Introduction

Rapidly urbanizing areas may represent sources of nitrogen to streams, rivers, and estuaries (Howarth et al. 1996; Grimm et al. 2008; Kaushal et al. 2008a). Both nitrogen concentration and export to rivers are strongly correlated to human population density and wastewater inputs (Peierls et al. 1991; Howarth et al. 1996). In the Chesapeake Bay region of the USA, increased nitrogen export associated with agriculture and urbanization in coastal areas has contributed to eutrophication and hypoxia, decreased plant diversity, enhanced formation of harmful algal blooms, and caused fish kills (Boesch et al. 2001; Kemp et al. 2005; Kaushal et al. 2008a).

Although human activities have greatly increased nitrogen inputs to watersheds, retention of nitrogen by headwater streams and riparian wetlands can be an important sink for anthropogenic nitrogen under a certain range of hydrologic conditions and N concentrations. Retention of nitrogen in streams can occur by several mechanisms such as temporary uptake by autotrophic algae and plants, heterotrophic uptake by microbes, storage in sediments, and also permanent removal via denitrification (Böhlke et al. 2004; Grimm et al. 2005; Kaushal et al. 2008b). Nitrate is a major form of nitrogen in many streams and rivers (Howarth et al. 1996; Groffman et al. 2004; Mayer et al. 2007), and there is considerable interest in its removal via denitrification. In situ measurements of denitrification using 15N isotopic techniques in streams have shown that denitrification can account for 15% of nitrate uptake in a forest stream with very low concentrations of nitrate (Mulholland et al. 2004), 50% of nitrate uptake in an agricultural stream with higher concentrations of nitrate (Böhlke et al. 2004), and a median of 16% of total nitrate uptake in a larger survey of 72 streams and 8 regions representing several biomes (Mulholland et al. 2008). Other studies have also recently suggested the importance of denitrification in influencing N flux in running waters (Alexander et al. 2000; Peterson et al. 2001; Seitzinger et al. 2002; Pribyl et al. 2005).

Stream degradation due to land use change may impair denitrification in streams by increasing channel incision, headwater stream burial, amplifying nitrate export during wet years, and decreasing the hydrologic “connectivity” between the stream channel, the hyporheic zone, and riparian areas (Groffman et al. 2002; Böhlke et al. 2007; Elmore and Kaushal 2008; Kaushal et al. 2008b). Furthermore, elevated nitrate loads may “saturate” the ability of headwater streams to attenuate nitrate pollution and contribute to increased nitrate uptake lengths and decreased nitrate retention in streams (Bernot and Dodds 2005; Grimm et al. 2005; Meyer et al. 2005; Mulholland et al. 2008).

There has been a growing interest in the use of stream restoration as a partial means to offset the water quality effects of urbanization and other forms of land use change as indicated by the completion and planning of over 37,000 river restoration projects in the United States (Bernhardt et al. 2005). The goals of these diverse projects are often to restore water quality, restore riparian zones, improve in-stream habitat, reduce channel erosion, promote bank stability, and protect and repair aging infrastructure (Bernhardt et al. 2005; Palmer and Bernhardt 2006; Doyle et al. 2008). Little work has been done to actually evaluate the effectiveness of different stream restoration designs and more research is necessary to understand variability in N cycling (Bernhardt et al. 2005). Decreased flow velocity, increased hydrologic “connectivity” with floodplain wetlands and hyporheic zones, and increased hydrologic residence time may influence N cycling in restored streams during baseflow (Bukaveckas 2007; Roberts et al. 2007; Kaushal et al. 2008b). There is still little data regarding denitrification rates in restored streams, however, and it is not certain how these rates may vary across different restoration designs.

A key question may be related to the efficacy of different stream restoration designs on stimulating ecosystem functions such as nitrate uptake and denitrification. Some experimental designs may choose not to strictly focus on “restoration” and “re-creation” of pre-disturbance conditions, but may attempt to improve strategies for stormwater management practices while repairing aging urban infrastructure along stream valleys (Kaushal et al. 2008b). The objectives of this study were to (1) conduct background measurements of nitrate uptake and sediment denitrification potential in two restored and two unrestored streams in Baltimore, Maryland, USA and (2) to quantify rates of in situ denitrification in a reach of restored urban stream using a 15N stable isotope addition. The study reaches of the restored streams, Minebank Run and Spring Branch, were restored primarily to repair aging infrastructure along the stream corridor (exposed sewer lines, a damaged bridge) and enhance stormwater management and dissipation of erosive force using either hydrologically “connected” floodplains or stormwater ponds and catch basins below storm drains (US EPA River Corridor and Wetland Restoration 2002; Kaushal et al. 2008b). The work described here was part of two long-term research projects: the Baltimore Ecosystem Study (BES, http://beslter.org), an urban Long-term Ecological Research (LTER) project supported by the US National Science Foundation (Groffman et al. 2004; Kaushal et al. 2005, 2008a), and a long-term study of a restored stream, Minebank Run, supported by the US Environmental Protection Agency (Mayer et al. 2003; Groffman et al. 2005; Kaushal et al. 2008b; Gift et al. 2009).

