Reduction of industrial iron pollution promotes phosphorus internalloading in eutrophic Hamilton Harbour, Lake Ontario, Canada☆

Stefan Markovic a, Anqi Liang a, b, Sue B. Watson c, David Depew b, Arthur Zastepa b,Preksha Surana a, Julie Vanden Byllaardt d, George Arhonditsis a, Maria Dittrich a, *

a Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1065 Military Trail, Scarborough, Toronto, ON, M1C 1A4, Canadab Environment and Climate Change Canada, Canada Centre for Inland Waters, Burlington, ON, L7S 1A1, Canadac Department of Biology, University of Waterloo, 200 University Ave. W, Waterloo, ON, N2L 3G1, Canadad Hamilton Harbour Remedial Action Plan, Canada Centre for Inland Waters, Burlington, ON, L7S 1A1, Canada

a r t i c l e i n f o

Article history:Received 16 January 2019Received in revised form22 May 2019Accepted 23 May 2019Available online 29 May 2019

a b s t r a c t

Diagenetic sediment phosphorus (P) recycling is a widespread phenomenon, which causes degradationof water quality and promotes harmful algal blooms in lakes worldwide. Strong P coupling with iron (Fe)in some lakes is thought to inhibit diagenetic P efflux, despite elevated P concentrations in the sediment.In these sediments, the high Fe content leads to P scavenging on ferric Fe near the sediment surface,which increases the overall P retention. Reduced external Fe inputs in such lakes due to industrialpollution control may lead to unintended consequences for sediment P retention. Here, we study sedi-ment geochemistry and sediment-water interactions in the historically polluted Hamilton Harbour (LakeOntario, Canada) which has undergone 30 years of restoration efforts. We investigate processes con-trolling diagenetic P recycling, which has previously been considered minor due to historically high Feloading. Our results demonstrate that present sediment P release is substantial, despite sediment Fecontent reaching 6.5% (dry weight). We conclude that the recent improvement of wastewater treatmentand industrial waste management practices has reduced Fe pollution, causing a decrease in diageneti-cally reactive Fe phases, resulting in the reduction of the ratio of redox-sensitive P and Fe, and thesuppression of P scavenging on Fe oxyhydroxides.

© 2019 Elsevier Ltd. All rights reserved.

1. Introduction

Excessive phosphorus (P) concentrations in aquatic ecosystemsfuel eutrophication, including algal blooms and bottom wateranoxia, and impact biodiversity (Smith and Schindler, 2009).Perpetuated sediment P release, or internal loading, is a commonissue in many lakes. However, in eutrophic lakes, P enriched sedi-ments and favorable biogeochemical conditions promote therecycling of the legacy P accumulated in the sediments at higherrates relative to P inputs and permanent P burial. Consequently,internal P loading can potentially delay the water quality im-provements gained by reducing the external P inputs. Therefore, todelineate the future ecosystem response to the remediation efforts,it is critical to understand the mechanisms controlling sediment P

release and retention. While elevated sediment Fe concentrationsare generally considered to promote P retention and reduce Prelease, due to P sorption on Fe oxyhydroxides, it is unclear howconcurrent Fe and P loading reduction impacts this FeeP coupling.This study focuses on one such unique system, where historicallyhigh Fe and P inputs were followed by a sharp reduction of Fe and Ploading due to the improvement of industrial and residentialwastewater treatment and management practices.

Sediment P recycling is controlled by multiple factors, whichinclude hypolimnetic oxygen levels, gross sedimentation fluxes ofreactive forms of organic matter, P and Fe forms, and lake watersulfate concentration (Katsev et al., 2006; Hupfer andLewandowski, 2008; Dittrich et al., 2013; Orihel et al., 2017). Theprimary factors controlling short-term summer P release areorganic matter mineralization, hypolimnetic oxygenation andFeeSeP coupling in surface sediments (Dittrich et al., 2009). Long-term sediment P dynamics are controlled by P chemical speciationof the input source material, diagenetic transformation and

☆ This paper has been recommended for acceptance by Joerg Rinklebe.* Corresponding author.

E-mail address: mdittrich@utsc.utoronto.ca (M. Dittrich).

Contents lists available at ScienceDirect

Environmental Pollution

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

https://doi.org/10.1016/j.envpol.2019.05.1240269-7491/© 2019 Elsevier Ltd. All rights reserved.

Environmental Pollution 252 (2019) 697e705

recycling of labile P forms and permanent burial of authigenic Pphases in anoxic sediments (Ruttenberg and Berner, 1993; Katsevand Dittrich, 2013). The scavenging of dissolved P on Fe oxy-hydroxides in surface sediments is critical for both short- and long-term P release and retention. It prevents diffusive P efflux out ofsediments into the overlying water column, while the eventualburial of Fe-bound P forms in the anoxic sediments promotes long-term P sequestration through the authigenesis of apatite (Slompet al., 1996) or vivianite (Rothe et al., 2015). Although sediment Fecontent is often invoked as a predictor of its P binding capacity(Geurts et al., 2008), the dynamics of FeeP coupling in lakes un-dergoing the remediation of Fe pollution have received lessattention.

Hamilton Harbour (Lake Ontario, Canada) is one of the sites inthe Laurentian Great Lakes that has had a record of culturaleutrophication and anthropogenic pollution since the 19th century.Historically, most P and Fe pollution can be traced to the rawdischarge from industrial and municipal wastewater facilities, witha significant contribution of urban runoff from the surroundingpopulation centers. Eutrophication and pollution issues were notedvery early in the 19th century (HHRAP, 1992), with serious waterquality degradation being present throughout the 20th century.The Harbour is identified as one of the most polluted regions inNorth America by theWater Quality Board of the International JointCommission (HHRAP, 1992).

The water quality has significantly improved since the controlmeasures to curb excessive pollution were implemented (1970-present). In the period between 1987 and 1997, the total phos-phorus (TP) concentrations decreased by 50% in response toexternal loading reduction (Hiriart-Baer et al., 2016). However,phytoplankton biomass proxies have remained unchanged in thepast 4 decades (Hiriart-Baer et al., 2016). The unchanged algalbiomass despite falling nutrient levels has been explained in termsof light limitation (Pemberton et al., 2007). However, other factors,such as the increased bioavailability of P, are likely supporting thepersistently high algal biomass and the development of harmfulalgal blooms. For instance, the ratio of the soluble reactive P (SRP)to TP in the water column has consistently increased concurrentlywith the large increase of hypolimnetic SRP accumulation rates (upto 0.5e1 mg/l/d in the past decade, Hiriart-Baer et al., 2016; up to8mg/m2/d normalized for the hypolimnion area, Fig. 1).

Despite these trends that are indicative of internal loading,previous studies have suggested that sediment P release is not asignificant issue (Barica, 1989; Mayer and Manning, 1990). High Feloading from nearby steel mills has generally been thought toprevent P release (Barica, 1989; Hiriart-Baer et al., 2016), yetexternal Fe loading has dramatically decreased from up to34,000 kg Fe/d in the 1970s to 500 kg Fe/d at present (Nriagu et al.,1983; Mayer and Manning, 1990; Harlow and Hodson, 1988;HHRAP, 2018, Fig. 1). The impact of up to 70-fold decline in Feloading on sediment P internal loading has not been examined.

This study aims to understand the interplay between thechanges in Fe and P loading and P diagenesis in the historicallyhighly polluted Hamilton Harbour. Considering the large reductionof Fe external pollution, we expected that sediment P retentionwasimpacted, and depending on the Fe speciation, redox conditionsand organic matter supply, historically low P efflux into the watercolumn may have increased. In order to prove this hypothesis, wecharacterized the Fe and P diagenetic processes and quantified thesediment P release, using a combination of long-term compre-hensive data on P and Fe external loading into the Harbour and ourexperimental data on sediment P binding forms and correspondingmetal contents, porewater chemistry and physicochemical gradi-ents across the sediment-water interface (SWI).

2. Methods

2.1. Site description

Hamilton Harbour is a triangular-shaped embayment located atthe western tip of Lake Ontario. The area of the Harbour isapproximately 21.5 km2, with a mean depth (Zmean) of 13 m and amaximumdepth Zmax of 25m (HHRAP,1992). Thewatershed area is~25 times the surface area of Hamilton Harbour and occupies

Fig. 1. a) Hypolimnetic SRP accumulation rate normalized for areal extent of hypo-limnion using long-termwater-quality record (Hiriart-Baer et al., 2016) (black squares),epilimnion Fe (MOE, 1985; 2008e2012 Debbie Burniston pers. comm.) (red circles) andgross external Fe loading in the period 1975e2016 (Harlow and Hodson, 1988; HHRAP,2018) (blue squares). b) External P loading (HHRAP, 2018) c) External Fe loading fromdifferent sources in the period 1995e2016 (HHRAP, 2018). Watershed tributariesrepresents combined inflows from Red Hill, Grindstone and Indian Creeks, and CootesParadise Marsh through the Desjardins Canal (see Section 2.1 in the main text). WWTPrepresents combined wastewater inputs (Woodward and Skyway WWTP). Steel millsrepresents combined inputs from Stelco and ArcelorMittal Dofasco. Combined seweroverflows (CSOs) represent a mix of untreated sewage and stormwater during intenserain events. Note that steel mils P inputs (b) represent net inputs. Negative inputs thusindicate that steel mils removed more P from the Harbour than they returned withprocessed waste water.

