Aluminum application to restore water quality in eutrophic lakes: maximizing binding efficiency between aluminum and phosphorus

ABSTRACT

Huser BJ. 2017. Aluminum application to restore water quality in eutrophic lakes: maximizing binding efficiency between aluminum and phosphorus. Lake Reserv Manage. 33:143–151.

Aluminum (Al), typically added to lakes to reduce internal cycling of legacy phosphorus (P) in sediment, was added to the littoral zone of Lake Harriet (Minnesota) to inhibit the uptake of sediment P by buoyancy regulating species of plankton. Analytical results from sediment collected over an 11-year period showed that the added Al (32 g/m²) started to move outside the treatment zone almost immediately after treatment, with <5% remaining within the treatment area after 6 months. Although the original treatment design failed, the application method was, unexpectedly, a success with respect to binding efficiency between Al and P in the sediment. As the Al drifted to deeper areas of the lake, internal P release in non-treated, deeper areas of the lake declined and the binding ratio between Al and Al-bound P decreased, reaching 2.1 (molar) in profundal sediments 10 years after treatment. The increased contact with available (mobile) sediment P increased binding efficiency, resulting in a 163–581% increase of P bound per unit Al compared to previous whole-lake aluminum treatments. The binding efficiency exceeded expectations showing that, in addition to the amount of Al added, treatment location and subsequent translocation of the Al floc can substantially affect binding efficiency and treatment effectiveness.

KEYWORDS:

• Alum
• aluminum sulfate
• internal loading
• mobile phosphorus
• sediment

Aluminum (Al) salts have been used to reduce phosphorus (P) in lakes around the world for more than 4 decades. The success of past treatments has varied greatly, with lasting water quality improvements ranging from just months to more than 40 years (Welch and Cooke 1999, Huser et al. 2016a). Because improvements to surface water quality are generally the goal for most Al treatments, changes in nutrient-related surface water quality parameters (total P, phytoplankton biomass, and water clarity) are often studied after treatment. Few studies have looked at in situ movement of the amorphous Al hydroxide (Al(OH)₃) layer after treatment and how this may affect binding efficiency between Al and P. Because long-term control of P release from lake sediment is needed for long-lasting improvements to water quality, P-binding efficiency and spatial distribution of the applied Al are key drivers for treatment effectiveness.

Of the data that exist for changes to sediment P pools after treatment with Al, substantial variation exists between studies with respect to binding efficiency. Rydin and Welch (1999) showed that P binding by Al in the sediment of Lake Delevan (Wisconsin) resulted in an Al to Al-bound P (Al:P_Al) ratio of 5.7:1 (molar). Rydin et al. (2000) examined 6 aluminum sulfate treated lakes in the state of Washington and found a constant relationship between Al added and P_Al formed, with Al:P_Al averaging 12.6 (range 11.3–13.1). Reitzel et al. (2005) found an Al:P_Al ratio of 7.7 in sediment collected 2 years after Al treatment of a polymictic lake in Denmark, whereas ratios between 5.6 and 14.5 were detected in sediment cores collected from 6 polymictic and dimictic lakes in Minnesota (Huser et al. 2011, Huser 2012). Although not a one-time whole-lake treatment, low volumetric doses (2 mg Al/L) added annually over a 16-year period in a shallow German lake resulted in the lowest reported Al:P_Al ratios in lake sediment in published literature (2.1–2.2; Lewandowski et al. 2003).

Surface complexation–sorption reactions are the primary binding mechanism for phosphate by Al hydroxides. Because Al salts are generally applied to the water column where hydrolysis reactions and Al(OH)₃-floc formation occur, interactions with other components in the water column may limit available binding sites, potentially reducing binding effectiveness. Al forms strong complexes with fluoride, sulfate (Roberson and Hem 1967), and organic matter (Lind and Hem 1975), potentially limiting P sorption. Although the availability of fluoride and sulfate is relatively low in fresh waters in relation to the amount of Al added, dissolved organic matter might be significant in, for example, humic-rich boreal lakes (Hessen and Tranvik 1998). de Vicente (2008b) showed silicates and dissolved organic carbon could compete with phosphate for sorption sites on Al(OH)₃ in lake water, whereas Huser (2012) did not find a significant relationship between organic matter and P-binding efficiency of Al in lake sediment.