Materials and methods

Study design

Background measurements of denitrification in stream sediments (similar to Groffman et al. 2005) and nitrate uptake using reach scale nutrient additions of nitrate (following the methods described in Stream Solute Workshop 1990) were conducted in two restored streams and two degraded urban streams in the Baltimore metropolitan area from June to August of 2006. These types of background measurements were collected to provide a context for interpretation of in situ denitrification rates at Minebank Run, also conducted during the summer of 2006, using 15N isotopic tracer techniques (following the methods described in Mulholland et al. 2004, 2008). All study sites were located in Baltimore County, Maryland, USA in the Gwynns Falls and Gunpowder Falls watersheds (Fig. 1).

Fig. 1
figure 1

Map of sites located within Baltimore County, Maryland. All sites are pictured within their 12 digit watershed. Minebank Run (MNBK) and Spring Branch (SPBR) are both part of the larger Gunpowder Falls watershed while Glyndon (GLYN) and the tributary of Dead Run (DR5) are both part of the larger Gwynns Falls watershed. Baltimore City is represented by the area highlighted in gray within the Baltimore County map

Full size image

Site descriptions

Minebank Run (MNBK) is a second order stream in the Gunpowder Falls watershed, a predominantly suburban watershed within Baltimore County, Maryland (Doheny et al. 2007; Fig. 1). The Lower Gunpowder watershed is approximately 11,828 ha with 30% agricultural, 32% forested, 18% urban, 19% suburban land cover, and 1% other land use. The section of Minebank Run chosen for this study was restored in 1998 and 1999 (Mayer et al. 2003; Kaushal et al. 2008b). Average baseflow discharge in this reach of Minebank Run during the study period was approximately 3.6 L s−1 but could show peaks of 20.4 L s−1 based on precipitation events (Ed Doheny and Jon Dillow, US Geological Survey unpublished results; Doheny et al. 2006). Restoration in the stream included techniques such as installing step-pool structures designed to reduce erosion. The primary restoration manipulation, however, was a reshaping of the stream banks to create hydrologically “connected” floodplains for improved stormwater management, dissipation of erosive force, and re-establishing riparian vegetation in contrast to high armored banks and channelized structures (Mayer et al. 2003; Groffman et al. 2005; Kaushal et al. 2008b) (Fig. 2a).

Fig. 2
figure 2

a Stream reach with hydrologically “connected” floodplain at Minebank Run where experimental nitrate injection and 15N tracer experiment were performed (Photo by P. Mayer). b Series of stormwater ponds below a storm drain in the headwaters of Spring Branch (Photo by T. Newcomer). c Spring Branch downstream of stormwater ponds and catch basins where experimental nitrate injection was performed (Photo by T. Newcomer)

Full size image

Spring Branch (SPBR), a restored first order stream in Baltimore County, Maryland, drains the suburban Loch Raven watershed, eventually emptying into the Loch Raven Reservoir, a major drinking supply for the Baltimore Metropolitan area (Fig. 1). Land use for the Loch Raven, 9,437 ha, watershed was 12% agriculture, 36% forested, 14% urban, 29% suburban, and 9% other. Routine discharge measurements were not available at Spring Branch during the study period, but baseflow discharge can be 2–3 L s−1 with variations based on precipitation events (T. Newcomer and G. Stanko, unpublished results). The goal of restoration at Spring Branch, completed in 1997, was to manage stream flow to control for erosion and floods (US EPA River Corridor and Wetland Restoration 2002). The primary restoration feature included upstream stormwater ponds below the storm drain outfall, plunge pools below storm pipe outfalls, catch basins to attenuate flow, and hydrologically “connected” floodplain access for bankfull discharges (US EPA River Corridor and Wetland Restoration 2002; Gift et al. 2009). The present study reach of Spring Branch was downstream from a series of shaded headwater stormwater ponds (Fig. 2b, c).