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705698

approximately 500 km2. The geology of the watershed comprisesPaleozoic shales, limestones and dolostones overlain with Quater-nary glacial till and glaciolacustrine sediments (McCann, 1987). Thewatershed is highly urbanized and receives municipal and indus-trial wastewater from the cities of Hamilton and Burlington. Directtributaries include Red Hill, Indian and Grindstone Creeks. To thewest is Cootes Paradise Marsh, which is connected to HamiltonHarbour through the Desjardins Canal. To the east, HamiltonHarbour is connected to Lake Ontario by the Burlington ship canal(BSC). The exchange through BSC represents ~80% of the total in-flows (Yerubandi et al., 2016). The hydraulic residence time inHamilton Harbour is highly variable, modulated by episodic large-scale intrusions of hypolimnetic waters fromWestern Lake Ontario(Barica, 1989; HHRAP, 1992; Yerubandi et al., 2016).

Two long-term monitoring sites were sampled during thesummer months of 2014e2016: station HH 9031 (43.27722 N,79.87833 W; depth - Zmax¼ 7 m, Fig. S1), which represents theshallow western part of the basin, and the centrally located stationHH 1001 (43.28750 N, 79.83833W; Zmax¼ 24m), which representsthe deepest part of the basin.

2.2. Sediment and porewater chemistry and calculation of diffusiveP fluxes

Samples were collected from the HH 1001 and HH 9031 stations,using a gravity corer (UWITEC, Austria) and polycarbonate coreliners 5.5 cm in diameter and ~70 cm in length. Sampling dates areshown in Table S1. Six or 7 cores per station were collected duringeach sampling. The sediment cores were sealed and transported tothe laboratory in a thermo-isolated box at ~4 �C.

One core per site was used to extract the porewater usingRhyzon samplers fitted with porous polymer 0.1 mm membranefilters. Upon extraction, the porewater was immediately pulled intoa syringe and divided into separate vials for analysis of Fe2þ and SRP(see SI). Samples were preserved following standard U.S. EPA (1982)guidelines. Porewater SRP and Fe (II) were analyzed at the Uni-versity of Toronto using the micro-spectrophotometric methodsdescribed in detail in Laskov et al. (2007).

Measurements of H2S, pH and redox potential were carried outex situ onshore, immediately after sampling using microelectrodes,on at least 1 core per site (see SI). Vertical profiles were measuredsimultaneously from the overlying water into the sediment. Prior tomicroelectrode measurements cores were kept tightly capped withno airspace between cap andwater overlying sediment. Intact coreswere exposed to the air only during measurements which did notexceed 40min.

Prior to sectioning, the sediment cores were first logged todifferentiate the visually distinct sedimentary layers, thensectioned within 24 h after sampling for the analysis of P bindingforms, metal contents in P fractions, porosity, dry weight, totalorganic matter and dissolved substances (see SI for more details).Composition of freeze-dried bulk sediments was analyzed on aPhilips PW2404 X-ray fluorescence spectrometer (see SI).

Sediment P release rates were estimated using porewaterdiffusive fluxes (JSRP), calculated using Fick's first law and diffusioncoefficients from Boudreau (1997) (See SI for more details).

Geochemical equilibrium calculations were performed usingPHREEQC version 3 with the thermodynamic database ‘phreeqc.-dat’ (Parkhurst and Appelo, 2013). The input values included porewater chemistry (Fe2þ, SRP), redox potential (pe), pH (varying be-tween 7 and 8.1) and hypolimnetic temperature at the time of thesampling (10e15 �C and 15e20 �C at HH1001 and HH9031,respectively). Thermodynamic calculations were used to simulateactivities of Fe2þ and SRP, exchange between solid-liquid phasesand the saturation state of vivianite in lake water and porewater

was calculated assuming solubility constant of vivianite ofKviv¼ 1� 10�36 from Nriagu (1972).

2.3. Sequential extraction for assessing phosphorus binding forms

Sediment P binding forms were extracted following the P frac-tionation technique from Psenner and Pucsko (1988), with modi-fications by Hupfer et al. (2009). This technique separates sedimentP into the following fractions: (a) loosely adsorbed P (total Pextracted with NH4Cl, NH4CleP); (b) redox-sensitive P bound to Feand Mn oxyhydroxides (total P extracted with bicarbonatedithionite (BD), BD-P); (c) P bound to aluminum hydroxide andclays (dissolved P extracted with NaOH, NaOH-SRP); (d) organicbound P (particulate P extracted with NaOH, NAOH-NRP); (e)calcium-bound P (HCleP) (total P extracted with HCl); and (f) re-fractory P (Ref-P) (Hupfer et al., 1995; Hupfer et al., 2009). Theconcentrations of themajor elements (including P, Fe, Al, Ca, Mn, Si)in the P extraction pools were measured by inductively coupledplasma mass spectrometry (ICP-MS).

3. Results and discussion

3.1. Dynamics of dissolved and solid P forms are closely coupledwith Fe cycling

The vertical pH and redox (Eh) profiles across the SWI showdecreasing pH (by 0.5e0.75 units) and redox potential(100e300mV) in the top 2 cm (Fig. S2). Sediments are highlyreducing with Eh< 100mV below 3 cm and present visual evidenceof gas, most likely methane or CO2 ebullition pockmarks, or both(Fig. S3). The measured Eh range in the top 2 cm (<200mV) atcircumneutral pH values is consistent with the shallow ironreduction zone (Sigg, 2000; Weber et al., 2006), contradictingearlier studies that suggested that redox potential near the SWIremains above the range expected for iron reduction and betweenthe range of denitrification and manganese reduction (MOE, 1985;Mayer and Manning, 1990).

Fe reduction is intimately linked to organic matter degradation,thus promoting the simultaneous release of P bound to organiccompounds and Fe oxyhydroxides. In the presence of sufficientamounts of Fe2þ, released P can precipitate in authigenic phases,such as vivianite Fe3(PO4)28H2O (Rothe et al., 2015). This takesplace only in environments where sulfide, a strong binding partnerof Fe2þ, is absent in porewater. Hamilton Harbour sediments arehighly enriched in organic matter (Fig. 2), which enhances S2�

production through dissimilatory sulfate reduction and organicmatter desulfuration (Rothe et al., 2015). However, sediment Fecontent is exceptionally high (5e6.5% Fe d.w. at the sediment sur-face, Table S2 and S3), indicating the presence of an almost inex-haustible pool of available Fe to bind excess sulfide. Indeed,porewater sulfide concentrations were below the microsensordetection limit (0.3 mM H2S) throughout the top 5e6 cm of sedi-ment, suggesting the presence of active sulfide sink, likely due tothe precipitation of Fe sulfides, as suggested by Loh et al. (2013).Alternatively, this may also reflect vigorous sulfur cycling bysulfide-oxidizing bacteria (MOE, 1985).

Dissolved Fe2þ accumulates in the porewater, reaching con-centrations of 20e50mg/L (HH 1001) and 3e28mg/L (HH 9031) inthe depths between 20 cm and 30 cm (Fig. 3). Concurrently, theporewater SRP concentrations mirror the increase of porewaterFe2þ, rising below the SWI and reaching maximum concentrationsbetween 2 and 3mg P/L at station HH 1001 and 2.5e5.5mg P/L atstation HH 9031 in the depths between 5 cm and 10 cm. From thismaximum SRP, the concentration decreases with depth, apart fromthe porewater sampled at station HH 1001 in July 2016, which

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705 699

shows increase with depth. The porewater ratio of Fe2þ to P reflectsavailability of reactive iron species and consequently P bindingcapacity of sediments (Smolders et al., 2001). A ferric Fe-rich layer,formed by the oxidation of Fe2þ at surface sediment can quantita-tively trap P when porewater ratio of Fe2þ:P> 1e3.5 (Smolderset al., 2001; Geurts et al., 2008). In Hamilton Harbour, this ratio islargely <1, indicating that the sediment surface contains an insuf-ficient amount of reactive Fe to bind all the P released duringdiagenesis (Fig. S4).

