In-stream sorption of fulvic acid in an acidic stream: A stream-scale transport experiment

Authors


Abstract

[1] The variation of concentration and composition of dissolved organic carbon (DOC) in stream waters cannot be explained solely on the basis of soil processes in contributing subcatchments. To investigate in-stream processes that control DOC, we injected DOC-enriched water into a reach of the Snake River (Summit County, Colorado) that has abundant iron oxyhydroxides coating the streambed. The injected water was obtained from the Suwannee River (Georgia), which is highly enriched in fulvic acid. The fulvic acid from this water is the standard reference for aquatic fulvic acid for the International Humic Substances Society and has been well characterized. During the experimental injection, significant removal of sorbable fulvic acid occurred within the first 141 m of stream reach. We coinjected a conservative tracer (lithium chloride) and analyzed the results with the one-dimensional transport with inflow and storage (OTIS) stream solute transport model to quantify the physical transport mechanisms. The downstream transport of fulvic acid as indicated by absorbance was then simulated using OTIS with a first-order kinetic sorption rate constant applied to the sorbable fulvic acid. The “sorbable” fraction of injected fulvic acid was irreversibly sorbed by streambed sediments at rates (kinetic rate constants) of the order of 10−4–10−3 s−1. In the injected Suwannee River water, sorbable and nonsorbable fulvic acid had distinct chemical characteristics identified in 13C-NMR spectra. The 13C-NMR spectra indicate that during the experiment, the sorbable “signal” of greater aromaticity and carboxyl content decreased downstream; that is, these components were preferentially removed. This study illustrates that interactions between the water and the reactive surfaces will modify significantly the concentration and composition of DOC observed in streams with abundant chemically reactive surfaces on the streambed and in the hyporheic zone.

1. Introduction

[2] In many catchment-stream systems the concentration and composition of dissolved organic material is a critical water quality characteristic. Dissolved organic carbon (DOC) is typically the most abundant form of carbon transported in lotic systems. Understanding the factors that control concentrations and composition of DOC in streams has been a long-standing goal of limnologists and biogeochemists.

[3] At the catchment scale, observed spatial and temporal variability in concentration and composition of DOC in oligotrophic streams can be controlled by autochthonous or allochthonous processes in three ways. Different landscape units may produce different amounts and qualities of DOC. Second, interactions between the active (surface) stream, the streambed, and the hyporheic zone may result in removal or modification of DOC. Finally, mixing of waters from differing subcatchments at stream confluences may induce reactions that affect DOC. Much previous work has been aimed at landscape controls [e.g., Foster and Grieve, 1982; Meyer and Tate, 1983; McDowell and Likens, 1984; Fiebig et al., 1990; Baron et al., 1991; Denning et al., 1991; Hedges et al., 1994; Houle et al., 1995; Boyer et al., 1997].

[4] The variation of concentration and composition of DOC in stream waters cannot be explained solely on the basis of soil processes in contributing subcatchments. Wallis [1979] suggested that the riparian zone is the landscape unit exerting main control of stream DOC concentrations. From a hydrologic perspective, riparian sediments are part of the hyporheic zone, and the overlying riparian soils and vegetation may act as proximate sources of particulate organic material and DOC to the stream and hyporheic zone [Pusch et al., 1998; Chafiq et al., 1999]. Uptake of DOC by microbial communities in hyporheic zone sediments also contributes to DOC losses in a range of stream ecosystems [Kaplan and Bott, 1985; Mulholland and Hill, 1997]. Experimental studies in small streams have shown that microbial respiration in the hyporheic zone is limited by DOC quantity and quality, with labile substrates such as algal leachate and acetate stimulating microbial activity more than leaf leachate [Jones, 1995; Baker et al., 1999, 2000]. In large rivers, reports have suggested that interactions with riparian sediments have a marked influence on DOC composition and concentration [e.g., Depetris and Kempe, 1993].

[5] Such interactions also may be critical in mountain streams. A strong riparian influence in mountain streams is plausible given that stream waters and hyporheic waters are continuously interchanging [e.g., Bencala et al., 1990]. Furthermore, although in-stream biological processes may affect concentrations of DOC in streams elsewhere [e.g., Mulholland and Hill, 1997], biological uptake in the water column has been found to be of much less importance than are in-stream abiological processes in streams in the Rocky Mountains, because the rates are slow compared with flow rates [McKnight et al., 1993].Thus, in mountain streams, hyporheic exchange may transport labile DOC in stream water to microbial communities in hyporheic zone sediments where longer residence times would promote uptake.

[6] Another reason that riparian influences may be important is that streambed and hyporheic substrates typically include oxides, formed through several biogeochemical processes, that can cause abiotic retention of DOC through sorption reactions [e.g., McKnight et al., 1992]. Abiotic retention in hyporheic sediments of dissolved free amino acids was found to last for periods of many weeks, and abiotic retention of labile DOC may help sustain microbial activity in hyporheic sediments [Fiebig, 1997]. In pristine streams, oxides can be formed at oxic/anoxic interfaces where reduced Fe and Mn species from reducing regions in the hyporheic zone are oxidized and precipitate. In streams receiving acidic inflows, precipitation of oxides at the higher pH of the stream can form an abundant flocculent deposit covering the streambed. Finally, photoreduction of streambed oxides during the day and subsequent reoxidation of the dissolved ferrous at night and precipitation of oxides can maintain the reactivity of oxide surfaces [McKnight et al., 2001].

[7] We postulate that in streams with abundant chemically reactive surfaces on the streambed and hyporheic substrates, interactions between the water and the reactive surfaces will modify significantly the concentration and composition of DOC observed in the stream. In Rocky Mountain headwater streams, conditions are such that these interactions may be the primary controls on in-stream DOC. First, the water in the surface stream is continually and rapidly interchanging with water in the hyporheic zone [Bencala et al., 1990]. Second, freshly precipitated iron, manganese, and aluminum oxyhydroxides are often found on the streambed in both pristine streams and streams receiving inflow of acid rock or acid mine drainage, such as in the Snake River in Summit County, Colorado. We further postulate that interactions with iron oxyhydroxide surfaces may have an observable effect on DOC in reaches as short as 100 m in many streams with abundant oxides.