Two degraded suburban/urban streams, which were routinely monitored as part of the National Science Foundation supported Baltimore Long-term Ecological Research (LTER) site were also studied, Glyndon and DR5. Glyndon (GLYN) is the first order headwater sub-watershed of the 19,000 ha Gwynns Falls (Groffman et al. 2004; Kaushal et al. 2005) (Fig. 1). The 4,607 ha Upper Gwynn Falls watershed that Glyndon is within consists of 7% agriculture, 24% forested, 50% urban, 17% suburban, and 1% other land cover. Average baseflow discharge in Glyndon during the study period was approximately 2.6 L s−1, but discharge can show very strong peaks of >100 L s−1 during storms (US Geological Survey, unpublished data). DR5 is a headwater tributary of the larger third order Dead Run stream in the lower Gwynns Falls watershed (Fig. 1). Land use for the 12,233 ha of the Lower Gwynns Falls watershed is 2% agriculture, 14% forested, 75% urban, 8% suburban, and 1% other. Routine discharge was not available at DR5 during the study period, but discharge can be 2–3 L s−1 during baseflow conditions and can show very strong peaks >100 L s−1 during storms (USGS, provisional data http://waterdata.usgs.gov/md/nwis/dv?referred_module=sw&site_no=01589312).

Streamwater chemistry at Minebank Run

Approximately 52 surface water samples from Minebank Run were collected routinely from 1 October 2005 to 1 October 2006. Time-series samples for nitrate concentrations were collected at USGS station 0158397925, Minebank Run at Intervale Court, Towson, Maryland, USA, since June of 2004 (further information on Minebank Run can be found at http://waterdata.usgs.gov/nwis/uv?0158397967). Samples were stored, filtered and analyzed as described below.

Denitrification enzyme assays (DEA)

Sediment samples were collected from each of the four study reaches in June 2006 to measure denitrification potential. The upper 10 cm of sediment was collected using a random sampling design from the measured center of the stream every 25 m (along the 200 m designated reach) and were refrigerated until analysis within a week of collection. Nine sediment samples were collected per each stream reach, and denitrification enzyme activity was measured using a short term denitrification enzyme assay on four replicate subsamples in the laboratory (Groffman et al. 2005). Briefly, we amended sieved sediments with NO3 −, dextrose, chloramphenicol, and acetylene, and incubated them in sealed flasks under anaerobic conditions for 90 min. We took samples from the air space of flasks at 30 and 90 min, stored them in evacuated glass tubes, and then analyzed them for N2O by electron capture gas chromatography (Groffman et al. 2005).

Stream nitrate injections

Nitrate injections were conducted once at all four sites during July and August of 2006. A solution of KNO3 − and NaBr was dripped at the upstream location of all four study reaches for 6–10 h, allowing enough time for concentrations at all stations to plateau. Concentrations of NO3 −N are typically between 1 and 2 mg L−1 in the urban streams of Baltimore (Groffman et al. 2004; Kaushal et al. 2008a), therefore the goal of the injection was to raise background stream concentrations of nitrate by at least 500 μg N L−1.

Stations were located in areas of the stream where the channel was constricted to ensure well-mixed samples at each station (LINX II 2004; Mulholland et al. 2008). Stream study reaches ranged from 74 to 212 m in length with 5 sampling stations located throughout the reach. Samples collected before nutrient additions were analyzed to determine background concentrations of nitrate and bromide in the stream at each station. Samples taken throughout the addition at each site (approximately every 30 min) were used to estimate the time of arrival of the nitrate and bromide at each station as well as to determine when concentration plateau or equilibrium had been reached at each site (Stream Solute Workshop 1990; Webster and Ehrman 1996; LINX II 2004; Mulholland et al. 2008). Samples of water were collected and transported to the University of Maryland Center for Environmental Science Appalachian Laboratory, Frostburg, Maryland for filtration and storage. Bromide analyses were performed using a Dionex 500 ion chromatograph. Analyses of nitrate, nitrite, ammonium and total nitrogen were performed on all pre and plateau samples using a Lachat Quick Chem 8000 autoanalyzer (Lachat Instruments, Milwaukee, WI, USA).

Because of low flow and difficulty in making accurate measurements with a flow meter, reach travel time, stream discharge, and average stream surface water velocity were estimated using channel measurements and analysis of the conservative tracer data. Channel measurements of wetted width and reach length were collected within the study reaches using a meter tape (Webster and Ehrman 1996). Reach travel time was calculated as the difference between the times at which the bromide breakthrough curves for the upstream and downstream stations reached the maximum rate of increase (Houser et al. 2005; Roberts et al. 2007). Average stream surface water velocity (u) was then calculated as reach length divided by reach travel time. Discharge (Q) was calculated as:

$$ Q = {\frac{{Q_{\text{pump}} \times [{\text{Br}}_{\text{inj}} ]}}{{({\text{Br}}_{\text{p}} - {\text{Br}}_{\text{b}} )}}} \, $$
(1)

where Q pump = the injection rate of the pump, Brinj = the concentration of bromide in the injection solution, Brp = the concentration of bromide at the station during plateau, Brb = the background concentration of bromide at the station (Webster and Ehrman 1996; Houser et al. 2005; Roberts et al. 2007).