At station HH 1001, the sediment P concentrations decreasefrom 3mg P/g d.w. in the surface sediment to 2mg P/g d.w. in thedepths between 15 cm and 25 cm, increasing to 2.5mg P/g d.w. inthe deepest sampled interval (Fig. 2). At station HH 9031, thesediment P concentration decreased from amaximum of 2.7mg P/g

d.w. in the surface sediment to 2e2.3mg P/g d.w. in the depthsbetween 10 cm and 25 cm (Fig. 2). The total concentrations ofsediment P measured using the XRF technique confirmed thesevalues, ranging between 0.33 and 0.23% d.w. (equivalent of3.3e2.3mg P d.w.) at station HH 1001, and 0.27 and 0.18% d.w.(equivalent of 2.7e1.8mg P d.w.) at station HH 9031 (Table S2 andS3).

The redox-sensitive bound BD-P fraction accounts for 33e35% ofthe total P in the surface sediment, decreasing rapidly with sedi-ment depth at both stations (Fig. 2). In contrast, the NaOH-SRPfraction, which ranges between 20% and 30% of the total P in thesurface sediment at stations HH 9031 and HH 1001 respectively,increases with depth, reaching 40% and 50% at stations HH 9031and HH 1001 respectively. This fraction represents P, which is

Fig. 2. Sediment P binding forms: (a) HH 1001 and (b) HH 9031. Note that potentially mobile P represents the sum of NH4CleP, BD-P and NaOH-NRP. LOI represents loss on ignition.

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705700

exchangeable with hydroxyl ion, e.g., P sorbed on Al hydroxides,clays and non-reducible inorganic P compounds. The accumulationof this fractionwith the sediment depth suggests that this P bindingform is the most important burial sink for P.

Both NaOH-NRP and HCleP concentrations decline with depthat both stations. While NaOH-NRP is generally thought to becomposed of diagenetically reactive organic P phases, HCleP is theimmobile fraction of P, composed mainly of Ca-bound phases,chiefly (hydroxy)apatite (Lukkari et al., 2007). However, HCleP issensitive to pH and thus it is possible that some apatite dissolutiontakes place during burial, as pH typically decreases with sedimentdepth (Fig. S3). Exchangeable P (NH4CleP) is in equilibrium withporewater and represents 1e2% of the total P and decreases withdepth from the maximum in the surface sediment to the minimumin the deepest interval.

The total Fe increases with depth from minima in the surfacesediments (4.98% Fe d.w. and 6.34% Fe d.w. at HH 9031 and HH1001, respectively, Table S2 and Table S3, Fig. S5), to maximumconcentrations in the deepest sediment layer (6.37% and 13.05% Fed.w. at HH 9031 and HH 1001, respectively). At present, surface Feenrichment at HH1001 (Fig. S5) is comparable to the tailings lakesimpacted by mining activity (>60mg Fe/g d.w., Kleeberg et al.,2013). The downward increasing Fe concentration reflects a his-tory of continually decreasing wastewater Fe discharges, whichdeclined from 34,000 kg Fe/d to 500 kg Fe/d at present (Nriaguet al., 1983; Harlow and Hodson, 1988; Mayer and Manning, 1990,Fig. 1). In comparison, the Ti and Al concentrations are constantwith depth, representing an inert detrital supply from watershederosion and resulting in a downcore increase of Fe/Ti and Fe/Al

ratios. However, sediment Fe concentrations are still considerablyhigher than watershed erosion inputs estimated to be 4.0e4.5% Fed.w. (Mayer and Manning, 1990). Excess Fe, despite diminishedexternal inputs (Fig. 1), leads to diagenetic Fe migration toward thesurface due to Fe reduction in deeper sediment, diffusive fluxes tothe sediment surface, subsequent oxidation to Fe3þ, and precipi-tation near the sediment surface.

Sediment Fe enrichment has a profound effect on P mobility.Jensen et al. (1992) have shown that P release is strongly sup-pressed in sediment with Fe/P ratio >15 (w%/w%) at a circum-neutral pH range (Jensen et al., 1992). The Fe:P ratios of the surfacesediments in Hamilton Harbour are higher (Fe:P> 18; Table S2 andS3) than this empirical cutoff, suggesting the presence of excess Fe,which can scavenge P. However, Fe enrichment itself does nottranslate into P retention, because sediments are comprised of amix of different Fe phases with variable reactivity toward P. Whilecomposition of historical Fe load cannot be reconstructed withcertainty, most of the Fe inputs can be traced to industrial effluentswith amorphous Fe3þ phases, hematite and wüstite from industrialwaste, and subordinate chlorite from the watershed, which com-bined represented ~95% of the total Fe load prior to the pollutioncontrol (Mayer and Manning, 1990). Mineralogical characterizationof Fe inputs using X-ray diffraction and M€ossbauer spectral deter-mination has shown that only up to 40% of the total sediment Feloading was in the form of reactive amorphous Fe oxyhydroxides atthe time when industrial waste inputs prevailed (Manning et al.,1980; Mayer and Manning, 1990). Considering that these indus-trial inputs have recently been reduced (Fig. 1), presently we findthat amorphous BD-reducible Fe oxyhydroxides make up to

Fig. 3. (a) Porewater P (black square) and Fe (red circle) concentrations at station HH 1001 - top left panel; (b) Porewater P (black square) and Fe (red circle) concentrations atstation HH 9031 - top right panel; (c) Diffusive fluxes of P (diagonal lines) and Fe (grey) at station HH 1001- left bottom panel; (d) Diffusive fluxes of P (diagonal lines) and Fe (grey)at station HH 9031 - right bottom. Note the different scales in the subplots of panels a) and b).

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705 701

0.6e0.93% and 0.6e0.95% sediment d.w. at stations HH 1001 andHH 9031, respectively. Diagenetically less reactive phases aremagnetite and hematite, as well as a portion of Fe bound in silicates(Raiswell et al., 1994). These less reactive phases are extracted bycold HCl in the present dataset (HCleFe acid extracted and re-fractory fraction, Raiswell et al., 1994) and account for a largerproportion of the total Fe.

The importance of P sorption on Fe oxyhydroxides is supportedby a strong correlation between BD-Fe and BD-P (Fig. S6). The ratioof BD-Fe:BD-P can be used to estimate the spare sorption capacityof Fe oxyhydroxides, the principal P binding partner (Anschutzet al., 1998). At station HH 1001, the BD-Fe:BD-P ratio is 4.2:1,considerably lower than the ratio of 6.7:1, which corresponds to thesupersaturation of the surface binding sites of nanoscale ferrihy-drite, the phase with the highest P sorption capacity (Sigg andStumm, 1981; Anschutz et al., 1998). While this suggests that theFe oxyhydroxides at station HH 1001 have a limited spare capacityto bind excess P, the low BD-Fe:BD-P ratios can also denote theprecipitation of ferric phosphate during the rapid oxidation of Fe2þ

in P-enriched water, which produces a phase with a maximum Feand P ratio of 2:1 (Thibault et al., 2009). Since station HH 1001experiences seasonal hypolimnetic anoxia, it can be suggested thatthe oxygenation of hypolimnion during turnover or periodic in-trusions of oxygenated Lake Ontario water results in the precipi-tation of iron oxyhydroxides and ferric phosphate. Conversely, BD-Fe:BD-P ratios at the shallow station HH 9031 are 7.9 and thus overthe saturation limit of 6.7:1 (Anschutz et al., 1998), indicating somespare binding capacity. This modest spare BD-P binding capacitycan be related to the local pollution and hydrodynamic patterns, asHH 9031 is located distally from major industrial and municipalwastewater plants and receives lesser inputs from those pointsources (see map at Fig. S1).

3.2. Sediment P release is a significant P source to the water column

The diffusive P fluxes during summer 2016 ranged between 2.0and 5.2mg P/m2/d and between 3.5 and 6.5mg P/m2/d at stationsHH 1001 and HH 9031, respectively (Fig. 3). These diffusive P fluxesare comparable with hypolimnetic P accumulation rates 3e8mg P/m2/d normalized for the area of hypolimnion estimated in theperiod of 2013e2016 (Fig.1; comparewith Fig. 3). Themagnitude ofP diffusive fluxes also matches well mass balance modeling1.2e3.8mg P/m2/d (Gudimov et al., 2010, 2011), and diffusive Pfluxes of 1.7mg P/m2/d by Azcue et al. (1998), which were the firstto suggest that sediment represents a potential P source. Incontrast, several previous studies have concluded that HamiltonHarbour sediments represent a net sink, rather than a net source, ofP (e.g., Barica, 1989; Mayer and Manning, 1990; Kellershohn andTsanis, 1999). High nitrate levels supporting nitrate reduction,and sporadic hypolimnion oxygenation, were thought to maintainthe sediment Eh potential above the range of iron reduction (MOE,1985; Barica, 1989; Mayer and Manning, 1990). At the same time,high Fe inputs provided enough Fe oxyhydroxides to trap P in theuppermost sediment (Barica, 1989; Mayer and Manning, 1990).