One main factor likely responsible for much of the variation in binding of P by Al in lakes once the Al(OH)₃ floc reaches the sediment surface is crystallization (or aging) of the Al floc. Aging of amorphous Al(OH)₃ may reduce binding efficiency, especially in the absence of P. Increases in crystallization of freshly precipitated Al(OH)₃ are accompanied by a decrease in surface area (Sims and Ellis 1983, Berkowitz et al. 2006) as well as a decrease in P adsorption (Berkowitz et al. 2006, de Vicente et al. 2008a). Al translocation, or focusing, may occur after treatment because of the high surface area and low density of the amorphous Al(OH)₃ floc, which Huser (2012) suggested as one of the factors controlling Al accumulation and binding efficiency in 6 aluminum sulfate-treated lakes in Minnesota. Although data are limited for Al focusing in lakes, Huser et al. (2011) showed littoral zone treatment resulted in an 85% reduction of internal P release from sediment in deeper, non-treated areas of the lake. They suggested this reduction was due to translocation of Al(OH)₃ after treatment occurred, but no P_Al data were presented in the study, and no Al data were provided to support this hypothesis.

In this study, temporal and spatial changes in sediment P and Al were examined in a lake treated with Al, with the main hypothesis that the amorphous Al(OH)₃ floc would slowly translocate to deeper areas of the lake. Sediment cores were collected from 3 locations in a transect from littoral to profundal zone in Lake Harriet (Minnesota) over an 11-year period to determine the potential for translocation of the added Al. P fractions and other basic sediment properties were also measured to determine changes to sediment P pools caused by the Al treatment over time. The results were then compared to those from other Al-treated lakes.

Methods

Lake Harriet is a glacial kettle lake with a mean depth of 9 m and an area of 138 ha (Table 1). Aluminum sulfate (32 g/m² of Al, based on sediment P mass in the upper 4 cm of sediment and a 75:1 Al dosing ratio) was applied to the littoral zone (3–8 m water column depth) in May 2001 (Fig. 1) to convert mobile (porewater, loosely sorbed, and reductant soluble) P to P_Al, thereby limiting P availability to buoyancy regulating plankton. The effective Al dose (using the entire lake area) was 11 g/m² of Al. A Willner gravity corer was used to collect sediment from Lake Harriet from April 2001 (pre-treatment) through October 2002 and then again in October 2011. Sediment cores were collected from shallow (4 m), mid-depth (11 m), and deep (24 m) areas of the lake. Sediment transport areas were estimated using average, lake-wide bed slope calculations according to Håkanson and Jansson (1983). Local sediment slope in sediment collection areas was calculated as the distance between each transect point divided by the difference in water column depth between the 2 points.

Table 1. Background information for Lake Harriet.

	Units
Surface area	ha	139
Mean depth	m	8.7
Maximum depth	m	25.0
Watershed area	ha	461
Residence time	y	4.3
Al dose	g/m^2	32
Treatment date		May 2001

Figure 1. Depth contours, coring location, depths, and approximate Al treatment zone for Lake Harriet. All depths are in meters.

Water quality analysis

Water quality samples for total P (TP; 0–2 m surface composite and profile grab samples) from 1991 to 2011 were collected once every two weeks (generally May–Sep growing season) with one sample collected in April (late) and October (early) by the Minneapolis Park and Recreation Board. On each sampling date, water clarity (Secchi disk depth [SD]) and vertical profiles for water temperature, conductivity, pH, and dissolved oxygen were collected every meter at the maximum water column depth of the lake. Internal release rates and loads during the growing season were estimated using in situ volume-weighted TP values and corresponding lake areas (partial net estimate; Nürnberg 2009). Sampling procedures, preservation, and handling times followed Standard Methods (Clesceri et al. 1998) procedures or US Environmental Protection Agency Methods (US EPA 1979).