Nitrate uptake length in the four stream reaches was estimated using the nutrient spiraling metrics equations (Newbold et al. 1981; Stream Solute Workshop 1990; Webster and Ehrman 1996). Plateau concentrations of both nitrate and bromide were corrected for background concentrations. Nitrate was also corrected for dilution using the ratio of nitrate to bromide (Stream Solute Workshop 1990; Webster and Ehrman 1996). Uptake length (S w) was calculated as the negative inverse slope of the regression line of the natural log of the dilution-corrected concentration of nitrate versus distance downstream (Stream Solute Workshop 1990; Grimm et al. 2005; Gücker and Pusch 2006). The slope of this line is the fractional rate of decline of the nitrate, k. Uptake rate and uptake velocity were calculated using the spiraling metric equations:

$$ S_{\text{w}} = - {\frac{1}{k}} = {\frac{QC}{Uw}} $$
(2)
$$ U = V_{\text{f}} C $$
(3)
$$ V_{\text{f}} = {\frac{Q}{{S_{\text{w}} w}}} $$
(4)

where S w = uptake length (m), U = uptake rate (μg m−2 s−1), V f = uptake velocity (mm h−1), Q = discharge (L s−1), C = concentration of the nutrient (mg L−1), and w = stream wetted width (m).

Isotope addition and denitrification in Minebank Run

15N–NO3− addition and in situ denitrification

A 15N–NO3 − isotope addition was conducted at Minebank Run from 16 to 17 August 2006, to provide a more intensive measurement of the three nutrient spiraling metrics and to quantify actual denitrification rates at the reach scale without raising the ambient concentration of nitrate in the stream. The length of the study reach was 220.5 m. A solution of 99% 15N labeled KNO3 − along with the conservative tracer Br− was dripped in the stream for 27 h to increase the δ15N of nitrate in the stream by approximately 20,000‰ (LINX 2004).

Methods were similar to LINX II (2004) and water samples were collected at each station before and at plateau for 15NO3 − analysis. Nitrate was reduced to ammonium using Devarda’s alloy and an ammonium alkaline headspace diffusion procedure was used to collect 15N from the sample for analysis (Sigman et al. 1997; LINX II 2004). Water samples were collected in plastic syringes for extraction of 15N labeled gases (Mulholland et al. 2004, 2008). Helium headspace was added to each syringe while under stream water to further guard against contamination of N2 from the atmosphere (Mulholland et al. 2004; LINX II 2004; Hamilton and Ostrom 2007; Mulholland et al. 2008). Equilibrated headspace was injected into evacuated exetainers (Labco, Buckinghamshire, England). Exetainers were stored in centrifuge tubes filled with water to prevent diffusion of N2 gas from the atmosphere (Mulholland et al. 2004; LINX II 2004; Hamilton and Ostrom 2007; Mulholland et al. 2008). Filters infused with 15N and gas samples were sent to the University of California-Davis Stable Isotope Laboratory to be analyzed for 15N/14N isotopic ratios using isotope ratio mass spectrometry, and precision was 0.3‰ for replicate analyses of reference standards (Cloern et al. 2002).

Reaeration rates and whole stream metabolism

Reaeration rate of O2 and whole stream metabolism were determined following methods described by Marzolf et al. (1994) and adapted by Young and Huryn (1998). The gas exchange rate of propane was determined by the slope of a regression of the natural log of the dilution-corrected concentration of propane versus distance downstream, and a reaeration rate (k 2) for dissolved O2, N2 and N2O was determined (Marzolf et al. 1994; Young and Huryn 1998; Mulholland et al. 2004).

15N mass flux calculations

Tracer 15N flux was calculated from measured δ15N using a series of equations developed by Mulholland et al. (2004). Briefly, δ15N values were converted to mole fraction ratios using the equation:

$$ {\frac{{{}^{15}{\text{N}}}}{{({}^{15}{\text{N}} + {}^{14}{\text{N}})}}} = {\frac{{[(\updelta {}^{15}{\text{N}}/1,000) + 1] \times 0.0036765}}{{1 + [(\updelta {}^{15}{\text{N}}/1,000) + 1] \times 0.0036765}}} $$
(5)

The 15NO3 − mass flux (15N flux i ) was then calculated as

$$ {}^{15}{\text{N flux}}_{i} = \{ {\text{MF}}_{i} \times [{\text{NO}}_{3}^{ - } -{\text{N}}] \times Q_{i} \} - \{ {\text{MF}}_{\text{bi}} \times [{\text{NO}}_{3}^{ - } - {\text{N}}] \times Q_{\text{i}} \} $$
(6)

where MF i  = the plateau mole fraction at the station, MFbi = is the background (or Pre) mole fraction at the station, [NO3 −-N] is the concentration of nitrate at the station and Q i  = stream discharge at each station. Q i was calculated from the same equation used for the nutrient additions (Eq. 1). The ln 15N flux was plotted against distance downstream to calculate fractional rate of decline of the nitrate (k total) and uptake length (S w). Uptake rate (U) and uptake velocity (V f) were then calculated similarly to the nitrate additions using Eqs. 3 and 4.

