JEQ Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 9 August 2005
Published in J Environ Qual 34:1610-1619 (2005)
DOI: 10.2134/jeq2004.0324
© 2005 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Morrison, M. A.
Right arrow Articles by Benoit, G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Morrison, M. A.
Right arrow Articles by Benoit, G.
Agricola
Right arrow Articles by Morrison, M. A.
Right arrow Articles by Benoit, G.
Related Collections
Right arrow Surface Water Quality
Right arrow Colloids
Right arrow Geochemical Processes
Right arrow Runoff
Right arrow Heavy Metals

TECHNICAL REPORTS

Surface Water Quality

Temporal Variability in Physical Speciation of Metals during a Winter Rain-on-Snow Event

Matthew A. Morrisona,b,* and Gaboury Benoita

a Yale School of the Environment, Yale Univ., New Haven, CT 06520
b Current address: National Risk Management Research Lab., USEPA, Cincinnati, OH 45268

* Corresponding author (morrison.matthew{at}epa.gov)

Received for publication August 19, 2004.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Particulate matter in urban rivers transports a significant fraction of pollutants, changes rapidly during storm events, and is difficult to characterize. In this study, the physical speciation of trace metals and organic C in an urban river and upstream headwaters site in Torrington, CT, were measured during a winter rain-on-snow event. In addition, a selective fractionation scheme, using membrane and tangential-flow ultrafiltration methods to separate suspended particulate matter into sand, silt, clay, and colloid fractions, was evaluated based on the appropriateness of the chosen size categories. During peak runoff at the urban river site, total-recoverable concentrations of the metals Cu and Pb increased 6- and 13-fold to 16.9 and 9.5 µg L–1, respectively, compared with baseflow concentrations. Concentrations of Cu and Pb reached only 0.9 and 0.86 µg L–1 at the headwaters site. For the measured storm event, the majority of metals were transported by the urban river in association with coarse silt (20–80 µm particle diam.) during peak runoff. During peak runoff at the urban site, organic C associated with the large colloid fraction (0.1–1.0 µm) increased from 5% (at baseflow) to 54% of the total C in transport, whereas dissolved organic C and that associated with smaller colloids decreased from 91.5% (at baseflow) to 41% of the total. Other elements that were monitored as part of the study were Na, K, Ca, Mg, Fe, Mn, Al, Cd, Cl–, NO3–, and SO2–4. The chosen fractionation scheme was useful to characterize pollutant transport during this event, but further testing should be undertaken to determine the most appropriate size range categories, and to ensure that the sizes measured are comparable to those used in other studies.

Abbreviations: BMP, best management practice • MWCO, molecular weight cut-off • NTU, nephelometric turbidity units • OC, organic carbon • TFU, tangential-flow ultrafiltration


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
THE CHARACTERIZATION of stormwater particulate matter is critical to understanding both toxicity and transport pathways for trace metals and other pollutants. During storm events in urban watersheds, a complex mixture of suspended sediment and particulate organic matter enters the water column of rivers and streams through runoff pathways, overbank flows, streambank erosion, and channel sediment resuspension. One of the major transport pathways for pollutants in freshwater is through adsorption or complexation to the organic and reactive mineral surfaces of suspended sediment and colloids (Grout et al., 1999; Ross and Sherrell, 1999; Sigg et al., 2000; Sinclair et al., 1989). Physical and chemical speciation of pollutants is important to the toxicity of stormwater, because complexation by organic matter decreases the toxicity of most metals (Novotny and Witte, 1997; Pitt, 1995). Characterization of the size distribution of particles to which stormwater pollutants are attached is likewise important to the performance of stormwater best management practices (BMPs), because the effectiveness of BMPs is expected to increase significantly with particle size (Liebens, 2001; Newberry and Yonge, 1996). Sand particles are effectively removed by many BMPs, but colloidal particles are likely to be attenuated only by BMPs that include infiltration or coagulation and filtration. Particles with sizes between those of sand and colloids are attenuated to varying degrees by stormwater BMPs (such as wet detention basins), but research on this topic is limited (Davies, 1995; Urbonas, 2001). While the composition of urban stormwater is likely to vary significantly with local land use, seasonal changes in terrestrial inputs, and geographical location, it is critical for researchers to establish a framework through which the analysis of stormwater particulate matter can be accomplished.