Sediment analysis

All sediment cores were sliced on site into 1.0 cm sections down to 10 cm sediment depth, and additional samples from the 2011 cores were analyzed down to 30 cm sediment depth at 2 cm intervals. Samples were stored in opaque containers at 4 C until analysis within ∼1 week of collection. Sediment P fractions were determined using a modified version (Hupfer et al. 1995) of the technique for fractionating P forms in sediment by Psenner et al. (1988). Water content was determined by freeze-drying sediment after storage at −70 C for 24 h, and sediment density was calculated according to Håkanson and Jansson (1983) after loss on ignition at 550 C for 2 h. All sediment P analyses were measured as soluble reactive phosphorus (SRP) according to Murphy and Riley (1962). Total sediment Al was determined by digestion with 50% HNO₃ at 120 C in an autoclave and analyzed by inductively couple plasma atomic emission spectrometry at a wavelength of 309.5 nm. Background concentrations for Al and P_Al were estimated using deeper sediment layers unaffected by Al treatment (i.e., >10 cm sediment depth). Background concentrations were then subtracted from Al and P_Al to determine the amount of formed P_Al and excess Al in the sediment due to treatment (Huser 2012). To determine if significant amounts of Al and P_Al from treatment were present, sediments collected after treatment were compared to samples collected before treatment. Al added and P_Al formed from treatment were found in the upper 10 cm of sediment for all cores collected in the study except those from 2011, and thus values given in grams per square meter are based on the sediment depth interval of 0–10 cm or 0–15 cm, respectively.

To account for mobile P released from sediment during sampling periods when deeper areas of the lake were exposed to anaerobic conditions (generally Jun–Sep), the amount of excess P in hypolimnetic water was calculated and assumed to be released from the mobile sediment P pool. Excess P was calculated by subtracting TP concentrations in the epilimnion from those in the hypolimnion for each sampling date. Total mobile sediment P (both released to the hypolimnion and remaining in the sediment) was then used to estimate potential internal P loading rates in the lake (Pilgrim et al. 2007). These values were then compared to release rates determined using increases in hypolimnetic TP during the growing season.

Results

Sediment cores collected from 4 and 11 m water column depth had elevated Al and P_Al concentrations shortly after treatment, although most of the Al and P_Al due to treatment was detected in the treatment zone (Fig. 2). Approximately 21 g/m² Al was detected at 4 m (Fig. 2a) and 11.6 g/m² was found outside the treatment area at 11 m water column depth (Fig. 2b) weeks after application. Sediment bed slopes surrounding the coring locations were 11 and 7% between water column depths of 3 and 9 m and 11 and 24 m, respectively. Mean slopes for these depth ranges for the whole lake were somewhat larger at 18 and 8%. After 6 months, there were no significant amounts of excess Al or P_Al compared to pre-treatment conditions in the treatment application area, and after 11 years, Al from treatment was only found in the deepest area of the lake (Fig. 2c). The transitory nature of the amorphous Al(OH)₃ can be seen as amounts varied temporally along the sampling transect, moving from the initial treatment area to the mid-depth site, and then eventually being detected only in the deeper, profundal area of the lake. Even though the treatment was not designed to lower internal P release from the sediment, internal loading rates declined in Lake Harriet following treatment (Fig. 3). The internal P release rate began to increase in the years following treatment, however, reaching pre-treatment conditions by 2011.

Figure 2. Aluminum (Al) added, Al-bound P (P_Al) formed, and the ratio between added Al and P_Al formed from cores collected at 4 (a), 11 (b) and 24 (c) m water column depth between April 2001 (pre-treatment) and October 2011. Dashed line represents the date of Al treatment.

Figure 3. Internal P release rate (Li) before and after Al treatment, indicated by the dashed line.

Sediment Al and P fractions

Above background (i.e., P bound by the added Al) P_Al increased in the surficial sediment in the littoral treatment area (4 m) after treatment to 0.58 g/m² (Fig. 2a). During the next 6 months, however, P_Al decreased and was not significantly different than natural pre-treatment concentrations of P_Al (i.e., no excess Al due to treatment) by the December 2001 sampling event. At the mid-depth location (11 m), P_Al also increased after treatment to a maximum of 1.7 g/m² in July 2001 (Fig. 2b). Similar to the sediment collected at 4 m water column depth, P_Al at 11 m depth then decreased but remained elevated compared to pre-treatment conditions until 2011 when the final coring was conducted.

Al changed similarly over time at the 4 and 11 m sample collection locations. Al from treatment increased directly after treatment at 4 m water column depth but then declined until December of the treatment year, when total sediment Al (excess plus background) was not significantly greater (P > 0.05) than the pre-treatment concentrations found in the lake (Fig. 2a). Just outside the treatment zone at 11 m water column depth, Al also increased in the first 2 weeks after treatment, although to a lesser extent compared to the treatment zone (Fig. 2b). It continued to increase to 21.7 g/m² during the following 2 weeks until beginning to decline and finally decreasing to just 0.2 g/m² in 2011, which was not significantly elevated compared to total sediment Al before treatment.