In order to determine the concentrations of 15N2 and 15N2O, the measured headspace δ15N values were first corrected for isotopic fractionation (Mulholland et al. 2004, 2008). N mass values were corrected for incomplete gas transfer into the headspace using the volumes of headspace and water for each sample along with the Bunsen coefficients for N2 and N2O at the same pressure and temperature the headspace equilibration was performed (Mulholland et al. 2004). Mole fraction values were calculated from the fractionation corrected δ15N using Eq. 5. Fluxes of 15N2 and 15N2O were then calculated using Eq. 6.

Production rates of N2 and N2O from denitrification were estimated by fitting a denitrification model, created by Mulholland et al. (2004), to the average tracer 15N flux data for N2 and N2O at each station. A least squares fitting technique was used with the model to estimate values for fractional rate of decline of the nitrate due to denitrification (k den) from the 15N mass flux data for N2 and N2O separately. Denitrification rates were also calculated as a nitrate mass removal rate per unit area for N2 and N2O using the uptake rate equation (Eq. 3) and the model predicted k den values for each (Mulholland et al. 2004). To quantify uncertainty, model simulations were run while varying the gas exchange rate of N2 and N2O (k 2) by 0.5 k 2 and 2 k 2 and by varying k den.

Statistical analysis

All statistical analyses were performed using SAS Analyst (version 9.1, SAS Institute, Cary, North Carolina, USA). Significance for all of the reported data was determined at α = 0.05. Regression analyses were used to examine the longitudinal pattern of N–NO3 − uptake within each stream and the produced slope (k) was used to calculate the uptake length (S w) for each stream. To evaluate the differences in uptake length, and the related uptake rate and uptake velocity, an analysis of covariance (ANCOVA) was used on the pooled regression data. Analysis of variance (ANOVA) followed by the Tukey–Kramer adjusted least square means test, was used to examine differences in denitrification potential and background [N–NO3 −] between sites. For comparisons of denitrification potential and uptake length, pairwise Pearson correlations were used on pooled data across all streams.

Results

Nitrate concentrations in the stream water at Minebank Run ranged between 0.76 and 1.36 mg L−1 for the months of June–August, with a mean concentration of 1.03 mg L−1 (0.12 SE) for the month of June, 1.04 mg L−1 (0.08 SE) for July, and 0.87 mg L−1 (0.04 SE) for August (Fig. 3). Nitrate concentrations varied significantly (p < 0.05) in the four streams; SPBR had the highest nitrate–N concentrations, DR5 had the lowest nitrate–N concentrations, and MNBK and GLYN were intermediate (Table 1).

Fig. 3
figure 3

Stream water nitrate concentrations at Minebank Run from 2005 to 2006. Arrow indicates date of 15N tracer experiment at Minebank Run

Full size image
Table 1 Stream chemistry and channel characteristics for study sites in the survey of restored and unrestored streams
Full size table

Denitrification rates in surface sediments of the stream bed were variable and showed no predictable pattern across sites or with stream water nitrate concentrations. MNBK, SPBR, and DR5 had significantly higher mean denitrification potential than GLYN (Table 2). DR5 had the highest mean denitrification potential in streambed sediments of the four streams and was significantly higher than GLYN and MNBK, but not SPBR. DR5 and SPBR had similar denitrification potentials in streambed sediments.

Table 2 Nitrate retention metrics and denitrification potential results of the nitrate additions for the survey study sites
Full size table

Surface water velocity measurements were low in all four streams, particularly the two restored sites MNBK and SPBR (2.1 and 2.0 cm s−1, Table 1). Travel times for the four reaches ranged from 1 to 3 h over distances of 74–212 m, with plateau at the final stations not occurring until several hours after additions began (Fig. 4a). Travel time for MNBK was especially slow, as it took 5.5 h to reach plateau at the furthest station, only 74 m from the injection site (Fig. 4a).

Fig. 4
figure 4

a An example of bromide concentrations and travel times for a nitrate addition at Minebank Run. b Relationship between stream surface water velocity, u, and nitrate uptake length, Sw, for the four urban streams; (R2 = 0.92, p < 0.05, n = 4). Restored streams (MNBK and SPBR) had the slowest stream water velocities

Full size image

MNBK had the shortest uptake length, 357 m, while DR5 had the longest, 1,341 m (Table 2), but there were no significant differences in uptake slopes (k, the fractional rate of decline of nitrate) among the four sites. There was a significant decline in the ln corrected concentration of N–NO3 − with distance downstream in DR5 and GLYN (R 2 = 0.84 and 0.98, respectively), but not in MNBK and SPBR. There was a significant relationship between uptake length (S w) and mean surface velocity (u) in the four streams (R 2 = 0.92, p < 0.05; Fig. 4b). There was no correlation between S w and Q or between S w and either background [N–NO3 −] or experimental [N–NO3 −]. There were significant correlations between background [N–NO3 −] and both U and V f.