[8] Previous work in the Snake River has shown that fulvic acid is an abundant and reactive fraction of DOC and that the more aromatic fulvic acid molecules will preferentially sorb to the oxyhydroxides, leaving the in-stream fulvic acid relatively depleted in aromatic character and less “sorbable” [McKnight et al., 1992]. These field results are consistent with laboratory experiments showing that sorption of simple aromatic organic acids is enhanced by multiple adjacent carboxylic and phenolic functional groups [Evanko and Dzombak, 1998, 1999] and that aromatic carboxyl-group moieties contribute to strong acid character in aquatic fulvic acid [Leenheer et al., 1995]. To investigate in-stream processes that control DOC, we injected DOC-enriched water, high in fulvic acid with aromatic sorbable moieties, into a reach of the Snake River that has abundant iron oxyhydroxides. The Snake River is typical of many stream systems in the region receiving acid rock or acid mine drainage, and previous study has shown that the iron oxyhydroxides are rapidly turned over by photoreduction [McKnight and Bencala, 1998; McKnight et al., 1988]. We found that the composition of the injected fulvic acid was altered as it was transported along the stream in response to interaction with the streambed. The sorbable fraction of injected fulvic acid was sorbed by streambed sediments at rates (kinetic rate constants) of the order of 10−4–10−3 s−1.

2. Site Description: Snake River and Deer Creek Catchments

[9] Our ideas about the controls on spatial and temporal variability of DOC in mountain streams stem from work over the past 15 years. The adjoining catchments of Deer Creek and the upper Snake River in Summit County, Colorado (Figure 1), afford the opportunity for study of significant differences in the natural chemistries of Rocky Mountain streams. On balance, the chemistry of Deer Creek is that of a pristine stream, while the chemistry of the upper Snake River reflects that of an acidic, metal-enriched stream. Although there has been limited mining activity in both catchments, the chemical characteristics of the streams predominantly result from natural processes. One specifically relevant difference in the two stream chemistries is the concentration of DOC. The DOC concentration in Deer Creek is typically higher than in the upper Snake River, often by a factor of 2 and occasionally by a factor of 4 (Figure 2).

Figure 1.

Location map of upper Snake River–Deer Creek catchments, Summit County, Colorado.

Figure 2.

Comparison of dissolved organic carbon (DOC) concentrations measured contemporaneously in the upper Snake River and Deer Creek. DOC concentrations in Deer Creek are typically higher than in the Snake River, by a factor of 2 to 4 during water years 1979–1985. (Data compilation by Boyer et al. [1999].)

[10] The concentration difference is consistently evident both in 6 years of monitoring data [McKnight et al., 1992] and in a year of intensive sampling focused on DOC variation [Boyer et al., 2000]. The catchments are of similar size (Snake River, 12 km2; Deer Creek, 11 km2), relief (∼900 m from their confluence at an elevation of 3300 m), and annual discharge range (20–1000 L s−1). Because pyrite is disseminated in the host rock of the upper Snake River catchment, the Snake River has concentrations of iron, aluminum, and sulfate that are significantly greater than in Deer Creek [McKnight et al., 1992]. The concentration and speciation of iron in the Snake River is also influenced by photoreduction of dissolved, colloidal, and particulate ferric iron [McKnight and Bencala, 1998]. There are some differences between the catchments that may account at least in part for the differences in DOC concentration. The Snake River catchment has a greater areal proportion of talus, a landscape unit that produces less soluble organic carbon than other units [Brooks et al., 1999]. In landscape units that include regions of disseminated pyrite, the acidic conditions limit carbon cycling, and these also produce less soluble organic carbon [Brooks et al., 1999]. Detailed sampling from soil lysimeters in the large fraction of both catchments not in either talus or pyritic regions has shown no characteristic differences in the time histories of DOC concentrations within the soil waters feeding the two streams [Boyer et al., 2000]. It is not clear that the differences in landscape controls of DOC can account for the severalfold differences in stream DOC observed in the Snake River and Deer Creek.

[11] We believe that the twofold to threefold lower DOC concentrations in the Snake River relative to Deer Creek may to a significant extent result from in-stream processes of chemical removal coupled with a higher degree of solute interaction with streambed materials. This chemical removal mechanism has been documented within the confluence zone of Deer Creek with the upper Snake River [McKnight et al., 1992]. Results from a 6-year sampling study indicate that approximately 40% of the DOC entering the confluence is removed, due to sorption onto freshly precipitating iron and aluminum oxides [McKnight et al., 1992]. The rate and spatial extent of the removal is influenced by the contact between solute and streambed materials, as is now well documented in tracer studies of upland streams [Harvey and Fuller, 1998; Runkel et al., 1999]. Along the length of the upper Snake River the streambed consists of gravel and cobble coated with iron oxides; our previous stream-scale studies of iron photochemistry have shown that these iron oxides are highly reactive [McKnight and Bencala, 1998, 1989]. On the basis of these earlier studies we hypothesize that similar removal processes (as observed at the confluence of Snake River with Deer Creek) may account for DOC concentrations being lower in Snake River than in Deer Creek.

[12] In previous work [Bencala et al., 1990] we established that the physical parameters controlling solute transport in the upper Snake River vary over distances of less than a kilometer. For example, estimated stream cross-sectional areas and volumetric lateral inflows vary (at late-summer low flow) by factors of approximately 3. The numerous minor inflows can accumulate to yield significant stream discharge increases and nonuniformity in stream solute concentrations. The roughness variations of the cobble streambed result in transient storage of solutes, which increases the contact time between the water and the reactive streambed sediment.

3. Methods

[13] To test our ideas about removal of DOC in the Snake River, we injected high-DOC water containing fulvic acid enriched in aromatic moieties (relative to the Snake River) into a reach of the river and observed concentrations as the pulse of injected water moved downstream. The expectation was that the more aromatic portions of the fulvic acid fraction of the injected material would preferentially be sorbed onto fresh iron oxyhydroxide surfaces on the streambed cobbles and that the DOC pulse would decline in magnitude due to removal and would change in character to become more like the background DOC in the Snake River, i.e., dominated by nonsorbable fulvic acid rather than by sorbable fulvic acids. We coinjected a conservative tracer (lithium chloride) to quantify the physical transport mechanisms.