We posit that the increased importance of sediment P recyclingsupported by our findings and recent empirical and modelingstudies (Gudimov et al., 2010; Hiriart-Baer et al., 2016) is consistentwith the FeeP decoupling due to dramatically decreasing Feloading. A significant portion of the earlier Fe loading was steelpickling solution waste, comprised mostly of Fe(OH)3 and ferricchloride solution, which have a strong capacity to bind P (e.g.,Smolders et al., 2001). Industrial Fe discharges have virtually ceasedin the past 3 decades (Nriagu et al., 1983; Harlow and Hodson,1988,Fig. 1), leading to reduced epilimnion Fe concentrations and coin-ciding with an increase in hypolimnetic SRP accumulation rates

(Hiriart-Baer et al., 2016, Fig. 1). Therefore, it can be concluded thatthe declining Fe pollution reduced the sediment capacity to bindexcess P. Simultaneously with declining Fe pollution, P externalinputs did not decline by a comparable rate during the last 20e30years and consequently P originating from organic matter nowa-days cannot be retained in the sediment as much as it did before,leading to increasing SRP hypolimnion accumulation (Fig. 1).

3.3. P immobilization is controlled by P scavenging on Feoxyhydroxides and FeeP authigenesis

An intriguing feature of the sediment P binding form dynamicsis sink switching from redox-sensitive BD-P to NaOH-SRP withsediment depth. The latter extraction pool has been linked to Psorbed on amorphous Al hydroxides, which is stable at reducingconditions and, therefore, not vulnerable to diagenetic recyclingduring burial (Kop�a�cek et al., 2005). The relative sorption capacityof available “free” Al hydroxides compared with available reducibleFe oxyhydroxides is reflected in the ratio of Al and Fe extracted inthe first 3 extraction pools (NH4Cl, BD, NaOH-SRP fraction). Al:Pratios indicate the overall availability of Al oxyhydroxides relativeto the amount of P bound in the diagenetically most reactive phases(sum of NH4CleP and BD-P). Kop�a�cek et al. (2005) showed that ifAl:Fe(NH4Cl, BD, NaOH-SRP) ratio >3 and Al(NaOH-SRP):P(NH4ClePþBD-P)>25, there are enough Al hydroxides to adsorb P released duringthe reductive dissolution of BD-P. This is not the case in the Ham-ilton Harbour sediments, where Al:Fe(NH4Cl, BD, NaOH-SRP) ratios are0.2e2 and Al(NaOH-SRP):P(NH4ClePþBD-P) vary between 0.4 and 18(Fig. S7), implying that Al hydroxide concentrations are too low toexplain diagenetic sink switching from BD-P to NaOH-SRP.

This poses the question of the nature of the NaOH-SRP pool,which is often termed “metal bound P” but includes a complex poolof P phases with a varying degree of diagenetic reactivity, includingP associated with Al hydroxide, vivianite, clays and some hydro-lyzed organic P (Lukkari et al., 2007; Rothe et al., 2015). Notably, inour dataset, the Fe and P in the NaOH-SRP extraction pool arecorrelated (albeit moderately, Pearson r¼ 0.76), unlike Al and P(Pearson r¼�0.33 and 0.06 at stations HH 9031 and HH 1001respectively, Fig. S8). The Fe:P ratios vary between 0.5 and 2.5(average~1), which is fairly similar to the molar ratio of vivianiteFe3(PO4)28H2O (Fe:P¼ 1.5). Therefore, similar to the Rothe et al.(2015) study, we propose that sink switching to the NaOH-SRPextraction pool is likely associated with vivianite authigenesis.

Rothe et al. (2015) have also identified the S:Fe molar ratio<1.1e1.5 (total S to reactive Fe measured by digestion with aquaregia) as an empirical threshold/indicator of vivianite formation.The basis of this indicator is the presence of an excess amount ofreactive Fe relative to the production of sulfide during sulfatereduction. Hence, vivianite formation is not restricted by the supplyof reactive Fe and Fe sulfide precipitation. While we did notquantify the aqua regia extractable Fe, we propose that the sum ofFe extracted by the BD reagent and diluted HCl (BD-Fe, HCleFe andRef-Fe) (Fig. S5) represent a reasonable approximation of the totalFe that is reactive toward sulfide (Rothe et al., 2015). Thus, theestimated reactive Fe varies between 0.50 and 0.72mmol Fe/g d.w.at station HH 1001 and between 0.41 and 0.72mmol Fe/g d.w. atstation HH 9031, giving Stotal:Fereactive ratios well below theempirical threshold value of 1.1 (Table S2 and S3).

Another indicator for possible authigenic vivianite formation isporewater supersaturation (Fig. 4). However, porewater supersat-uration in respect to vivianite alone is not enough to prove vivianitepresence, which must be demonstrated through mineralogicalidentification (Rothe et al., 2016). Manning et al. (1980) have sug-gested that due to high concentrations of porewater Fe and P, viv-ianite would be expected to precipitate in Hamilton Harbour;

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705702

however, their M€ossbauer spectroscopic analysis of the bulk sedi-ment could not conclusively identify its peaks, likely due to the lowoverall content. Subsequent X-ray diffraction (XRD) measurementsof bulk sediment samples also could not discern vivianite presence(Mayer and Manning, 1990). If we assume that NaOH-SRP reflectsthe vivianite content of sediments, which is unlikely given the non-selective and operational nature of extraction (Rothe et al., 2015),the maximum vivianite concentration is at most 0.81% sedimentd.w (Fig. 2 and Fig. S5). This is considerably less than a minimum of5% d.w of vivianite which was demonstrated to be necessary fordirect XRD identification in the bulk sediment samples (Rothe et al.,2014). Although sediment and porewater chemistry, sediment Pbinding forms and S:Fe molar ratios indicate vivianite authigenesis,furthermineralogical investigation is necessary to confirmvivianitepresence and quantify its concentrations in Hamilton Harboursediments.

3.4. Future trends of internal P loading and retention

The historically high Fe loading in Hamilton Harbour

accumulated a large amount of excess Fe, which still modulates Pcycling through sorption on reducible Fe oxyhydroxides and theprecipitation of vivianite in anoxic sediment. However, the dimin-ished industrial Fe inputs reduce the availability of reactive Fespecies relative to P and S. G€achter and Müller (2003) have shownthat in anoxic sediments, the permanent P retention is sustained ifthe molar ratio of reactive Fe:P is> 2. If we assume that the entireFe and P external loading in Hamilton Harbour is composed ofreactive species, the threshold value for Fe:P in the external loadingsources has been breached around the year 2000 (Fig. 1).

However, a gross external loading Fe:P ratio may overestimatethe availability of the reactive Fe species relative to P. The sources ofFe and P are distinctly divergent, with present-day P originatingmostly from wastewater treatment, and Fe representing erosionalsources from thewatershed (Fig.1). It has been demonstrated that Poriginating from wastewater treatment effluents is mostlybioavailable (e.g., Qin et al., 2015). Watershed Fe inputs are pre-dominantly composed of clays (Manning et al., 1980; Mayer andManning, 1990), which are less reactive during diagenesis. Thus,it is likely that recent sediments are receiving much lower inputs ofreactive Fe than suggested by the total Fe loading. This decreasesthe sediment ability to bind excess P.

The reduced external Fe inputs relative to the historical valuesdecreases the inventory of reactive Fe. The current Fe pool in thesurface sediment (top 1 cm, dry bulk density ~0.2 g/cm3) isapproximately 100 and 127 g/m2 (stations HH 9031 and HH 1001,respectively). The decrease rate of the Fe inventory can be calcu-lated based on the total Fe profiles (Fig. S5; Table S2 and S3) and thesediment accumulation rates (assuming 0.15 g/cm2/yr; Nriagu et al.,1983; Jonlija, 2014) is 2 and 12 g/m2/yr at stations HH 9031 and HH1001, respectively. Therefore, assuming the current sedimentaccumulation rates, within the next 5 years surface sediments willreach Fe concentration in watershed erosion inputs (4.0e4.5% Fed.w., Mayer and Manning, 1990) which is equivalent to total Fe of80e90 g Fe/m2).

