Influences of seasonal flow regime on the fate and transport of fine particles and a dissolved solute in a New England stream



[1] Fine particles are necessary for biogeochemical cycling but pose a risk to water quality if they are present in excess or carry sorbed contaminants. The objective of this work is to elucidate the effects of flow on the transport and retention of fine particles within a mountain stream and to compare particulate matter transport processes to those of a conservative solute tracer. We measured the migration of bromide (a conservative solute tracer) and micrometer-sized titanium-dioxide particles under three seasonal flow regimes in second-order stream in Connecticut. A one-dimensional transport model based upon the transient storage model was applied in inverse mode to the measured data to quantify solute and particle transport processes. Rates of particle movement by advection and longitudinal dispersion matched those of bromide. Transient, or temporary, storage influenced conservative solute transport, while particle storage was irreversible on the time scale of our experiments and likely dominated by deposition within the sediments of the hyporheic zone. Both solute transient storage and particle deposition increased with decreasing discharge.

1. Introduction

[2] Fine particulate material in streams is a mixture of organic and inorganic particles originating from the watershed hillslope and within the channel. The organic particles consist of microorganisms (e.g., bacteria, protozoa, viruses) and plant detritus, while the inorganic particles include silt- and clay-sized minerals and amorphous precipitates. Regardless of their composition and origin, suspended fine particles are small (100 nm to 63 μm diameter) and often assumed to move downstream with the current. The transport of fine particles contributes to the functioning of stream ecosystems by transmitting nutrients necessary for growth and by distributing particulate organic carbon throughout fluvial systems [Bungartz et al., 2006]. Although fine-particle transport plays several beneficial roles, it can also be detrimental to stream habitat and water quality. In particular, fine particles can affect primary production by reducing light penetration [Bilotta and Brazier, 2008], adversely impact fish and filter-feeding insects [Henley et al., 2000], and lower hyporheic exchange by altering streambed properties [Bilotta and Brazier, 2008; Rehg et al., 2005]. Furthermore, suspended mineral particles can adsorb and increase the mobility of stream water contaminants, such as heavy metals [Walling et al., 2003], radionuclides [Malmon et al., 2005], pesticides and herbicides [Owens et al., 2005], excess nutrients [Haygarth et al., 2006], and microbial pathogens [Characklis et al., 2005; Searcy et al., 2005].

[3] The transport of suspended particles in surface waters is governed in part by advection and dispersion [Saiers et al., 2003; Huang et al. 2008]. Advection accounts for the transport of waterborne constituents at the average fluid velocity, while longitudinal dispersion accounts for mixing of the constituents due to local differences in fluid velocity. In streams, this mixing arises from molecular diffusion, turbulent diffusion, and shear flow dispersion [Fischer et al., 1979].

[4] Transient storage also influences the movement and stream water concentrations of fine particles [Minshall et al., 2000; Newbold et al., 2005; Paul and Hall, 2002], although knowledge of this process is derived largely from experiments with dissolved solutes [Bencala and Walters, 1983; Castro and Hornberger, 1991; Gooseff et al., 2005; Harvey et al., 1996]. We define transient storage as the reversible exchange of an aqueous constituent between the active stream channel and any adjacent, relatively stagnant storage zone. Storage zones are usually divided into two categories: surface storage in stagnant pools and hyporheic storage in the comparatively slow-moving pore water of the streambed. Traditional approaches approximate transient storage as a first-order mass transfer reaction between the active channel, where stream water constituents move by advection and dispersion, and a single storage zone, where advective transport is negligible and stream water constituents are perfectly mixed [Bencala and Walters, 1983; Paul and Hall, 2002; Stofleth et al., 2008]. Several studies have extended this framework by introducing two or more storage zones, each characterized by their own storage capacity, mass transfer, and chemical reaction parameters [Briggs et al., 2009; Choi et al., 2000; Gooseff et al., 2004; Newbold et al., 2005]. Analyses of stream tracer data suggest that these approaches may be most appropriate for simulating surface storage in stagnant pools, within which solute residence times tend to be exponentially distributed, but that hyporheic zone storage may be better approximated by power law distributions to accommodate a wider range in exchange time scales [Gooseff et al., 2005; Haggerty et al., 2002; Marion et al., 2008; Worman et al., 2002].