3.1. DOC Injectate

[14] On July 31, 1997, approximately 23,000 L of water with a DOC concentration of 58.8 mg C L−1 was collected from the Suwannee River, Georgia, into a tanker truck and transported to our site at the Snake River in Colorado. We chose the Suwannee River because its waters are highly enriched in fulvic acid; the Suwannee River is the standard reference solution for aquatic fulvic acid and has been well characterized [Averette et al., 1989]. We characterized the sorptive properties of the Suwannee River DOC in a laboratory experiment using a 30-L subsample from the tanker truck. A concentrated solution of ferric chloride was added and a large mass of iron oxide floc formed, removing most of the brown color. The floc was settled, supernatant was removed, and the floc was dissolved at pH 2 and diluted. Fulvic acid was isolated from the supernatant and from the dissolved floc solution following the methods of Thurman and Malcolm [1981]. The fulvic acid dissolved in the supernatant is referred to as the “nonsorbable fraction,” and the fulvic acid from the floc is referred to as the “sorbable fraction.”

3.2. Lithium and DOC Injections

[15] On August 5, 1997, a solution of lithium chloride was injected into the Snake River. The injection began at 0700 LT and continued until 1130 LT. (The lithium is used to quantify transient storage in the stream reach. To ensure that a steady concentration was reached in the stream, the injection was begun several hours prior to injection of the DOC-enriched water.) Battery-powered metering pumps delivered the lithium at a rate of 0.45 g s−1. For the analysis of this experiment, [Li+] is assumed to be a conservative tracer. This assumption is supported by previous studies in the Snake River and in another acidic, metal-enriched stream [Bencala et al., 1990; Broshears et al., 1993; Zellweger, 1994]. In a comment, Stewart and Kszos [1996] raised the possibility of lithium toxicity to certain aquatic biota. Tate et al. [1996] responded and concluded that there have been no published studies that have indicated that [Li+] has an effect on algal-microbial processes.

[16] The Suwannee River water was transferred from the tanker to four holding tanks plumbed to allow water to be released into the river. The injection water was routed through a Parshall flume to measure the volumetric injection rate. Injection of the DOC-enriched water commenced at 1000 LT on August 5 and continued until 1130 LT. The injection rate varied but was held at approximately 4.2 L s−1. The discharge in the Snake River was 0.37 m3 s−1, and the upstream DOC concentration was 1.6 mg C L−1.

3.3. Field Sampling

[17] Sampling stations were established at five locations downstream of the injection site (Figure 3). The locations were at distances 10, 34, 141, 386, and 681 m downstream of the injection point. The two most upstream stations were selected because preliminary calculations suggested that there was a possibility that the injected DOC would react so rapidly with the sediments in the hyporheic zone that the transported pulse would be missed at the stations farther downstream. This was not the case, however. Grab samples were collected at 10- to 15-min intervals from the center of the stream. Conductivity, pH, and temperature of samples were measured in the field. Samples for cation and metal analysis were collected in 125-mL acid-washed plastic bottles, while samples for DOC analysis were collected in 125-mL amber-colored, precombusted (250°C overnight) glass bottles. Sample bottles were rinsed with filtered stream water prior to sample collection. All samples were promptly filtered through 0.4-μm Nuclepore membranes and stored at 4°C until analysis. (The use of trade names is for identification purposes only and does not constitute endorsement by the U.S. Geological Survey.)

Figure 3.

Sketch depicting sampling station locations along the upper Snake River. The injection location (site 0 m) was approximately 2217 m upstream of the confluence of the upper Snake River with Deer Creek.

[18] In addition to the frequent grab samples, we collected large-volume (60–100 L) samples at all sites before the DOC injection and during the estimated midpoint of the DOC pulse. These samples were collected in rinsed 5-gal. (18.9 L) cubitainers and were kept chilled prior to processing. Because photoreduction typically prevents any colloidal or particulate iron oxides from persisting in the water column during the day [McKnight and Bencala, 1998], the large-volume samples were not filtered. Analysis of iron speciation confirmed that ferrous iron accounted for almost all of the dissolved iron. Fulvic acids were isolated from the samples following the method of Thurman and Malcolm [1981].

[19] For the reactive solute transport analysis we present data only for the sites at 141, 386, and 681 m because tracer data from the sites at 10 and 34 m indicated that the injected solutes had not mixed thoroughly at the point in the stream where the samples were collected. (The concentrations from the sampling location in the stream at the first two sampling sites were too low as judged by mass-balance calculations. Prior to the experiment we had estimated that it would take almost 100 m to achieve thorough mixing, but we chose to sample at the sites at 10 and 34 m in case the sorption reactions were extremely fast.) Because the large-volume samples were collected over a significant portion of the channel, we believe that these samples are less prone to problems caused by incomplete mixing and provide representative integrated samples at all site locations. Thus we present the data from the large-volume samples for all sampling locations.

3.4. Microbial Processing of DOC During the Injection

[20] Numerous water samples for analysis of suspended microbial abundance were collected at four sites: 0, 34, 141, and 386 m. These samples were preserved by addition of acridine orange for direct counts (AODC). To assess microbial growth rate, bottle bioassays were done by inoculating bacteria-free water collected from sites 0, 34, and 141 m during the injection with a dilute inoculum containing nutrients and a previously studied bacterial consortium from Paine Run in Virginia. A sample of the Suwannee River injectate diluted 30-to-1 was also studied. Duplicate bottles were incubated in the dark at 20°C for 3 days. Growth rate was determined for the period of exponential growth which followed a lag phase of 1.5–2 days. A second bioassay was done using bacteria-free water from site 0 m and a dilute inoculum of nutrients and bacteria from the Snake River incubated at 4°–6°C for 3 days. This bioassay more closely matched conditions in the stream and also exhibited a lag phase. Bacterial counts for the bioassay experiments were done by AODC. The standard deviations for the growth rates determined for duplicate bottles were 10–20%. Analyses had a standard deviation of 12–23%.