P_Al formed from treatment was found in sediment collected from the deepest area of the lake weeks after treatment (Fig. 2c). In contrast to the sediments from the shallow (4 m) and intermediate depth zones (11 m), P_Al continued to increase over the next 7 months, reaching 3.2 g/m². The reduction of mobile P (Fig. 4), caused either by conversion to P_Al in the deeper sediment accumulation zone of the lake and/or conversion of mobile P to P_Al in shallower areas of the lake that was then transported to deeper areas, resulted in a sharp decline in internal P release from sediment (Fig. 3). Internal P release rate (Li), based on increases in hypolimnetic TP, ranged from 7.8 to 10 mg/m²/d prior to treatment (1991–2000), decreased to 0 mg/m²/d during the treatment year, and then began to increase slowly during the following decade, reaching 8.2 mg/m²/d in 2014. Mobile P concentration in 2011 was 90% of that found in the 2001 pre-treatment core (1.37 vs. 1.52 mg/g; Fig. 4), and, using the relationship developed by Pilgrim et al. (2007) relating mobile P to Li, the estimated potential Li was calculated to be 8.1 mg/m²/d. This value was close to the average Li of 7.2 mg/m²/d calculated in situ using hypolimnetic TP for 2011, the final year of coring.

Figure 4. Mobile and Al-bound P (P_Al) concentrations (dry weight) at the deep coring location (24 m) from April 2001 to October 2011. Dashed line indicates date of Al treatment.

Binding efficiency between added Al and P_Al formed

The binding ratios between Al added and P_Al formed in the sediment varied substantially over time at both the 4 and 11 m core collection locations (Fig. 2). Just after treatment, the ratio between added sediment Al and P_Al formed (Al:P_Al) varied between 27 and 40 at 11 and 4 m water column depths, respectively. As the binding sites on the recently formed amorphous Al(OH)₃ continued to fill, the Al:P_Al ratio (i.e., only the above background Al added from treatment and the P bound to it) started to decrease through July of the treatment year (Fig. 2a and b). After July, significant amounts of added Al and P_Al formed from treatment were not detected in the treatment zone. The ratio of added Al to P_Al was initially low (3) at the deep sampling site (24 m water column depth), and it continued to decline until reaching a minimum of 2.1 (molar basis) in 2011. This value was lower than the predicted binding ratios of 5.9 (32 g/m² for the treatment area alone) or 4.4 (effective Al dose of 11 g/m² over total lake area; see methods) between Al added and P_Al formed using the model for estimating P binding efficiency of Al in Al-treated lakes by Huser (2012). Peaks of Al and P_Al were detected at 6.5 cm sediment depth in the core collected from 24 m water column depth 10 years after treatment (Fig. 5), corresponding well to a previous estimation of sediment rate in Lake Harriet (0.6 cm/yr; Engstrom and Swain 1997).

Figure 5. Concentrations (dry weight) of aluminum (Al) and Al-bound P (P_Al) by depth at 24 m water column depth in October 2011. Vertical dashed and solid lines denote background concentrations for Al and P_Al, respectively.

Discussion

The Al applied to Lake Harriet in spring 2001 covered sediments with water column depths of ∼3–8 m. The treatment was designed to decrease algal production in littoral areas by limiting the amount of P available in the sediment to plankton (mainly cyanobacteria) able to use this P to become buoyant and accumulate within macrophyte beds in the littoral region. Analysis of the sediment data showed that the Al added and P_Al formed after treatment began to translocate toward the center within 2 weeks after Al application (Fig. 2), reflecting the transitory (i.e., transport bottom) nature of the sediments within the Al treatment zone in Lake Harriet. Thus, although Al was applied to sediment in transport bottom areas of the lake, it eventually had an effect on sediment P in deeper lake areas by decreasing the amount of mobile sediment P via conversion to P_Al (Fig. 4).