Isotope addition and in situ measurements of denitrification

Background physical and chemical properties of Minebank Run during the isotope addition experiment are presented in Table 3. A significant longitudinal linear decrease was observed for the tracer 15NO3 − flux versus distance downstream, R 2 = 0.89, p < 0.05 (Fig. 5a). An uptake length (S w) of 556 m was calculated from the slope of this line (k tot). Uptake rate (U) was calculated to be 1.75 μg m−2 s−1 and uptake velocity (V f) was 1.80 × 10−3 mm s−1 (Table 4). When examining patterns at the first three stations, the portion of the reach also used for the nitrate additions, uptake length (S w) was shorter at 204 m (R 2 = 0.99 and k tot = 0.0049). The corresponding uptake rate (U) and uptake velocity (V f) were 5.0 μg m−2 s−1 and 4.7 mm s−1, respectively.

Table 3 Stream chemistry and channel characteristics of Minebank Run for the 15N isotope addition
Full size table
Fig. 5
figure 5

a The natural log of the tracer 15NO3− flux versus distance downstream for the 15N isotope addition at Minebank Run; (R2 = 0.89, p < 0.05, n = 7). b15N mole fraction values versus distance downstream for N2 at Minebank Run

Full size image
Table 4 Results of NO3 − uptake from the 15N isotope addition at Minebank Run
Full size table

The air water exchange rate of propane for the entire stream reach was 0.0223 m−1. This rate was then used to calculate an exchange rate (k 2) for N2, 0.0292 m−1, and N2O, 0.0286 m−1. A hump shaped curve, similar to those reported by other researchers (Mulholland et al. 2004), was found for the values of 15N mole fraction for N2 versus distance downstream (Fig. 5b). In addition, the propane addition and dissolved oxygen measurements were used to estimate a gross primary production (GPP) of 1.12 g O2 m−2 day−1 during the time of the isotope tracer addition and community respiration (CR24) of 9.26 g O2 m−2 day−1 (calculated similarly to Marzolf et al. 1994; Young and Huryn 1998). Net ecosystem metabolism (NEM), calculated as community respiration (CR24) subtracted from GPP, NEM = GPP − CR24 (Marzolf et al. 1994; Young and Huryn 1998; Meyer et al. 2005) was −8.14 g O2 m−2 day−1, indicating that metabolism during the isotope addition was dominated by heterotrophic activity (which was conducive to denitrification) in the subsurface hyporheic zone versus autotrophic activity in the stream channel.

Denitrification rates were determined using the Mulholland et al. (2004) denitrification model, which was separately fitted to the 15N flux data for both N2 and N2O. Due to detection problems with N2O, the model was fit to data from only the first 50 m of the reach for N2O. The best fit k den for N2 production was 0.0016, 89% of the k total (0.0018). N2O production was considerably less and the best fit k den was 1.96 × 10−4, approximately 11% of k den. The mass flux rates of N2 and N2O production per unit area (U den-N2 and U den-N2O) were 136.6 and 16.5 mg N m−2 day−1, respectively, totaling 153 mg N m−2 day−1 removed by denitrification (Table 4). When adjusting values of the gas exchange rates (k 2) and k den for uncertainty analysis, k den for N2 production ranged from 0.0010 to 0.0025 m−1, and ranged from 1.10 × 10−4 to 2.95 × 10−4 m−1 for N2O.

Discussion

Our results suggest that nitrate uptake and denitrification processes in sediments and at the stream reach scale may be important in influencing nitrate dynamics in some unrestored and restored streams in urban Maryland, USA, and that hydrologic residence time during baseflow may be an important variable. Previous work at some of the present study sites has shown that low-density suburban catchments export total nitrogen and nitrate loads mostly at relatively low flows, whereas more urbanized sites export total nitrogen and nitrate at higher and less frequent flows (Shields et al. 2008). Therefore, variations in export distribution should be considered when targeting stream channel restoration efforts at the watershed scale (Shields et al. 2008) and when considering variability in the efficacy of different stream restoration designs.

We investigated the importance of nitrogen transformations in several different ways including denitrification enzyme assays, nitrate solute injections, and use of a 15NO3 − tracer injection at one of the sites. Denitrification enzyme activity from surface sediments in the stream bottom showed no obvious patterns among streams and was highly variable. Nonetheless, results showed that substantial denitrification in surface sediments could be occurring in Minebank Run, Spring Branch, and the tributary of Dead Run at rates similar to ranges reported for urban streams in Baltimore (Groffman et al. 2005; Gift et al. 2009). Urban streams may have sediment microzones with low oxygen and high organic matter content that function as “hot spots” of denitrification (McClain et al. 2003). Denitrification studies in stream sediments have found rates to be highly variable and “patchy” both within and among surface sediments in streams contributing to the variability observed in the present study (Groffman et al. 2005). For example, previous work in streams of the present study have shown significant variations in denitrification rates in debris dams, pools, riffles, vegetated gravel bars, and mucky gravel bars (Groffman et al. 2005). More research is needed regarding the effects of restoration and watershed land use on the spatial distribution of stream benthic habitats and implications for N processing.