3.5. Laboratory Analyses

[21] Lithium analyses were made by flame atomic absorption spectroscopy using a Perkin-Elmer AAnalyst 100. Replicate analyses indicate a precision of ±0.1 mg L−1 at the 1 mg L−1 level. DOC samples were analyzed using a Dohrmann DC-190 carbon analyzer. DOC analyses were completed within 2 weeks of sample collection except for some repeat analyses that were completed within 1 month of sample collection. The precision of the instrument was estimated by repeating the analysis of 24 samples. The standard deviation of absolute errors was 0.25 mg L−1. A subset of the DOC samples was analyzed on an Oceanographic International carbon analyzer, which agreed within about 0.2 mg C L−1 of the initial analyses. Absorbance of the DOC samples at 254 nm was determined on a Pye-Unicam 8800 spectrophotometer. The specific ultraviolet absorbance (SUVA, (mg C)−1) of the isolated fulvic acid samples was determined by measuring the UV absorbance of solutions of the isolated fulvic acids at pH 4.0 and dividing the measured absorbance by the DOC concentrations of the solutions. The concentrations of total dissolved iron and ferrous iron were determined using the 2′2′-bipyridine method, and the reagents were added immediately after sample filtration and the samples analyzed within 2 days.

[22] Lyophilized fulvic acid samples were analyzed for elemental content by Huffman Laboratories [Huffman and Stuber, 1985]. For determination of 13C, lyophilized fulvic acid was combusted to CO2 at 550°C in vacuum-sealed Pyrex glass tubes with CuO and Ag wire. Ratios of 13C/12C were measured on a Prism isotope ratio mass spectrometer (V.G. Isogas) and reported in δ‰ relative to Peedee belemnite (PDB). Analytical errors were less than ±0.2‰. In order to determine the range of δ13C characteristic of upper Snake River DOC, we averaged δ13C values from three samples: the one obtained directly above injection site 0 m, at Snake River Site 386 m (before the injection commenced), and from an archive (1985) sample collected above the confluence with Deer Creek (2217 m downstream of site 0 m). For other fulvic acid analyses the values determined for the sample from above the injection site are used as representative of the upper Snake River. Samples were analyzed by quantitative natural abundance 13C-NMR spectroscopy with a solid-state Chemagnetics CMX spectrometer. Cross-polarization magic angle spinning (CP/MAS) 13C spectra were measured at 50.3 MHz with a 7.5-mm-diameter probe. The spinning rate was 5000 Hz, with a contact time of 2 ms and a pulse delay of 1 s. A line-broadening of 100 Hz was applied in Fourier transformation of the free induction decay data. Chemical shifts were calibrated with hexamethylbenzene (methyl carbon at 17.35 ppm). The spectral regions of the functional groups were designated as AL-I (0–62 ppm) aliphatic carbons; AL-II (62–90 ppm) alcohol carbons in carbohydrates and other aliphatic alcohol and ethers; AL-III (90–110 ppm) primarily aromatic carbons in carbohydrates; AR (110–160 ppm) aromatic and olefinic carbons; C-I (160–190 ppm) carboxylate and amide carbons; and C-II (190–230 ppm) carboxyl carbons in aldehydes and ketones. The δ15N values were determined using a Carlo-Erba (Milan, Italy) NA1500 elemental analyzer coupled to a VG Isochrom stable isotope ratio mass spectrometer (Micromass, Inc., Manchester, England).

3.6. Data Analysis

[23] The one-dimensional transport with inflow and storage (OTIS) model [Runkel, 1998, 2000] was used to interpret the data. Allan [1995] described the significance for ecological stream function of the transient storage of water and solute represented in these equations. In separate studies of comparative stream systems, Valett et al. [1996] and Mulholland et al. [1997] have demonstrated the use of the concept in characterizing hydrological factors related to nutrient retention. J. Harvey and coworkers [Harvey and Bencala, 1993; Harvey et al., 1996; Harvey and Fuller, 1998; Harvey and Wagner, 2000; Choi et al., 2000] have established that the parameters embodied in the transient storage concept represent the basic hydrometric characteristics of hyporheic exchange flows occurring at the rapid timescales similar to in-stream transport. Finally, we note that the transient storage concept, while mathematically distinct, attempts to represent physical processes similar to those of the aggregated dead-zone model (see e.g., the application by Castro and Hornberger [1991] and the comparative analysis by Less et al. [2000]. The equations in OTIS are

equation image
equation image

where A is the main channel cross-sectional area [L2 ], As is the storage zone cross-sectional area [L2], c is the main channel solute concentration [M L−3], cL is the lateral inflow solute concentration [M L−3], cs is the storage zone solute concentration [M L−3], D is the dispersion coefficient [L2 T−1], Q is the volumetric flow rate [L3T−1 ], qL is the lateral inflow rate [L3T−1 L−1], t is time [T], x is distance [L], α is the storage zone exchange coefficient [T−1], and λ is a first-order, irreversible sorption rate coefficient [T−1]. Within the OTIS code package, these coupled equations are solved by the technique presented by Runkel and Chapra [1993] based on the Crank-Nicolson algorithm. Parameters for the simulations are determined by trial and adjustment to achieve a good visual alignment of measured concentrations and simulated values. The parameter determination sequence follows the approach discussed by the Stream Solute Workshop [1990].

4. Results

4.1. Chemical Characteristics of Injectate DOC and Snake River DOC

[24] The DOC concentration of Suwannee River water was 58.8 mg C L−1. Of this, 82% (48.2 mg C L−1) was fulvic acid that sorbed to iron oxides in the fractionation experiment, referred to as sorbable fulvic acid. Another 7% (42 mg C L−1) was fulvic acid that did not sorb, referred to as nonsorbable fulvic acid. Finally, 11% (6.4 mg C L−1) was nonhumic material (i.e., dissolved organic material that did not sorb to XAD-8 resin at low pH and also did not contribute to the measured UV absorbance of the Suwannee River water).