One important consequence of decreasing Fe inventory is adecline of the sediment capacity to scavenge P on Fe oxyhydroxidesin the surface sediments. Still, there is another feedback mecha-nism, namely that Fe oxyhydroxides are known to enhance thepreservation of organic matter in sediments (Lalonde et al., 2012).Thus, reduction of Fe oxyhydroxides inventory reduces organicmatter preservation thereby promoting degradation and release oforganic P, whose sedimentation rates likely did not change signif-icantly over the past 2 decades, considering that algal biomassduring this period has remained at present levels (Hiriart-Baeret al., 2016). Furthermore, following the argument by Rothe et al.(2015), lake productivity and organic matter export are the pri-mary controls of sulfide production through sulfate reduction anddesulfuration of organic S. Hence, it can be presumed that sedimentsulfide production will be sustained if primary productivity doesnot change, as is the case in Hamilton Harbour (Hiriart-Baer et al.,2016). Since Fe sulfides preferentially precipitate over vivianite(G€achter and Müller, 2003; Rothe et al., 2015), lower amounts of Fewill be available to bind with P and, consequently, it will lead todecreasing P retention and act as a positive feedback on watercolumn eutrophication.

Presently, we estimate that a considerable amount of legacy P inthe sediment is being recycled. The sediment P diffusive fluxesduring summer are estimated to be 2e6.5mg P/m2/d. Extrapolatingthese values across the surface of sediment overlain by a hypo-limnetic layer leads to a P flux of ~200± 100 kg P/day in the range ofall combined external inputs (Fig. 1). However, this is a worst-casescenario and likely an overestimate, as this calculation does nottake into account flushing from Lake Ontario and the fact thatlegacy P is not uniformly concentrated throughout the Harbour.

Fig. 4. Ion activities of phosphate and iron saturation in the porewater, calculatedusing PHREEQC (Parkhurst and Appelo, 2013): (a) HH 1001 and (b) HH 9031. The di-agonal red line represents the vivianite solubility constant Kviv¼ 1� 10�36 given byNriagu (1972). Ion activities of pore water samples collected in June 2016 (filledsquares), July 2016 (empty circles), August 2016 (up-pointing triangle) and September2016 (down-pointing triangle).

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705 703

Nonetheless, one can expect that this P release from the sedimentto the water column will deteriorate the water quality of HamiltonHarbour during prolonged periods of stagnation.

One uncertainty about this adverse impact comes from thesporadic hypolimnion exchange with Lake Ontario (e.g., Yerubandiet al., 2016). Considering that hypoxic episodes may persist(Yerubandi et al., 2016), the possibility of extensive surface water Penrichment originating from sediments should, therefore, beconsidered and factored into the expectations of the area recoveryby the water managers. Understanding sediment impact on thewater quality is particularly important in the context of climatechange, which has a potential to alter the hydrodynamic regimewithin the Harbour, likely reducing the frequency of Lake Ontariointrusion events and increasing the duration of the stratificationperiod (Yerubandi et al., 2016), thus exacerbating the impact of thehypolimnetic P accumulation and feedback associated withreduced Fe pollution.

4. Conclusions

In this study, we delineated the processes controlling the sedi-ment P dynamics in eutrophic Hamilton Harbour, a unique systemundergoing the reduction of Fe industrial pollution. We found thatthe external loading and diagenetic processes are responsible forthe redistribution of mobilized P and Fe within the anoxic sedi-ment. These processes drive an effective sink switching to a stableauthigenic Fe(II) phase, most likely vivianite, and contribute to Pretention. However, reducing Fe sediment levels limit the surfacesediment capacity to bind newly deposited P, which emphasizesthe importance of external P loading reduction.

Our results demonstrated that the reduction of Fe loading hasaccelerated the P recycling processes, which contrasts with earlierstudies that concluded that historically high Fe levels effectivelytrap P in sediment. The Hamilton Harbour case is, therefore, anexample of a systemwhere controlling industrial pollution leads toan unintended increase of sediment diagenetic P recycling.

Currently, the diagenetic P efflux from sediment into the watercolumn has considerable impact on the overall P budget in eutro-phic Hamilton Harbour. In the worst-case scenario, 2e6.5mg P/m2/d of sediment P is released into the hypolimnion, which is withinthe range of external P loading during summer months. Remedia-tion plans should consider that these internal P subsidies will exerta growing influence over nutrient balance and productivity anddelay recovery process.

Acknowledgements

This project was funded by the Great Lakes Sustainability Fundand NSERC Discovery Grant to MD. We are thankful for thegenerous support of Environment and Climate Change Canada, theHamilton Harbour Remediation Action Plan and the Ontario Min-istry of the Environment and Climate Change. Debbie Burniston isacknowledged for providing water-quality monitoring data forFig. 1. Furthermore, the field and laboratory assistance of Elaine Luis kindly acknowledged.

Appendix A. Supplementary data

Supplementary data related to this article can be found athttps://doi.org/10.1016/j.envpol.2019.05.124.

References

Anschutz, P., Zhong, S., Sundby, B., Mucci, A., Gobeil, C., 1998. Burial efficiency ofphosphorus and the geochemistry of iron in continental margin sediments.

Limnol. Oceanogr. 43, 53e64. https://doi.org/10.4319/lo.1998.43.1.0053.Azcue, J.M., Zeman, A.J., Mudroch, A., Rosa, F., Patterson, T., 1998. Assessment of

sediment and porewater after one year of subaqueous capping of contaminatedsediments in Hamilton Harbour, Canada. In: Water Science and Technology,pp. 323e329. https://doi.org/10.1016/S0273-1223(98)00214-5.

Barica, J., 1989. Unique limnological phenomena affecting water quality of HamiltonHarbour, Lake Ontario. J. Gt. Lakes Res. 15, 519e530. https://doi.org/10.1016/S0380-1330(89)71507-0.

Boudreau, B.P., 1997. Diagenetic Models and Their Implementation. ModellingTransport and Reactions in Aquatic Sediments. Springer, New York.

Dittrich, M., Chesnyuk, A., Gudimov, A., McCulloch, J., Quazi, S., Young, J., Winter, J.,Stainsby, E., Arhonditsis, G., 2013. Phosphorus retention in a mesotrophic lakeunder transient loading conditions: insights from a sediment phosphorusbinding form study. Water Res. 47, 1433e1447. https://doi.org/10.1016/j.watres.2012.12.006.

Dittrich, M., Wehrli, B., Reichert, P., 2009. Lake sediments during the transienteutrophication period: reactive-transport model and identifiability study. Ecol.Model. 220, 2751e2769.

G€achter, R., Müller, B., 2003. Why the phosphorus retention of lakes does notnecessarily depend on the oxygen supply to their sediment surface. Limnol.Oceanogr. 48, 929e933. https://doi.org/10.4319/lo.2003.48.2.0929.

Geurts, J.J.M., Smolders, A.J.P., Verhoeven, J.T.A., Roelofs, J.G.M., Lamers, L.P.M., 2008.Sediment Fe:PO4 ratio as a diagnostic and prognostic tool for the restoration ofmacrophyte biodiversity in fen waters. Freshw. Biol. 53, 2101e2116. https://doi.org/10.1111/j.1365-2427.2008.02038.x.

Gudimov, A., Stremilov, S., Ramin, M., Arhonditsis, G.B., 2010. Eutrophication riskassessment in Hamilton Harbour: system analysis and evaluation of nutrientloading scenarios. J. Gt. Lakes Res. 36, 520e539. https://doi.org/10.1016/j.jglr.2010.04.001.

Gudimov, A., Ramin, M., Labencki, T., Wellen, C., Shelar, M., Shimoda, Y., Boyd, D.,Arhonditsis, G.B., 2011. Predicting the response of Hamilton Harbour to thenutrient loading reductions: a modeling analysis of the “ecological unknowns.J. Gt. Lakes Res. https://doi.org/10.1016/j.jglr.2011.06.006.

HHRAP (Hamilton harbour remedial action plan), 1992. Remedial Action Plan forHamilton Harbour Stage I Report, second ed.

HHRAP, 2018. Contaminant Loadings and Concentrations to Hamilton Harbour:2008-2016 Update. Produced for the HHRAP Technical Team.

Harlow, H.E., Hodson, P.V., 1988. Chemical contamination of Hamilton HarbourCanada a review. Can. Tech. Rep. Fish. Aquat. Sci. 0, 1e91.

Hiriart-Baer, V.P., Boyd, D., Long, T., Charlton, M.N., Milne, J.E., 2016. HamiltonHarbour over the last 25 years: insights from a long-term comprehensive waterquality monitoring program. Aquat. Ecosys. Health Manag. 19, 124e133. https://doi.org/10.1080/14634988.2016.1169686.