Changes to sediment and water chemistry after treatment

The reduction of mobile sediment P in deeper, pelagic areas of the lake (Fig. 4) led to a mean reduction of internal P loading from the sediment of 85% during the first 2 years after treatment. Internal loading has since returned (Fig. 3), but this would be expected because of the low Al dose relative to the amount of mobile P in the sediment (Huser et al. 2011). Because of the treatment design, substantial increases in water quality in Lake Harriet were not expected; however, SD increased and TP and chlorophyll a concentrations declined substantially after treatment, all of which were primarily driven by the decrease in internal P loading after treatment (Huser et al. 2016b). Because of the low effective Al dose, however, any improvements in water quality would be expected to be short lived. Based on a 50% reduction of mean growing season TP concentration in the epilimnion, the effects of Al treatment on TP reduction lasted 4 years, close to the 5-year longevity Huser et al. (2016a) predicted with their model based on Al dose, watershed area to lake area ratio, and the Osgood Index (mean depth/square root [lake area]).

The most significant (and unexpected) result from this study was the high binding efficiency detected between Al and P in the sediment of Lake Harriet (Al:P_Al of 2.1 molar). Al:P_Al binding ratios (molar) after whole-lake treatments with Al have generally ranged from ∼5 to 11 (Rydin et al. 2000, Reitzel et al. 2005, Huser et al. 2011, Jensen et al. 2015) and in Minnesota from 5 to 15 (Huser 2012). The amount of P bound per unit Al in Lake Harriet represents an increase of 163 to 581% compared to previous studies of sediment from lakes treated with one-time, whole-lake doses of Al. The only other study with a binding efficiency similar to that detected in Lake Harriet was conducted by Lewendowski et al. (2003), who found an Al:P_Al binding ratio of 2.1 (molar). Theirs was not a one-time, whole-lake treatment, however, but instead the German Süsser See was treated with small Al doses annually (2 mg/L) to strip P from the water column. This dose is relatively low, especially in comparison to more recent treatment doses that have typically ranged from 12 to 30 mg Al/L (Welch and Cooke 1999, Huser et al. 2016a). Although the areal dose was moderate in Lake Harriet (32 g/m² based on treatment area), it was low considering the entire lake area and volume (11 g/m²).

The binding ratio in Lake Harriet sediment is also low considering the time between treatment and core collection. Huser and Pilgrim (2014) showed that the Al:P_Al ratio in the upstream Lake Calhoun, also treated with Al in 2001, averaged 80 across the treated area of the lake 1.5 years after treatment. The lake-wide average for Al:P_Al in lake Harriet sediment dropped from 45 ∼1 month after treatment (the original dosing ratio of Al to mobile sediment P was 75:1) to 3.7 only 7 months after treatment. This large difference in binding efficiency has likely decreased as available binding sites on the Al(OH)₃ floc in Lake Calhoun have become saturated. Nonetheless, the greater Al doses now being used in response to improved dosing methods based on mobile sediment P may result in lower binding efficiency, especially in lakes with steep slopes prone to translocation of the added Al (Huser 2012, Egemose et al. 2013).

To achieve greater binding efficiency with respect to Al treatment in lakes, mobile sediment P must be available in adequate amounts relative to the amount of Al added so that the availability of P does not limit binding. As the amount of sorbate decreases relative to sorbent, chemical equilibrium (i.e., Le Chatelier's principal) dictates that binding efficiency will decrease. This is the basic principal behind collision theory (Trautz 1916) that, for example, the less P there is relative to Al, the lower the chances are that Al will come into contact with available P. In lake Harriet, the low relative dose (11 g/m²), as well as the physical movement of the Al floc to different sediment areas of the lake, both increased the chance for collision between Al and P and was likely the reason for the high binding efficiency of treatment. The most efficient type of binding between Al and P is the formation of Al phosphate (AlPO₄), giving a 1:1 binding ratio, but this reaction is only favored when PO₄ is >10 mg/L in water (Jenkins et al. 1971). At PO₄ concentrations <10 mg/L, competition from hydroxide formation limits direct binding of P by Al due to the higher metal to P ratio (Stumm and Morgan 1996). Porewater P concentrations rarely reach concentrations of 10 mg/L or greater except in extremely eutrophic (i.e., hypereutrophic) systems (Enell and Löfgren 1988, Lewandowski et al. 2002), and thus adsorption to Al hydroxides is the primary binding mechanism for Al in most lakes treated with Al salts. The similar binding efficiency reported in this study and that by Lewandowski et al. (2003) might therefore be close to a maximum binding efficiency for Al treatment of lakes where most P binding occurs via sorption by Al(OH)₃. More research is needed, however, to determine the maximum binding efficiency achievable under these types of treatment conditions.