[25] The Suwannee River sorbable and nonsorbable fulvic acid had distinct chemical characteristics, which can be seen by comparing the 13C-NMR spectra (Figure 4 and Table 1). As expected, on the basis of previous laboratory and field studies, the sorbable fulvic acid had a greater aromaticity (AR) than the nonsorbable (20% versus 12%). The upstream Snake River sample had an intermediate aromaticity. The carboxyl (C-I) content showed the same pattern (17% versus 13%, with 14% of the Snake River fulvic acid). For the upstream Snake River sample, the values of aromaticity (AR) and carboxyl (C-I) content are lower and the aliphatic (Al-I) content is higher compared with the results of the previous analyses of Snake River fulvic acids using liquid-state 13C-NMR [McKnight et al., 1992], which probably reflects a difference in optimization for obtaining quantitative spectra between the two methods. Nonetheless, these results confirm that the same pattern of differences in the core carbon structure and reactive functional groups would be expected to result from fractionation by iron oxide sorption in the field-scale experiment.

Figure 4.

The 13C-NMR spectra of fulvic acid characterizing input (Snake River background, Suwannee River sorbable and nonsorbable) and first mixed site (Snake River at 34 m). Spectral bands are C-I, 160–190 ppm; Al-III and AR, 90–160 ppm; Al-II, 60–90 ppm; and Al-I, 0–60 ppm.

Table 1. Chemical Characteristics of Suwannee River Fulvic Acid Injected Into the Snake River, in Comparison With Those of the Snake River Fulvic Acid From the Upstream Site
SampleSUVA,a (mg C)−1δ13C, ‰δ15N, ‰Elemental Content Percent by WeightCarbon Distribution Percent by Weightb
CHONSAl-IAl-IIAl-IIIARC-IC-II
  • a

    Specific ultraviolet absorbance.

  • b

    The functional groups in the carbon distribution are as designated in section 3.5. The ∼1% difference from 100% of the percent distribution summations represents round-off error.

  • c

    For the δ13C analysis a representative background value is presented, characteristic of upper Snake River dissolved organic carbon. This value is the average from three samples obtained directly above injection site 0 m, at Snake River site 386 m (before the injection commenced), and from an archive (1985) sample collected 2217 m downstream of site 0 m.

Suwannee River sorbable fraction0.049−27.5−4.145.64.244.10.70.353517820172
Suwannee River nonsorbable fraction0.025−28.9−5.551.45.239.40.50.364916612133
Snake River background, 0 m0.019−25.c−0.147.24.839.41.10.684415716153

[26] The structural differences evidenced in the 13C-NMR spectra are reflected in the elemental analysis (Table 1). Specifically, the higher oxygen content of the sorbable fulvic acid is consistent with the greater carboxyl content, and the lower hydrogen content of the sorbable fulvic acid is consistent with a greater abundance of sp2-hybridized (primarily aromatic) carbon atoms. The N and S contents of the two Suwanee River fractions are lower than the N and S contents of the background Snake River fulvic acid. The N content of the sorbable fraction was slightly higher than that of the nonsorbable fraction, which is consistent with the involvement of N-containing functional groups in surface complexation with iron oxides. However, there was no difference in the S content between the sorbable and nonsorbable fractions.

[27] The specific absorbance at 254 nm of Suwannee River water was 0.0425 (mg C)−1 [McKnight et al., 1994], and specific absorbances of the sorbable and nonsorbable fulvic acid were 0.0492 and 0.0253 (mg C)−1, respectively. The higher specific absorbance of the sorbable fulvic acid was again consistent with the greater aromaticity. The specific absorbance for the whole river water was only slightly less than that of the sorbable fulvic acid because the sorbable fulvic acid was a large portion of the DOC. The upstream Snake River fulvic acid had a specific absorbance that was even lower than that of the nonsorbable fulvic acid, despite a higher aromaticity. This low specific absorbance may reflect the influence of in-stream photochemical processes.

[28] The δ13C values of the Suwannee River fulvic acid fractions were lighter than the average δ13C value for the Snake River fulvic acid, which was based on analysis of samples from the upstream site, from site 386 m prior to the injection, and from a site just above the confluence with Deer Creek collected during snowmelt in 1985 [McKnight et al., 1992]. These differences in δ13C values provide an additional chemical tool for tracking the fate of the injected fulvic acid in the experiment. Similarly, the δ15N values were lighter for the Suwanee River fulvic acid fractions than for the Snake River background fulvic acid, which provides a chemical tool for tracking the fate of the N-containing fraction of the fulvic acid relative to the fate of the bulk fulvic acid.

4.2. Simulation of DOC Injection Experiment

[29] Analysis of lithium data using OTIS was used to select appropriate physical parameters for the reaches bounded by the sampling sites (Table 2, Figure 5). At approximately 1100 LT on the day of the experiment, a steady rain started and persisted until about 1500 LT. Results from OTIS simulations for the lithium tracer assuming constant discharge indicate that the rain had only a slight effect on transport because a significant increase in stream discharge would have resulted in a drop in the concentration of lithium (Figure 5).

Figure 5.

Reach-scale transport simulations of lithium tracer.

Table 2. Reach-Scale Transport Parameters for the Upper Snake River Determined by Simulating Injection of Lithium Tracer
ReachReach Length, mA, m2qL, m3 s−1 m−1As, m2D, m2 s−1α, s−1
Injection to 141 m1410.550.000390.100.901 × 10−5
141 m to 386 m2451.250.000130.200.901 × 10−5
386 m to 681 m2951.550.000130.200.901 × 10−5

[30] For the reactive transport simulation of the DOC injection, simulations were performed in two steps. First, absorbance was simulated as a surrogate for the transport of fulvic acid. Second, DOC concentrations were calculated from these absorbance simulation results. Measurements of the absorbance of samples at 254 nm showed a very clear signal in the stream measurements. We used the absorbance as an indicator of the transport of fulvic acid in the following way. We assumed that the sorbable fraction experienced irreversible, kinetically controlled sorption to the streambed and that the other fractions were transported conservatively. We also assumed that the specific absorbances of the different fractions of fulvic acid (Table 1) would remain unchanged and that the background absorbance in the Snake River remained constant at 0.030. Using these assumptions, the downstream transport of fulvic acid as indicated by absorbance was simulated using OTIS with a first-order kinetic sorption rate constant applied to the sorbable fulvic acid and with the remainder of the DOC, corresponding to the sum of the nonsorbable fulvic acid and the nonhumic material, treated as conservative (Table 3 and Figure 6). (Note that injection of DOC could not be held absolutely constant; measured temporal variations in the injection rates are reflected in the simulations and in the observed data.)