Hupfer, M., G€achter, R., Giovanoli, R., 1995. Transformation of phosphorus species insettling seston and during early sediment diagenesis. Aquat. Sci. 57, 305e324.https://doi.org/10.1007/BF00878395.

Hupfer, M., Lewandowski, J., 2008. Oxygen controls the phosphorus release fromlake sediments e a long-lasting paradigm in limnology. Int. Rev. Hydrobiol.https://doi.org/10.1002/iroh.200711054.

Hupfer, M., Zak, D., Robberg, R., Herzog, C., P€othig, R., 2009. Evaluation of a well-established sequential phosphorus fractionation technique for use in calcite-rich lake sediments: identification and prevention of artifacts due to apatiteformation. Limnol Oceanogr. Methods 7, 399e410. https://doi.org/10.4319/lom.2009.7.399.

Jensen, H.S., Kristensen, P., Jeppesen, E., Skytthe, A., 1992. Iron:phosphorus ratio insurface sediment as an indicator of phosphate release from aerobic sedimentsin shallow lakes. Hydrobiologia 235e236, 731e743. https://doi.org/10.1007/BF00026261.

Jonlija, M., 2014. Assessment of Toxic Cyanobacterial Abundance at HamiltonHarbour from Analysis of Sediment and Water. Univesity of Waterloo.

Katsev, S., Dittrich, M., 2013. Modeling of decadal scale phosphorus retention in lakesediment under varying redox conditions. Ecol. Model. 251, 246e259. https://doi.org/10.1016/j.ecolmodel.2012.12.008.

Katsev, S., Tsandev, I., L'Heureux, I., Rancourt, D.G., 2006. Factors controlling long-term phosphorus efflux from lake sediments: exploratory reactive-transportmodeling. Chem. Geol. 234, 127e147. https://doi.org/10.1016/j.chemgeo.2006.05.001.

Kellershohn, D.A., Tsanis, I.K., 1999. 3D eutrophication modeling of HamiltonHarbour: analysis of remedial options. J. Gt. Lakes Res. 25, 3e25. https://doi.org/10.1016/S0380-1330(99)70713-6.

Kleeberg, A., Herzog, C., Hupfer, M., 2013. Redox sensitivity of iron in phosphorusbinding does not impede lake restoration. Water Res. 47, 1491e1502. https://doi.org/10.1016/j.watres.2012.12.014.

Kop�a�cek, J., Borovec, J., Hejzlar, J., Ulrich, K.U., Norton, S.A., Amirbahman, A., 2005.Aluminum control of phosphorus sorption by lake sediments. Environ. Sci.Technol. 39, 8784e8789. https://doi.org/10.1021/es050916b.

Lalonde, K., Mucci, A., Ouellet, A., G�elinas, Y., 2012. Preservation of organic matter insediments promoted by iron. Nature 483, 198e200. https://doi.org/10.1038/nature10855.

Laskov, C., Herzog, C., Lewandowski, J., Hupfer, M., 2007. Miniaturized photo-metrical methods for the rapid analysis of phosphate, ammonium, ferrous iron,and sulfate in pore water of freshwater sediments. Limnol Oceanogr. Methods 5,63e71. https://doi.org/10.4319/lom.2007.5.63.

Loh, P.S., Molot, L.A., Nürnberg, G.K., Watson, S.B., Ginn, B., 2013. Evaluating re-lationships between sediment chemistry and anoxic phosphorus and iron

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705704

release across three different water bodies. Inl. Waters 3, 105e118. https://doi.org/10.5268/IW-3.1.533.

Lukkari, K., Hartikainen, H., Leivuori, M., 2007. Fractionation of sediment phos-phorus revisited. I: fractionation steps and their biogeochemical basis. LimnolOceanogr. Methods 5, 433e444. https://doi.org/10.4319/lom.2007.5.433.

Manning, P.G.G., Jones, W., Birchall, T., 1980. M€ossbauer Spectral studies of iron-enriched sediments from Hamilton Harbour, Ontario. Can. Mineral. 18,291e299.

Mayer, T., Manning, P.G., 1990. Inorganic contaminants in suspended solids fromHamilton Harbour. J. Gt. Lakes Res. 16, 299e318. https://doi.org/10.1016/S0380-1330(90)71423-2.

McCann, S.B., 1987. Physical landscape of the Hamilton region. In: Dear, M.J.,Drake, J.J., Reeds, L.G. (Eds.), Steel City: Hamilton and Region. University ofToronto Press, pp. 13e33.

MOE (Ontario Ministry of the Environment), 1985. Hamilton Harbour TechnicalSummary and General Management Options. Ministry of the Environment,Toronto, Ontario, Canada.

Nriagu, J.O., 1972. Stability of vivianite and ion-pair formation in the systemFe3(PO4)2-H3PO4H3PO4-H2O. Geochem. Cosmochim. Acta 36, 459e470. https://doi.org/10.1016/0016-7037(72)90035-X.

Nriagu, J.O., Wong, H.K.T., Snodgrass, W.J., 1983. Historical records of metal pollutionin sediments of Toronto and Hamilton Harbours. J. Gt. Lakes Res. 9, 365e373.https://doi.org/10.1016/S0380-1330(83)71908-8.

Orihel, D.M., Baulch, H.M., Casson, N.J., North, R.L., Parsons, C.T., Seckar, D.C.M.,Venkiteswaran, J.J., 2017. Internal phosphorus loading in Canadian fresh waters:a critical review and data analysis. Can. J. Fish. Aquat. Sci. 1e25. https://doi.org/10.1139/cjfas-2016-0500.

Parkhurst, D.L., Appelo, C.A.J., 2013. PHREEQC (version 3)-A computer program forspeciation, batch-reaction, one-dimensional transport, and inverse geochemicalcalculations. Model. Tech. B. 6, 497 https://doi.org/Rep. 99-4259.

Pemberton, K.L., Smith, R.E.H., Silsbe, G.M., Howell, T., Watson, S.B., 2007. Controlson phytoplankton physiology in Lake Ontario during the late summer: evidencefrom new fluorescence methods. Can. J. Fish. Aquat. Sci. 64, 58e73. https://doi.org/10.1139/F06-166.

Psenner, R., Pucsko, R., 1988. Phosphorus fractionation: advantages and limits of themethod for the study of sediment P origins and interactions. Adv. Limnol. 30,43e59.

Qin, C., Liu, H., Liu, L., Smith, S., Sedlak, D.L., Gu, A.Z., 2015. Bioavailability andcharacterization of dissolved organic nitrogen and dissolved organic phos-phorus in wastewater effluents. Sci. Total Environ. 511, 47e53. https://doi.org/10.1016/j.scitotenv.2014.11.005.

Raiswell, R., Canfield, D.E., Berner, R.A., 1994. A comparison of iron extractionmethods for the determination of degree of pyritisation and the recognition ofiron-limited pyrite formation. Chem. Geol. 111, 101e110. https://doi.org/10.1016/

0009-2541(94)90084-1.Rothe, M., Frederichs, T., Eder, M., Kleeberg, A., Hupfer, M., 2014. Evidence for viv-

ianite formation and its contribution to long-term phosphorus retention in arecent lake sediment: a novel analytical approach. Biogeosciences. https://doi.org/10.5194/bg-11-5169-2014.

Rothe, M., Kleeberg, A., Grüneberg, B., Friese, K., P�erez-Mayo, M., Hupfer, M., 2015.Sedimentary sulphur:iron ratio indicates vivianite occurrence: a study fromtwo contrasting freshwater systems. PLoS One 10. https://doi.org/10.1371/journal.pone.0143737.

Rothe, M., Kleeberg, A., Hupfer, M., 2016. The occurrence, identification and envi-ronmental relevance of vivianite in waterlogged soils and aquatic sediments.Earth Sci. Rev. https://doi.org/10.1016/j.earscirev.2016.04.008.

Ruttenberg, K.C., Berner, R.A., 1993. Authigenic apatite formation and burial insediments from non-upwelling, continental margin environments. Geochem.Cosmochim. Acta. https://doi.org/10.1016/0016-7037(93)90035-U.

Sigg, L., 2000. Redox potential measurements in natural waters: significance, con-cepts and problems. In: Redox: Fundamentals, Processes, and Applications,pp. 1e12.

Sigg, L., Stumm, W., 1981. The interaction of anions and weak acids with the hy-drous goethite (a-FeOOH) surface. Colloids Surface. 2, 101e117. https://doi.org/10.1016/0166-6622(81)80001-7.

Slomp, C.P., Epping, E.H.G., Helder, W., Raaphorst, W. Van, 1996. A key role for iron-bound phosphorus in authigenic apatite formation in North Atlantic continentalplatform sediments. J. Mar. Res. 54, 1179e1205. https://doi.org/10.1357/0022240963213745.

