190 allowed to sorb to POC, HFO, and HMO in the sediment and precipitate as carbonates and/or hydroxides. 191 For the anoxic sediment simulations, copper binding to HFO and HMO was not considered. Copper was 192 allowed to sorb to POC and precipitate as sulfides, carbonates, and/or hydroxides. 193 To model the formation and dissolution of copper sulfide solid, the following reaction and solubility product 194 from Dyrssen and Kremling (1990) were used initially: 195 Simpson et al. (2000) suggest Cu2S is an important copper solid phase in anoxic sediments. The solubility 197 of this species has been described with the following solubility product in (Dyrssen and Kremling, 1990): 198 Cu2S(s) + H + = 2Cu + + HS − log *Ks = -34.65 (SI-3) 199 Simulations were made with both copper-sulfide solid species. 200 Water column results for the anoxic simulation indicate that around day 24 of the simulation, the total and 203 dissolved copper concentrations leveled off. Similar behavior was noted for the water column runs ( Figure  204 SI-2). For the remainder of the 365-day simulation, water column copper concentrations decreased only 205 very slightly. Between day 24 and day 365, the ranges of water column copper concentrations were 140 to 206 160 ng/L (total) and 56 to 62 ng/L (dissolved) respectively. These dissolved values are greater than 150 207 times lower than the 70% removal concentration of 10.5 μg/L. The relatively flat copper response between 208 day 24 and 365 in the water column was the result of a local equilibrium established between the settling 209 flux of copper (directed into the sediment) and the resuspension flux (directed out of the sediment). The 210 local equilibrium can be referred to as a pseudo steady-state. The smaller burial flux slowly depleted copper 211 from the system moving the water column / sediment system toward the final steady-state in which the 212 water column and sediment copper concentrations were zero. Detailed output from anoxic and oxic 213 simulations is shown in  (0)) to the maximum total concentration during the quasi-steady-state period (Max QSS CT). This is meant to give an indication of where sustained water column concentrations lie relative to the 70% removal benchmark.

215
A series of sensitivity analyses were conducted to assess effect of (i) low AVS (1 μmol/g versus the base 216 case value of 9.1 μmol/g), (ii) variation in water column/sediment pH (6/7, 7/7, and 8/7.5 versus the base 217 case combination of 7.07/7.56), (iii) low sediment solids concentration (150 g/Lbulk versus the base case 218 value of 500 g/Lbulk), (iv) variation in hardness (factor of 2 about the base value of 516 mg/L as CaCO3), 219 (v) variation in resuspension rate (0.1, 1, 3.2, and 10 times the default rate of 2.44 cm/yr), and (vi) variation 220 in copper loading (initial copper concentrations of 10, 100, and 1000 µg/L). 221 Model results from the anoxic simulation with AVS of 1 μmol/g were identical to those with an AVS of 222 9.1 (Table SI-9). This is because AVS was present in excess of the total sediment copper concentration. 223