Existing research employs conflicting categories for particle-size classes and typically presents an incomplete evaluation of the entire particle-size spectrum, which ranges from sand to organic colloids (Douglas et al., 1999). Many studies continue to separate dissolved and particulate fractions using a 0.45-µm membrane filter, a practice that is questionable in light of the artifacts associated with the technique (Horowitz et al., 1996; Morrison and Benoit, 2001, 2004) and the fact that freshwater colloids have an upper size limit of at least 1.0 µm diameter (Gustafsson and Gschwend, 1997). Most studies focus on either suspended sediment or colloids, typically lacking detail on one or the other and neglecting the interface between sediment and colloidal particles (Characklis and Wiesner, 1997; Grout et al., 1999; Tanizaki et al., 1992). Aquatic particulate matter is isolated and characterized using many techniques, including membrane filtration, sieving, centrifugation, and tangential-flow ultrafiltration (Buffle and Leppard, 1995; Buffle et al., 1992; Lead et al., 1997). The over-arching need in the separation of aquatic particles is to use existing methods in a way that isolates meaningful particle classes based on particle characteristics, rather than to simply use conventional methods or to allow the limitations of a given technique to dictate the size range of isolated particle-size classes (e.g., the lower size limit of particles isolated by centrifugation).

This paper presents the results of intensive stormwater sampling and particle-size distribution analysis of trace metals and organic carbon (OC) for a rain-on-snow event at an urban river site and an upstream headwaters site. The goals of the research were twofold: (i) to present changes in the concentrations and physical speciation (i.e., particle-size class) of trace metals and OC in an urban river and upstream headwaters site during a winter rain-on-snow event; and (ii) to evaluate a method for collecting detailed, meaningful data on particle sizes for trace metals (Fe, Al, Mn, Cu, Pb, Cd) and OC. Evaluation of the fractionation method was based on two criteria: (i) Do the fractions contain significant concentrations of metals and OC? (ii) Do the fractions provide useful information concerning physical speciation by responding independently during storm runoff?


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Sampling Locations
Two sampling sites on the Naugatuck River, one above and one just below the city of Torrington, CT (population ~34500; Litchfield County) were sampled during a rain-on-snow event in March of 2001. The upper site, located on the East Branch Naugatuck River (41°50'3.8'' N lat; 73°07'14.8'' W long), receives drainage from 24.0 km2 (first-order stream) (Nosal, 1997) of largely undeveloped land. The lower site, on the Naugatuck River (41°46'38.1'' N lat; 73°07'7.4'' W long), receives drainage from 147 km2 (second-order stream) including the city of Torrington (downtown is located 1.9 km upstream of the sampling location), its associated industries, and the town's wastewater treatment plant (located approximately 0.75 km upstream of the sampling location). Precipitation data obtained from the National Climatic Data Center (NCDC) for the Danbury, CT, and Hartford, CT, weather stations (unedited, hourly observations) were used to estimate rainfall at the Torrington sampling locations. The Naugatuck River is of interest because it has historically high levels of metal pollution from industrial sources, and because there is a significant effort underway to restore fish habitat within the river corridor and to improve access to headwaters stream reaches.

Field Methods
In situ sampling of pH, conductivity, temperature, depth, and turbidity was performed by automated water-quality dataloggers (YSI 6920; YSI, Yellow Springs, OH), which were deployed at each sampling location. Dataloggers were set to record all of the above parameters at 15-min intervals and were tested and re-calibrated as necessary on a weekly basis. Automated water samplers (Compact Sampler; ISCO, Lincoln, NE) were used to collect fixed-interval stormwater samples at each of the field sites. Plastic tubing—Teflon, 9.5 mm (0.375 inch)—was used for sampling water from the river, silicone pump tubing was used within the peristaltic pump, and 1.0-L bottles (ISCO; polypropylene) were used to collect samples. Inlet tubing, pump tubing, and sample bottles were cleaned in accordance with clean techniques as described below. Inlet tubing was rinsed with sample (two times, automated) before the collection of each individual sample. The automated sampler was set to collect hourly composite samples with a subsample collection frequency of four samples per hour; each 1.0-L autosampler bottle contained a composite of four 200- to 250-mL subsamples representing 1 h of streamflow. The capacity of the ISCO Compact Sampler is 24 bottles. Samples were collected from a fixed location during the entire sequence, and therefore may not be representative of the entire stream cross-section. During sampling, ice was placed inside the autosampler to preserve sample integrity before processing; samples were removed from the automated sampler for laboratory analysis typically every 10 to 15 h, but at least once every 24 h.