Nitrate uptake in Minebank Run relative to the other streams suggested that travel times can be long in both the restored and degraded urban streams during summer baseflow conditions as indicated by the long time for conservative tracers to reach a stable plateau after beginning continuous additions. We did not observe a strong relationship between uptake length, S w, and discharge, Q, as is often found in stream nutrient addition studies (Peterson et al. 2001; Grimm et al. 2005). We did observe an increase in uptake length (S w) with increasing surface water velocity (u), however. Interpretation of these results should be considered with caution given the small number of measurements of uptake length. Given the restoration strategies of hydrologically “connected” floodplains at Minebank Run and upstream flow through a series of stormwater ponds at Spring Branch, it is not surprising that restoration could alter flow conditions. Minebank Run and Spring Branch, the two restored streams, had the lowest surface water velocities as well as the shortest uptake lengths of the four sites, but there was no significant difference in other uptake metrics between restored and degraded streams. Increasing hydrologic contact with sediments may influence N cycling under certain stream restoration conditions (Kasahara and Hill 2006; Bukaveckas 2007; Roberts et al. 2007; Kaushal et al. 2008b; Craig et al. 2008). For example, previous work at Minebank Run has shown a strong relationship between mass removal of nitrate and hydrologic residence time (Kaushal et al. 2008b), and integrating restoration approaches with stormwater management in uplands, watersheds, and stream valleys to slow the flow of water may be critical to maximizing N removal (Kaushal et al. 2008a).

Although many studies have focused on reporting uptake length, S w, there can be variability in S w among streams due to both biogeochemical and hydrogeomorphic effects (Doyle et al. 2003). Therefore, there is also an interest in uptake velocity, V f, the rate at which a nutrient travels vertically before being taken up by the stream bed, because it normalizes for the effects of stream hydrogeomorphology and is more strongly related to biogeochemical changes (Davis and Minshall 1999; Doyle et al. 2003). Our uptake velocity values were lower than those typically reported (Davis and Minshall 1999; Grimm et al. 2005; Gücker and Pusch 2006) and were lower than the average V f value for first order streams (168 mm h−1) determined in a comprehensive review of 52 reported literature values by Ensign and Doyle (2006). Uptake rate and length, when compared to other streams, would indicate that Minebank Run is highly retentive, yet uptake velocity indicated otherwise. We speculate that low discharges in Minebank Run and the three other study sites during baseflow conditions may have contributed to the estimates of low uptake velocity when compared to other studies. V f represents the vertical velocity of the movement of nutrient molecules through the water column towards the benthos (Ensign and Doyle 2006). Another way of expressing this metric may be u × h ÷ S w, where u = stream velocity, h = depth, and S w = uptake length (Ensign and Doyle 2006). Therefore, very low values of stream velocity in the present study could have contributed to low estimations of V f.

Nitrate uptake length measured from the 15N isotope tracer addition was 556 m in Minebank Run. When comparing our isotope spiraling metric results to other studies that used the 15N isotope tracer technique, uptake length for Minebank Run was longer than the forested Walker Branch (35.7 m) studied by Mulholland et al. (2004) and the arid desert Agua Fria River (36 m) studied by Grimm et al. (2005) but shorter than the agricultural streams (800–9,600 m) studied by Bernot et al. (2006). Minebank Run was similar to and within the mid-range for the three urban arid streams also studied by Grimm et al. (2005). Surprisingly, we found no relationship between nitrate uptake and background nitrate concentrations at Minebank Run and the other three stream sites from unenriched additions. This may have been due to small sample size, small range in background concentrations, or the importance of hydrologic factors during baseflow conditions.

According to the Mulholland et al. (2004) model, the best fit fractional NO3 − removal rate due to denitrification (k den) was 0.0018 m−1, with an uncertainty of approximately ±0.001 m−1. Thus, the model suggests 100% of uptake was explained by denitrification, with the uncertainty being ±40%. Production of N2 made up approximately 89% of denitrification while N2O production made up the remaining 11%. The uncertainty was relatively high and may have been due to analytical factors such as incomplete gas transfer (although these samples were adjusted based on equations by Mulholland et al. 2004).