SI-20
Simulations associated with items (ii) -(iv) use an anoxic simulation with Cu2S(s) and total initial copper 224 of 100 µg/L. Detailed results can be found in Table SI-10. For the relative pH variation sensitivity analysis, 225 the largest change in time required to achieve 70% removal of total copper was for the 8/7.5 simulation; 226 the predicted increase in removal time was from 3.5 days to 3.9 days. Variation of pH water column / 227 sediment pH values had a relatively minor impact on the magnitude of the pseudo steady-state water column 228 copper concentrations (Table SI-10). Total copper concentrations at the start of pseudo steady state were 229 within 10% of the base case. Dissolved concentrations from the pH 8/7.5 simulation were slightly higher 230 than the base case (i.e., 7.07/7.56). However, the total copper concentrations at the start of pseudo state-231 state conditions were at least 60 times smaller than the concentration representing 70% removal. The mass 232 balance results were similar to the base case. In all pH variation runs, particulate copper speciation in the 233 sediment was dominated by the formation of Cu2S(s). Removal time, pseudo steady-state water column 234 copper concentrations, and mass balance results showed no sensitivity to a factor of 2 variation in sediment 235 hardness. Though the simulations with the decreased sediment solid concentration showed some departures 236 from the base case values for pseudo steady-state water column copper concentrations as well as settling 237 and resuspension fluxes, key outcomes remain similar to the base case. The 70% removal time was still 238 around 3.5 days, the total copper concentrations at the start of pseudo state-state conditions was still 239 significantly smaller than the concentration representing 70% removal, the integrated diffusive flux was 240 directed into the sediment and particulate copper speciation in the sediment was dominated by the formation 241 of Cu2S(s). 242 Simulations associated with item (v) used an anoxic simulation with AVS = 9.1 µmol/g, CuS(s), and total 243 initial copper of 35 µg/L. Model runs with the sediment resuspension rate set at 0.1, 1, 3.2, and 10 times 244 the default rate (item v) were made to examine the impact of resuspension on water column copper 245 concentrations in more detail. The resuspension rate was multiplied by factors of 0.1, 3.2, and 10 keeping 246 the settling velocity and burial rate constant. The total and dissolved copper concentration in the water 247 column at day 15 increased as the resuspension rate increased ( Figure SI-3). However, even at the highest 248 SI-21 resuspension rate (i.e., 10 times the default value), total and dissolved copper in the water remained more 249 than 6 times lower than the concentration representing 70% removal of 10.5 μg/L. 250 Simulations associated with items (vi) used an anoxic simulation with AVS = 9.1 µmol/g and CuS(s). The 252 results indicate that total dissolved pseudo steady-state copper concentration and mass flux/balance values 253 (except the diffusive flux) varied linearly with the initial copper concentration (Table SI-11). For the 254 sensitivity analysis simulation with initial copper concentration at 1,000 µg/L and an AVS of 1 µmol/g, the 255 capacity of the sediment to bind copper as CuS(s) was exceeded. Sediment copper in excess of AVS was 256 bound by POC. This results in a decreased sediment KD, increased pore water copper concentration and an 257 integrated diffusive flux directed out of the sediment. However, this change in sediment speciation does 258 not impact the amount of time required for 70% removal in the water column nor does it influence the 259 pseudo steady-state concentration to an appreciable extent (compare last two columns in Table SI- copper loading sensitivity analysis simulations show 70% removal times that are less than 6 days and 261 pseudo steady state water column total copper concentrations that are more than 50 times lower than that 262 representing 70% removal. 263 Empirical KD sediment simulations indicate total and dissolved copper concentrations at the onset of pseudo 264 steady-state conditions of 0.31 and 0.21 µg/L, respectively, which are more than 30 times lower than the 265 70% removal concentration of 10.5 μg/L. Although the water column log KD was greater than the sediment 266 log KD, the integrated diffusive flux was directed into the sediment. 267

SI-23
269 a Base-case simulation parameters: water column pH 7.07; sediment pH 7.56; anoxic sediment with AVS = 1 μmol/g, settling velocity 2.5 m/d; initial Cu concentration = 0.1 mg/L; Cu2S is the potential copper sulfide precipitate b Ranges and average are based on data from the quasi-steady state period of the simulation. c This number is the diffusive flux integrated over the entire 365-day simulation. Negative diffusive flux values are directed out of the sediment and positive diffusive flux values are directed into the sediment. d This quantity is the ratio of the total Cu concentration representing 70% removal (0.3×CT(0)) to the maximum total concentration during the quasi-steady-state period (Max QSS CT). This is meant to give an indication of where sustained water column concentrations lie relative to the 70% removal benchmark.  This quantity is the ratio of the total Cu concentration representing 70% removal (0.3×CT(0)) to the maximum total concentration during the quasi-steady-state period (Max QSS CT). This is meant to give an indication of where sustained water column concentrations lie relative to the 70% removal benchmark.

Parameter Selection 273
Where possible, physical and chemical parameters serving as input for the TICKET-UWM were specified 274 based on direct measurements provided in (van Hullebusch et al., 2003a;van Hullebusch et al., 2003b) 275 (Table SI-12). Since settling rate, burial rate, and sediment solids concentration were not measured for 276 the reservoir, these parameters were set to regional values from the EUSES model lake (EC, 2004;277 European Chemicals Agency, 2010) ( Table SI- (van Hullebusch et al., 2003a;van Hullebusch et al., 2003b) b Data from EUSES model lake (EC, 2004;European Chemicals Agency, 2010). Surface area and volume values quoted refer to a regional scale assessment. They do not influence the removal simulations. c Calculated using the settling velocity, suspended solids concentration, sediment bulk solid concentration, and the burial (net sedimentation) rate shown in the table using a steady-state solids balance (Chapra, 1997). d EUSES pore water side mass transfer coefficient. Mass transfer resistance is assumed to be all on the sediment side (Di Toro et al., 1981). e 50 th percentile AVS concentration for a Flemish dataset (Vangheluwe et al., 2005) f This was calculated by adding the theoretical dose value to the measured pre-dose background copper concentration of 1.28 μg/L.