Determination of Sampling Frequency
Water-quality dataloggers were deployed at each sampling site before collection to establish the appropriate sample-collection frequency. During the first hour following the onset of runoff at the Torrington location, turbidity increased by nearly 300 NTU and conductivity increased by 220 µS cm–1. The sampling theorem states that the sampling frequency must be at least twice that of the highest signal frequency containing chemical information to ensure accurate data acquisition (Strobel and Heineman, 1989). While the context of that remark concerns instrumental analytical techniques, the application of the sampling theorem to environmental sampling is appropriate. Environmental signals, such as diurnal temperature change and sediment-runoff pulses, occur at often repeatable frequencies, and environmental sampling must occur at a frequency that is high enough to accurately represent the effects of those events on water quality. The period of the conductivity peak (baseline to baseline) was 2.5 h, which results in a minimum sampling interval of 1.25 h. The turbidity signal contains isolated peaks of higher frequency, but the sampling interval necessary to characterize these isolated signals, which are based on a single turbidity measurement and may be spurious, is impractical. For the research presented in this paper, automated samplers were set to collect 250-mL subsamples of water every 15 min; four subsamples were combined into hourly 1.0-L composite samples within the autosampler. While the use of hourly composite samples may slightly distort the response hydrograph for trace metals, the inclusion of higher-frequency subsamples within hourly composite samples is less likely to miss high-frequency chemical information.

The identical timing and sample-collection frequency that was used for the Torrington location was also used for the East Branch headwaters sampling location. Conductivity and turbidity data shown in Fig. 1b , however, demonstrate that a lower-frequency sampling regime would have adequately characterized stream chemistry for this site. For the East Branch, the 100 µS cm–1 change in conductivity from baseflow to stormflow peak occurred over 6 h (i.e., 12-h period), which, using the same criteria as above, could have been adequately characterized using a 6-h sample collection interval.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 1. Changes in water-quality parameters and depth for the (a) urban and (b) headwaters locations. Complete datasets for YSI 6920 dataloggers include depth, conductivity, temperature, pH, and turbidity; pH is excluded for simplicity of presentation. Precipitation is presented along the top of the graph (maximum value is 6.4 mm h–1) and was interpolated (both timing and intensity) from Hartford, CT, and Danbury, CT, weather stations.

 
Filtration Methods
Laboratory filtration and analyses were performed as described previously (Benoit, 1994; Morrison and Benoit, 2001), and will not be repeated here, except for variations in detail. Observed changes in flow rate during filtration were used to indicate membrane clogging, and membrane filtrations were ended before, or immediately following, a noticeable decrease in filtrate flow rate. All samples were filtered using a 47-mm membrane with a 0.45-µm cutoff (Durapore; Millipore, Billerica, MA) to provide a conventionally defined dissolved phase concentration that is comparable to the work of other researchers. In addition, a selected number of samples were filtered to remove particles in six size categories, using the specified membrane and filtration method as outlined in Table 1. Tangential-flow ultrafiltrations (TFU) were performed on approximately 125 mL of sample, using a 0.1-µm cutoff hollow-fiber cartridge (MiniKros; Spectrum, Rancho Dominguez, CA) and a 3000-MWCO (molecular-weight cutoff) hollow-fiber cartridge (Amicon; Millipore, Billerica, MA) (Morrison and Benoit, 2004). Effective final concentration factors for the tangential-flow ultrafiltrations were between 2.5 and 4, which may lead to an overestimate of the colloidal fraction (Guo et al., 2000). All filtrations were performed using the whole water sample, except for the 1.0-µm polycarbonate membrane and 0.1-µm and 3000-MWCO tangential-flow ultrafiltrations, which were pre-filtered through a 20-µm nylon membrane filter. Percent composition calculations, as for the size distribution data, were performed by difference, using the unfiltered water sample concentration as the total and then using the successive filtrate concentrations to determine the concentration in each particle size category.


View this table:
[in this window]
[in a new window]
  		Table 1. Particle classes, size ranges, and fractionation methods used for detailed physical speciation of selected water samples. All membrane filters used were 47 mm in diameter. Tangential-flow ultrafiltration (TFU) cartridges were both hollow-fiber bundles.{dagger}

 
Laboratory Analyses
Major metal analyses were performed via inductively coupled plasma atomic emission spectrometry (Optima 3000; PerkinElmer, Boston, MA), which provides a method detection limit of 5 to 15 µg L–1 for Al, Fe, and Mn, and 0.5 to 1.0 mg L–1 for Na, Mg, K, and Ca. Samples for metals analysis were heated for 1 h to 95°C [following acidification to pH 2 with concentrated HNO3 (Seastar Chemicals, Sidney, BC)] in Teflon beakers on a hotplate to estimate total-recoverable metals in water and suspended particulate matter. Graphite-furnace atomic absorption spectrophotometry (HGA 600; PerkinElmer, Boston, MA), following evaporative preconcentration (20 times, in PFA beakers) was used for Cu, Pb, and Cd analyses (method detection limits of 36, 24, and 6 ng L–1, respectively). When suspended sediment concentrations were high, particularly during peak storm events, these techniques did not completely digest particulate matter, so that reported "total concentrations" are likely to be less than the actual values. Total OC analyses were performed via high-temperature catalytic oxidation (TOC 5000-C; Shimadzu, Columbia, MD) with a quantitation limit of 0.45 mg L–1, and anion analyses (Cl, NO3, SO4) via ion chromatography (Dionex, Sunnyvale, CA).