Despite the large uncertainty, denitrification still appeared to comprise a majority of NO3 − uptake within Minebank Run during the isotope addition. The denitrification rate for Minebank Run from the isotope addition, expressed as removal per unit of area, was 153 mg N m−2 day−1. Even at our highest uncertainty estimate, denitrification in Minebank Run comprised up to 60% of the NO3 − uptake within the reach and could potentially account for removal of 40% of the daily nitrate load during summer over the 220.5 m study reach. The importance of denitrification at Minebank Run is supported by results from the propane addition and dissolved oxygen measurements which allowed us to estimate a gross primary production. Results from gross primary production estimates indicated that the metabolism during the isotope addition was dominated by heterotrophic activity (e.g., denitrification) in the subsurface hyporheic zone versus autotrophic activity in the stream channel. This reach of Minebank Run has a densely shaded riparian canopy that covers the stream and reduces light penetration. Therefore, denitrification may have been the predominant process controlling N uptake due to riparian canopy shading (Mulholland et al. 2008).

Denitrification enzyme activity (DEA) in the surface sediments and in situ denitrification rates using the 15N tracer approach both showed that denitrification was occurring at Minebank Run. The two measurements use different methodologies, and their measurements target different locations (i.e., DEA measured potential denitrification activity in surface sediments as opposed to in situ denitrification at the stream reach scale using the 15N addition). Although denitrification rates in surface sediments could have been important, results from the 15N tracer addition also suggest that subsurface hydrologic flowpaths may have also been important in influencing denitrification at Minebank Run (Kaushal et al. 2008b). In situ denitrification at Minebank Run was within the range reported elsewhere using similar methodology (Mulholland et al. 2004, 2008; Böhlke et al. 2004). Recently, Mulholland et al. (2008) reported that denitrification accounted for a median of 16% of nitrate uptake in a large study of 72 streams across biomes. Other studies have shown that denitrification accounted for 16% of uptake within Walker Branch, a forested stream in Tennessee (Mulholland et al. 2004), and 52% of uptake within Sugar Creek, an agricultural stream in Indiana (Böhlke et al. 2004).

Despite the relatively high denitrification rates during baseflow conditions at Minebank Run, the rate of N loading in urban streams can be larger than biotic demand, thus exceeding the potential for removal via denitrification (Mulholland et al. 2008). Furthermore, high flow conditions during “flashy” storm events may greatly compromise urban stream and watershed retention processes (Booth 2005; Walsh et al. 2005). Stormwater management strategies that increase hydrologic residence time in watersheds (uplands, wetlands, and streams) may be extremely critical for managing N during high flow conditions (Booth 2005; Walsh et al. 2005; Kaushal et al. 2008a). Our isotopic study was conducted in the summer under baseflow conditions and high temperature, which are optimal conditions for maximum denitrification. Denitrification performance across differing restoration designs and hydrologic conditions is likely to be highly variable and contain large uncertainties. Because there can be large variability in the effectiveness of different designs (Kaushal et al. 2008b), stream restoration by itself is currently not adequate to mitigate for excess N inputs or to compensate for stream destruction and degradation. Certain forms of stream restoration aimed at increasing hydrologic residence time and hydrologic “connectivity” with denitrification “hot spots” may have the potential to complement comprehensive watershed management strategies when also focusing on reducing effective impervious surface coverage in uplands (Walsh et al. 2005) and repairing aging infrastructure (Doyle et al. 2008). This is particularly important in Baltimore, Maryland, where many restorations have been conducted while repairing urban infrastructure such as sewer lines, and watersheds were urbanized before current stormwater practices could be implemented. Future studies should investigate the effects of varying nitrate loads, hydrologic residence times, hydrological “connectivity,” and seasonality on denitrification rates in restored streams. Future studies should also include the use of models, such as the OTIS and OTIS-P models (Gooseff et al. 2003; Runkel 2007), which are transport-based, time-series approaches that provide estimates of U and V f independent of hydrologic effects, allowing for separation of hydrologic and non-hydrologic processes. Finally, more studies are necessary to evaluate the efficacy of various stream restoration designs on denitrification rates under different land uses and stream flow conditions (Tague et al. 2008).

In conclusion, stream restoration approaches that increase hydrologic “connectivity” with hyporheic sediments and increase hydrologic residence times may influence denitrification at the stream reach scale. Although stream restorations are likely to continue or grow in the future (Bernhardt et al. 2005; Palmer and Bernhardt 2006), very little work has been done on measuring nitrogen uptake and denitrification in restored streams (Bukaveckas 2007; Roberts et al. 2007; Kaushal et al. 2008b). Stream restoration is still a highly experimental field and designs that improve biogeochemical function may not be necessarily related to recreating pre-disturbance conditions (Palmer 2009). More empirical research combining water quality monitoring and measurements of ecosystem functions is necessary across different sites and under varying hydrologic conditions to evaluate the effectiveness of different stream restoration designs.