[5] The concentrations of fine particles suspended within storage zone waters, as well as within the active channel, are affected by deposition. Deposition accounts for processes that remove fine particles from aqueous suspension and includes settling and hyporheic zone filtration. This removal can be reversible or essentially irreversible depending on the deposition mechanism and flow conditions.

[6] Deposition by settling occurs when individual particles or flocs fall under the influence of gravity to the streambed surface. Organic and inorganic materials collide to form floc, which is larger than an individual particle, and therefore settles more rapidly [Droppo, 2001]. Formation of floc, and the resulting increase in settling rates, is regulated by stream chemical properties, such as ionic strength and the presence of extracellular polymeric substances (EPS) [Leppard et al., 2003], and by stream water turbulence [Bungartz et al., 2006].

[7] Deposition also accounts for filtration, whereby particles are trapped within the pore spaces of the hyporheic zone [Packman et al., 2000a]. Water exchanges between the channel and hyporheic zone due to pressure gradients over the streambed surface [Elliott and Brooks, 1997; Hester and Doyle, 2008]. During this advective-exchange process, suspended particles are susceptible to filtration by adhesion (physiochemical attachment of fine particles to the surfaces of porous medium) and straining (retention of fine particles within portions of the porous medium that are too narrow to transmit particles) [Bradford et al., 2002]. Rates of filtration have been linked to water chemistry [Ren and Packman, 2002] and surface properties of the fine particles [Rehg et al., 2005]. Particle or floc size also affects filtration such that larger particles deposit more readily than finer particles [Ren and Packman, 2007]. Filtration has a larger impact than settling on the deposition of fine particles [Huettel et al., 1996].

[8] Filtration is irreversible or slowly reversible while the streambed remains immobile. Streambed evolution or the movement of bed forms can lead to the remobilization of previously filtered fine particles [Packman and MacKay, 2003; Rehg et al., 2005]. Particle remobilization occurs physically with the movement of streambed features under flood conditions [Benda and Dunne, 1997] and the erosion of previously deposited particles from floodplains [Lauer and Parker, 2008]. The mobilization rates of streambed sediments are sensitive to discharge as well as critical shear stress and bed material grain size [Singer and Dunne, 2004]. Additionally, flood events can cause larger bed conglomerates to disintegrate, thereby adding fine suspended particles to the stream water column [Benda and Dunne, 1997]. Finally, biological perturbation of the streambed by fish and aquatic invertebrates can lead to remobilization of settled and filtered fine particles [Hassan et al., 2008; Petticrew et al., 2007].

[9] Although the processes that influence fine-particle transport have been identified, knowledge of how these processes respond to changes in stream properties has not been well demonstrated. It is reasonable to assume that information gleaned from experiments on the transport of dissolved solutes can be used to make at least semiquantitative inferences on the advection, dispersion, and transient storage of fine particles [Hart et al., 1999; Harvey et al., 1996, 2003; Paul and Hall, 2002; Wondzell, 2006; Zarnetske et al., 2007]; however, the few observations on fine-particle transport that are available suggest that this assumption may not always be valid [Newbold et al., 2005; Paul and Hall, 2002; Wanner and Pusch, 2000]. Additional field investigations and evaluations of models are needed to constrain inferences on fine-particle movement in streams and to address, in a satisfactory way, issues of particle transport that have implications to surface water quality and biogeochemical cycling.

