We hypothesized that C. fluminea could remove anthropogenic N at a rate sufficient to affect particulate N in the water column of North Buffalo Creek, North Carolina, USA, a 4th-order stream that receives treated urban wastewater. We used stable-isotope analysis to evaluate trophic relationships between seston and C. fluminea and conducted field sampling and laboratory experiments to evaluate the potential qualitative and quantitative effects of C. fluminea on seston. Corbicula fluminea δ15N was 3 to 5‰ enriched compared to seston along a longitudinal transect downstream of the wastewater treatment plant (WWTP), a result consistent with use of seston N. However, seston δ13C declined and C. fluminea δ13C showed no pattern with distance downstream, a result that was inconsistent with a trophic relationship between seston and C. fluminea. In 2 laboratory experiments designed to measure filter-feeding rate and qualitative effects on seston, seston ash-free dry mass and chlorophyll a data indicated that C. fluminea either was not filtering or was filtering at a rate insufficient to affect seston concentration over the course of the experiments. δ15N data showed that the sediment was an N source for C. fluminea, but δ13C and C∶N data from the same experiments indicated that C. fluminea probably affected seston quality by suspending benthic algae and returning settled algae to the water column. These results illustrate that the food sources for C. fluminea and implications of C. fluminea activity in stream ecosystems should be evaluated more fully.
Anthropogenic N addition to terrestrial systems is increasing worldwide and, in turn, is increasing N loading to freshwaters (Peterson et al. 2001, Mulholland et al. 2008). Most urban streams receive both point and nonpoint sources of anthropogenic N. Point sources, such as wastewater treatment plants (WWTP), are highly regulated, but they continue to affect urban streams by contributing 50 to 90% of their annual nutrient loading (Haggard et al. 2001, 2005). WWTP effluent has a long-lasting effect on urban-stream ecosystems by altering their natural ability to process N effectively. Increased anthropogenic NO3− transported from streams and rivers to downstream coastal waters accelerates coastal eutrophication, reduces biodiversity, and causes the formation of hypoxic zones and algal blooms (Rosenzweig et al. 2008, Weijters et al. 2009). As N input increases, N-processing efficiency and proportional denitrification decreases, resulting in greater N export to downstream reaches (Peterson et al. 2001, Mulholland et al. 2008). For example, denitrification rate was high in an urban stream receiving WWTP effluent, but denitrification was unimportant compared to total N loads, and therefore, large quantities of N were exported downstream (Lofton et al. 2007). If a stream is unable to retain N, then the downstream system will receive higher N loads (Grimm et al. 2005). N loading from upstream areas is a particularly serious problem in North Carolina estuaries where increased N is causing water-quality problems (Burkholder et al. 2006).
Stable isotopes of N are useful for detecting anthropogenic N in aquatic systems (Ulseth and Hershey 2005). Stable-isotope ratios can be used to trace organic matter through food webs because consumers typically are enriched 3 to 5‰ in 15N and 0.5 to 1‰ in 13C relative to their food sources (Peterson and Fry 1987). Sewage-derived N below WWTPs is readily traced through aquatic food webs because it has a higher δ15N value than most natural N sources (Wayland and Hobson 2001, DeBruyn and Rasmussen 2002, Ulseth and Hershey 2005, Nishikawa et al. 2009).
Corbicula fluminea (Asian clam) is an invasive species in many freshwater ecosystems. It has a high potential to alter seston quality and concentration through filter feeding. Stable-isotope analysis has shown that C. fluminea diet is very similar to the diet of the invasive zebra mussel Dreissena polymorpha (Delong 2010), which reduces phytoplankton abundance in lakes (Mackie 1991). In lakes and large rivers, C. fluminea also decreases phytoplankton abundance (Cohen et al. 1984, Newell et al. 2005) and zooplankton density through filter feeding, thereby increasing water clarity (Sousa et al. 2009). The filtering rate of C. fluminea has been estimated to be sufficient to clear the water column in the Chowan River, North Carolina, every 1.6 d (Lauritsen 1986); a reach of the Potomac River, Virginia, in 3 to 4 d (Cohen et al. 1984); and Meyers Branch, South Carolina, in 21 h (Leff et al. 1990).