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Hydrology and Water Quality Parameters
Water quality dataloggers provide important information on changes in background chemistry and physical conditions (e.g., turbidity, pH, conductivity, depth, and temperature) during storm events (Fig. 1). Figure 1a shows changes in water-quality parameters and depth for the Naugatuck River below Torrington, CT, while Fig. 1b shows changes in water-quality parameters and depth for the East Branch location, which is a headwaters tributary of the Naugatuck River above Torrington. Values for pH were omitted from Fig. 1 for clarity, and because the values did not change significantly during the storm event (urban site range: 6.6–6.9; East Branch site range: 6.55–6.7). Precipitation data are shown in Fig. 1, but a values axis is not included for simplicity of presentation. The maximum hourly precipitation for the storm was 6.4 mm, and the cumulative total for the first part of the storm (13 Mar. 2001) was 17.1 mm. The rapid response of streamflow depth and water quality parameters to rainfall at the Torrington location (Fig. 1a) is typical of urban runoff (change in depth to the first storm peak = 0.2 m). The presence of road salt in runoff is the obvious cause of increased conductivity, and the occurrence of rain on snow (with preservation of the majority of the snowpack throughout the storm event) combined with diurnal cycling causes the observed changes in stream temperature. Data are missing from the East Branch site (Fig. 1b) due to an instrument malfunction. The peak in conductivity at the East Branch location is also indicative of the presence of road salt, but the delayed response of depth (change in depth to peak = 0.09 m) and conductivity to rainfall suggests that event precipitation enters the East Branch stream mainly through subsurface flow or snowmelt, not runoff pathways. This assertion is supported by the lack of a significant change in turbidity for the East Branch.

Soluble Base Cations and Anions
Figures 2 and 3 depict changes in base cations (Na+, K+, Mg2+, Ca2+) and acid anions (Cl–, NO–3, SO2–4) during the storm event for the Torrington location on the Naugatuck River and the East Branch headwaters location, respectively. Each graph contains a trace of conductivity for direct comparison. Data in these charts is taken from the 0.45-µm filtrate fraction, and all of these constituents are expected to be truly-dissolved ions or soluble complexes. As expected for a winter storm in the Northeast USA where road salt is applied, Na+ and Cl– concentrations increased dramatically during storm runoff at the Torrington location (Fig. 2), tracking changes in overall conductivity. The K+, Mg2+, and Ca2+ concentrations (Fig. 2b) at Torrington were all diluted significantly during the initial phase of runoff by the input of fresh water, but increased to pre-event concentrations by the end of the sampling period. In contrast, the East Branch location (Fig. 3) data showed no dilution of base cations during the storm event. As at the Torrington site, Na+ and Cl– concentration increases were responsible for changes in overall conductivity at East Branch, but it is clear that the increase in conductivity was not due to direct, high-volume runoff. The NO–3 concentration was largely unchanged during the storm event at either location, while SO2–4 concentration (Fig. 2a) was diluted modestly during the initial phase of runoff only at the Torrington location.



View larger version (41K):
[in this window]
[in a new window]
  		Fig. 2. Changes in (a) anions and (b) base cation concentrations, compared with conductivity, during the March storm event at the Torrington location (urban). Data shown for 0.45-µm filtrate.

 


View larger version (38K):
[in this window]
[in a new window]
  		Fig. 3. Changes in (a) anions and (b) base cation concentrations, compared with conductivity, during the March storm event at the East Branch location (headwaters). Data shown for 0.45-µm filtrate.

 
Total Recoverable Metals
Figure 4 shows the changes in total recoverable metal (Fe, Al, Mn, Pb, Cu, and Cd) concentrations with time over the course of the March 2001 storm event for the Naugatuck River below Torrington. Hourly composite data (Fig. 4) show a dramatic increase in total recoverable metal concentrations during stormwater runoff, which closely follow changes in water depth (shown) and turbidity (compare with Fig. 1a). Total Pb, Cu, and Cd (Fig. 4b) exhibit dramatic increases in concentration over the course of the storm, which indicates their presence as adsorbed constituents on suspended particles (Harrison and Wilson, 1985; Hewitt and Rashed, 1992; Wang et al., 1997). Total Pb increased 12.9 times from a baseflow concentration of 0.74 to 9.53 µg L–1 during peak stormflow. Total Cu increased 5.9 times from a baseflow concentration of 2.9 to 16.9 µg L–1 during peak stormflow, and total Cd increased 3.4 times from a baseflow concentration of 0.14 to 0.48 µg L–1 during peak stormflow.