[10] The overall goal of this study is to improve our current understanding of fine particle transport in streams. In particular, we seek (1) to quantify the deposition rates of micrometer-sized particles and the responses of these deposition rates to seasonal changes in stream water discharge; (2) to identify a form of a mathematical model suitable for describing fine-particle transport in streams; and (3) to test the assumption that the advection, dispersion, and transient storage rates of micrometer-sized particles match those of a conservative solute tracer. To accomplish these objectives, we compared calculations of a transient storage model to measurements of solute and particle transport made in tracer experiments conducted in a second-order stream during three seasonal flow regimes.

2. Study Site

[11] The tracer experiments were conducted in Wangum Brook, a second order stream in Litchfield County, Connecticut. The stream channel consists of step-pool geomorphology and has an average gradient of 0.05. Stagnant surface water zones within the stream include ponded areas behind log jams and other channel obstructions. Channel substrate primarily consists of cobble and boulders with isolated areas of sand and gravel. Our study reach is located within a deciduous, mixed hardwood forest managed for timber production. Active timber harvesting has not taken place in the immediate area of our study reach in several decades.

[12] Our study stream is ungauged. Discharge measurements were performed by salt dilution and with a flowmeter multiple times between 1 June 2007 and 31 May 2008. Steady discharge (e.g., discharge not during storm events) varied between 0.008 and 0.45 m3s−1, with the lowest discharge measured in early October 2007 at the end of a drought period, and the highest discharge measured in the early spring of 2008. Unsteady discharge during storm events appeared to exceed the highest measured discharge.

3. Experimental Methods

[13] The tracer injection experiments were performed on 25 July 2007, 5 October 2007, and 6 May 2008 under different seasonal flow conditions. The stream was at base flow condition in July 2007 and spring (high)-flow condition in May 2008. Stream discharge was lowest for the October 2007 injection, which was conducted during a drought period prior to senescence. In each experiment, conservative solute and 0.45 μm diameter particle tracers were added to the stream over 1.5 to 3.5 h. Downstream sampling began at the time of injection and continued for several hours afterward. During the high flow and base flow conditions, samples were withdrawn in multiple locations over 418 m and 280 m of the study reach, respectively, while under drought flow conditions, the reach was shortened to 93 m due to decreased stream velocity (Figure 1). Sodium bromide (NaBr) was used as the conservative solute and titanium dioxide (TiO2) was used as the particle tracer.

Figure 1.

Location of study reaches. The injection site remained the same for all experiments. During drought flow (October 2007), samples were taken at 0 m, 47 m and 93 m. During base flow (July 2007), samples were taken at 0 m, 93 m and 280 m. During high flow (May 2008), samples were taken at 0 m, 93 m, 280 m, and 418 m. The upstream boundary used in the model simulations was located at 0 m for the drought flow and base flow experiments and at 93 m for the spring flow experiment. As a geographic reference, the 0 m sample station is located at 41.953 degrees latitude and −73.279 degrees longitude.

[14] Titanium dioxide was chosen as the particle tracer due to its similarity in size to natural clays and its use as a clay particle surrogate in previous groundwater and wetland studies [Ryan et al., 2000; Saiers et al., 2003]. As a white pigment, the TiO2 was readily visible in the stream during the injection experiments. TiO2 was obtained as slurry from DuPont Chemical Corporation and diluted in stream water prior to its injection into the stream. During the experiment, the TiO2 injectate reservoir was constantly homogenized with an electric mixer. Despite the low settling velocity of individual TiO2 particles (ω = 0.318 μm s−1 by Stokes Law), we found the constant mixing was necessary to maintain a constant slurry concentration within the reservoir, where concentrations of TiO2 exceeded 500 g/L.

[15] The TiO2 suspension and NaBr solution were contained in separate reservoirs and applied to the stream with peristaltic pumps. To minimize flocculation of the TiO2 particles (which occurs at elevated ionic strengths), the NaBr solution was introduced 5 m upstream of the particles so that the electrolyte solution was diluted prior to reaching the location of the TiO2 injection. The reservoir concentrations and the injection rates were adjusted between experiments in order to achieve well mixed stream concentrations of approximately 30 mg L−1 for Br and 45 mg L−1 for TiO2 at the upstream sampling station.