Corbicula fluminea has the potential to have direct and indirect effects on its sediment habitat. Reduction of phytoplankton abundance by filter feeders can lead to nutrient burial through biodeposition (Newell et al. 2005). Biodeposits of C. fluminea have a high N content relative to those of native mussels (Atkinson et al. 2011). Several investigators have suggested that C. fluminea uses both filter feeding and deposit feeding to meet nutritional needs (Yeager and Cherry 1994, Hakenkamp and Palmer 1999, Hakenkamp et al. 2001, Vaughn and Hakenkamp 2001, Cummings and Graf 2010). Pedal feeding can result in an estimated sediment organic matter consumption of 50 mg individual−1 d−1 (Hakenkamp and Palmer 1999).
We evaluated the potential for C. fluminea to reduce particulate anthropogenic N in transport via filter feeding in North Buffalo Creek, an urban stream in Greensboro, North Carolina (USA). We hypothesized that C. fluminea would be able to filter at a rate sufficient to substantially reduce anthropogenic particulate N in the water column and, therefore, to provide an ecosystem service. We used field campaigns and laboratory experiments: 1) to assess whether suspended particulate organic matter (seston) appeared to be the dominant food source for C. fluminea in North Buffalo Creek, and 2) to estimate the effect of C. fluminea filtering activity on quality and quantity of seston.
North Buffalo Creek originates in Greensboro, North Carolina, a city with a population of 260,083 (Greensboro Convention and Visitors Bureau 2012), and receives both point and nonpoint pollution from the city. The stream is the headwater of the Cape Fear River Basin and is 4th order in the section we studied. The major N point source is the North Buffalo WWTP (Ulseth and Hershey 2005, Lofton et al. 2007). The WWTP primarily treats residential sewage from the northern ½ of the city and contributes ∼50% of the baseflow discharge to the stream. Since 1980, the capacity of the North Buffalo WWTP has been 60,567 m3/d (City of Greensboro Water Resource Department 2011). North Buffalo WWTP uses a process ending with sand filters and the addition of sodium hypochlorite to mitigate pathogenic bacteria. However, nutrient removal is not part of the wastewater treatment process (City of Greensboro Water Resource Department 2011). North Buffalo WWTP is permitted to discharge 8 mg/L NH4+ to the river in winter and 4 mg/L in summer (as cited by Lofton et al. 2007). Previous investigators (Ulseth and Hershey 2005, Lofton et al. 2007) found elevated levels of dissolved inorganic N (DIN) throughout the study reach. Elevated levels of DIN are a consistent occurrence in streams receiving urban runoff and wastewater (Grimm et al. 2005, Lofton et al. 2007). Average nutrient concentrations in water at site B, Rankin Mill Road (Fig. 1), were ∼0.13 mg/L NH4+ and ∼5.75 mg/L NO3− and ∼0.2 mg/L NH4+ and ∼9.5 mg/L NO3− in previous studies (Lofton et al. 2007, Ulseth and Hershey 2005, respectively). A site upstream of the WWTP appeared to serve as an N sink where ∼46% of NO3-N was removed via denitrification, whereas at downstream sites only ∼2.3% of NO3-N was removed through denitrification (Lofton et al. 2007). δ15N of seston at site B was ∼8.3‰ (Ulseth and Hershey 2005), a value consistent with values from other streams receiving WWTP input (Mayer et al. 2002).
North Buffalo Creek has no permitted discharge points upstream of the WWTP (as cited in Ulseth and Hershey 2005). Upstream of site A (Fig. 1), land use is virtually all urban except in forested areas in urban parks and in narrow buffer zones. Runoff is diverted directly to the stream. Between sites A and B, land uses include residential development, forest, a landfill, other municipal and county facilities, and the WWTP. Below the WWTP, land use is primarily forested with some rural pastureland that has forested buffer zones (Lofton et al. 2007). North Buffalo Creek has been studied for a number of years as a part of a US Environmental Protection Agency (EPA) mandated Total Maximum Daily Load (TMDL) project targeting fecal coliform bacteria (NCDENR 2004). North Buffalo Creek has been used as a research site for studies of urban stream food webs, restoration effects, incorporation of wastewater-derived N into aquatic food webs, and the capacity of the stream to process N (Ulseth and Hershey 2005, Northington and Hershey 2006, Kalcounis-Rüppell et al. 2007, Lofton et al. 2007, Hines and Hershey 2011).
Site A, Summit Avenue has a narrow forested buffer and is the only site upstream of the WWTP that was sampled during our study. Site B is ∼8 stream km downstream of the North Buffalo WWTP and is outside Greensboro city limits (Lofton et al. 2007). Site D, McLeansville Road, ∼7 stream km downstream of site B is just above the confluence with South Buffalo Creek. Site C, Creekview Road, is at the only road crossing between sites B and D (Fig. 1).