View larger version (35K):
[in this window]
[in a new window]
  		Fig. 4. Changes in total recoverable (a) trace and (b) heavy metal concentration compared with changes in turbidity for the March storm on the Naugatuck River below Torrington, CT.

 
Total recoverable metal concentrations at the East Branch location (Fig. 5) also increased in response to discharge, as shown in relation to water depth. The magnitude of change in concentration for trace metals was much smaller (ca. 10 times, as reflected in the y axis scales) than the increase in total recoverable metals concentration for the Naugatuck River urban location. The maximum concentrations for Pb, Cu, and Cd at the East Branch location were 0.86, 0.94, and 0.1 µg L–1, respectively. Total Fe, Al, and Mn concentrations (Fig. 5a) did not increase substantially. The increase in total recoverable metals for the East Branch site may have been due solely to the resuspension of bed sediments in response to increased discharge.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 5. Changes in total recoverable (a) trace and (b) heavy metal concentrations compared with turbidity, for the East Branch above Torrington, CT.

 
Metals in the 0.45-µm Filtrate
Figure 6 shows changes in depth and 0.45-µm filtrate metal concentrations with time over the course of the storm event for the Naugatuck River below Torrington. Trace metals in the <0.45-µm fraction are often defined conventionally as the dissolved phase, but some metals may be associated with colloidal particles in this fraction (Horowitz et al., 1996; Morrison and Benoit, 2004). Iron and Mn (Fig. 6a) exhibit significant differences when compared with the event chronology for total metals (Fig. 4a), which closely mirrors changes in depth. Iron in the 0.45-µm filtrate decreased in response to stormflow runoff peaks, but increased rapidly to pre-event concentrations thereafter. Manganese decreased slightly in response to the initial phase of runoff, but otherwise shows small peaks that appear to correspond roughly with increases in conductivity (not shown). Copper, Cd, and Pb concentrations for the Naugatuck River (Fig. 6b) generally tracked each other and appear to coincide with changes in Mn, except that the overall trend in their concentration during the storm event is downward. The complexity of these signals (i.e., changes in <0.45-µm metal concentration) cannot adequately be explained by the limited amount of data collected for this study. Al concentration in the <0.45-µm fraction was close to the detection limit under most conditions, including stormflow, and is not depicted in Fig. 6a.



View larger version (31K):
[in this window]
[in a new window]
  		Fig. 6. Changes in (a) trace and (b) heavy metals found in the <0.45-µm filtrate compared with depth for the March storm on the Naugatuck River below Torrington, CT.

 
The observed change in the concentration of <0.45-µm fraction trace metals at the East Branch location was minimal, and is thus not presented in graphical form. Iron concentration did decrease slightly and Mn concentration increased slightly in response to increased flow at the East Branch location (baseflow concentrations were 61 ± 6 and 22 ± 2 µg L–1 for Fe and Mn, respectively). Changes in Pb, Cu, and Cd were likewise minimal (baseflow concentrations of 0.019 ± 0.007, 0.26 ± 0.03, and 0.005 ± 0.003 µg L–1 for Pb, Cu, and Cd, respectively).

Particle-Size Associations for Metals
Figure 7 (a, b, and c) depicts changes in the percent of total recoverable concentration for selected particle-size classes for Fe, Mn, and Al (respectively) for the Naugatuck River below Torrington. An overlay of total recoverable metal concentration in each case shows the storm event samples that were chosen for size-distribution analysis (highlighted by square boxes). The first sample in each time series represents pre-event baseflow conditions. Figure 7 shows what appears to be a large influx of sand (>80 µm) and coarse silt (20–80 µm) particles to the river during the measured storm event. There was a small increase in fine silt (1–20 µm) particles, and very little other change in the particle-size distributions of Fe, Mn, and Al. The Fe graph shows the large influx of 20- to 80-µm size particles; sand particles (>80 µm), which account for nearly half of the baseflow concentration of sediment-associated Fe, increased during storm runoff, but were <5% of the total during peak runoff. Fine silt and clay-associated Fe (1.0–20 µm) was not detected under baseflow conditions, but was a measurable fraction of Fe concentration in stormflow samples (post peak).



View larger version (32K):
[in this window]
[in a new window]
  		Fig. 7. Changes in the size distribution of particles associated with selected trace metals for the Naugatuck River, below Torrington, CT. Total recoverable metal concentration is shown for comparison, and boxes indicate samples chosen for detailed size distribution analysis.