[16] Following the experiments, bromide concentrations in filtered samples were measured using ion chromatography (IC) standard methods. TiO2 concentrations were measured using a UV/Visible spectrophotometer with readings taken at a wavelength of 562 nm. Standard curves were created for each injection experiment using stream water collected on the date of each experiment prior to the injection. Background stream water turbidity was monitored at each sampling station and found to contribute less than 1 mg L−1 to the measured TiO2 concentration at each station during each injection experiment. Background turbidity measurements were subtracted from the measured TiO2 concentration for each sample.

3.1. Measurements of Stream Water Chemistry

[17] Water chemistry affects the stability of fine suspended particles [Elimelech et al., 1995] and has been shown to vary seasonally in the streams of deciduous forests [Likens and Buso, 2006]. In order to evaluate the seasonal differences in water chemistry, stream water samples were collected at least monthly during rainfall-free periods between June 2007 and June 2008. All stream water chemistry analyses were performed according to standard methods at Yale University. Concentrations of total organic carbon (TOC) were measured by combustion catalytic oxidation. Base cation concentrations were measured using Inductively Coupled Plasma Atomic Emission Spectroscopy. Anion concentrations were measured using ion chromatography.

3.2. Batch Stability Experiments

[18] The stability of TiO2 in the stream water was assessed in batch experiments. In these experiments, the concentrations of TiO2 suspended in still water were measured over time by light scattering with a spectrophotometer. A decrease in light scattering reflects a decrease in TiO2 concentrations, which we attribute to particles settling out of suspension. Four different waters were used in the batch stability experiments: distilled-deionized water and stream waters collected on the days of the injection experiments. Aliquots of these waters were amended with bromide in varying concentrations (0, 15, 30, and 60 mg L−1) and with 45 mg L−1 TiO2. The initial TiO2 concentration used in the batch experiments approximated the maximum, well-mixed stream concentration in our field experiments. In all experiments, samples were collected over an 11 h period and analyzed immediately for TiO2 concentrations.

4. Mathematical Model

[19] We quantified the processes affecting the transport of bromide and TiO2 particles by fitting solutions of a one-dimensional form of a transient storage model to the measured breakthrough curve data. The equations that govern this model are

equation image
equation image

where C and Cs are the tracer concentration (mg L−1) in the active channel and storage zone, respectively, v is stream water velocity (m s−1), x is stream distance (m), D is the longitudinal dispersion coefficient (m2 s−1), A is active stream cross-sectional area (m2), As is storage zone cross-sectional area (m2), α is the storage exchange coefficient (s−1), and λ is the deposition coefficient (s−1).

[20] Similar to other transient storage models (e.g., Bencala and Walters, 1983), our transport model assumes steady flow in the active stream channel and relatively slow or no flow in the adjacent storage zone. The processes of settling and filtration are quantified together in one deposition coefficient and treated as irreversible on the time scales of our experiments. To simplify the parameterization, we tested the suitability of the assumption that deposition rates in the active channel and storage zone can be quantified by the same value of λ.

[21] We fit solutions of equations (1) and (2) to the data on the transport of NaBr to estimate the following parameters: the longitudinal dispersion coefficient (D), active channel area (A), storage zone area (As), and storage exchange coefficient (α). These parameters were assumed to be uniform over the 280 m reach during the base flow experiment and over the 418 m reach during the spring (high)-flow experiment. It was necessary to relax this assumption during the drought flow experiment, and A, As, and α were let vary between two measured subreaches (0–43 m and 43–93 m). In this case, length-weighted averages of the fit parameters are presented for the 0–93 m reach.

[22] We ran multiple simulations of TiO2 transport. Initially we assumed negligible particle deposition (λ = 0) and set D, A, As, and α equal to those values estimated from the solute transport simulations. We then compared this forward simulation of TiO2 transport to results of inverse simulations in which one or a combination of model parameters were estimated from the TiO2 breakthrough data.