Trophic relationships between seston and C. fluminea in North Buffalo Creek
To evaluate objective 1, we sampled C. fluminea and seston for stable-isotope analyses in summer and winter. Summer sampling occurred on 23 June 2009 at site B and on 26 June 2009 at sites A, C, and D. Discharge conditions were 0.65 m3/s and 0.67 m3/s on those dates (US Geological Survey [USGS] gage 02095500 at site B), respectively. Discharge at site A (not measured) would have been much lower because it is above the WWTP. Winter sampling occurred on 15 December 2009 (discharge = 1.27 m3/s at site B) at ∼500-m intervals along the longitudinal transect from sites B to D, which we accessed by canoeing the reach. We used a global positioning system (GPS) to measure longitudinal stream distances.
We held C. fluminea individuals from each site overnight in deionized water and then dried them at 60°C for ≥24 h. We ground dried C. fluminea tissue to a powder, which we placed in 4 × 6-mm pressed Sn capsules and sent to the University of California (UC)-Davis Stable Isotope Laboratory for measurement of δ15N and δ13C.
We collected water for seston analyses in 4-L cubitainers and returned it to the laboratory. We vacuum-filtered subsamples (400 mL) onto 6 precombusted glass-fiber filters (0.7-µm pore size) per site. We dried all filters at 60°C for ≥48 h and prepared 3 of the 6 filters/site for stable-isotope analysis. We put ¼ of each filter for stable-isotope analysis in a Sn capsule and sent the samples to UC-Davis Stable Isotope Laboratory for measurement of δ15N and δ13C. We calculated seston C∶N for each sampling site with the data received from the UC-Davis Stable Isotope Laboratory. We used the 3 remaining filters from each site to estimate seston ash-free dry mass (AFDM) after combustion in a muffle furnace at 500°C.
Evaluation of the effect of C. fluminea filtering on seston quality and quantity
To evaluate objective 2, we assessed field density on 2 sampling dates and conducted 2 laboratory experiments. We estimated field density to evaluate whether C. fluminea was abundant enough to affect seston. We used the laboratory experiments to measure effects of C. fluminea filtering activity on seston quantity and quality with and without sediment present at realistic field densities. We quantified the effect of C. fluminea on seston quantity as change over time in seston AFDM. We assessed the effect of C. fluminea on seston quality by measuring change in seston δ15N, δ13C, C∶N, and chlorophyll a (chl a) over time.
We estimated C. fluminea density with a Surber sampler (0.093 m2) when we sampled individuals for isotope analyses. Corbicula fluminea density is higher in summer than in winter (Cohen et al. 1984). We collected 3 replicate samples in pool, riffle, and run habitats at each sampling site on the summer sampling dates to assess whether C. fluminea density was related to habitat. In winter, we collected 3 replicate Surber samples at pools sampled for stable isotopes. We sampled only pools so that we could complete the sampling during daylight hours.
We conducted experiment 1 on 23 September 2009 with C. fluminea, sediment, and stream water collected from site B (Fig. 1). This experiment was designed to measure effects of C. fluminea on seston quality and quantity when sediments were present. We used 38-L aquaria (0.125 m2 bottom area) as experimental mesocosms. Three control mesocosms (sediment + water) contained sediment (1.5-cm depth) and unfiltered stream water (34.1 L). Three treatment mesocosms (C. fluminea + sediment + water) contained 4 C. fluminea (mean length = 21 mm), sediment (1.5-cm depth), and unfiltered stream water (34.1 L). Corbicula fluminea density (32 C. fluminea/m2) was at the low end of that observed in North Buffalo Creek (see Results). Based on previous work (Lauritsen 1986), the 4 C. fluminea in each mesocosm should have been able to filter the entire water column during the 12-h experiment.
We filled the mesocosms with sediments and water 36 h before initial sampling (t = 0) and addition of C. fluminea and housed them in an environmentally controlled room at 23.4°C, the temperature of North Buffalo Creek when water, sediment, and C. fluminea were collected. We added 100 mL of a solution of 1 g of 99% 13C-NaHCO3 (13C-dissolved inorganic C [DIC]) in deionized water 24 h before t = 0 to each mesocosm to enrich the seston in 13C via algal uptake of 13C-DIC. Our goal was to increase the likelihood of seeing a change in δ13C caused by removal of algal components of seston by C. fluminea during the 12-h experiment. We aerated the mesocosms via aeration tubes secured halfway up the side walls to minimize settling of seston and to prevent the water from becoming anoxic. We held the mesocosms in full light for 24 h to allow the algae to incorporate the 13C label, then turned the lights off at t = 0 and allowed the experiment to proceed in the dark. We sampled 1.2 L of water from each mesocosm without replacement at t = 0 and at 2-h intervals (in low light during sampling) until t = 12 h. We did not replace water to avoid diluting the experimental water and the isotope tracer. The final volume was 25.7 L.