 
Figure 7b shows changes in the size distribution of Mn-associated particles over the course of the storm event for the Naugatuck River below Torrington. In sharp contrast to the size-class associations of Fe, baseflow Mn concentration consisted almost exclusively of the <0.45-µm fraction. Manganese associated with sand particles accounted for a significant fraction of the total particulate load (15%) only during the initial stages of runoff, but at that time, Mn distribution was dominated by 20- to 80-µm particles. In contrast to Fe and Mn, the size distributions for Al (Fig. 7c) show that Al in the initial phase of runoff was split evenly between sand (>80 µm) and silt (20–80 µm) particles. The size associations for Al with suspended sediments did not change substantially during the runoff event, except for a small increase in the sand fraction during peak runoff and an increase in the fine silt fraction (1–20 µm) during the latter part of the storm event.

Figure 8 (a, b, and c) shows the size distributions for particles associated with Fe, Mn, and Al at the East Branch location, which are very similar to the baseflow distributions for these metals at the Torrington location. The distribution data support the notion that 1.0- to 80-µm particles (and the metals they carry) in the river below Torrington were contributed by urban runoff. In contrast to the Torrington location, the overall character of the size distributions for trace metals in the East Branch changed little with discharge. The 20- to 80-µm fraction was significant for Fe and Al, but did not dominate the size distribution at any time during the storm event. Manganese, in contrast, was present almost entirely associated with the sand fraction and the <0.45-µm filtrate fraction.



View larger version (30K):
[in this window]
[in a new window]
  		Fig. 8. Changes in the size distribution of particles associated with selected trace metals for the East Branch, above Torrington, CT. Total recoverable metal concentration is shown for comparison, and boxes indicate samples chosen for detailed size distribution analysis.

 
Figure 9 (a, b, and c) depicts changes in the size distribution of particles associated with Pb, Cu, and Cd (respectively) for the Naugatuck River below Torrington. Because these solid-phase metals are expected to be present as adsorbed contaminants on the reactive surfaces of suspended sediments (Wang et al., 1997; Warren and Zimmermann, 1994a), the size distributions of particles associated with Pb, Cu, and Cd will be compared to the size distributions of particles associated with Fe, Mn, and Al discussed in the previous paragraphs. The size distributions in Fig. 9 exhibit the same, not unexpected, dramatic concentration increase in the 20- to 80-µm size fraction in the initial phase of runoff as Fe, Mn, and Al (compare with Fig. 7). The difference in the distributions of Pb and Cu is that they were present in measurable amounts in the clay and fine silt fraction (8.3 and 14% of the total concentrations, respectively). Both Pb and Cu are particle reactive, and thus, the larger surface area of particles in the smaller silt fraction should explain the relative enrichment of Pb and Cu in the 1- to 20-µm size range. In contrast, Cd was not found in the 1- to 20-µm fraction, and only 51% of the total was present in the 20- to 80-µm coarse silt fraction. This difference may be due to the weaker adsorption of Cd to suspended sediments (Erel et al., 1991; Wang et al., 1997; Warren and Zimmermann, 1994b).



View larger version (30K):
[in this window]
[in a new window]
  		Fig. 9. Changes in the size distribution of particles associated with selected heavy metals for the Naugatuck River, below Torrington, CT. Total recoverable metal concentration is shown for comparison, and boxes indicate samples chosen for detailed size distribution analysis.

 
Figure 9a also shows that Pb is present in the colloidal and dissolved phase, but that these fractions are relatively minor. Figure 9b shows the importance of the colloidal and dissolved phases to the physical speciation of Cu, which changes little during the storm event. This result is supported by the fact that Cu speciation in freshwater systems is typically controlled by complexation to colloidal and dissolved organic matter (Donat et al., 1994; Ross and Sherrell, 1999; Rozan and Benoit, 1999; Sigg et al., 2000), which would be found in the <0.45-µm fraction. Figure 9c shows an increase in the larger colloid fraction (0.45–1.0 µm) for Cd during the initial phase of runoff. The concentration of colloidal Cd in the larger fraction is significant for all samples, which is in contrast to the conventional notion that freshwater Cd is largely present in the smaller colloidal and dissolved phases (<0.1 µm), although research does show that Cd complexation to OC is significant in many freshwaters (Pham and Garnier, 1998; Xue et al., 2000).

Particle-size distributions for Pb, Cu, and Cd at the East Branch location did not change substantially during the sampled storm event, and were very similar to the baseflow size distribution for each element at the urban Torrington location. The results are not shown graphically, in part due to the error associated with deriving the individual fraction concentrations by difference for very small total element concentrations. Lead was predominantly found in the 20 to 80-µm fraction, but it was also consistently present in the sand-sized fraction (>80 µm). Copper was mostly present in the dissolved phase, and Cd was mostly in the larger colloid fraction (0.45–1.0 µm) and the dissolved fraction.