[23] During the May 2008 experiment, TiO2 concentrations were the lowest at the most upstream sampling location (Figure 1; x = 0 m). Problems with the operation of a peristaltic pump impeded the even distribution of the particle tracer over the stream width at the injection location, which, in turn, led to incomplete mixing of the particles at x = 0. Complete mixing is requisite to the appropriate application of the one-dimensional transport model. Therefore, we prescribed the upstream boundary on the basis of the measurements made at x = 93 m (where the particle tracer was well mixed), and simulated particle transport through the 93–280 m and 280–418 m subreaches of the stream.

[24] In all inverse simulations, best fit values of the parameters were determined by using a Levenberg-Marquardt nonlinear least squares algorithm to minimize the sum-of-the-squared residuals between measured and modeled tracer concentrations. The goodness of model fit was quantified through calculation of R2 values. In addition, we computed the coefficient of variation for each fitted parameter, which expresses the ratio of the standard error of the parameter estimate to best fit parameter value. We considered well fit parameters to have a coefficient of variation below 0.3.

[25] We calculated nondimensional parameters from the best fit model parameters to facilitate comparison of the results from the experiments conducted under different streamflow conditions. The Damkohler numbers relate the time scale of advection to those for transient storage (DaIα) and deposition (DaIλ), and DaIα can be used to evaluate the appropriateness of the tracer experiment for quantifying transient storage [Harvey and Wagner, 2000]:

equation image
equation image

where L is a characteristic reach length (m).

5. Results

5.1. Batch Turbidity

[26] The total decrease in TiO2 concentrations in the four bromide-free waters ranged from 15 to 20% over the 11 h period (Figure 2). The rate of concentration decrease was similar between the stream waters from all three collection dates and the distilled-deionized water; thus, settling appeared insensitive to seasonal differences in background water chemistry. These stream waters varied in TOC from 2.2 to 4.9 mg L−1. Ionic composition also varied between seasons (Table 1). Most notably, concentrations of Ca2+ and Mg2+ were lower during the high discharge. Furthermore, no significant difference in settling rate was found across the four bromide concentrations in any of the four waters evaluated; therefore, we infer that the addition of the NaBr tracer did not affect the stability of the TiO2 particles in the stream tracer experiments. Settling of TiO2 in batch turbidity experiments, although small, likely exceeded settling in the stream tracer experiments because turbulent in-stream conditions would promote the suspension of the TiO2 particles and because the in-stream residence times were less than the duration of the batch experiments.

Figure 2.

TiO2 concentrations measured in batch turbidity experiments with bromide-free stream waters collected on the day of each tracer experiment and with distilled deionized (DDI) water.

Table 1. Measured Concentrations of Anions, Cations, and Total Organic Carbon Prior to Each Injection Experiment
 Drought Flow (mg L−1)Base Flow (mg L−1)High Flowa (mg L−1)
  • a

    High flow is in the spring.


5.2. Solute Transport

[27] Modeled and measured concentrations of the conservative solute tracer (Br) agree reasonably well for each of the three injection experiments (Figure 3). The R2 values for the three model fits exceed 0.97 (Table 2). Stream discharge varied by two orders of magnitude between our experiments, and the values of the model parameters that govern advection, dispersion, and transient storage are sensitive to this variation in stream flow. Dispersion was a significant solute transport process under all three flow regimes, and the best fit value of D increases 10 fold as discharge increases from 0.002 to 0.32 m3s−1 (Table 2). Transient storage could be neglected without jeopardizing the description of solute transport under spring (high)-flow conditions, but influenced solute transport under drought flow and base flow conditions. As discharge increases, storage zone area (As) decreases threefold and the exchange coefficient (α) increases by nearly an order of magnitude (Table 2). The average transport distance before entry into storage (=v/α) was much shorter during drought flow (34 m) than base flow (795 m), and once in the storage zone, the solute had a shorter average residence time (=As/αA) under the drought flow (3.2 h) than under the base flow conditions (6.5 h). Computations of DaIα made by setting L in equation (4) to 93 m are 3.4 and 0.23 for the drought flow and base flow experiments, respectively, indicating that ratio of the storage zone exchange rate to the advective-transport rate was 15 times greater during drought flow than base flow.