We filtered samples onto precombusted glass-fiber filters to measure AFDM, stable isotope ratios of C and N, C∶N, and chl a (n = 3). The procedures for AFDM and stable isotope sample collections were the same as previously described. For the chl a analysis, we filtered 400 mL of each sample onto separate glass-fiber filters and froze the filters in a light-tight container until analysis. We used 10 mL of 90% high performance liquid chromatography (HPLC)-grade acetone to extract the chl a (APHA 1998). We were unable to obtain chl a data for experiment 1 because of spectrophotometer failure. We collected bulk sediment samples from each mesocosm at t = 0 and t = 12 h, dried and weighed them, packed them in Sn capsules, and sent the samples to the UC-Davis laboratory for stable-isotope analysis.
We ran a 2nd laboratory experiment (experiment 2) on 10 March 2010 because we found no evidence of filtering activity by C. fluminea in experiment 1 (see Results). We used a different combination of treatments (C. fluminea + sediment + water, C. fluminea + water, water only) (n = 3 replicates/treatment). The change in treatments was designed to test whether a C. fluminea filtering effect would be detected in the absence of sediment resources and whether that filtering activity affected seston quality and quantity. We included the C. fluminea + sediment + water treatment because that treatment mimicked the natural environment. In each replicate of the 2 treatments containing C. fluminea, we used 5 bivalves (mean length = 21 mm) per 38-L mesocosm and 32.1 L of unfiltered water with or without sediment (1.5-cm depth) from site B. Mesocosms in the water-only treatment contained 32.1 L of unfiltered water from site B. We expected the increase in C. fluminea density and decrease in initial water volume to enhance our ability to assess whether C. fluminea was filtering. We conducted experiment 2 in the same manner as experiment 1 except for the following: 1) 100 mL of a solution of 900 mg of 99% 13C-DIC in 900 mL of deionized water was added to each mesocosm, 2) δ13C and δ15N of C. fluminea were measured in treatments containing C. fluminea, and 3) a fluorometer was used to measure chl a concentration (Wetzel and Likens 2000). The final volume in each mesocosm was 23.7 L.
We conducted all statistical analyses in SPSS (version 16.0; IBM, Armonk, New York). We examined all data sets (each transect variable and all response variables from experiments 1 and 2) for normality with Kolmogorov–Smirnov (KS) tests. We ln(x)-transformed transect AFDM data to correct for non-normality, but no other variable required transformation. We evaluated the difference in C. fluminea density between different habitats (summer only) and between seasons (pools only) with 1-way analyses of variance (ANOVAs). We pooled replicates within sites to provide a single observation per site. We used independent-sample t-tests to compare δ15N and δ13C values of C. fluminea and seston at each sampling site (June 2009 sampling) and to compare mean δ15N and δ13C values of Corbicula between treatments in experiment 2. We used linear regression with ANOVA to test whether AFDM, δ15N, and δ13C changed with distance along the longitudinal transect (December 2009 sampling). We did not apply isotope-mixing models to C. fluminea stable-isotope data because the only potential food source sampled was seston.
We used repeated measures ANOVA with time, treatment, and the time × treatment interaction effect to test for changes in AFDM, δ15N, δ13C, C∶N, and chl a (experiment 2 only) of seston through time. We compared initial and final measures of seston response variables with t-tests for experiment 1 and 1-way ANOVA for experiment 2. In experiment 1, we used a paired t-test to evaluate the effect of C. fluminea presence on sediment δ13C and a t-test to compare sediment δ15N between treatments at t = 12. A paired t-test could not be applied to δ15N data because we dropped ½ of the initial samples from the data set because they had highly enriched values indicative of contamination with enriched 15N, which was in use in the laboratory in association with another project. We considered p-values < 0.05 significant.