Organic Carbon
Figure 10 shows changes in OC over the course of the storm event for the Naugatuck River location. The total OC graphs contain size-distribution data that are not present on the trace metal charts. The four samples chosen for intensive size-distribution analysis, as discussed above for metals, were further fractionated via tangential-flow ultrafiltration to 0.1 µm and 3000 MWCO (note: the fourth sample in Fig. 10a was only filtered to 0.1 µm). These size cutoffs provide a clearer distinction between large colloids (0.1–1.0 µm), colloids (3000 MWCO to 0.1 µm), and dissolved species (<3000 MWCO).



View larger version (30K):
[in this window]
[in a new window]
  		Fig. 10. Changes in the size distribution of particles associated with (a) OC and (b) OC concentration vs. time for the March storm on the Naugatuck River below Torrington, CT. (a) Size distribution graph contains two colloid fractions and a truly dissolved fraction, except for the final point. (b) Concentration graph shows data for total OC, 0.45-µm filtrate, and 0.1-µm filtrate, and indicates points chosen for size distribution analysis.

 
Total and dissolved OC concentrations (Fig. 10b) were both diluted during the initial phase of runoff, but they returned to pre-event concentrations rapidly thereafter. The source of dissolved OC, and the reason that OC concentrations are initially as high as 8.0 mg L–1, is the Torrington wastewater treatment plant located <1 km upstream from the sampling site. Wastewater treatment plants typically contribute colloidal and dissolved OC to waterbodies (Markich and Brown, 1998; Sigg et al., 2000), so it is to be expected that the additional points in Fig. 10b for the 0.1-µm TFU filtrate showed a significant dilution of colloidal and dissolved OC during peak runoff. Particle-size distributions (Fig. 10a) confirm that baseflow OC was almost entirely dissolved and colloidal and that these fractions were diluted at peak runoff. Figure 10a also shows that stormwater runoff was a significant source of large colloidal OC at this time and location. Colloidal OC could be present in runoff as oil and grease, and as an adsorbed constituent on particles, or it could be that OC present in the water column aggregated or adsorbed rapidly to larger colloidal particles in surface runoff. Organic C concentration at the East Branch location changed little during the storm event, and is thus not shown graphically. Only 2% of the OC at the East Branch location was associated with particles >1.0 µm, 9.5% was present as large colloids (0.1–1.0 µm), 54.5% was present as colloids between 3000 MWCO and 0.1 µm, and 34% was dissolved (<3000 MWCO). The results for the size distribution of OC, in both the headwaters and urban location, are typical of Connecticut streams (Rozan and Benoit, 1999).

Comparison of Colloidal Fraction Changes for the Urban River
Figure 11 (a, b) shows the relative concentrations of colloids and dissolved species for OC, Fe, Mn, Pb, Cu, and Cd in the first two fractionated samples for the Naugatuck River below Torrington. Figure 11a shows pre-event baseflow conditions (12 Mar. 2001, 1300 h), and Fig. 11b shows the peak runoff sample (13 Mar. 2001, 0700 h). Each element is normalized to the total concentration in the 1.0-µm filtrate; the concentration of the 0.45-µm filtrate is not considered here because it falls in the middle of the 0.1- to 1.0-µm size range. Aluminum is not included in Fig. 11 because 100% of the <1.0-µm Al concentration in both samples was associated with large colloids (0.1–1.0 µm).



View larger version (33K):
[in this window]
[in a new window]
  		Fig. 11. Size distribution for the <1.0-µm fraction (colloids and dissolved species) for organic carbon (OC) and trace and heavy metals. Graph compares pre-event baseflow and first runoff peak stormflow distributions for the March storm on the Naugatuck River below Torrington, CT. Note: 00 and 06 in axis labels refer to the sample number in the storm sequence.

 
The larger colloidal size fraction (0.1–1.0 µm) in natural waters is thought to be composed of fine clay-layer silicates (Al is a surrogate for clays) and Fe (oxyhydr)oxides. This assertion is supported by the colloidal size distribution of particles associated with Al noted above and by the distribution of Fe (67% large colloidal) shown in Fig. 11a. At baseflow, significant fractions of Fe and Mn were present in the smaller colloidal fraction (3000 MWCO to 0.1 µm), which may be due to the association of these metals with OC from natural sources and from the wastewater treatment plant effluent (Boyle et al., 1977; Perret et al., 1994; Pham and Garnier, 1998). Dissolved Mn may also be associated with dissolved OC (Roitz et al., 2002; Ross and Sherrell, 1999; Sholkovitz and Copland, 1981). The majority of the colloidal Pb (53%) and Cu (46%) concentration was found in the larger colloidal size range at baseflow, with significant concentration of each metal in the smaller colloidal and dissolved fractions. This result is expected, because Pb and Cu are both particle-reactive, and in addition, Cu and Pb bind strongly to OC, which is typically present on the surface of clay particles and as smaller colloidal aggregates and macromolecules. Cadmium is more strongly associated with OC than with metal oxides (Erel et al., 1991; Laxen and Harrison, 1981; Ross and Sherrell, 1999) and was preferentially found in the smaller colloidal and dissolved fractions at baseflow, as shown in Fig. 11a.