Figure 3.

Measured and modeled tracer breakthrough curves for (top) drought flow, (middle) base flow, and (bottom) spring high flow conditions. The x axis is reported as a dimensionless ratio of time since initiation of the tracer injection to time scale for advection: T = vt/L, where L/v is the time scale for advection and, for these cases, L = 93 m. Note the different x axis scale for the drought flow breakthrough curves.

Table 2. Measured and Modeled Conservative Solute Transport Parametersa
 Injection Duration (h)Q (m3 s−1)A (m2)v (m s−1)D (m2 s−1)As (m2)α (s−1)R2
  • a

    The coefficient of variation is given in parenthesis with each model-fitted parameter. NA means not applicable.

Drought flow1.50.0070.68 (0.05)0.010.013 (0.13)2.20 (0.11)2.82 × 10−4 (0.10)0.99
Summer base flow3.50.0210.61 (0.01)0.0340.784 (0.06)0.62 (0.16)4.33 × 10−5 (0.08)0.97
Spring high flow20.3271.9 (0.01)0.1721.158 (0.13)NA0.00.97

5.3. Particle Transport

[28] Simulations that used the values of v, D, As, and α from the bromide experiments, but that ignored deposition (λ = 0) described TiO2 transport reasonably well under the spring (high)-flow condition (R2 = 0.95), but could not mimic transport under drought flow or base flow conditions (R2 = 0.75 for drought flow; R2 = 0.35 for base flow). Accounting for deposition (by fitting λ) while using values of v, D, As, and α estimated from the Br inversions improved the model description of TiO2 transport for all three treatments, but the description of TiO2 transport during drought flow remained poor (R2 = 0.76). Allowing D, As, and α to vary with λ in the model inversion did not improve the model-data agreement for the base flow and spring (high)-flow treatments, and the large standard errors associated with the estimates of As and α indicated that these parameters could not be constrained in inverse simulations of drought flow transport. Because the storage-related parameters could not be constrained by the drought flow data, we simplified the model by neglecting the transient storage process. A variant of the transport model that ignored transient storage (α = 0), accounted for deposition, and that used values of D and v from the bromide experiments accounts for 95% of the variation in the TiO2 data collected under drought conditions (Table 3 and Figure 3). This approach also led to excellent description of transport during the base flow (R2 = 0.96) and high-flow (R2 = 0.96) experiments (Table 3 and Figure 3). These results suggest that a model that accounts for the processes of advection, dispersion, and first-order kinetics deposition, but that neglects transient storage (as traditionally formalized), is suitable for approximating the transport of micrometer-sized particles under the three flow regimes evaluated in this study.

Table 3. Measured and Modeled Particle Transport Parameters for Best Fit Simulations With TiO2 Particle Tracera
 v (m s−1)D (m2 s−1)λ1 (s−1)λ2 (s−1)λ3 (s−1)R2
  • a

    The coefficient of variation is given in parenthesis with each model-fitted parameter. Italicized parameters are taken directly from the inverse simulations of Br transport. The parameter λ1 corresponds to the 0–93 m reach, λ2 corresponds to the 93–280 m reach, and λ3 corresponds to the 280–418 m reach.