Smith, V.H., Schindler, D.W., 2009. Eutrophication science: where do we go fromhere? Trends Ecol. Evol. https://doi.org/10.1016/j.tree.2008.11.009.

Smolders, A.J.P., Lamers, L.P.M., Moonen, M., Zwaga, K., Roelofs, J.G.M., 2001. Con-trolling phosphate release from phosphate-enriched sediments by addingvarious iron compounds. Biogeochemistry 54, 219e228. https://doi.org/10.1023/A:1010660401527.

Thibault, P.J., Rancourt, D.G., Evans, R.J., Dutrizac, J.E., 2009. Mineralogical confir-mation of a near-P:Fe ¼ 1:2 limiting stoichiometric ratio in colloidal P-bearingferrihydrite-like hydrous ferric oxide. Geochem. Cosmochim. Acta 73, 364e376.https://doi.org/10.1016/j.gca.2008.10.031.

U.S. EPA (United States Environmental Protection Agency), 1982. Handbook forSampling and Sample Preservation of Water and Wastewater. EPA-600/4-82-029. United States Environmental Protection Agency, Cincinnati, OH.

Weber, K.A., Achenbach, L.A., Coates, J.D., 2006. Microorganisms pumping iron:anaerobic microbial iron oxidation and reduction. Nat. Rev. 4, 752e764. https://doi.org/10.1038/nrmicro1490.

Yerubandi, R.R., Boegman, L., Bolkhari, H., Hiriart-Baer, V., 2016. Physical processesaffecting water quality in Hamilton Harbour. Aquat. Ecosys. Health Manag. 19,114e123. https://doi.org/10.1080/14634988.2016.1165035.

S. Markovic et al. / Environmental Pollution 252 (2019) 697e705 705

1

Supplementary Information

Reduction of industrial iron pollution promotes phosphorus

internal loading in eutrophic Hamilton Harbour, Lake Ontario, Canada

Stefan Markovic1, Anqi Liang1.2, Sue B. Watson3, David Depew2, Arthur Zastepa2, Preksha Surana1,

Julie Vanden Byllaardt4, George Arhonditsis1, Maria Dittrich1*

1Department of Physical and Environmental Sciences, University of Toronto Scarborough,

1065 Military Trail, Scarborough, Toronto, ON, M1C 1A4, Canada

2Environment and Climate Change Canada, Canada Centre for Inland Waters, Burlington, ON, L7S

1A1, Canada

3Department of Biology, University of Waterloo, 200 University Ave. W., Waterloo, ON, N2L 3G1,

Canada

4Hamilton Harbour Remedial Action Plan, Canada Centre for Inland Waters, Burlington, ON, L7S 1A1,

Canada

*Corresponding author: mdittrich@utsc.utoronto.ca

2

2014 05.06 27.08

2015 30.07 10.09

2016 14.06 28.07 31.08 28.09

Table S1. Sampling dates. Note that sediment samples were collected on all sampling dates, but data for

porewater used in this paper were sampled during 2016.

Analysis of dissolved H2S, pH and redox potential

The pH and redox potential (reduction potential measured in mV) were measured with a Unisense

needle electrode relative to Radiometer Analytical reference electrodes. The pH electrodes (Unisense pH-

N, Denmark) were calibrated with commercial pH buffers (pH 4, 7 and 9, HACH). The offset between the

reference electrode and the standard hydrogen electrode redox potential was calibrated prior to

measurement using freshly prepared quinhydrone redox buffers with known redox potentials. The

porewater H2S(g) was measured using a Unisense amperometric ferricyanide microsensor. Since

ferricyanide electrolyte is sensitive to light, sensor was calibrated in a darkened calibration chamber and all

measurements were conducted in the dark. A 3-point calibration in the range of 0 to 100 mol H2S was

conducted using oxygen-free standards prepared from 10 mM Na2S stock solution.

Each pH and redox profile presented hereafter represent the average of the replicate measurements from

the same sediment core. The standard deviation of the repeated pH measurements was <5%, while it was

<10% for the H2S and redox measurements.

Analysis of chemical and physical properties of sediments and porewater

Sediment water content was measured by drying at 105°C for 60 h. Loss on ignition (LOI) was

calculated as the loss of weight during ignition at 550°C for 4 h. The total carbon has been estimated as OC

= LOI550/2.5 (i.e., assuming organic matter composition as CH2O, Burdige 2006). Dry weight (water

3

content), porosity (calculated using moisture loss during drying) and loss on ignition represent the average

of 2 samples from the same layer.

For bulk sediment analysis, freeze-dried sediment samples were compressed in pellets and analyzed on a

Philips PW2404 X-ray fluorescence spectrometer. XRF analysis provides quantitative results for major and

certain minor elements, with detection limits down to 1 ppm. Analytical precision of XRF measurements

was evaluated through repeat measurements averaging ± 1% for major elements and ±1.5% for trace

elements. To ensure measurement accuracy, the international reference standards were analyzed along with

the samples. The samples were run in triplicate, and the relative standard deviation for the different

measured variables was under 10%. The data quality assurance and control were ensured through the use of

standard procedures, including method blanks and blind reference samples.

Estimation of P release from sediment and Fe diffusive fluxes

The SRP diffusive flux was calculated from concentration gradients across the sediment-water interface

(SWI) using Fick’s first law (Berner 1980):

𝐽𝑆𝑅𝑃 = 𝐷𝑠𝑤𝜑

𝜃2

𝜕𝐶

𝜕𝑥

where is the porosity, 𝜕C represent SRP concentration, 𝜕C/𝜕𝑥 represent concentration gradient in the

depth interval over which gradient is calculated (uppermost 1-2cm), and represents the tortuosity, or the

degree of deviation around particles;factor was calculated using the empirical relationship 𝜃2 = 1 −

𝑙𝑛(𝜑2), where φ is porosity (after Boudreau 1997), and Dsw is a molecular diffusion coefficient of PO4 in

the pore water, assuming ambient hypolimnion temperature (values after Boudreau 1997). Similarly,

diffusive fluxes of reduced iron were calculated from their respective measured concentration gradients in

pore water. The depth interval used for calculation of concentration gradient was the top 1-2cm. Diffusion

coefficients for Fe2+ correspond to ambient temperature freshwater estimates after Boudreau (1997).

4

Hypolimnetic SRP accumulation rates

A long-term water-quality monitoring dataset for Hamilton Harbour was used to estimate SRP mass

accumulation rates in the hypolimnion (MOE 1981, 1985&1986; Nriagu et al. 1983; Harlow and Hodson

1988; Hiriart-Baer et al. 2016). Weekly water column profiles of water temperature and depth (binned to 1

m strata) from station HH 1001 were used to define the depth of the hypolimnion for each sampling date in

a given year. The top of the hypolimnion was chosen as 1 m below the inflection point of the thermocline,

and hypolimnetic volumes (m3) were calculated using 1 m depth intervals and 1 m interval bathymetric

data for Lake Ontario (www.ngdc.noaa.gov/mgg/greatlakes/ontario.html). Concentrations of SRP collected

from the hypolimnion (depths of 19 to 22 m) were assumed to be representative of SRP concentrations

throughout the hypolimnion. The mass of SRP in the hypolimnion (SRPH) at the respective sampling time

was calculated as:

𝑆𝑅𝑃𝐻 = 𝑆𝑅𝑃 × 𝑉𝐻

where SRP is the concentration of SRP in the hypolimnion at the time of sampling (mg/m3) and VH is

the volume of the hypolimnion (m3). SRP mass accumulation rates were calculated via linear regression of

SRPH over day of year during the stratified summer period for each year with available records. To

facilitate comparison with the sediment SRP diffusive flux (JSRP) and long-term P recycling rates, obtained

from sediment P depth profiles (described in the main text), hypolimnetic volume SRPH mass accumulation

rates (Hiriart-Baer et al. 2016) were normalized to areal rates by dividing by the average hypolimnetic area

during the stratified period.

5

Figure S1. Sampling site location.

6

Figure S2. Porewater pH and Eh depths profiles at the sediment-water interface a) HH1001 b) HH9031

7

Figure S3. Photographs of vertical core splits from core collected on June 5, 2014: Deep site HH 1001

(left) and shallow site HH 9031 (right).

Depth Si Ti Al Fe Mn P S Stot:Fe Pb Cr Ni Fe/P Fe/Ti Fe/Al

8

Tabl

e S2.

Bulk

sedi

ment

com

posit

ion at station HH 1001 (on dry weight basis).