In June 2009, C. fluminea density in North Buffalo Creek was 141 ± 146 (mean ± SD), 32 ± 48, and 46 ± 40 individuals (ind)/m2 in pool, riffle, and run habitats, respectively (Table 1). Density did not differ significantly among habitats. Corbicula fluminea density in pools was significantly lower in December (41 ± 35 ind/m2) than in June (t = −2.34, df = 14, p = 0.035).
Mean ±1 SD (n) Corbicula fluminea density (individuals/m2) in 3 habitats and 2 seasons in North Buffalo Creek. Riffles and runs were not sampled in December.
Seston δ15N (9.32 ± 0.43‰) was significantly greater than C. fluminea δ15N (6.80 ± 0.11‰) at site B, but site B was the only location where the seston was more 15N enriched than C. fluminea (Fig. 2A–D, Table 2). At sites C and D, δ15N of seston was significantly lower than δ15N of C. fluminea (Fig. 2C, D, Table 2). At all sites except site B, the expected shift of 3 to 5‰ N from resource to consumer was observed, but the comparison was not significant at site A.
Comparison of stable-isotope signatures between seston and Corbicula fluminea samples taken in June 2009 from North Buffalo Creek.
Seston δ13C was significantly greater than C. fluminea δ13C at all sites in June (Table 2, Fig. 3A–D). The difference in δ13C between seston and C. fluminea was greatest at site A (4.6‰) followed by sites D (2.7‰), C (2.5‰), and B (1.5‰). The δ13C and δ15N biplot (Fig. 4) for C. fluminea and seston at the 4 sites showed that the isotopic signatures did not correspond to the expected pattern based on a trophic shift from resource to consumer.
Seston AFDM declined exponentially with distance from sites B to D in December (R2 = 0.33, F1,41 = 19.495, p < 0.001; Table 3, Fig. 5A). A decline in seston AFDM was the expected pattern if C. fluminea was removing WWTP particulate organic matter from North Buffalo Creek. However, no significant linear regression model was found for changes in seston C∶N (R2 = 0.01, F1,42 = 0.434, p = 0.514; Fig. 5B). Seston δ13C decreased exponentially with distance downstream (R2 = 0.28, F1,42 = 16.138, p < 0.001; Table 3, Fig. 5C), but C. fluminea δ13C showed no clear longitudinal pattern (F1,42 = 0.688, p = 0.412; Fig. 5C). These results suggest that C. fluminea was not using seston as its primary C source. Both seston and C. fluminea δ15N increased with distance downstream (Fig. 5D). The regression model for seston δ15N was significant but weak (R2 = 0.14, F1,42 = 6.632, p = 0.014; Table 3), whereas distance downstream was a stronger predictor of C. fluminea δ15N (R2 = 0.29, F1,44 = 17.804, p < 0.001; Table 3). At each site along the stream transect, C. fluminea δ15N was 4 to 5‰ greater than seston δ15N.
Regression models for longitudinal transect data downstream of a wastewater treatment plant (in m) on North Buffalo Creek (15 December 2009). Seston and Corbicula fluminea samples were analyzed for δ15N and δ13C, and seston was further analyzed for ash-free dry mass (AFDM).
Seston AFDM did not differ between the sediment + water and C. fluminea + sediment + water treatments at t = 0 (t = −0.405, p = 0.706; Fig. 6A, Table 4) or t = 12 (t = −1.001, p = 0.374). The repeated measures ANOVA did not indicate a significant effect of time, treatment, or time × treatment over the course of the experiment (Table 5).
Comparison of seston variables at the beginning (t = 0) and end (t = 12) of experiment 1. Independent t-tests were used to compare seston ash-free dry mass (AFDM), isotopic signatures, and C∶N between Corbicula fluminea + sediment + water and sediment + water treatments.
Repeated measures analysis of variance comparing seston ash-free dry mass (AFDM), isotopic signatures, and C∶N between Corbicula fluminea + sediment + water and sediment + water treatments.
The 13C-DIC enriched the seston by ∼13‰ compared to stream seston δ13C. δ13C of the seston did not differ between the C. fluminea + sediment + water treatment and the sediment + water treatment at t = 0 or at t = 12 h (Fig. 6B, Table 4), but the repeated measures ANOVA showed a significant time effect (F1,4 = 28.212, p = 0.006; Table 5), and a significant time × treatment interaction effect (F1,4 = 9.711, p = 0.036). These results showed that seston δ13C changed differently between treatments over the course of the experiment. The ANOVA results showed that the presence of C. fluminea altered the δ13C of the seston, such that seston δ13C became more enriched through time in the C. fluminea + sediment + water treatment but did not change in the sediment + water treatment (Fig. 6B).