Drought0.010.024 (0.12)2.4 × 10−4 (0.055)0.95
Summer base flow0.0340.7841.8 × 10−4 (0.027)7.4 × 10−5 (0.12)0.96
Spring high flow0.1720.594 (0.22)1.0 × 10−4 (0.14)0.00.96

[29] The mass of TiO2 deposited varied between experiments, increasing with decreasing stream discharge. The maximum relative concentration of TiO2 equaled 0.15 at the bottom of the 93 m reach under drought flow, compared with 0.48 at the bottom of the 280 m reach under base flow and 0.90 at the bottom of the 418 m reach under spring (high) flow (Figure 3). Integration of the TiO2 breakthrough curves reveals that 97.5% of the injected mass was retained with the 93 m reach under drought flow, 58% was retained in the 280 m reach under base flow, and 4% was retained within the 418 m reach during spring (high) flow.

[30] Deposition rates, quantified by the fitting parameter λ, increased as stream discharge decreased (Table 3). When more than one reach was examined during a single experiment (i.e., base flow and high flow), larger estimates of λ were associated with the more upstream reach. Significant, nonzero deposition coefficients (λ), varying from 7.4 × 10−5 s−1 to 2.4 × 10−4 s−1, were estimated from the drought flow and base flow experiments (Table 3). Under the high flow conditions, only one of the two reaches examined (93–280 m) had a significant nonzero deposition coefficient (λ = 1.0 × 10−4 s−1). Computations of DaIλ, made by using the best fit values of λ in equation (5) range from 2.2 for drought flow to 0.031 for spring (high) flow (for L = 93 m), which reveals that the rate of deposition relative to the rate of advection increased with decreasing stream discharge.

6. Discussion

[31] The model parameters that govern the advection and dispersion of the bromide tracer are sensitive to seasonal changes in stream flow. Average stream water velocity (v) varies by a factor of 17 over the three flow conditions examined (Table 3), and the dispersion coefficient increases in a linear fashion with the logarithm of stream velocity (D = 0.91 log(v) +1.9; R2 = 0.92). Our estimates of bromide dispersion fall within the range of published values of D for solute transport in mountain streams at low and base flow [Paul and Hall, 2002; Scott et al., 2003; Wondzell, 2006]. The values of v and D estimated from the bromide experiments could be used to simulate the movement of the TiO2 particles, suggesting that the advective-dispersive transport of micrometer-sized particles mimics that of a conservative solute tracer over a broad range of discharge (seasonal flow) conditions.

[32] Both the solute and particle tracers were stored within the stream reach, but only solute transport was affected by transient storage, which has been traditionally defined as a two-way (reversible) exchange process. In accordance with published observations [Harvey et al., 1996; Paul and Hall, 2002; Scott et al., 2003], the transient storage characteristics of bromide within Wangum Brook depended on stream discharge. Although transient storage during high flow was negligible, storage zone exchange rates and storage zone area increased as stream discharge declined so that transient storage effects on bromide transport were greatest during drought flow. The inverse relationship between storage zone area and discharge could reflect expansion of the hyporheic zone with decreasing streamflow [Harvey et al., 2003; Wondzell, 2006; Zarnetske et al., 2007] and the presence of more pools within the stream channel. Under drought conditions, much of the water within the active channel was in contact with the bed, and pools also occupied a larger portion of the channel. The presence of pools could have increased transient storage in multiple ways because these stagnation zones can both store waterborne constituents and enhance hyporheic exchange by increasing advective pumping [Mutz et al., 2007].

[33] Hyporheic exchange is driven by the hydraulic gradient across the streambed and is influenced by the hydraulic conductivity of the streambed material [Anderson et al., 2005]. Hyporheic exchange and advective pumping across the bed surface has been associated with bed form structures, such as dunes [Packman et al., 2000a, 2000b], and is promoted by the presence of woody debris [Mutz et al., 2007]. Field investigations have demonstrated that vertical hydraulic gradients regulating hyporheic exchange increase with decreasing stream discharge [Storey et al., 2003]. Based on these published findings, we infer that contributions of hyporheic exchange to the storage of bromide and TiO2 particles within the streambed sediments of Wangum Brook increased as the stream discharge declined. Thus, under drought flow conditions in particular, hyporheic exchange may have dominated the transient storage response of bromide and may have led to the transmission of large quantities TiO2 particles into the streambed, where they were removed from the water by porous medium filtration. This conclusion does not imply that bromide and TiO2 exchange with stagnant surface water zones was unimportant, but rather that this mass transfer was fast (in comparison to hyporheic exchange) and could be approximated as a Fickian process that contributed to longitudinal dispersion, quantifiable by the parameter D [Briggs et al., 2009].