Figure 6.

Reach-scale transport simulations of absorbance.

Table 3. Reach-Scale Kinetic Rate Coefficients for Irreversible Sorption of the Sorbable Fraction of the Injected Suwannee River Water
ReachReach Length, mλ, s−1
Injection to 141 m1411.0 × 10−3
141 m to 386 m2451.5 × 10−4
386 m to 681 m2952.5 × 10−4

[31] Simulated DOC concentrations were calculated directly from the absorbance results by summing the simulated absorbance of the sorbable fraction divided by its specific absorbance, the simulated absorbance of the nonsorbable fraction divided by its specific absorbance, the conservatively transported nonhumic fraction of the injected Suwannee River DOC, and the Snake River background DOC (taken to be constant at 1.6 mg L−1). Whereas absorbance measurements are quite precise, measures of DOC at the concentrations observed have errors that are significant. Using the standard deviation of the absolute errors in repeat measurements as an error bar, the OTIS simulations for DOC are reasonable inasmuch as almost all of the simulated values for the 141- and 386-m sites were within the confidence limits (Figure 7).

Figure 7.

Reach-scale transport simulations of DOC.

4.3. Chemical Characteristics of DOC in the Snake River During the Injection

[32] The analysis of the tracer experiment indicated significant removal of sorbable fulvic acid within the first 141 m of the experimental reach. This result is based on analysis of samples collected throughout the experiment, whereas the large-volume samples were collected over only a short period of time during one point of the plateau concentration and do not correspond to a set of synoptic samples. However, the chemical characteristics of the dissolved fulvic acid allow us to evaluate the hypotheses we used in analyzing the tracer experiment, specifically that the sorbable fulvic acid experienced irreversible, kinetically controlled sorption to the streambed and that the other fractions were transported conservatively.

[33] The δ13C values progressed to heavier values in the downstream direction from the injection site, going from −26.7‰ at site 10 m to −25.3‰ at site 141 m (Table 4). This is the reach where the transport analysis showed that the first-order removal rate was highest by an order of magnitude. Using the measured δ13C values for the sorbable and nonsorbable Suwannee River fulvic acid and for the upstream Snake River fulvic acid, we can calculate the distribution of these two sources required to yield the measured δ13C values at sites 10, 34, and 141 m. To account for the small contribution of the nonsorbable Snake River fulvic acid, we assumed that at 10 m the ratio of nonsorbable to sorbable Snake River fulvic acid was the same as in the injectate (0.09) and that at 34 m the ratio of nonsorbable Snake River fulvic acid to Suwannee River fulvic acid was the same as at the 10-m site (0.16). The sorbable fulvic acid accounted for 61% of the total dissolved fulvic acid at site 10 m, 40% at site 34 m, and 10% or less at site 141 m. The δ13C value at site 681 m was the same as that at site 141 m, but the value at site 386 m was slightly greater. This slight discrepancy may result from the samples not having been collected in truly synoptic manner following a water parcel. The δ13C values support the interpretation of more rapid removal by sorption in the first reach, but the uncertainty of ±0.5‰ in the value for the background Snake River fulvic acid prevents using these data to learn about the slower rates of sorption in the reaches farther downstream.

Table 4. Chemical Characteristics of Fulvic Acid Collected in the Snake River During the Injection of the Suwannee River Water
Sample (Time of Collection During Injection)SUVA,a (mg C)−1δ13C, ‰δ15N, ‰Elemental Content Percent by WeightCarbon Distribution Percent by Weightb
CHONSAl-IAl-IIAl-IIIARC-IC-II
  • a

    Specific ultraviolet absorbance.

  • b

    The functional groups in the carbon distribution are as designated in section 3.5. The ∼1% difference from 100% of the percent distribution summations represents round-off error.

  • c

    For the δ13C analysis a representative background value is presented, characteristic of upper Snake River dissolved organic carbon. This value is the average from three samples obtained directly above injection site 0 m, at Snake River site 386 m (before the injection commenced), and from an archive (1985) sample collected 2217 m downstream of site 0 m.

Suwannee River sorbable fraction0.049−27.5−4.145.64.244.10.70.353517820172
Suwannee River nonsorbable fraction0.025−28.9−5.551.45.239.40.50.364916612133
Snake River background, 0 m0.019−25.c−0.147.24.839.41.10.684415716153
Snake River site 10 m (1110 LT)−26.7−3.345.75.71.00.66
Snake River site 34 m (1030 LT)−26.3−0.949.95.31.40.64371581916
Snake River site 141 m0.024−25.2−3.142.75.80.840.62411491715
Snake River site 386 m0.027−26.0−1.344.05.91.20.48371481817
Snake River site 681 m0.021−25.342.16.11.30.74421481716

[34] The changes in SUVA and in δ13C downstream of the injection site fall along a mixing line (Figure 8). The nonsorbable fulvic acid is chemically distinct, representing an outlier. The sorbable Suwanee River fulvic acid and the Snake River background fulvic acid both contribute to the fulvic acid in the stream during the experiment. Because the nonsorbable fulvic acid was only 7% of the DOC in the injectate, the nonsorbable fulvic acid has no discernable influence on the position along the mixing line of the downstream samples. It should be noted that the position of the samples along the mixing line does not match with the location sequence in the stream; however, the samples from site 386 m are closer to the sorbable end-member than the samples from sites 141 and 681 m.

Figure 8.

Change in specific ultraviolet absorbance (SUVA) relative to changes in δ13C for the four Snake River sample sites during the Suwannee River DOC injection.