Recent studies have shown that that efficiency of mobile P conversion to P_Al decreased as the ratio of metal to mobile P increased in sediment (Huser and Pilgrim 2014, James and Bischoff 2015). Thus, if Al doses required to inactivate mobile sediment P are added as one-time only treatments, there is a greater risk that the binding efficiency will be lower and limit the effectiveness of treatment. In addition, binding efficiency of amorphous Al(OH)₃ decreases as it ages and crystalizes (Berkowitz et al. 2006), especially in the absence of P (de Vicente et al. 2008b). A lower chance for collision between the added Al and sediment P will increase crystallization of the Al and lead to lower binding efficiency. Thus, the minimum amount of Al needed to have a positive effect on water quality should be used during a single treatment. Models predicting the short-term decrease in sediment P release after treatment (Huser and Pilgrim 2014) can be combined with methods to estimate Al binding efficiency (Huser 2012) to maximize long-term binding efficiency between Al and P while still improving water quality in the short-term.

Translocation of Al, treatment efficiency, and targeted P pools

Because the sediment layers affected by Al treatment generally have similar or slightly lower densities compared to untreated sediment from the same area (Huser 2012), the processes leading to movement or translocation should affect treated and non-treated sediment similarly under the same conditions. Thus, if Al is added to the sediment transport zones of a lake, it should move along the lake bottom along with low density, organic, and P-rich sediment until it reaches the accumulation zone (Håkanson and Jansson 1983). Given that sediment P in accumulation areas originates from both vertical sedimentation and horizontal translocation, it should not matter whether the conversion of mobile P to P_Al occurs in the originating areas or in the destination area (i.e., the accumulation zone). The primary goal of an Al treatment is to reduce a known amount of mobile P in the sediment and return the in-lake cycling of P to a pre-impact state. Whether this P comes from recent inputs or the historical loads contained in the lake sediment is of no concern; it is the total mass of mobile P remaining that contributes to release of P from the sediment (Pilgrim et al. 2007). Thus, to improve the chances for collision and binding between Al and P (either in sediment or P-rich hypolimnetic bottom water), sediment bottoms in littoral lake areas can be treated because both the organic-rich sediment and Al(OH)₃ floc will be transported to deeper areas of the lake. These conditions likely led to a more than doubling of binding efficiency (Al:P_Al = 2.1, measured) compared to that predicted (Al:P_Al = 4.4, estimated) using efficiency equations developed by Huser (2012) based on Al dose and sediment bed slope. Improving binding efficiency in this manner is dependent on lake morphology and bed slope, but in lake bottoms with direct transport paths from littoral to profundal lake areas where anoxic conditions occur, treatment of sediment in transport areas will improve the chance of collision between Al and P and limit crystallization of the added Al before it comes into contact with available P in sediment.

Additional considerations

Using the littoral zone for Al treatment of lakes has additional benefits. If treatment areas are expanded to include transport lake bottoms, the alkalinity contained in the water mass of the additional treated areas can be used to limit the amount of buffer needed in cases where pH depression is a concern. Concurrently, the addition of Al to a greater sediment area will increase the chances for contact with P and increase binding efficiency. For example, if the treatment applied to Lake Harriet is compared to the next most efficient treatments in the same region (Rydin and Welch 1999, Huser et al. 2011), more than 2.5 times the amount of P was bound per unit Al. Adding together both the cost savings achieved with the decreased need for a buffer during treatment and the improved P binding efficiency of Al suggests that applying Al not only to the areas with greatest mobile P in the sediment, but also to those that contribute mobile sediment P to these areas, could substantially decrease the cost of such treatments. Note, however, that the results of this study are a limited assessment of treatment not originally designed to maximize binding efficiency, and thus further assessments, including mechanisms related to sediment transport, should be conducted.

Acknowledgments

The author is grateful to Mike Perniel, Rob Brown, and other staff at the Minneapolis Park and Recreation Board for providing sampling support, lake water quality data, and GIS information for Lake Harriet; Kevin Menken at Barr Engineering for collecting and analyzing sediment collected in 2011; and 3 reviewers for their comments that helped improve the manuscript.

Funding

This study was partially funded by FORMAS Swedish Research Council and USEPA section 319 grants.