Seston δ15N did not differ between treatments at t = 0 or t = 12 h (t = −1.222, p = 0.287, and t = 1.692, p = 0.166, respectively; Fig. 6C, Tables 4, 5). The repeated measures ANOVA did not indicate a significant time (F1,4 = 4.18, p = 0.117), treatment (F1,4 = 0.69, p = 0.693), or interaction (F1,4 = 5.28, p = 0.083) effect on seston δ15N.
Seston C∶N increased in the sediment + water treatment over time. Initial C∶N did not differ between treatments (t = 0.443, p = 0.681; Fig 6D, Table 4), but the final C∶N differed significantly between treatments (t = 3.767, p = 0.020). The repeated measures ANOVA showed significant time (F1,4 = 8.756, p = 0.042) and treatment (F1,4 = 11.986, p = 0.026; Table 5) effects, but the interaction was not significant (F1,4 = 6.939, p = 0.058). These effects indicate that the C∶N of seston changed unless C. fluminea was present.
Bulk sediment δ15N ranged from 17.45 ± 1.21 to 19.32 ± 0.46 and did not differ between treatments (t = −1.445, p < 0.222; Fig. 7A). Sediment δ13C ranged from −25.9 ± 0.73 to −23.02 ± 1.05 and did not differ between treatments (t = −0.309, p < 0.772; Fig. 7B).
Seston AFDM did not change over time (Fig. 8A, Tables 6, 7), and initial (F2,8 = 2.074, p = 0.201) and final (F2,8 = 0.726, p = 0.522) AFDM did not differ among treatments. The repeated measures ANOVA showed no time (F1,6 = 0.024, p = 0.883), treatment (F2,6 = 1.289, p = 0.342), or interaction (F2,6 = 1.822, p = 0.241) effect. Thus, C. fluminea did not filter at a rate sufficient to change seston AFDM regardless of whether sediment was present.
Comparison of seston variables at the beginning (t = 0) and end (t = 12) of experiment 2. Analyses of variance were used to compare seston ash-free dry mass (AFDM), isotopic signatures, C∶N, and chlorophyll a (chl a) between water only, sediment + water, and Corbicula fluminea + sediment + water treatments.
Results of a repeated measures analysis of variance comparing seston ash-free dry mass (AFDM), isotopic signatures, C∶N, and chlorophyll a (chl a) for initial and final sampling events among water only, Corbicula fluminea + water, and C. fluminea + sediment + water treatments.
Initial mean seston δ13C values were −12.28, −11.55, and −9.51‰ in the water only, C. fluminea + water, and C. fluminea + sediment + water treatments, respectively (Fig. 8B). Initial (F2,8 = 0.734, p = 0.519; Table 6) and final (F2,8 = 0.138, p = 0.874) δ13C values did not differ among treatments. δ13C increased significantly over time in all treatments (F1,6 = 55.82, p < 0.001; Table 7), but the treatment (F2,6 = 0.321, p = 0.737) and interaction (F2,6 = 0.014, p = 0.986) effects were not significant. Thus, seston δ13C changed in the same way in all treatments over time.
Seston δ15N increased over the course of the experiment in all treatments (Fig. 8C, Tables 6, 7). Initial δ15N values differed among treatments (F2,8 = 6.770, p = 0.029). Seston δ15N differed between the C. fluminea + water and C. fluminea + sediment + water treatments (F2,8 = 6.770, p = 0.026). Final seston δ15N values did not differ among treatments (F2,8 = 1.674, p = 0.264). Repeated measures ANOVA showed a significant time effect (F1,6 = 76.52, p < 0.001), but the treatment (F2,6 = 4.595, p = 0.062) and interaction (F2,6 = 4.373, p = 0.071) effects were not significant. Thus, seston δ15N changed in the same way in all treatments over time.
Seston C∶N declined over time (Fig. 8D, Tables 6, 7), but neither initial (F2,8 = 0.283, p = 0.763) nor final (F2,8 = 2.728, p = 0.144) C∶N differed among treatments. The decrease in C∶N was greatest in the C. fluminea + water treatment (2.2), followed by the C. fluminea + sediment + water (1.8), and water only (1.1) treatments. Repeated measures ANOVA showed that treatment effect was not significant (F2,6 = 0.540, p = 0.609), but the time (F1,6 = 294.3, p < 0.001) and interaction (F2,6 = 10.02, p = 0.012) effects were significant, indicating that C. fluminea decreased seston C∶N relative to changes in the control treatment.