The colloidal size distribution for the stormflow sample (Fig. 11b) exhibits relative dilution of the smaller colloidal (3000 MWCO to 0.1 µm) and dissolved fractions for all elements and an increase in the larger colloid fraction (0.1–1.0 µm). This shift is likely due to inputs from runoff sources, but it may be enhanced by the adsorption and/or aggregation of smaller colloidal and dissolved OC (and associated metals) to form larger colloidal particles. Aggregation processes may increase during peak runoff as a result of increasing ionic strength from road salt, although Warren and Zimmermann (1994b) suggest that increasing NaCl concentration may also lead to higher dissolved-phase concentrations for metals (Cd, Cu, and Zn). Even with this dramatic shift, the smaller colloidal and dissolved phase concentrations remain significant for OC, Mn, Pb, Cu, and Cd, and the posited association between these elements and OC is strengthened.

Evaluation of the Particle-Size Fractionation Scheme
This research tested an analytical framework for assessing hydrologic response and characterizing changes in trace metal and OC particle-size associations during a storm event. The fractionation method for stormwater particulate matter did meet the basic criteria set for the study. The individual fractions each contained measurable concentrations of trace metals and OC, and the fractions did appear to respond differently during storm runoff. However, the fractionation method, while providing valuable information, should not be standardized as used in this research because it is subject to significant limitations. The most apparent, and difficult, of these is the length of time it takes to complete a single tangential-flow ultrafiltration, coupled with the time-sensitivity of natural water samples containing particles. To further complicate this part of the critique, the colloidal size range covers three orders of magnitude, from 1 nm (ca. 1000 MWCO) to 1.0 µm, and, it could be argued, the colloidal size range should be divided into more than two fractions. The lower limit for colloids used in this study might be more appropriately set at 500 or 1000 MWCO, so that organic macromolecules are more fully excluded from the dissolved phase. The rough size categories used for suspended sediments were more appropriate, but could be shifted slightly to more accurately reflect sand, silt, and clay size cutoffs. This would be accomplished by using a 62.5 to 70-µm membrane for the lower sand cutoff, and a 2 to 5-µm membrane for the lower silt cutoff, because these values more closely match the Wentworth scale used by the U.S. Geological Survey for suspended and bedded sediment. Instrumental particle-size distribution (e.g., laser-diffraction based instruments) should be used in conjunction with separation techniques, but cannot replace isolation and characterization of particle-size fractions. Researchers should continue work to standardize particle-size classes and separation techniques and to improve inter-comparability of studies of aquatic particulate matter.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
The paired headwaters and urban site data show the magnitude of trace metal concentrations contributed to the Naugatuck River by urban runoff at Torrington, CT, in comparison to natural background levels. Detailed size-distribution data for suspended sediments show that urban runoff, in this case, preferentially contributed silts and clays (including large colloids) to surface waters, and that metals were predominantly transported by large silts (20–80 µm) during the monitored storm event. Colloid fractionation provided a more complete picture by demonstrating that the dissolved and small colloid fractions were diluted (or reduced due to coagulation or adsorption processes) during the storm in favor of the larger colloid fractions. Dissolved and small colloid fractions did, however, remain important to trace metal speciation during the storm event, and the data presented in this study showed that their concentrations may recover quickly after dilution by peak runoff. Urban runoff data from this study highlights the importance of large colloids and fine clays to the speciation of trace metals and OC, and demonstrates the type of information gained by using fractionation techniques and cutoffs other than conventional 0.45-µm membrane filtration.


   ACKNOWLEDGMENTS
 
The authors would like to thank Michael Thompson and three anonymous reviewers for thoughtful and constructive comments that greatly improved the manuscript. This material is based on work supported by the National Science Foundation under Grant no. 0132428.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 


Related articles in JEQ:

This Issue in Journal of Environmental Quality

JEQ 2005 34: v. [Full Text]  




This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via ISI Web of Science (1)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Morrison, M. A.
Right arrow Articles by Benoit, G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Morrison, M. A.
Right arrow Articles by Benoit, G.
Agricola
Right arrow Articles by Morrison, M. A.
Right arrow Articles by Benoit, G.
Related Collections
Right arrow Surface Water Quality
Right arrow Colloids
Right arrow Geochemical Processes
Right arrow Runoff
Right arrow Heavy Metals

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Soil Science Society of America Journal Journal of Plant Registrations The Plant Genome