Table S3. Bulk sediment composition at station HH 9031 (on dry weight basis).

reactive

cm [% d.w.] molar ppm %/%

0-2 23.28 0.39 5.14 6.34 0.31 0.33 1.04 0.49 128 189 46 19.21 16.26 1.23

2-3 23.84 0.40 5.51 6.81 0.33 0.30 0.96 0.46 138 195 49 22.70 17.03 1.24

3-5 24.26 0.39 5.56 7.02 0.33 0.27 0.88 0.38 146 180 52 26.00 18.00 1.26

5-7 24.45 0.39 5.66 7.16 0.34 0.25 0.80 0.49 158 192 54 28.64 18.36 1.27

7-11 25.05 0.40 5.66 7.90 0.38 0.24 0.92 0.57 188 211 60 32.92 19.75 1.40

11-12 23.46 0.39 5.77 9.12 0.40 0.24 1.00 0.54 270 275 71 38.00 23.38 1.58

12-26 22.62 0.38 5.51 10.25 0.41 0.23 1.36 0.71 415 362 94 44.57 26.97 1.86

26-28 22.34 0.38 5.51 11.11 0.45 0.26 1.56 0.77 509 411 103 42.73 29.24 2.02

28-33 21.36 0.37 5.56 13.05 0.47 0.25 1.64 0.75 664 536 112 52.20 35.27 2.35

Depth Si Ti Al Fe Mn P S Stot:Fe

reactive Pb Cr Ni Fe/P Fe/Ti Fe/Al

cm [% d.w.] molar ppm %/%

0-1 22.06 0.39 5.51 4.98 0.17 0.27 0.48 0.21 80 97 39 18.44 12.77 0.90

1-2 22.72 0.40 5.61 5.01 0.18 0.26 0.44 0.25 86 104 41 19.27 12.53 0.89

2-6 23.00 0.40 5.66 5.11 0.20 0.26 0.52 0.34 85 114 42 19.65 12.78 0.90

6-10 23.93 0.40 5.98 5.39 0.22 0.20 0.48 0.37 102 117 46 26.95 13.48 0.90

10-13 23.75 0.40 5.98 5.67 0.22 0.20 0.56 0.34 116 126 49 28.35 14.18 0.95

13-15 23.56 0.40 5.93 5.96 0.22 0.21 0.60 0.36 146 141 55 28.38 14.90 1.01

15-24 23.70 0.41 6.04 6.01 0.23 0.20 0.52 0.32 175 172 60 30.05 14.66 1.00

24-28 23.51 0.40 5.93 6.37 0.24 0.18 0.60 0.36 257 203 73 35.39 15.93 1.07

9

Figure S4. Molar ratios of dissolved Fe and SRP in porewater: (a) HH 1001 and (b) HH 9031. The

lower panels represent expanded shaded blue areas in the upper panels. The porewater molar Fe:SRP ratio

>3.5 (black solid line) indicates a low risk of internal loading (Geurts et al. 2008). Smolders et al. (2001)

similarly suggested a low risk of P release when porewater molar Fe:SRP ratio >1 (gray dashed line).

10

Figure S5. Fe and Al extracted with P binding forms: a) Total Fe (see Table S2) and Fe extracted in

different P extraction pool at HH 1001; b) Total Fe (see Table S3) and Fe extracted in different P extraction

pool at HH 9031; c) Al extracted in different P extraction pool at HH 1001 d) Al extracted in different P

extraction pool at HH 9031

11

Figure S6. Molar ratio between BD-extractible Fe and associated BD-P (HH 1001 - filled black

squares; HH 9031 - empty red circles). At station HH 1001 molar Fe:P ratio is lower than 6.7,

suggesting the supersaturation of surface binding sites of ferrihydrite (Sigg and Stumm 1981) or the

coprecipitation of metastable P-ferrihydrate assemblage, which has a maximum Fe:P ratio of 2:1

(Thibault et al. 2009), or both

12

Figure S7. Upper panel: Molar ratios of Al vs. Fe against the sum of loosely sorbed (NH4Cl-P) and

redox-sensitive (BD-P) fractions, which are the diagenetically most reactive P phases. Kopáček et al

(2005) ratio cutoff is indicated as a red line. Lower panel: Sediment molar ratios of Al (associated with

NaOH-SRP) vs. P[NH4Cl+BD-P] against the sum of loosely sorbed (NH4Cl) and redox-sensitive (BD-

P) fractions. Kopáček et al. (2005) dataset ratio cutoff is indicated as a red line. Note that all ratios in (a)

and (b) represent surface sediments (upper 5 cm).

13

Figure S8. Scatter matrix showing the correlation between P (NaOH-SRP), Al (Al NaOH-SRP) and

Fe (Fe NaOH_SRP) in the NaOH-SRP extraction pool at station HH 1001 (a) and station HH 9031 (b).

14

References

Berner, R.A., 1980. Early Diagenesis: A Theoretical Approach, Princeton Series in Geochemistry.

Boudreau B.P., 1997. Diagenetic Models and Their Implementation. Modelling Transport and

Reactions in Aquatic Sediments, Springer; New York.

Burdige, D.J., 2006. Geochemistry of Marine Sediments, Princeton University Press.

https://doi.org/10.1086/533614

Dittrich, M., Chesnyuk, A., Gudimov, A., McCulloch, J., Quazi, S., Young, J., Winter, J., Stainsby, E.,

Arhonditsis, G., 2013. Phosphorus retention in a mesotrophic lake under transient loading

conditions: Insights from a sediment phosphorus binding form study. Water Res. 47, 1433–1447.

https://doi.org/10.1016/j.watres.2012.12.006

Geurts J.J.M., Smolders A.J.P., Verhoeven J.T.A., Roelofs J.G.M. and Lamers L.P.M., 2008. Sediment

Fe:PO4 ratio as a diagnostic and prognostic tool for the restoration of macrophyte biodiversity in

fen waters, Freshw. Biol. 53, 2101–2116 https://doi.org/10.1111/j.1365-2427.2008.02038.x.

Harlow, H.E., Hodson, P.V., 1988. Chemical Contamination of Hamilton Harbour Canada a Review.

Can Tech Rep Fish Aquat Sci 0, 1–91.

Hiriart-Baer, V.P., Boyd, D., Long, T., Charlton, M.N., Milne, J.E., 2016. Hamilton Harbour over the

last 25 years: Insights from a long-term comprehensive water quality monitoring program. Aquat.

Ecosyst. Heal. Manag. 19, 124–133. https://doi.org/10.1080/14634988.2016.1169686

15

Kopáček, J., Borovec, J., Hejzlar, J., Ulrich, K.U., Norton, S.A., Amirbahman, A., 2005. Aluminum

control of phosphorus sorption by lake sediments. Environ. Sci. Technol. 39, 8784–8789.

https://doi.org/10.1021/es050916b

MOE (Ontario Ministry of the Environment), 1981. Hamilton Harbour Study 1977 Vol 1. Water

Resources Branch, Ontario Ministry of the Environment, Toronto, Ontario.

MOE (Ontario Ministry of the Environment), 1985. Hamilton Harbour Technical Summary and

General Management Options. Ministry of the Environment, Toronto, Ontario, Canada.

MOE (Ontario Ministry of the Environment), 1986. Hamilton Harbour Trace Contaminants 1982-83

Loadings to, and concentrations in the Harbour. Water Resources Branch, Ontario Ministry of the

Environment, Toronto, Ontario.

Sigg L. and Stumm W., 1981. The interaction of anions and weak acids with the hydrous goethite (α-

FeOOH) surface, Colloids Surface. 2, 101–117 https://doi.org/10.1016/0166-6622(81)80001-7.

Smolders A.J.P., Lamers L.P.M., Moonen M., Zwaga K. and Roelofs J.G.M., 2001. Controlling

phosphate release from phosphate-enriched sediments by adding various iron compounds,

Biogeochemistry 54, 219–228 https://doi.org/10.1023/A:1010660401527.

Thibault P.J., Rancourt D.G., Evans R.J. and Dutrizac J.E., 2009. Mineralogical confirmation of a

near-P:Fe = 1:2 limiting stoichiometric ratio in colloidal P-bearing ferrihydrite-like hydrous ferric

oxide, Geochem. Cosmochim. Acta 73, 364–376 https://doi.org/10.1016/j.gca.2008.10.031.

Nriagu, J.O., 1972. Stability of vivianite and ion-pair formation in the system Fe3(PO4)2-H3PO4H3PO4-

H2O. Geochim. Cosmochim. Acta 36, 459–470. https://doi.org/10.1016/0016-7037(72)90035-X

Nriagu, J.O., Wong, H.K.T., Snodgrass, W.J., 1983. Historical Records of Metal Pollution in

Sediments of Toronto and Hamilton Harbours. J. Great Lakes Res. 9, 365–373.

https://doi.org/10.1016/S0380-1330(83)71908-8