Seston chl a decreased after t = 0 in all treatments (Fig. 8E, Tables 6, 7), probably because the lights were off. Initial chl a values were highly variable and did not differ among treatments (F2,7 = 2.292, p = 0.197). However, final values differed among treatments (F2,8 = 8.864, p = 0.016). Chl a was significantly higher in the C. fluminea + sediment + water treatment (p = 0.033) than in the water only and C. fluminea + water treatments (p = 0.020). Repeated measures ANOVA showed neither a significant time (F1,6 = 3.283, p = 0.130), treatment (F2,6 = 5.428, p = 0.056), or interaction (F2,6 = 0.729, p = 0.527) effect.
Corbicula fluminea δ15N was significantly lower in the C. fluminea + water treatment than in the C. fluminea + sediment + water treatment (t = −2.074, df = 4, p = 0.050; Fig. 9A). Mean C. fluminea δ13C (n = 3) did not differ between the 2 treatments with C. fluminea (t = −1.306, df = 4, p = 0.13; Fig. 9B). Thus, the presence of sediment had a qualitative effect on C. fluminea N composition but not on C composition.
The C. fluminea + sediment + water treatment was the only treatment repeated in both experiments. We compared patterns in this treatment qualitatively between experiments. Seston AFDM changed very little and seston 13C increased over time in both experiments. The pattern of change in seston δ15N differed qualitatively between the experiments. In experiment 1, seston δ15N remained fairly consistent over time, whereas in experiment 2, seston δ15N increased over time. Initial seston δ15N values were lower in experiment 2 than in experiment 1. Values were more typical of a stream receiving WWTP effluent in experiment 1 than in experiment 2. NO3− produced from microbially processed wastewater has δ15N values that range from +7 to +20‰ (Mayer et al. 2002). Seston C∶N was affected by C. fluminea in both experiments. Proportional N content was greater in the presence than in the absence of C. fluminea.
Our study was designed to test the hypothesis that C. fluminea affects seston quality and quantity in an urban stream through its filtering activity. At densities ranging from 350 to 1400 ind/m2, C. fluminea populations were estimated to clear the water column of streams and large rivers (Lauritsen 1986, Cohen et al. 1984, Leff et al. 1990). However, our laboratory experiments did not support the hypothesis that filter feeding by C. fluminea affects seston concentration in North Buffalo Creek, although C. fluminea activity significantly affected seston quality. Experimental densities (32–40 ind/m2) were comparable to those used in previous estimates of filtering rate (Lauritsen 1986), so if filtering occurred, we should have been able to detect it.
Furthermore, our field studies showed that seston was not the primary food source for C. fluminea in North Buffalo Creek. Evaluation of the stable-isotope signatures of seston and C. fluminea at multiple sites did not support a trophic link between these foodweb components. The 15N enrichment of seston reflected a WWTP point-source input. Treated wastewater has a δ15N value ranging from +7 to + 20‰ (Mayer et al. 2002), and the δ15N of seston in the North Buffalo WWTP effluent was within that range (∼8.3‰) when measured by Ulseth and Hershey (2005). Greater 15N enrichment of seston than C. fluminea at site B during summer 2009 illustrates that seston could not have been an important food source at that site.
At the other 3 sites sampled during the summer of 2009, C. fluminea 15N was enriched compared to seston, suggesting that C. fluminea could have assimilated primarily seston N. However, C. fluminea was 13C depleted rather than enriched compared to seston, indicating that seston was not the primary C source. Data from the longitudinal transect also did not support a trophic link between seston and C. fluminea. The progressive decrease in seston δ13C indicated that processing of sewage-derived C occurred along the longitudinal transect to approach the approximate δ13C value of terrestrial leaf litter (−28‰). The observed decline in AFDM was consistent with the seston δ13C results. Corbicula fluminea was 3 to 5‰ enriched in 15N compared to seston along the transect, results commensurate with a trophic shift from seston to C. fluminea (Peterson and Fry 1987), but C. fluminea δ13C did not show a pattern comparable to that of seston. Thus, the combination of stable-isotope and seston AFDM data suggest that the seston dynamics were controlled by processes other than filter feeding by C. fluminea. These results also suggest that an alternate food source or sources that were not evaluated must have been present. Corbicula fluminea can pedal feed (Vaughn and Hakenkamp 2001), so the sediments probably served as a major food source.