[34] Storage of TiO2 particles transmitted into the hyporheic zone was irreversible on the time scale of our experiments and sensitive to changes in stream discharge and corresponding changes in stream water velocity. For the conditions tested in this study, TiO2 storage could be approximated as a kinetics reaction quantified by first-order deposition coefficient (λ). Values of λ decreased with increasing stream discharge and could be expressed as a power law function of stream velocity (λ = 5 × 10−5v−0.3; R2 = 0.89). Given that deposition dynamics were consistent with an irreversible, first-order kinetics process, the transport distance over which particles concentrations would be reduced by one half can be computed as L1/2 = 0.693*v/λ. Based on our best fit values of λ, calculations of L1/2 equal 29 m, 183 m, and 2073 m for drought, base, and high flows, respectively. These estimates reveal that submicrometer particles that are capable of traveling only tens of meters during late summer drought flow conditions can travel in excess of two kilometers under high-flow conditions.

[35] Particle flocculation, followed by gravitational settling of the flocs, may have contributed in a small way to overall particle deposition; however, changes in particle flocculation rates associated with seasonal variation in stream water chemistry cannot account for the differences in particle deposition rates observed between the field experiments. Results of the batch experiments, which constrain the maximum settling rates, indicate a decrease in TiO2 concentration of 15–20% over the time scale of our experiments. A larger decrease was seen in the concentration of TiO2 in the stream under the lower flow conditions, with 97.5% of the TiO2 remaining in the reach under drought conditions and 58% during base flow. Furthermore, the seasonal differences in chemical composition of our dilute stream waters did not correspond to any differences in settling during the batch experiments. Thus, it is unlikely that seasonal differences in stream water chemistry would alter the particle stability enough to produce the observed differences in deposition between experiments.

[36] Few studies have been aimed at quantifying the mobility of micrometer-sized particles in stream water environments. Most studies that are available have used particulate organic matter (POM) as the tracer, which is larger and of lower density than the TiO2 particles used in this study. Downstream movement of both POM and TiO2 positively correlates with average channel velocity [Hunken and Mutz, 2007]; however, differences exist between the in-stream storage characteristics of POM and TiO2. Storage of TiO2 was irreversible, while the transport of POM tracers could be approximated assuming reversible, transient storage [Minshall et al., 2000; Newbold et al., 2005; Paul and Hall, 2002] or with deposition and subsequent resuspension [Hunken, 2006]. In cases of irreversible POM deposition, hyporheic water exchange and advective transport to the streambed have been suggested as possible mechanisms of particulate removal [Hunken and Mutz, 2007; Thomas et al., 2001]. Our experiments with TiO2 suggest that colloid-sized mineral particles can be removed from main channel flow despite low calculated settling velocities. Our work also reveals the considerable sensitivity of particle transport and deposition to seasonal changes in stream discharge and illuminates the need to quantitatively resolve multiple deposition mechanisms to advance predictive understanding of fine particle transport in stream water environments.


[37] We wish to thank the Great Mountain Forest Corporation for access to the Wangum Brook study site. Field work was graciously performed by Syeda Absar, Agha Akram, Noel Aloysus, Devorah Ancel, Rebecca Barnes, Maura Bozeman, Gerald Bright, David Butman, Natalie Ceperley, Tao Cheng, Heather Clark, Ted Elwartowski, Sara Enders, Troy Hill, Meredith Sattler, Lu Sun, Annika Walters, Jason Weiner, and Na Xu. We are also grateful for efforts of Carolyn Oldham and three anonymous reviewers who provided comments and suggestions that led to improvement of this paper.