Distribution Coefficients -Empirical KD 295
According to measured total and dissolved copper and suspended particulate matter (SPM) concentrations 296 in the reservoir, log KD values spanned the range 3.66 to 4.87 (i.e., KD = 10 3.66 to 10 4.87 L/kg). The Empirical 297 KD simulation used the average log KD value of 4.56. Based on the average SPM concentration over the 298 study period of 18.0 mg/L, the average fraction particulate was 0.39. The distribution coefficient in the 299 sediment of the reservoir was not measured. Therefore, for the Empirical KD simulation the sediment log 300 KD was set at 4.39 based upon the copper partition coefficient review of Heijerick and Van Sprang (2005). 2003b); POM is 100% active with 50% HA and 50% FA (Lofts and Tipping, 2000); DOM is 63.5% 310 active with 0% HA and 100% FA (Bryan et al., 2002;Lofts and Tipping, 2011). 311 Concentrations of HFO and HMO in the water column were estimated from measured total sediment 312 particulate iron and manganese concentrations (Saint Germain les Belles Reservoir), and the water column 313 suspended solids concentrations for the reservoir (Table SI-12). The following specifications were used to 314 determine the amount of HFO and HMO from measured total iron and manganese, respectively. 315 • Approximately 40% and 18% of the total particulate iron and manganese in the water column were 316 specified to be HFO and HMO, respectively (HydroQual and Manhattan College, 2010) 317 • HFO-iron was assigned a formula weight of 89 g HFO/mol Fe (Dzombak and Morel, 1990) SI-29 • HMO-manganese Mn was assigned a formula weight of 119 g HMO/mol Mn (Tonkin et al., 2004). 319 The AVS of the sediment in the reservoir was not measured, so it was set at the 50 th percentile AVS value 320 of 8.8 µmol/g dry weight from Vangheluwe et al. (2005). 321 322 SI-30  (Schäfers, 2001). Water and sediment were taken from a reference site within the Euro-Ecole project. The study included all important community elements (phyto and zooplankton, macrophytes, benthic invertebrates), which were observed for an exposure period of about 3 months. Copper was tested at permanent nominal concentrations of 5, 10, 20, 40, 80, and 160 μg/L, which were maintained by three treatments weekly.

REMOVAL OF COPPER FROM THE WATER COLUMN
• The averaged added copper concentrations measured one day after treatment were in the range of 70 -90 % of the respective nominal concentrations. Within the two or three days between treatments, the dissipation of total copper from the water column was about 25%.

SI-31
• The exposure period covering spring/early summer up to autumn conditions and the oligotrophic conditions as well as the low hardness and DOC values of the used water can be regarded as a worse case scenario for assessing biological effects.
• Actual primary production of the systems, dominated by macrophytes and indicated by water parameters pH and oxygen content, showed distinctly less increase in spring/early summer at nominal copper concentrations higher than 10 μg/L. At the end of the study, biomass of macrophytes was significantly lower than controls at 40 μg/L (NOEC = 20 μg/L).
• Phytoplankton in total showed higher densities compared to controls at treatments of 20 μg/L and higher (NOEC = 10 μg/L), mainly caused by picoplankton and green algae. Some taxa (Diatomea and Cryptophyta) temporarily showed slightly lower abundances at the same concentrations.
• Daphnia longispina and Phyllopoda in general were identified as the most sensitive zooplankton taxa with a NOEC of 20 μg/L. Copepoda and Rotatoria seemed not to be affected or only at the highest treatment level.
• No effects could be detected for benthic macroinvertebrates neither in sediment toxicity tests with microcosm sediments after study termination.
• In total, added copper concentration up to 20 μg/L showed no relevant direct effects on structural endpoints and only slight and temporary effects on functional endpoints represented by oxygen concentrations and pH values. The No Observed Ecologically Adverse Effect Concentration (NOEAEC) of the study can be derived with 20 μg/L, and can be regarded as representative of plankton-dominated waters.

Dissolved Copper Measurements
Dissolved copper measurements were made using anodic stripping voltammetry (ASV) with a Metrohm hanging mercury drop electrode interfaced with a Metrohm 757 Computrace, with a Metrohm Dosimat attached in order to provide accurate automated additions of copper standard. Table SI-14 below contains the dissolved copper results used in present study.

Total Copper Measurements
Total copper measurements were made following the international standard method ISO 11885 / DEV E22.
A Liberty II ICP-OES from Varian (Eschborn) was used.