Changes in seston C∶N and δ13C in experiment 1 probably were caused by C. fluminea activity in the sediments. This activity apparently suspended 13C-enriched particles (resulting from the 13C-DIC enrichment) from the sediment surface into the water column. Corbicula fluminea activity also could have resuspended 13C-enriched algae that settled from the water column during the experiment. In the absence of C. fluminea (sediment + water treatment), settling algae could have caused the observed increased C∶N (Fig. 6D) because algae and other microbes typically have lower C∶N than terrestrial detrital particles (Caraco et al. 1998, Frost et al. 2002, Dodds et al. 2004). Resuspension of settled algae by C. fluminea could maintain, but not increase, seston δ13C, so this mechanism alone cannot explain the observed differences in δ13C and C∶N. However, a combination of suspension of 13C-enriched benthic algae and resuspension of settling algae could increase seston δ13C and maintain C∶N, consistent with our results.
Experiment 2 provided additional supporting evidence that C. fluminea affects seston quality and showed that microbial processing in the water column affects seston quality independently of C. fluminea activity. The increase in seston δ15N and decrease in seston C∶N (although uneven across treatments) over time in all treatments probably was caused by microbial processing of N in the water column. The effect of microbial processing of inorganic N on seston δ15N is not well studied. However, available experimental evidence shows that activity of heterotrophic microbes increases δ15N and decreases C∶N (Caraco et al. 1998, our study).
In experiment 2, C. fluminea could have become more 15N-enriched in the C. fluminea + sediment + water treatment only by feeding on a sediment N source because seston δ15N was isotopically ∼5‰ lighter than C. fluminea. The bulk sediments were slightly enriched in 15N compared to seston in experiment 1 and >10‰ enriched compared seston in experiment 2 (Fig. 7A compared to Figs 6C and 8C, respectively). Seston δ15N at site B, where materials for experiments were collected, is temporally highly dynamic, reflecting WWTP input and runoff effects that vary with discharge (Ulseth and Hershey 2005). However, δ15N of sediment is unlikely to be as temporally variable as seston. The effect of sediment feeding on seston quality was to increase the proportion of seston N relative to C content, as shown by the proportionally greater seston N content in the C. fluminea + sediment + water treatment than in the water only treatment in experiment 2 and sediment + water treatment in experiment 1.
Deposit feeding can be an important method of feeding for juvenile bivalves (Cummings and Graf 2010). Several investigators have suggested that C. fluminea uses both suspension and deposit feeding (Yeager and Cherry 1994, Hakenkamp and Palmer 1999, Hakenkamp et al. 2001, Vaughn and Hakenkamp 2001, Cummings and Graf 2010), and that pedal feeding affects stream C dynamics via the return of buried organic matter to the water column (Hakenkamp and Palmer 1999). We could find no evidence of significant filter-feeding activity, but we did find that the sediment was an important source of N for C. fluminea and that C. fluminea activity increases proportional N content of seston at this N-enriched site.
Our study was designed to estimate how much seston N C. fluminea removes from North Buffalo Creek via filter feeding. We observed a decline in AFDM along a 7-km transect in North Buffalo Creek downstream of the WWTP, but our laboratory experiments demonstrated that the effect could not be attributed to C. fluminea. Furthermore, although C. fluminea δ15N was 4 to 5‰ enriched compared to seston δ15N along the longitudinal transect, δ13C values indicated that seston was not the primary food source for C. fluminea. Laboratory data showed that C. fluminea δ15N shifts away from that of seston when sediments are present, indicating that sediments provide a food source. Corbicula fluminea activity in the sediments also affected seston C∶N and δ13C. The most parsimonious mechanism for the observed qualitative changes in seston is that C. fluminea activity in sediments redistributes benthic algae and may also resuspend settling algae to the water column. However, further studies are needed to understand the food sources and feeding behavior of C. fluminea and the effect of its behavior on stream seston.
We thank Jennifer Repa for her help in the field and the laboratory, Mike Bullard, Kaira Wagner, Danielle Hayes, Laura Johnjulio, and Chris Jones for help gathering samples for the laboratory experiments, Matina Kalcounis-Rüppell and Parke Rublee for helpful insight into the project, Sat Gupta for statistical advice, and Mary Beebe Hall-Brown for creating a map of the study sites. We appreciate the advice and comments of 3 anonymous referees and Associate Editor Scott Larned on our manuscript. This work was supported by the University of North Carolina Greensboro Department of Biology and Graduate School and the Julia Taylor Morton Endowment.