[35] The 13C-NMR spectra and the distribution among the major carbon moieties generally support the transport analysis in that they indicate that the sorbable “signal” of greater aromaticity and carboxyl content decreased downstream (Figure 9 and Table 4). Changes in the percent of aliphatic (Al-1), aromatic (AR), and carboxyl (C-1) carbon relative to the changes in δ13C for the downstream samples also follow mixing lines (Figure 10). As was shown for SUVA, the nonsorbable sample is an outlier in these diagrams, and the sorbable fulvic acid and the Snake River background both contribute to the fulvic acid in the stream during the experiment. The data for the fulvic acid from site 34 m are closest to the sorbable end-member. As for the SUVA diagram, the fulvic acid from site 386 m is positioned closer to the sorbable end-member than the fulvic acid from sites 141 and 681 m.

Figure 9.

The 13C-NMR spectra of fulvic acid collecte at four Snake River sample sites during Suwannee River DOC injection. Spectral Bands are C-I, 160–190 ppm; Al-III and AR 90–160 ppm; Al-II, 60–90 ppm; and Al-I, 0–60 ppm.

Figure 10.

Change in aliphatic, aromatic, and carboxyl carbon contents relative to changes in δ13C for the four Snake River sample sites during the Suwannee River DOC injection.

[36] The results of the elemental analysis for C and H for the downstream sites show subtle differences that are least consistent with the 13C-NMR results. More significantly, the elemental analysis shows that the N and S contents of fulvic acids from the downstream sites do not follow a simple mixing of the N- and S-poor sorbable fraction with the N- and S-rich Snake River background. The N and S contents of the downstream samples were typically significantly greater than the elemental content of the sorbable fulvic acid, and at some sites even greater than the elemental content of the Snake River background sample.

[37] The measurements of δ15N provide additional data for understanding the results for the N-content analysis. The N contents of the downstream samples do not fall along the mixing line between the sorbable fulvic acid and the Snake River background sample (Figure 11). Rather, three of the four downstream samples are shifted toward higher N contents, indicating an enrichment of the fulvic acid pool in the water column with N-rich Snake River fulvic acid during the experiment. Although there are no isotopic data to indicate that the increase in S content was from addition of S-rich Snake River fulvic acid, because the S content in the background Snake River sample is much greater than that in the sorbable Suwanee River sample, an addition from a Snake River fulvic acid source is likely. As in the diagrams comparing the major carbon moieties, the nonsorbable fulvic acid is an outlier with the lightest δ15N value and the lowest N content.

Figure 11.

Change in nitrogen content relative to changes in δ15N for the four Snake River sample sites during the Suwannee River DOC injection.

4.4. Microbial Processes Involving DOC in the Snake River During the Injection

[38] An assumption in our analysis of the transport experiment is that the dominant reactions were abiotic chemical processes. This assumption was supported by the results of the direct bacterial counts and the bioassay experiments. The bacterial abundance (Table 5) in the Suwannee River injectate (2.9 × 106 cells mL−1) was much greater than that at all sites in the Snake River (1.8–2.4 × 105 cells mL−1). Using an estimate of 20 fg of C per cell, these values correspond to 0.3–0.5 uM of C, which was much less than the DOC. Thus bacterial growth at rates typical of natural waters would not have influenced DOC concentrations during the injection. The maximum growth rates measured in the bioassay experiments using the Paine Run inoculum at 20°C ranged from 0.125 to 0.183 h−1, corresponding to generation times of 3.8–5.6 hours. In the bioassay with the 30-to-1 dilution of the Suwannee River injectate the maximum growth rate was 0.059 h−1. The maximum growth rates measured in the bioassay experiments using the Snake River inoculum at 4°–6°C were slowest at 0.018 h−1, corresponding to a generation time of 64 hours. If these growth rates occurred during the experiment, which assumes that the lag phase observed in the bioassays did not occur in the stream, these generation times are nonetheless much slower than the travel time through the experimental reach, indicating that not even one doubling of the microbial biomass would have occured. Using an exponential growth equation and an assumed growth efficiency of 30% further shows that less than 1uM of organic carbon would have been assimilated.

Table 5. Bacterial Counts and Growth Indicators From Bioassay Experiments
SampleBacterial Abundance cells mL−1 × 106NaMaximum Growth Rate, h−1Minimum Generation Time, Hours
  • a

    N is the number of distinct samples collected at different times at the identified site.

Suwannee River injectate2.923
Snake River site 0 m0.18120.1255.6
Snake River site 34 m0.19290.1833.8
Snake River site 141 m0.28190.1664.2
Snake River site 386 m0.2416

[39] Although these results indicate that microbial uptake of DOC in the water column did not influence DOC concentrations during the injection, higher rates of microbial uptake could have occurred in the hyporheic zone. However, analysis of the tracer experiments using OTIS with first-order uptake occurring only in the water column, and not in the hyporheic zone, yielded a simulation which matched the data well within the uncertainty of the data. The rate of abiotic chemical removal of fulvic acid in the water column was so rapid as to overwhelm any changes in DOC associated with microbial uptake in the hyporheic zone. Thus the direct analyses of microbial biomass and growth rate in the water column support the assumption that the sorption of fulvic acid by iron oxides on the streambed was the dominant reaction.

5. Discussion

[40] The results of our experiment injecting Suwannee River water into the Snake River show clearly that stream-streambed interactions where abundant iron oxyhydroxides are present can be a strong controlling process for the concentration and composition of water-column DOC. A “loss rate” coefficient of about 10−4–10−3 s−1 is equivalent to a half-life of about 700–7000 s.With flow velocities of about 0.3–0.7 m s−1, a reach of several hundred meters to several kilometers would be indicated for removal of half of sorbable fulvic acid input to the stream. The sorbable fulvic acid fraction of soil water DOC flushed into streams like the Snake River during snowmelt would be depleted significantly during transit along the stream. DOC concentrations in soil lysimeters in the Snake River catchment typically are 10–50 mg C L−1 early in the snowmelt period. This high-DOC water is flushed into the stream, raising concentrations in the stream to a peak of 2–3 mg C L−1 (background is ∼1 mg C L−1) before the hydrograph peaks. In the stream and in lysimeter samples, fulvic acid accounts for about 50% of the DOC. The recession time constant for DOC in the stream is about 50–100 days, whereas the recession time constants for DOC in lysimeters is about 10–25 days [Boyer et al., 1997]. The difference in recession times has been attributed to asynchronous melting of the snowpack, with lower-elevation parts of the catchment contributing water to the stream first and high-elevation portions contributing last [Boyer et al., 2000]. The sorption process by the streambed that we report here may contribute to the time course of this annual dynamic. That is, dissolved fulvic acid in the water flushed into the stream at the upper end of the catchment must flow about 5 km to the confluence with Deer Creek. Our results suggest that at least half and possibly much more of this DOC would be “lost” to sorption on streambed sediments during transit. Thus the declining concentrations on the receding limb of the hydrograph may be due in part to the interaction between the stream and its bed.

[41] The simulations using OTIS indicate that the irreversible sorption occurs right at the streambed and not deep within the hyporheic zone. Our results do not indicate the ultimate fate of the sorbed DOC, however. If the aromatic fractions of DOC are sorbed on freshly precipitated iron oxyhydroxides at the surface of the streambed, then photoreduction could liberate at least some of this material during clear days on unshaded portions of the stream and DOC sorption could be greater at night. However, intense sunlight and photoreduction could produce photoproducts that react with DOC and cause the released DOC to be oxidized to CO2 or to be altered in character. Another possible fate of sorbed DOC is microbial uptake. The sorbed DOC would be expected to be less bioavailable than other DOC fractions because the sorbed DOC is primarily the aromatic fulvic acid fraction, which has been demonstrated to have low quality as a substrate for stream microbial communities [Moran and Hodson, 1990; Leff and Meyer, 1991]. Finally, sorbed DOC on streambed iron oxides may be lost from the headwater stream ecosystem during snowmelt when high flows scour the streambed.

[42] One conceptual model for describing downstream changes in the composition of DOC is that the more reactive component of the injected Suwannee River water competitively displaces more refractive fractions already sorbed on the bed. The results of Gu et al. [1996], who report that strongly binding organic compounds will be competitively adsorbed and displace weakly bound organic compounds, are consistent with this conceptual model. The model indicates that specific absorbance in streamwater would go down as the injected pulse moves downstream but that the fulvic acid concentration would remain relatively unchanged because more refractive material (with a lower specific absorbance) sorbed to the streambed sediments would be displaced during the sorption of the sorbable fulvic acid of the injected water. Most of the chemical characterization results are inconclusive with respect to a possible competitive displacement of previously sorbed fulvic acid. Our results for loss of adsorbance and change in isotopic signature clearly show sorption of the fulvic fraction. Because the displaced fulvic acid probably had a similar isotopic signature as the fulvic acid dissolved in the Snake River, however, competitive displacement cannot be resolved from the isotopic data.

[43] A finding, which cannot be explained simply by irreversible sorption without displacement, is that at sites 10, 34, and 386 m the N and S contents of the fulvic acids are higher than that which would result in a mixture of sorbable Suwannee River fulvic acid and Snake River fulvic acid. Previously study of fulvic acid sorption in the Snake River showed that sorbable fulvic acid was enriched in N and S. Analysis of amino acid residues in the sorbable and nonsorbable fulvic acids confirmed that this trend was caused by preferential binding by strong iron-binding groups such as glutamic and aspartic acid [McKnight et al., 1992]. Thus the sorbable Suwannee River fulvic acid with low N and S contents could be displacing previously sorbed Snake River sorbable fulvic acid, causing a rise in the N and S of the mixture.

[44] In the simulations we showed that the first-order uptake coefficient in the first reach (1.0 × 10−3) was an order of magnitude greater than in the two downstream reaches (1.5 × 10−4 and 2.5 × 10−4). When considered with the chemical characterization results, this significant difference in loss rate coefficient can be attributed to chemical heterogeneity within the sorbable fulvic acid fraction. Because the abundance of oxides on the streambed and the hydrologic conditions are generally uniform over the entire experimental reach, a downstream change in the interactions with streambed hydrous iron oxides is not a plausible explanation for the decrease in the loss rate coefficient.

[45] The sorption of fulvic acid by ion oxide on the streambed can be represented as

equation image

where A- corresponds to the ionized fulvic acid and FeOH= corresponds to the iron oxide surface site. An assumption of irreversible sorption corresponds to assuming that k2 is negligible compared with k1. In this case, the rate of formation of the fulvic acid–ferric surface complex is k1(FeOH+=)(A-), and λ, the loss rate coefficient, can be thought of as a representative value for k1(FeOH+=) at the stream reach scale. If we assume that the effective concentration of available fulvic acid sorption sites is uniform longitudinally, then the order of magnitude decrease in λ values between the first 141 m and the downstream reaches would correspond to an order of magnitude decrease in k1, or sorption affinity, between the most reactive fraction of the sorbable fulvic acid and the remainder of the sorbable fulvic acid. Such an order of magnitude range in reactivity of sorbable fulvic acid is well within the range observed in studies of proton and metal binding by Suwannee River fulvic acid [e.g., Averette et al., 1989]. We conclude that for the purpose of understanding fulvic acid sorption at the stream scale we can represent the variation in reactivity well enough with two components having very fast and fast loss rate coefficients (corresponding to very sorbable and sorbable fulvic acid). Because of general similarities between the chemical characteristics of Suwannee River fulvic acid and other fulvic acids derived from plant litter and soils, we suggest that this approach could be used in a reactive solute transport model to represent the reactivity of catchment fulvic acid in lateral inflow to a stream.

Acknowledgments

[46] This work was funded under grants in hydrologic science from the NSF and by the U.S. Geological Survey Toxic Substances Hydrology Program. Numerous individuals from the offices of the U.S. Geological Survey and INSTAAR (University of Colorado) participated, often creatively, in the significant logistics of the injection of approximately 23,000 L of water collected from the Suwannee River, Georgia. We particularly acknowledge the efforts of Greg O'Neill in coordinating the assistance of U.S. Geological Survey Lakewood, Colorado, Subdistrict Office personnel. We thank Ishi Buffam and Steve Macko for providing the microbial analyses and the carbon isotopic analyses, respectively.

Ancillary